1 LISTENER AND SPEAKER EFFECTS ON DOMINANT LANGUAGE PERCEPTION AND LANGUAGE RATINGS AMONG HERITAGE SPEAKERS IN NEW MEXICO By VALERIE J. TRUJILLO A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA I N PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Valerie J. Trujillo
3 To Don, Michael and Lillian
4 ACKNOWLEDGMENTS I would like to thank my dissertation committee chair, Dr. Gillian Lord for her direction and guidance in this study, as well as for serving as my wonderful thoughtful mentor for the past few years. I would also like to thank the members of my dissertation committee, past and present for their input and direction : Dr. Jason Rothman, Dr. Jessi Aaron, Dr. Ana de Prada Prez, Dr. Caroli n e Wiltshire, and Dr. Diana Boxer In addition, I would like to express my gratitude to the University of Florida Department o f Spanish and Portuguese Studies for granting me the 2011 Doctoral Student Summer Scholarship Award and in so doing, ensuring that I recruited a great number of quality participants. I am indebted to my former professors who have influenced me greatly in m y graduate work: Dr. Mara Dolores Gonzales, Dr. Enrique Lamadrid, and Dr. Tey Diana Rebolledo. I wish to acknowledge my grandmother, Adelecia Gallegos, without whom, through a twist of fate, I would have never entered the field of academia and research. I would also like to thank my fa mily for their help and support; especially my parents, Joe and Juliana for helping me with my data entry for this and many other projects, and my sister Lisa for helping me meet potential participants for this and other stu dies. I would like to thank my wonderful children, Michael and Lillian, for the sacrifices they have made for me in their few years and for inspiring me each and every day. Finally, but most importantly, I would like to thank my amazing husband Don for his support and encouragement, and for pushing me, pulling me, and carrying me towards the finish line when necessary; and for teaching me, once and for all, how to eat an elephant sandwich.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TA BLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 10 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 13 General Overview ................................ ................................ ................................ ... 13 The Research Problem ................................ ................................ ........................... 15 Motivation of the Study ................................ ................................ ..................... 22 Significance of the Study ................................ ................................ .................. 24 Orga nization ................................ ................................ ................................ ........... 29 2 BACKGROUND INFORMATION AND PRIOR RESEARCH ................................ .. 32 The Linguistic Situation in New Mexico ................................ ................................ ... 32 Historical Background ................................ ................................ ...................... 32 Language Policies that Shaped NMS ................................ ............................... 35 Language Attitudes ................................ ................................ .......................... 37 Characteristics of NMS ................................ ................................ ..................... 42 Research on Native Language Perception ................................ ............................. 44 Her itage Speakers ................................ ................................ ................................ .. 58 Heritage Language Phonology ................................ ................................ ......... 61 Heritage Speakers and Gender Agreement ................................ ..................... 64 Summary ................................ ................................ ................................ ................ 68 3 METHODS ................................ ................................ ................................ .............. 71 Research Questions and Hypotheses ................................ ................................ ..... 71 Research Design ................................ ................................ ................................ .... 76 Participants ................................ ................................ ................................ ....... 76 Language background questionnaire ................................ ......................... 80 Independent measure of proficiency ................................ .......................... 81 Tasks and Materials ................................ ................................ ......................... 83 Audio sample ................................ ................................ ............................. 83 Rating sheet ................................ ................................ ............................... 93 Language attitudes assessment ................................ ................................ 96
6 Data Analysis ................................ ................................ ................................ ... 98 Factors in assessing language dominance and accent ratings .................. 98 Correlation between region of origin perception and accent ratings ........ 101 Correlation between accent rating and language attitudes ...................... 102 Summary ................................ ................................ ................................ .............. 103 4 R ESULTS ................................ ................................ ................................ ............. 104 Quantitative Results ................................ ................................ .............................. 104 Factors Involved in Language Dominance Identification ................................ 104 Factors Involved in Assigning Language Ratings ................................ ........... 115 Language Ratings and Perceived Variety of Spanish ................................ .... 124 Language Ratings and Language Attitudes Towards NMS ............................ 132 Qualitative Results ................................ ................................ ................................ 143 Rating Sheet ................................ ................................ ................................ ... 143 Language Attitudes Assessment ................................ ................................ .... 155 Summary ................................ ................................ ................................ .............. 160 5 DISCUSSION AND CONCLUSIONS ................................ ................................ .... 164 Speaker Effects of Linking and Gender Agreement ................................ .............. 164 Research Question 1 ................................ ................................ ...................... 164 Research Question 2 ................................ ................................ ...................... 166 Qualitative Analysis ................................ ................................ ........................ 168 Implications ................................ ................................ ................................ .... 169 Listener Effects of Language Attitudes and Perceived Region of Origin ............... 169 Research Question 3 ................................ ................................ ...................... 169 Research Question 4 ................................ ................................ ...................... 171 Qualitative Analysis ................................ ................................ ........................ 172 Implications ................................ ................................ ................................ .... 175 Limitations and Directi ons for Future Research ................................ .................... 176 Summary ................................ ................................ ................................ .............. 178 APPENDIX A LANGUAGE BACKGROUND QUESTIONNAIRE ................................ ................. 182 B ELICITATION TASK ................................ ................................ ............................. 185 C RATING SHEET ................................ ................................ ................................ ... 188 D LANGUAGE ATTITUDES ASSESSMENT ................................ ............................ 189 E PARAMETER ESTIMATES TABLES FROM SPSS FOR MULTINOMIAL LOGISTIC MODEL (RQ1) ................................ ................................ ..................... 191 F RATING SHEET ANSWERS ................................ ................................ ................ 194
7 G COMMENTS PROVIDED ON LANGUAGE ATTITUDES ASSESSMENT ............ 200 LIST OF REFERENCES ................................ ................................ ............................. 203 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 222
8 LIST OF TABLES Table page 3 1 Distribution of speakers within audio sample. ................................ ..................... 86 3 2 Sentenc e variants in audio sample including distracter variants ........................ 87 3 3 Sentence variants in audio sample ................................ ................................ ..... 98 4 1 Sentence variants in au dio sample ................................ ................................ ... 105 4 2 Odds ratios and likely outcome categories ................................ ....................... 106 4 3 Estimate threshold s for condition and listener proficiency level. ....................... 115 4 4 Logit regression coefficients and likely outcome rating categories ................... 117 4 5 rating categories ................................ 119 4 6 Response category totals for low proficiency distractor sentences. ................. 123 4 7 Estimate thresholds for pe rceived region of origin, listener proficiency level .... 125 4 8 Logit regression coefficients and likely outcome rating categories ................... 126 4 9 Total n of LAVs rendered by proficiency level group ................................ ........ 13 3 4 10 Estimate thresholds for perceived region of origin, LAV, & proficiency level .... 134 4 11 Logit regression coefficients and likely outcome rating categories ................... 135 4 12 Sample of comments provided on rating sheet ................................ ................. 145 4 13 Percentage of answers provided on rating sheet per category. ........................ 146 4 14 Rating sheet comments referencing delivery of speech ................................ ... 151
9 LI ST OF FIGURES Figure page 3 1 Sample slide of PowerPoint presentation as presented to listeners ................... 92 4 1 Final logit re gression coefficient calculation formula with interaction ................ 116 4 2 Final logit regression coefficient calculation formula no interaction .................. 134 4 3 Types of open ended answers provided on rating sheet ................................ .. 148
10 LIST OF ABBREVIATIONS LAV Language attitudes value L1 First language L2 Second language NMS New Mexico Spanish RQ Research question SHL Spanish as a H eritage Language TB Transitional bilingual VOT Voice onset time
11 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 LISTE NER AND SPEAKER EFFECTS ON DOMINANT LANGUAGE PERCEPTION AND LANGUAGE RATINGS AMONG HERITAGE SPEAKERS IN NEW MEXICO By Valerie J. Trujillo August 2013 Chair: Gillian Lord Major: Romance Languages This study evaluates the effect of the presence or absence of connected speech gender agreement on the belief that a speaker is Spanish dominant, and the effect of these variables In addition, this study examines the rel ationship between these language ratings and the perceived region of origin of the speaker and also examines the relationship between language attitudes toward New Mexico Spanish (NMS) and language ratings Three listener groups were examined: second gene ration heritage speakers of high Spanish proficiency, third generation heritage speakers of intermediate Spanish proficiency, and third generation heritage speakers of low Spanish proficiency. Participants listened to a speech sample of read aloud speech spoken by four speakers, with stimulus sentences containing tokens of +/ connected speech and +/ gender agreement Participants were then asked to speculate on the dominant language and region of origin of the speaker and provide a rating of that speaker Spanish. In addition, pa rticipants completed a language attitudes questionnaire to assess their attitude toward NMS.
12 Logistic regression analyses found that only the high proficiency group yielded results that were consistent with the presence or absenc e of the variables in question, demonstrating that speakers must employ both connected speech and gender agreement to be believed to be Spanish highly proficient heritage speakers of Spanish. Intermediate and low proficiency listeners appeared to base their language ratings on factors external to this study. With regard to the relationship between language ratings and the perceived region of origin of the speaker, intermediate and high proficiency listeners parti cularly those who demonstrated less positive attitudes toward NMS were found to be more likely to rate Spanish speaking country than if they believed the speaker was fro m New Mexico. This study also provides a qualitative analysis of the answers provided by participants on the language rating sheet an d on the language attitudes questionnaire in order to contextualize the results of the quantitative analyses.
13 CHAPTER 1 I NTRODUCTION General Overview The processes by which speech is deemed as non native by an interlocutor are complex and not fully understood ; however they have been explored extensively within the literature of foreign accent perception and fluency perceptio n. While the majority of research focuses on the judgments of native and second language (L2) speakers ( Alba Salas, 2004; Buell Hill, 2001; Derwing, Rossiter, Munro, & Thompson, 2004; Flege, 1984; Leiken, Ibrahim, Eviatar, & Sapir, 2009; Munro, Derwing & M orton, 2006; Neufeld, 1980; Scales, Wennerstrom, Richard & Wu, 2006 ), only a handful of known studies have investigated heritage speaker performance in assessing foreign accent (Au, Oh, Knightly, Jun, & Romo, 2008; Oh, Jun, Knightly, & Au, 2003 ; Park, 2009 ). T he study of heritage speaker performance with regard to phonological perception is an important point of inquiry within the field of heritage language studies with implications for in group affiliation and language identity of the speaker. For example, heritage speakers who display a lower phonological competence may view themselves or be viewed by others Moreover, the current study will provide information on heritage speaker populations with regard to which languag e features, whether phonological or morphosyntactic, prove to be more salient to heritage speakers of various proficiency levels In addition to providing information on heritage speaker processes, the current study will also inform other populations as we ll. Performance by heritage speakers on the perception tasks in this study will provide a better understanding of how accent judgment is generally rendered by listeners and the tasks completed by heritage speaker participants in this study can be applied t o native and L2
14 speakers of Spanish in future studies. For these reason s the study of the listener and speaker effects that contribute to heritage speaker production and perception of accent is of particular interest to the field of heritage language stud ies as well as to the field of foreign accent and language perception. This study investigates phonological and morphosyntactic factors involved in assigning language ratings and in determining the languag e dominance of a given speaker, and evaluates whet her listeners provide judgments in accordance with the factors under investigation The primary linguistic factors under investigation are syllabic restructuring and errors in article noun gender agreement, which are examined in conjunction w ith language attitudes and language variety. Three listener groups are examined: second generation heritage speakers of high Spanish proficiency, third generation heritage speakers of intermediate Spanish proficiency, and third generation heritage speaker s of low Spanish proficiency. Two broad questions form the basis of this study: 1. How are the speaker effects of linking and the presence or absence of errors in article noun gender agreement related to language ratings and to the belief that a speaker is S panish dominant? 2. How are the listener effects of positive or negative language attitudes toward New Mexico Spanish (NMS) associated with the ratings these listeners assign to speakers they believe to be speakers of NMS and of other varieties? These quest ions are of particular interest because they may provide information on whether heritage speakers of varying proficiency levels render language judgments that are more accurately in accordance with linked speech than gender agreement, as expected, which wo uld indicate that listeners perceive a speaker to be Spanish dominant or to have excellent Spanish as long as their speech is fluid, regardless of the presence or absence of gender agreement.
15 This chapter presents the research problem and the motivating fo rces behind this study, as well as the significance of the study and its potential contributions to the fields of foreign language perception, heritage language development, and studies on NMS. It then concludes with a description of the organization of th e dissertation. The Research Problem The current study builds upon the existing knowledge base on the perceptive phonological competence of heritage speakers and focuses on the role of linking in comparison with the acceptability of gender agreement errors among the aforementioned groups of heritage speakers of Spanish In other words the study seeks to determine if heritage listeners will perceive a speaker to be Spanish dominant as long as his/her speech is presented in fluid, connected sentences, even i f the speech contains errors in gender agreement Likewise, it investigates if heritage listeners will assign higher ratings on a four point scale (excellent, good, fair, or poor) to speech that is fluid but contains gender agreement errors than to speech that is not fluid but contains no gender agreement errors This study also explores the nature of the relationship between language attitudes toward NMS and language ratings by asking if listeners who demonstrate positive attitudes toward NMS assign higher language ratings to speakers they believe to be from New Mexico and/or if they assign lower ratings to speakers they believe to be from outside their speech community Additionally we seek to understand if listeners of differing proficiency levels revea l similar tendencies in their assessment of speech samples The answers to these questions will not only shed light on the role of the specific variables in question but will also help lead to a greater overall understanding of the processes by which herit age speakers of varying generations and proficiency levels assess language dominance and
16 performance of a speaker The practical application of what this study offers is an insight into whether speakers across proficiency levels and language experiences ut ilize similar mechanisms in assessing the speech of others by way of universal perception tendencies or whether language experience accounts for not only what listeners perceive but also how they perceive it. Furthermore, i f there are differences found amo ng participants of varying generations, this study will serve to lend support to the argument that heritage speakers are not a homogenous group but rather a group whose many facets, among them generation, must be taken into account in any program aiming to serve this population. F luency has been notoriously difficult to define and operationalize as evidenced by studies that seek to explore and define these three measures (Kormos & Dnes, 2004; Norris & Ortega 2009; Riggenbach, 1991; Sakuragi, 2011; Sheppar d, 2004; Skehan, 2009) Traditionally scholars have operationalized fluency by invoking varying combinations of complexity and accuracy, ranging from speech rate and mean length of utterance to phonation time ratio and the number of ( stressed ) words produc ed per minute ( e.g., Ejzenberg, R., 2000; Hieke, 1984; Kormos & Dnes, 2004; Riggenbach, 1991; Wenners tro m, 2000). As will be described below, the present study examines one aspect of fluency known as connected speech or, alternatively, linking (Hieke, 198 4, 1985; Kormos & Dnes, 2004; Riazantseva, 2001; Riggenbach, 1991; Vanderplank, 1993) It also adopts a measure of accuracy by examining article noun gender agreement. Both of t hese variables form part of the complexity, accuracy, and fluency (CAF) constr uct ( Sk ehan, 1989) which have figured as major research variables in contemporary applied linguistic research on SLA and L2 pedagogy (Housen & Kuiken,
17 2009; Kormos & Dnes, 2004; Larsen Freeman, 2006; Norris & Ortega, 2003; Towell, 2007; Van Daele et al, 2007). 1994, p. 27). Linking occurs anytime the coda of one syllable is followed by an aton ic vowel onset and is re syllabified as the onset of the following syllable. In this study, tokens of linking are restricted to instances of the coda [s] followed by an onset of [a] across plural article noun word boundaries. An example of this is the phra los animal e s which would be syllabified in independent words as /los.a.ni.ma.les/, is resyllabified as /lo.sa.ni.ma.les/ in dynamic speech to produce a more fluid and unified phrasal unit. Linking was selected as a variable in the present study on the basis of feedback from participants of a pilot study, many of whom indicated whether or not 1 to be a factor that led them to identify speakers as either native or non native speakers of Spanish. In addition, the pilot study showed linking to be the only phonological factor 2 to have a high correlation with accent ratings among all four listener groups studied, thus suggesting that this particular feature merited further study. In order to provide a point of comparison for the factor of linking a factor within the domain of grammatical accuracy was sought since segmental phonological factors were shown to have weak or negligible correlation with language ratings in the pilot study. Alt to any number of 1 Examp les of answers given to an open w 2 Linking, although itself not a segmental factor, was studied among various segmental phonological factors. Temporal factors of fluency such as speech rate were not included in the pilot study.
18 prescriptive syntactic rules ( Hammerly, 1991; Housen & Kuiken, 2009 ), it is operationalized in the present study specifically as accuracy in article noun gender agreement. Studies have sh own that gender agreement is a grammatical domain that is largely mastered by around age 3 in Spanish monolingual children (Hernndez Pina, 1984; Prez Pereira, 1991) but causes persistent difficulty in adult L2 acquisition (Fernndez, 1999; Franceschina 2005; Hawkins & Franceschina, 2004 ; McCarthy, 2008; Montrul, 2004; Montrul, Foote, Perpin, 2008), thus suggesting that the presence of gender agreement errors within the current speech sample may serve as an indicator of non native speech. More pertinent to the purpose of the present study than the difficulties non native speaker s experience in the acquisition and production of this structure is the fact that at the level of language processing, gender agreement errors have been shown to elicit involunta ry brain signatures not only in L1 Spanish speakers but also in L2 Spanish speakers of high, intermediate, and low Spanish proficiency in electroph ysiological experiments (Gabriele, Fiorentino, & Alemn Ban, 2013). This result suggests that gender agreem ent violations are perceptible enough to cause a parsing failure for grammatical detection and brings about the question of the extent to which these agreement violations are perceptible for various proficiency levels; a question the current study aims to address. T he selection of gender agreement as the grammatical comparison factor in the present study also maintain s consistency with the factors studied in Au et al. (2002 2008 ) performance was measured in com parison with their phonological performance with regard to perception and production. The researchers chose to study the areas of phonology and morphosyntax because these aspects of language seem easy for
19 children to acquire and difficult for adults to mas ter, thus suggesting they are good candidates for revealing long term effects of overhearing a language during childhood (Au et al. 2002, 2008). Evaluating accuracy and identifying errors can be a thorny issue in any syntactic area, for example in determ ining whether criteria should be tuned to prescriptive standard norms as embodied by an ideal native speaker of the target language or to non standard usages acceptable in some social contexts or in some communities (Ellis, 2008; James, 1998; Polio, 1997). However, although s peakers of NMS have been shown to display variation in the use of determiners and gender assignment to English origin nouns in code switching discourse ( Clegg & Waltermire, 2009; Torres Cacoullos & Aaron, 2003), there is no documentatio n known to this author on variation in gender agreement in Spanish origin nouns in New Mexico. Furthermore, there is no known documentation on linking in NMS. NMS and the langua ge ratings they assign to speakers. One of the most salient aspects of speech is accent either dialectal differences attributable to region or class, or phonological variations resulting from L1 influence on the L2 (Derwing & Munro, 2009). Phonological v ariation has been shown to correlate with peer culture affiliation and in group identity (Derwing & Munro, 2009; Gatbonton, Trofimovich, & Magid, 2005; Slomanson & Newman, 2004) and may trigger positive or negative language attitudes toward the speaker and the variety of speech being used (Galindo, 1995; Gatbonton et al. 2005; Ryan, Carranza, & Moffie, 1977; Scales et al., 2006). The relationship between foreign accent perception and language attitudes has been studied at length and
20 indicates an intricate b ond between the two (Brennan & Brennan, 1981; Scales et al., 2006; Smit, 1996; Zuengler, 1988). Although previous studies on this relationship have viewed accent as an impetus to language attitudes and utilized matched guise techniques to elicit language a ttitudes toward stimuli (Galindo, 1995; Ryan & Carranza, 1977; Smit, 1996), the current study views the two as contemporaneous, and examines whether existing attitudes toward one variety of Spanish (NMS) are a predictive factor or otherwise associated with language ratings towards individuals believed to be speakers of this variety. Early research on evaluative reactions to language has distinguished two major dimensions in observing bilingual communities: status and solidarity When two stable language var ieties are present in a community, the standard speech variety is usually associated with status and influence, while the second speech variety is the language of friendship, intimacy, and solidarity, with which listeners are able to distinguish in group a ffiliation (Ryan & Carranza, 1977). The concept of in group affiliation is of particular interest to the current study, given that participants will be asked to identify speech as sh (NMS in group ); 2) a speaker of other native varieties of Spanish (non NMS out of group ); or 3) a non native speaker of Spanish (out of group ) In addition, participants will be asked to provide information on their attitudes toward: 1) their own vari ety of Spanish (NMS) and 2) all native varieties of Spanish that are not their own (non NMS), as a singular group. Grouping all non NMS varieties as a singular group ensure s a binary classification of language attitudes and allows juxtaposition between att itudes toward NMS versus a non NMS generic category to exemplify in group and out of group distinction. Such
21 grouping also refrains from assuming the participants have adequate experience with other varieties of Spanish needed to evaluate individual variet ies independently. Further details on the methodology will be given in Chapter 3 As stated previously the participants for this study are grouped by both operationalized as Mills and Villa (2009a) as the distance from contact generation ancestors, regardless of place of birth, wherein the contact generation is monolingual in Spanish and comes into contact with English speakers aft er the age of 15. This notion of linguistic generation is distinct from that of a biological generation and was designed by Rivera Mills and Villa specifically to accommodate the complex linguistic environment of the Southwest United States, which will be discussed in further detail in Chapter 2. The participant groups for this study consist of second and third generation heritage speakers of Spanish : the second generation participants have at least one contact generation parent, and the third generation p articipants have at least one contact generation grandparent. An understanding of the participant groups as determined by linguistic generation and proficiency level is important given that the participant group constitutes an independent variable in this study. It is predicted that there will be no difference in the results across all three participant groups. However, a finding contrary to this hypothesis would provide information on differences that may exist among the groups with regard to perceptually salient features. For example, gender agreement may prove to be more perceptually salient to second generation, high proficiency participants than to third generation intermediate and low proficiency participants. Given the importance in understanding
22 the participants of this study, the participant groups will be described in further detail in Chapter 3. Motivation of the S tudy Within the literature on heritage language processes and development, lit tle has been written on Spanish heritage speaker competen ce with regard to phonological perception. One possible reason for the deficit in studies in this area may be that heritage speakers are anecdotally described as having native like phonological production (Au et al., 2002; Montrul, 2010; Oh et al., 2003; P olinsky & Kagan, 2007) By extension, m any researchers may assume that native like phonological perception is certain in heritage speakers and unworthy of investigation One notable exception is a study by Au et al. (2008), which evaluated the phonological and morphosyntactic production and perception of two groups of heritage speakers 3 of Spanish of distinct proficiency levels in comparison with adult L2 learners of Spanish Their study included four different experiments, each focusing on a different area : phonolog ical production of stop consonants /p, t, k, b, d, g/; phonological perception of sentences in noise ; morphosynta ctic production of number, gender, person, tense/aspect and case marking; and morphosynta ctic perception in the form of a grammatical ity judgment task. The re searchers compared the results of each test to determine that both heritage groups outperformed the L2 group in both phonology tasks with little difference between the two heritage groups; however the high proficiency heritage grou p reliably outperformed both the low proficiency heritage speakers and the L2 speakers in both morphosyntax tasks. 3 Au et al. experiences as 1) adult learners of Spanish who had spoken Spanish as their native language before age 7 and only minimally, if at all, thereafter until t hey began to re learn Spanish around 14 years and 2) childhood.
23 The present study utilizes the work of Au et al. (2002) as a starting point for a more thorough investigation into both listener and speaker effects on language ratings. In spite of the advancements made by the scarce literature on heritage speaker phonological perception, a few questions remain unanswered, mainly, whether heritage listeners are more influenced by phonological or morphosyntact ic factors in rendering language judgments. Unlike the perception and production study of Au et al. (2002) this study focuses solely on the perceptive competence of heritage speakers with regard to both phonology and morphosyntax, and replaces the L2 grou p with a third heritage group: third generation heritage speakers of intermediate proficiency that are currently enrolled in Spanish classes. In addition, the present study deviates from that of Au et al. by evaluating both phonological perception and the acceptability of gender agreement errors in direct comparison within one another within the same stimuli, presented to participants within various conditions in order to tease apart the effect of both factors individually Finally, the present study explor es the listener related variable of language Motivated by the work of Au et al., the aforementioned pilot study was designed that evaluated the role of various segmental and supr a segmental factors in the assignment of language ratings. The pilot study (Trujillo, 2011) upon which the current study is based, investigated the following linguistic factors: measurement of voice onset time in voiceless stops /p, t, k/; lenition of voi ced stops /b, d, g/; articulation of laterals /l/; articulation of rhotics / / vs. / /; and average vowel length. In addition to these segmental factors, the pilot study also evaluated the role of linking and of the final pitch contour of interrogative sentences. These features were selected for analysis on the basis of
24 feedback from the participants who indicated these to be factors that led them to identify speakers as either native or non native speakers of Spanish. The statistical analysis found that only linking was highly correlated with the language ratings. In add ition, participants of the pilot study provided notable information on the rating sheet when learly pronounced each word with small pauses between words, sounded non native suggesting that for at least some participants, the phenomenon of linking was salient enough to describe and may play a significant role in the judgments of the listeners. The feedback from the participants as well as the findings of the statistical analysis suggested the feature of linking merited further study. Significance of the Study Understanding the ways in which listeners perceive and respond to accented speech has important social educational and linguistic implications which have been addressed in prior research as well as in the current study. For example, in social interactions, speaking with a foreign accent has both positive and negative consequences for L2 users (Flege, 1988 b ; Varonis & Gass, 1982). On the positive side, native listeners who hear foreign accented speech may assume that their interlocutor has limited L2 proficiency and m a y helpfully adjust their own linguistic output to accommodate the inter locutor and ensure comprehension (Munro et al. 2010). On the negative side, individuals who speak with a foreign accent may be discriminated against (Dvila Bohara, & Saenz, 1993; Lippi Green, 1997; Munro, 2 003) or simply avoided (Derwing, Rossiter, & Mun ro, 2002).
25 Furthermore, a nalyzing the contribution of various phonetic factors to the perception of global foreign accent is important for the second language acquisition process and has implications for the teaching of a second language as well (Magen, 1 998). The analysis of these factors is particularly important in establishing valid priorities for teaching pronunciation to second language learners (Anderson Hsieh, 1992). For some language educators, the aim of language teaching is to foster fluent spea kers. K nowing which measures contribute most fluency and which factors are use d to distinguish fluent and non fluent speakers can en hance their fluency (Kormos & Dnes, 2004). While the preceding discussion regarding the advantages to studying foreign accent and fluency perception focused on L2 speakers, there are advantages to understanding heritage speaker phonological processes as well. Heritage speakers are typically described as having good phonology, especially when they are compared to adult L2 learners of similar morphosyntactic proficiency (Au et al., 2002; Montrul, 2010; Oh et al., 2003) In addition, conventional wisdom erro neously states that even low proficiency heritage speakers tend to sound native like (Polinsky & Kagan, 2007) at least where phonological production is concerned However heritage speakers have also been found to differ significantly from native speaker 4 control groups in studies on Spanish voice onset time (VOT) values ( Amengual, 2012; Au et al., 2002) and 4 The native speaker control group used by Au et al. (2002) was also bilingual, yet classified as background data provided by the In addition, two native Spanish speaker colleagu es of the researchers listened to speech samples of the self identified native speakers and confirmed the self reported native speaking competence (Au et al., 2002, p. 239).
26 segmental articulation and global pronunciation of Korean (Oh et al., 2003; Yeni Komshian, Flege, & Liu, 2000), suggesting that heritage speakers also display some non native phonological features (Montrul, 2010). The latter point is in line with monolinguals, but rather most bilinguals should pronounce one language b etter than the other. There is also evidence that heritage speakers of varying proficiency levels differ in terms of degree of accent (Godson, 2003, 2004 ). Although some L2 learners may strive to sound native for travel purposes or to avoid the negative ef fects of speaking with a foreign accent ( as described above ) the stakes can be quite different for heritage speakers, for whom speaking with an accent that deviates from that of the heritage speech community may be a question of in group identity (Scott S henk, 2007). For example, Scott Shenk (2007) observed English dominant Mexican origin university students in California in informal conversation and described the occurrence of a less fluent Spanish speaker mispronouncing the word jueves, referring to the day of the joves cy one variety o f Spanish or no Spanish at all can have their identity as a Latino called into question ( Potowski, 2012 ). Although proficiency in Spanish is not a requirement in order to identify oneself as Hispanic (Pease Alvarz, 2002; Potowski & Matts, 2008; Rivera
27 Mil ls, 2000 ), proficiency in the heritage language has been shown to contribute (Caldas & Caron Caldas, 1999; Maloof, Rubin, & Miller, 2006; Phinney, Romero, Nava, & Huang, 2001). Although it is certainly possi ble that the level of competence in phonological production does not necessarily equate the level of competence in phonological perception, the investigation of the perceptive competence of heritage speakers is nonetheless important to investigate. After a ll if heritage speakers of varying generations and proficiency levels are shown in this study to vary in terms of phonological competence in perception as they have been in previous studies on phonological production, then those heritage speakers with less er competence may also view themselves or be viewed by others in the community as as stated previously, the study of factors that contribute to heritage speaker production and perception of accent is of particular interest to the field of heritage language studies as they pertain to third generation heritage speakers is particularly essential given that the shift to monolingualism in English is nearly complete by the third or four th generation (Bills & Vigil, 1999; Bills, Hudson, & Hernndez Chvez, 2000; Fishman, 1964; Garca & Otheguy, 1988; Potowski, 2004; Rivera Mills, 2001; Zentella, 1997). Knowledge of these factors could potentially be useful in heritage language classrooms serving third generation learners that are striving to gain proficiency in their heritage language and strengthen their ethnic identity. key concept in this study is the de termination of which factors lead heritage speakers to determine in group affiliation and identify speakers of their own variety of Spanish ( and
28 in so doing, distinguish in group speakers from out of ) For the purpose of this st within the fields of second language acquisition and foreign language learning ( Asher & Garcia, 196 9; Doughty & Long, 2003) as well as foreign accent perception (Flege, 1981; Spanis regularity within their own community. from that of the local variety and the impact of that difference on speakers and listener (Derwing & Munro, 2009). In summary, this study st ands to contribute to the field of foreign language perception by providing information on the salience of gend er agreement errors embedded within fluid sentences and in so doing, will provide a more thorough understanding of the speaker effects of phonological and morphosyntactic accuracy on language ratings and dominant language judgments. In addition to providin g information on perceptual processes, this study also has applied implications for L2 and heritage speech production. If interlocutors are unaware of errors in morphosyntax but aware that speech lacks fluidity, is accuracy on the morphosyntactic inflectio ns of gender even necessary among L2 and heritage speakers as long as they produce fluid, connected speech? Should L2 and heritage speakers spend more time perfecting the delivery of their speech rather than on gender agreement accuracy? On the other hand,
29 interlocutors may be more aware of gender agreement errors than the delivery of the speech or they may provide results that show that both linguistic factors must be present or that neither factor is necessary in order for speech to receive high ratings. The outcome s of this study will provide more information w ith regard to heritage language processes, given that the interlocutors in question in this study are heritage speakers of varying generations and proficiency levels This study will provide informa tion on the differences that may exist between heritage speakers of different linguistic generations with regard to language perception. This would have important implications for understanding heritage language phonology and would provide evidence that th e generalization that heritage speakers possess native like phonological competence does not necessarily apply to speakers of all generations and proficiency levels. Additionally this study will provide more information on the relationship between languag e attitudes toward NMS and language ratings Although researchers anecdotally acknowledge that New Mexicans tend to view their own variety of Spanish (NMS) negatively (Bills & Vigil, 2008), very few empirical studies (Durn Urrea, 2011; Hannum, 1978) have provided evidence of this statement. The current study will remedy this deficiency by empirically exploring whether New Mexican participants demonstrate positive or negative attitudes toward NMS and whether the these attitudes have an effect on the languag e ratings they assign to individuals believed to be speakers of NMS. Organization This dissertation examines the role of linking the presence or absence of errors in gender agreement, and language attitudes in assigning language ratings and in determinin g the language dominance of a given speaker among second and third
30 generation heritage speakers in New Mexico. Due to the unique linguistic history of the region, New Mexico provides the ideal setting for investigating these factors and the role they may play in the perceptive competence of third generation heritage speakers. These ideas are developed in the following chapters. Chapter 2 reviews the existing literature within the fields of New Mexican Spanish, native language perception, and Spanish herita ge speaker processes A discussion of the effects of language policy and language attitudes on the linguistic landscape of New Mexico provides a basis of knowledge for understanding the participants of this study. This discussion is followed by an examinat ion of existing studies on foreign accent and fluency perception, fields that contribute to the issue at hand, which is dominant language perception. The chapter closes with a review of the literature on Spanish as a heritage language and heritage speaker processes beginning with moving on to operationalize the term for the current study. It takes a closer look at studies that have evaluated heritage speaker performance within the real ms of phonology and morphosyntax and identifies a void in the literature that the current study fills. Chapter 3 details the methodology of this project. It introduces the research questions and hypotheses, and then presents a description of the research d esign, including the participants, materials, and tasks. The chapter closes with a description of the methods for data analysis. Chapter 4 presents the results of the study by examining the quantitative data relating to the research questions according to the themes of speaker and listener effects F irst, the role of speaker effects of linking and gender
31 agreement in language ratings and dominant language identification are discussed followed by the listener effects of language attitudes toward NMS on lang uage ratings of NMS vs. non NMS varieties. Next, the chapter presents the qualitative answers provided by the participants to the open ended questions on the research tasks These answers serve as a basis for a qualitative assessment of the relation of the se comments to the quantitative data. Finally, Chapter 5 discusses the results in light of the main ideas guiding the study as outlined in the present chapter, and presents the conclusions and contributions of the study as well as directions for future res earch.
32 CHAPTER 2 BACKGROUND INFORMATION AND PRIOR RESEARCH This chapter presents a review of previous literature published in the fields pertaining to the current study: New Mexican Spanish, native language perception, and Spanish heritage speaker proce sses The chapter opens with a presentation of background information on the linguistic situation in New Mexico that shaped the variety of Spanish known as New Mexican Spanish (NMS) and its speakers. Next, a review of research on the factors that contribut e to native language perception is presented. At that point an overview of the state of the field of heritage language acquisition and maintenance is provided, along with an introduction of the various definitions of the term ationalization of the term for this research. Finally, the chapter concludes with a presentation of the research that has been conducted on heritage speakers within the areas relevant to the current investigation: phonology and gender agreement. The Lingui stic Situation in New Mexico Historical B ackground The linguistic situation in New Mexico is such that English and Spanish have been in contact since 1846 when the United States invaded t he then territory of New Mexico. The Spanish language first appeared in the territory in 1598 with the presence of Spanish conquistadores who colonized the area in the name of Spain (Espinosa, 1917/1975). The settlers who came to New Mexico in 1598 brought with them a sixteenth century Spanish that was fundamentally rural Castilian, mixed with the speech of the Andalusians, Asturians, Basques, and Galicians (Cobos, 2003). By most accounts, Spaniards who settled in the coastal areas and large cultural centers such as
33 Mexico City were exposed to linguistic changes faster than those who had settled further north, and as a result the physical and linguistic isolation of rural communities promoted the maintenance of archaic and non standard forms of Spanish (e.g., Rosado, 2005). New Mexico was a Spanish colony from 1693 1 until N ew Spain gained independence from Spain in 1821 and became the independent nation of Mexico. The Mexican period in New Mexico was short (1821 1847) and quickly followed by the annexation of the Mexican territory west of Texas by the United States in 1848, at which point Anglo Americans and their language poured into the Southwest (Bills & Vigil, 2008). Due to the California Gold Rush, the English speaking population rapidly overwhelmed the Hispanics and Native Americans in numbers in California a fact that helped California achieve statehood more than 40 years before New Mexico (Schmid, 2001). In New Mexico, which had no gold, English speakers did not out number the Spanish speaking population for another three decades (Schmid, 2001) New Mexico remaine d a territory of the United States from 1848 1912. Repeated rejected petitions for statehood established that in order for New Mexico to be considered for statehood, a strict Americanization process was necessary. Schmid cial congressional committee recommended that Hernndez panish was the normal language in all 1 Spanish presence in New Mexico has been continuous since 1598 with the exception of a twelve year period following the Pueblo Revolt of 1680, during which time the Spanish inhabitants fled 300 miles south to present day El Paso. The New Mexico colony was permanently reestablished in 1693 (Bills & Vigil, 2008, Sanz & Villa, 2011 )
34 Unlike other states, New Mexico waited sixty four years for statehood. Approximately fifty petitions were made before New Mexico finall y became a state. Schmid (2001) notes that he U.S. government refused to grant statehood to New Mexico until 1912, when Anglos finally outnumbered Hispanics, thereby putting an end to any challenge by Spanish to the preeminence of English in American life and to the possibility of b ilingualism at the state level. By the time New Mexico was admitted to statehood in 1912, the potent and successful Americanization process was well underway and its effects were evidenced by the indoctrination of the Spanish spe aking inhabitants into the English language in the public domain. As early as 1917, Espinosa reported: As to the language, not one in a hundred is found who has entirely abandoned the use of Spanish and taken up English in his home . . With the new gene ration, however, and especially with the new Spanish population of the cities and towns where the Spanish and American inhabitants are evenly divided, the problem is becoming fundamentally different. The Spanish school children of the predominantly America n cities and towns like Roswell, Albuquerque, East Las Vegas, etc. speak English as well as the English speaking people and speak very poor Spanish. (1917/1975, p. 101) The generational shift to Spanish in which older generations spoke primarily Spanish w hile younger generations were being fully indoctrinated into the English language at school and within other spheres of the public domain (1964) theoretical framework of language shift. The Americanization movement and necessity of th e younger generation to communicate with elders within the community and within the home led to a period of diglossia for the younger generation between the 1920s 1960s and beyond. During this time it was not uncommon for New Mexicans to speak Spanish at h ome and English within the public domain. This situation however,
35 led to the view of Spanish as the language of lesser prestige and thus resulted in further displacement of Spanish. This complex political history of New Mexico is deeply intertwined with i ts linguistic history. These historical circumstances have led to the present course of Spanish language development in New Mexico and throughout the Southwest, and the creation of a particular dialect known as New Mexican Spanish (NMS) Language P olicies that S haped NMS New Mexico has a rich linguistic history that encompasses language maintenance and shift, diglossia, and multilingualism (Pealosa, 1980). Although the Spanish language has maintained a presence in New Mexico for over 400 2 years, Snchez (1 983) asserts that the continued presence of the Spanish language in the Southwest is not the result of an American language policy which strives for the maintenance of minority languages but rather the result of a historical process initiated in the early part of the nineteenth century with the westward movement of the U.S. population into Mexican territory B y 1848 the Southwest was part of the United States. In the process the small dominant Mexican population residing in the Southwest was engulfed and re legated to a subordinate status (Snchez, 1983). Part of t the Spanish language within the public domain In 1910 Congress presented New Mexico with the Enabling Act, or model constituti on that was a requisite for statehood. One article of this act was aimed specifically at New Mexico lawmakers, and read: 2 W ith the exception of 1680 1693, during which time the Spanish inhabitants fled to present day El Paso following the Pueblo Revolt (Bills & Vigil, 2008, Sanz & Villa, 2011 )
36 interpreter shall be a necessary qualificatio n for all State officers and members of the Berry, 2000, p. 173). Although the intention of Congress was clear that Spanish was not welcome in the government domain, New Mexico lawmakers were unhappy with this proposal and fou ght back, amending the clause to allow for the publication of all legal notices in Spanish and English. This article remained in effect for thirty years (Gonzales Berry, 2000). New Mexico lawmakers were not as successful, however, in the protection of the Spanish language within the educational system, which served as the principle vehicle for Americanization shall be made for the establishment and maintenance of a system of public schools, which shall be open to all children of said state and free from sectarian control, and that said schools shall always be conducted in English (Gonzales Berry, 2000, p. 174). Although some schools did continue to permit the use of Spanish, in others Spani sh Mexican American children in the Southwest as late as the 1960s (Schmid, 2001). Such actions on the part of school officials reinforced the conditions of the resulting diglossia in which Spanish was not to be spoken in public. The diglossic culture of New Mexico throughout the twentieth century and the language policies that generated this culture a lso set into effect the propagation of language attitudes toward English and toward NMS that will be discussed in the next section.
37 Language A ttitudes In addition to language policy, language attitudes towards Spanish in general and NMS in particular have had an effect on the shape of the language. Speakers of NMS have been subject to language attitudes from speakers of English on the one side and speakers of Mexican and other varieties of Spanish on the other side. These attitudes in addition to New Mexico further diminish its usage. In their linguistic atlas of the Spanish of New Mexico and southern Colorado, Bills 1 7). This attitude is not only a result of having been (in the case of some students, as mentioned above) punished for speaking Spanish in school, but also a direct result of a recognition of the social inequalities associated with speaking Spanish. Bills a nd Vigil note that a typical Spanish speaker manifests less valued material trappings and lower to conclude that the English language is superior to the Spanish language. Another myth that shapes the attitudes of New Mexican speakers according to standard Spanish is myth are people from Mexico or other Spanish speaking countries who have been
38 educated in Spanis h and imbued with the strong prescriptivist tradition of the Real Academia Espaola (Royal Academy of Spanish). Bills and Vigil argue that these speakers tend to have strong feelings about what is proper Spanish and find the local Spanish to be lacking, ev en to the point of being a corrupt and degenerate means of communication, and state that Anglos who have acquired a good command of standard Spanish may share those attitudes for the same reasons (Bills & Vigil, 2008). At the same time, p ositive attitudes toward NMS have also played a role in its maintenance and perseverance. In 19 65 the Chicano movement surge d, lasting more than a decade. The civil rights movement began as a response to the disenchantment us within the United States. Under the Chicano movement, the Spanish language became a symbol of ethnic pride and affirmation (Maciel & Pea, 2000). One direct result of the Chicano movement was the promotion of Chicano scholarship and the quest for a defi nition of the Chicano language experience In 1975, Pealosa reported that t he keynote of the Chicano movement wa s self defi nition and self determination, and noted that in line with the ideology of the movement, Chicano scholars had begun the process of i dentifying, outlining, and defining the field of Chicano sociolinguistics Indeed, his own work titled Chicano S ociolinguistics: A B rief I ntroduction (1980) and the anthology El Lenguaje de los Chicanos by Hernndez Chavez, Cohen, & Beltramo (1975) as wel Chicano D iscourse: Socio H istoric P erspectives (1983) serve to illustrate the results of this labor. In spite of the symbolic use of Spanish within the Chicano movement, New Mexico has been witness to a generational language shift from Span ish to English as
39 the dominant language. Southwest Hispanics over the past half century are shifting to English and abandoning Spanish rivals the loss of the ethnic mother tongue by practically any ethnic group in Cobos (2003) states that NMS is losing its struggle for oriented environment where all facets of daily living (commerce, education, entert ainment, local (p.xvi). In the 1983 introduction to A Dictionary of New Mexico & Southern Colorado Spanish, (2003) Cobos notes that Hispanos in this region of the Southwest are inevitably shifting to English as their most important medium of communication, and that in this region, most young Hispanic parents in their twenties and thirties are no longer kno w Spanish themselves, they find it very difficult and inconvenient to transmit it to their offspring however deficient 3 Consequently, m ic to whom Cobos referred now find themselves in a situation of language revitalization and maintenance with their own children. The New Mexico Association for Bilingual Education has even issued a pamphlet with suggestions for parents interested in maintaining the Spanish language with their children (Douglas, 1999). T he pamphlet informs parents: I t is important that we keep in mind that the child will be immersed in an English language environmen t in many of his/her classrooms, on the 3 s, 2003).
40 playground and by the media. This leads us to realize that the child will learn English even if it is not stressed in the home . It is therefore essential to place the most emphasis on learning Spanish. If not, t he language will be lost in spite of a healthy bilingual home environment (p. 3) The endeavor of language revitalization and maintenance in New Mexico has been helped by the continuing influx of Mexican immigrants to the region. Snchez (1983) asserts th at i n a society where Spanish has been relegated to a subordinate position by the dominant language, the future of bilingualism lies with the continued presence of first generation Mexican immigrants a Spanish speaking population not yet totally as similat ed to the English language. Snchez cites the growing availability in the (1983). Exposure to different varieties of Spanish can occur in various ways: through exposure to writte n Spanish, face to face interactions, via radio, television, and other media (Bills & Vigil, 2008). This exposure can be independent or in the classroom. However the revitalization of Spanish into New Mexico occurs at the expense of the traditional NMS dia lect. In the past, the formal study of Spanish has had a considerable impact on the native dialect of some New Mexican speakers, as Bills and Vigil explain that Spanish teachers, by virtue of their training, have become educated Spanish speakers, and many of them even native speakers of New Mexican Spanish tend to Hispanic students (2008). In Spanish classes today, especially at the college level but also in high s chool courses, there is a growing emphasis on accepting and nourishing the Spanish skills that the heritage speaker brings to the classroom, but this has not always been the case (Bills and Vigil, 2008).
41 In addition to explicit correction by Spanish teache rs, other varieties of Spanish have impacted NMS implicitly simply by their existence and exposure to speakers of NMS, making New Mexicans aware of alternative addition, the contact between speakers of different varieties can create negative attitudes toward NMS as mentioned above. Although anecdotal documentation (Bills & Vigil, 2008) states that New Mexicans hold negative attitudes toward NMS, only one known empirical study has been conducted to test this account. This study evaluated the attitudes of bilingual speakers in a small New Mexican community towards NMS, Mexican Spanish, and other varieties of Spanish or Spanish in general (Durn Urrea, 2011). This study evaluated information about language attitudes coded from soc iolinguistic interviews with 16 participants. Durn positive, with 75% of the participants expressing positive attitudes and 25% expressing negative attitudes toward NMS. However, the percentage of participants who expressed positive attitudes toward Mexican Spanish was slightly higher at 83%. In addition, Durn Urrea found that just 67% of the participants expressed positive attitudes toward other varieties of Spanish or Spanish in general. Durn caution, however, as they were retrieved from sociolinguistic interviews and a lower number of participants broached subjects in which they expressed attitudes toward Mexican Spanish (N=6) and other varietie s of Spanish or Spanish in general (N=6) than did participants who discussed NMS (N=12). Thus the results of this study would likely be different had more traditional methods of attitudes elicitation been used.
42 Characteristics of NMS The extended period of contact between NMS and English, Native American languages, and Mexican Spanish has created a particular dialect that includes characteristics of code switching, calquing, and regu larization (Sanchez, 1983). The lexicon of NMS includes influences from a v ariety of sources. The physical and linguistic isolation of rural communities promoted the development of archaic and non standard forms of Spanish, evident in the morphology, phonology and lexicon of NMS (Cobos, 2003). At the phonological level, the Spani sh of New Mexico is characterized by metathesis 4 and by the reduction processes of synaeresis 5 apheresis 6 apocope 7 and syncope 8 (Sanchez, 1983) while the lexicon of NMS includes features such as archaisms and contains items with a wide variety of sources such as Nahuatl, Native American languages, and English (Cobos, 2003). Bills and Vigil (2008) note that acknowledge that archaisms heard in NMS are also heard in the speech of the common people I other part s of the Spanish speaking world Bills and Vigil (2008) allude to the fact that people with more pr estige are no longer utilizing these forms in their language. This fact is directly related to the attitudes toward the use of archaisms and beliefs about the people that use these forms as being of 4 Metathesis is the transposition of two sounds within a word, such as pader for pared (Eng: wall) 5 Synaeresis is the diphthongization of two strong vowel sounds such as tualla for toalla (Eng: towel ) 6 Apheresis is the deletion of the initial element of a word, such as toy for estoy (Eng: I am) 7 Apocope is the deletion of the final element of a word, such as pa for para (Eng: for) 8 Syncope is the deletion of an element in the interior of a word, such as alredor for alrededor (Eng: around)
43 lesser prestige and/or education. Some examples of archai sms in the NMS lexicon 9 include semos for the modern somos we are seigo for soy I am puela for sartn frying pan and nodriza for enfermera nurse ( Cobos, 2003; Bills & Vigil, 2008). As stated above, o ther influences on NMS include indigenous lan guages such as Nahuatl, adding words to the NMS lexicon such as tecolote owl and zacate grass and a few words from Native American languages such as osh wild celery ( Cobos, 2003) and cunques coffee grounds or crumbs ( B ills & Vigil, 2008). Bills and Vigil (2008) explain that so few Native American loanwords were adopted into NMS because of the unidirectional type of bilingualism between the Native Americans and the Hispanics interactions between these groups tended to be conducted in Spanish, a nd people generally do not borrow words from languages they do not speak. Accordingly anglicisms have had the largest influence on the NMS lexicon, for example the phonologically adapted words Crismes Christmas espauda yeast powder and suera sweater ( Bills & Vigil, 2008). This diverse set of linguistic influences and characteristics helps to set NMS apart from other varieties of Spanish. In summary, New Mexico is defined by an extraordinary linguistic situation in which English and Spanish have been in contact for over 160 years. NMS has developed under peculiar conditions, isolated from both Spain and Mexico in its early history and subject to strong North American influence thereafter. The Americanization movement that helped New Mexico achieve st atehood had steep consequences for the existing culture and language of the New Mexicans. The middle of the twentieth century ushered in a period of diglossia and a generational language shift from Spanish to English as the dominant language. This process of language shift has been well 9 Many of the archaisms found in NMS are also found in rural Mexican varieties (Martnez, 1998)
44 documented ( Bills, Hernndez Chvez & Hudson, 1993; Hudson Edwards & Bills, 1982; Ortiz, 1975; Pease Alvarez, 1993; Sol, 1990) and has been described as occurring with such rapidity that it rivals the loss of the eth n ic mo ther tong ue by practically any ethic group in documented history (Bills, 1997). In addition, the language shift that has occurred in this region has had a profound effect on the expansive definition of a heritage speaker in New Mexico, as will be discussed ahead in this chapter. Nevertheless, many New Mexicans today find themselves in an effort to maintain the Spanish language, an endeavor that has been helped by the continuing influx of Mexican immigrants to the region. The dialectal contact between NMS an d other varieties of Spanish is resulting in a change in progress that defines a new period for NMS in the twenty first century. The extensive history of contact between English and Spanish as well as the language shift and maintenance that have occurred i n New Mexico make this region a compelling area to study with potentially exceptional results. Research on Native L anguage P erception Nonnative speakers of a language are often recognized as such because of their pronunciation, and in many cases their spec ific L1 backgrounds can be identified, even by casual interlocutors (Munro, 2008, p. 193 ). Although l isteners are highly sensitive to speech patterns that are different from those typically used in their speech community (Derwing & Munro, 2008) m uch of th e research on foreign accent has focused on factors that contribute to foreign accent production (Flege, 1984; Flege, Munro & McKay, 1995; Munro & Derwing 1998 ). In these studies, researchers utilize native speaker judgments as a qualitative assessment of foreign accent via listening and ratings tasks, in which a listener group listens to sentences, words, or segme nts produced by a speaker group and rates them for foreign accentedness These studies
45 tend to utilize phonetically untrained listeners in asses sing both fluency (Derwing, Munro, & Thompson, 2004; Rossiter, 2009) and foreign accent (Asher & Garcia, 1969; Flege, 1984; Scovel, 1981) because they may provide insight into how understandable L2 speakers are when they interact with other members of thei r community (Munro, 2008). While a number of studies have utilized native speaker judgments as a by which nonnative speaker perception is accomplished are not fully unde 626). He adds that perceptual cues signaling nonnative speaker status exist at multiple levels in the speech signal, and counts among the markers of L2 speaker status segment level errors, which are perceived by the listener as phone mic substit utions, insertions, and deletions. Munro acknowledges that L2 speech is typically delivered more slowly (Munro and Derwing, 2001; Raupach, 1980) and less fluently (Derwing et al., 2008; Riggenbach, 1991) than native produced speech, and infers that this is because L2 speakers access and implement lexical and grammatical knowledge less efficiently than native speakers (Munro 2010) Listeners are highly sensitive to divergences from native speaker patterns of speech production such that they can recognize sec ond language speakers even from very short stretches of speech (Munro, 2010). Flege (1984) found that even listeners unfamiliar with the French accent in English were able to distinguish native from non native English speakers on the basis of acoustic diff erences confined to a single syllable. This suggests that since they were unable to identify known characteristics of French accented English, they detected foreign accent by comparing salient features in
46 the speech samples to the phonetic norms of their o wn variety of English Such accent perception Listeners of varying degrees of proficiency have been found to utilize different features of speech in making perceptu al judgments. One study found that native English speakers utilized segmental deviation (particularly /r/ and /l/) in detecting foreign accent by Japanese speakers in English while non native speakers relied on non segmental parameters such as intonation, fluency and speech rate (Riney, Takagi, & Inutsuka, 2005). Studies have shown that listeners can detect a foreig n accent even in languages they do not speak (Major, 2007) and from content masked (backwards) speech (Munro, Derwing, & Burgess, 2010). In str iving to define and understand the processes of foreign accent perception, erception of a foreign accent derives from differences in pronunciation of a language by native and non p. 445). However, he caut ioned that t his does not necessarily mean that perception of a foreign accent is based just on overtly detectable mispronunciations of sound segments. Listeners are more likely to base a judgment of foreign accent on some combination of segmental, sub segm ental, and supra segmental differences which distinguish the speech of native from that of non native speakers ( Flege, 1981 ) Buell Hill (2001) echoes this assertion, noting accen This means that global foreign accent, defined as t he overall impression on whether or not and to what degree a p erson sounds native or nonnative, (Major, 2001) overrides any one segment or feature in foreign accent perception.
47 Given that foreign accent is fundamentally a perceptual phenomenon (Munro & Derwing, 1998) a number of studies have focused on the accuracy of native and non native listener groups have in detecting foreign accent (Alba Salas, 2004; Buell Hill, 2001; Derwing, Rossiter, Munro, & Thompson, 2004 ; Flege, 1984; Leiken, Ibrahim, Eviatar, & Sapir, 2009; Munro, Derwing & Morton, 2006; Neufeld, 1980; Park, 2009; Scales, Wennerstrom, Richard & Wu, 2006). Most of these studie s have determined that speakers with varying degrees of proficiency produce different results with regard to determining nativeness of other speakers. F or example in a study on the competence of native English speakers who learned French as adults in dete cting foreign accent in their L2, Neufeld (1980) found that advanced bilingual speakers were on par with native (monolingual) speakers. Similarly, Munro et al. (2006) examined listeners of various language backgrounds (10 native speaker s of English, a n d 30 non native speaker s of English from Cantonese, Japanese, and Mandarin language backgrounds, all advanced L2 learners of English) in rating short English sentences p roduced by other non native speaker s of English They found that the non native listeners c ould detect the foreign accent of the L2 speech reliably and that their rating patterns were not very different from those of the native listeners. On the other hand, Park (2009) investigated native English speakers, early Korean/English bilinguals, and la te Korean/English bilinguals listening to a speech sample of relatively short English stimuli spoken by 4 Korean English bilinguals and 2 NSs of American English, and found that native listeners were more sensitive than both early and late bilinguals to su btle differences denoting foreign accent such as the articulation of the vowel / found no significant
48 difference between the performance of early and late bilingual listeners in detect ing foreign accent. The current study examines the phenomenon of connected speech known as Riazantseva, 2001; Riggenbach, 1991; Vanderplank, 1993). Linking is generally seen as a m arker of fluency (Hieke, 1984; Simes, 1998 ) al though f luency has been notoriously difficult to define, as evidenced by studies that seek to explore and define this concept ( Kormos & Dnes, 2004; Norris & Ortega 2009; Riggenbach, 1991; Sakuragi, 2011; Sheppard, 2004; Skehan, 2009) and has been inconsis tently operationalized P revious studies have considered fluency as varying combinations of speech rate, the mean length of utterance, phonation time ratio and the number of stressed words produced per minute ( Ejzenberg, 2000; Hieke, 1984; Kormos & Dnes, 2004; Riggenbach, 1991; Wenners tro m, 2000). Kormos and Dnes (2004) assert that there are two types of definitions of fluency: one which considers fluency as a temporal phenomenon, and one that regards it as spoken language competence. In a study that exp lored which variables predict native and non native speaking teachers' perception of fluency and distinguish fluent from non fluent L2 learners of English, they found that fluency is not only a temporal phenomenon, as raters do not only look at speed and p ace when intuitively judging a it also has to do with other variables that are closely correlated with proficiency such as accuracy and lexical diversity (Kormos & Dnes, 2004) This finding led the researchers to conclude that fluen cy is best conceived of as fast, smooth and grammatically accurate performance.
49 There is no consensus on the (Chambers, 1997). For this reason it is important to note that fluency is operationalized within the present study simply as connected speech or, as stated above, linking Although temporal cues such as linking are generally seen as a marker of fluency and not necessarily of foreign accent, studies have identified a connection between temporal cues with t he perception of foreign accent (Chen 2010; Flege, 1988; Munro, 1995; Munro et al., 2010; Munro & Derwing, 2001; Tajima, Port, & Dalby, 1997). In a study on factors affecting degree of foreign accent in English sentences by native Chinese speakers, Flege ( 1980) discussed the relationship between fluency and foreign accent number, location, and duration of pauses, prolongations, and repetitions in sentences. Flege referenced the oral interview test used by the Foreign Service Institute to assess foreign language proficiency which utilized five variables: grammar, vocabulary, comprehension, ac cent, and fluency, and pondered whether accent and fluency indeed represent separate perceptual dimensions in which the degree of one may influence the degree of the other, as the test implies, or whether perceived degree of fluency had no contribution to foreign accent judgments. To examine this relationship, Flege investigated whether removing pauses from sentences spoken by non native speakers would improve the global foreign accent scores accorded to their sentences. He found that removing pauses had li ttle effect on the pronunciation scores, which suggested either that fluency judgments do not influence degree of perceived foreign accent or that fluency cannot be perceived independently from the segmental and supra segmental
50 di mensions which undermine a ccent, and suggested that perhaps re moving p ause s w ould h ave had a significan t e ffect h ad the nonnative speakers' segmenta l ar ticulation b een m ore accurate (Flege 1980) However Chen (20 10 percept ions of foreign accents: syllable duration, vowel reduction, pause duration, CV linking, consonant cluster simplification, and speech rate in Taiwanese learners of English. Although Chen found speech rate to be the primary predictor in determining native l resultant combination of the other five acoustic variables (Chen, 20 10 ). For example, linking can cause a change in utterance timing and thus speed up or slow down the speech r ate. Chen found that when speech rate was removed as a predictor, vowel reduction and linking duration became the two most heavily weighted variables, each with a very strong positive correlation with the global foreign accent ratings. This finding led her to propose a temporal perception model of foreign accents, according to which native listeners would perceive non native utterances through the prism of a foreign accent. Under this model, speech rate would be detected first, followed by the combination o f CV linking and vowel reduction, followed by the remaining timing variables under investigation in her study (Chen, 2010) Although Chen did not formally evaluate consonants and values in this study, she places them last on the perception model in citing prior studies that have shown that deviance in segments rather than prosody was found to be least significantly related to global accent rati ngs (Anderson Hsieh, Johnson, & Koehler, 1992; Magen 1998; Major 1986). The placement of segmental factors as the f inal candidate group to influence foreign accent ratings within
51 is consistent with the pilot study upon which the current study is based, as discussed in Chapter 1. Hieke (1984) notes that temporal pressures that are genera ted by the exigencies of speaking can lead to considerable alterations in pronunciation. Among these alterations is the phenomenon of linking, which Hieke explains as words strung together in citation form clearly differ from the way they are ultimately re alized as phonated sound sequences in running speech. He focused on CV linking and described this type of linking as occurring where final consonants become attached to the following syllable of that begins with a vowel and explained that the overriding ra tionale for linking is the avoidance of hiatus where possible (Hieke, 1984). In a comparison between native speakers of American English with German learners of English as an L2, Hieke found that native English speakers generated occurrences of linking at a rate of 77.63% (actual linked tokens to potential link points) whereas non native speakers generated occurrences of linking at a rate of 53.5%. Hieke interprets the substantial difference in actualized tokens of linking between the two groups to signify consequently be profitably considered one of the features that differentiate native from non speech can serve as a variable in fluency assessment, and suggested that further research evaluate the role of linking on input and listening comprehension (1984). Studies that focus on the effect of syllable structure on the perception of foreign accented speech show that supra segmental factors including syllab le structure Hsieh, et al., 1992; Magan, 1998). Magan (1998) evaluated the ratings assigned by native speakers of
52 American English on the speech of two native Spanish speakers in English, and a coustically edited the elicited speech produced by the speakers to make their Spanish accented speech to more closely resemble native American English presented both the edited and unedited stimuli to the listeners. She found that listeners showed a signif icant sensitivity to factors affecting syllable structure in comparison with segmental factors such as vowel reduction, vowel tense laxness, fricative voicing, stop voicing, and final /s/ deletion (Magan, 1998). Magan concluded that supra segmental factors likely future studies will find that factors that affect syllable structure will contribute heavily to foreign accent. Similarly, Anderson Hsieh et al. (1992) looked at the relationship between native speaker judgments of non native pronunciation and deviance in segmentals, prosody, and syllable structure in L2 speakers of English from 16 different language backgrounds including Spanish, Chinese, Arabic, and Korean. T he researchers took audio samples of the read aloud portion of the SPEAK Test, which has been used in evaluating the speaking proficiency of international teaching assistants at universities throughout the United States (Anderson Hsieh et al., 1992). The a udio samples were rated by three experienced ESL teachers who were familiar with the SPEAK test. The researchers found that all three variables, including syllable structure showed statistically significant influence on the pronunciation ratings, although the prosodic variable proved to have the strongest effect. Adding to the list of factors that may influence foreign accent perception, s tudies have shown a correlation between pronunciation judgments and non phonological properties such as grammar errors (Munro & Derwing, 1995; Varonis & Gass, 1982).
53 Varonis and Gass examined the impact of pronunciation and grammar errors on comprehensibility by asking native English speakers to measure comprehensibility of grammatical and ungrammatical sentences read by native speakers and non native speakers of English. They inferred from their results that pronunciation judgments can be influenced by non phonological properties such as grammar errors Although the findings of Munro and Derwing seemed to confirm those of Varonis and Gass, they cautioned that the relationship between grammatical error and accent scores could simply indicate that speakers who make pronunciation errors also tend to make grammatical errors. Indeed, the relationship between pronunciation and g rammatical accuracy is complex, and both can provide an interlocutor with clues as to the proficiency of heritage and L2 speakers and assist in determining the language dominance of a speaker. Consequently, the current study strove to control for grammatic al errors by utilizing using read aloud speech, in which the only grammatical errors present within the stimuli were the errors in gender agreement that were expressly under investigation. While studies on perceptual fluency tend to utilize elicited speech in the form of narrative tasks (Derwing et al., 2004; Kormos & Dnes, 2004; Riazantseva, 2001; Rossiter, 2009; Skehan & Foster, 1999), perceptual studies on foreign accent tend to follow the practice of utilizing laboratory (read aloud) speech in the stim uli (Anderson Hsieh et al., 1992; Chen, 2010; Flege et al. 1995; Mackay, Flege, & Imai, 2006; Magan, 1998; Munro et al., 2010; Scales et al., 2006; Tajima et al., 1997 ). Task based research has shown that having to produce different types of content places a different cognitive load on speakers, which, in turn, influences the fluency of the production (Kormos &
54 Dnes, 2004; Skehan, 1998) By providing fixed content in the form of read aloud sentences the influencing factor of content as well as temporal fa ctors resulting from problems with lexical access or other cognitive could be eliminated or reduced Flege (1984) found that accent was detected equally well in speech produced by non native speakers in a phrase reading task and in a spontaneous speech tas k. This finding suggests that "attention to speech" 10 does not affect the authenticity with which non native speakers produce phonetic segments or syllables in a foreign language 704). This finding was supported by Flege & Hillenbrand (1984) who conc luded that studies on foreign accent perception have continued to use full sentences f or speech stimuli (Flege & Fletcher, 1992; Flege, Munro & MacKay, 1995; Munro & Mann, 2005; Piske, MacKay, & Flege, 2001). Studies have found that listeners are able to detect non native speech in samples as small as a single phoneme (Alba Salas, 2004; Fle ge, 1984, Flege & Munro, 1994,) while other studies have shown that non native speech can be usually be detected much more easily in longer stretches of speech (Major, 2001). Alvord (2007) justifies the use of read aloud speech in his comparative phonolog ical study on intonation in contact in Miami Cuban bilinguals. He cites the practicality of being able to control the number of utterances, whereas the number of naturally occurring examples of any one type of utterance or phoneme can be very low. 10 ent production of variants found in the prestige dialect
55 Major (2 001) maintains that the number of occurrences of a given type of utterance in narrative or spontaneous speech can be low due to the ability of a speaker to avoid a number of segmental and prosodic phenomena. Furthermore, laboratory speech can help ensure a homogenous sample of occurrences across a variety of speakers. Derwing and Munro (1998) argue that because stereotyping and evaluative reactions Brennan, 1981; Cunningham Ander son 1993; Llurda 2000; Zuengler, 1988 ) perceptual research on accented speech must control sociolinguistic variables as much productions. The use of read aloud speech is o ne way to control such variables. The obvious caveat to the use of speech elicitation, however, is that the speech sample obtained cannot be assumed to be representative of spontaneous speech (Face, 2003). Southwood and Flege (1999) note that global rating s of foreign accent assist in making predictions about the particular variables that influence foreign accent and permit the identification of particular acoustic variables that infl judgments of degree of perceived foreign accent. They add that p erceptual measures are important because they corroborate the infl uence of these particular acoustic degree of perceived foreign accent and verify accent (Southwood & F lege, 1999). In an investigation on which type of scaling method is most appropriate in measuring the degree to which listeners perceive the accents of non native speakers to diverge from that of native speakers of English, Southwood and Flege (1999) compa red two types of scaling methods to determine whether degree of foreign accent is a metathetic
56 continuum amenable to linear partitioning such as in an equal appearing interval scale like a four point or a seven point scale, or a prothetic continuum hich would be resistant to linear partitioning. In comparing the accuracy of two different tasks in which listeners rated degrees of perceived accent of native Italian speakers of English, they found their data suggested that accentedness is a metathetic c ontinuum and thus that an interval scale is appropriate for scaling the accentedness of Italian speakers of English, although they stop short of endorsing a particular interval size. However, Alba Salas (2004) justifies the use of a four point scale in t hat it captures a qualitative distinction in terms of degree of goodness of the accent, although it allows for a straightforward comparison of positive and negative responses by collapsing the selection from four to two. Furthermore, a four point scale eli minates the possibility that the listener will select a neutral response out of indecisiveness or an attempt to remain impartial, which would thus disallow a binary classification of positive and negative ratings. In summary, listeners have been shown to be highly sensitive to speech patterns that differ from those typically used in their speech community or otherwise denote non native speech. Existing studies on the perception of non native speech have sought to better understand the processes by which no nnative speaker perception is accomplished. These studies have focused on fluency perception and perception of foreign accent. Although temporal factors such as pausing and linking have generally been considered in studies on fluency perception (Hieke, 198 4; Kormos & Dnes, 2004; Simes, 1998), there are also studies that have considered syllable structure and linking in foreign accent perception (Anderson Hsieh, et al., 1992; Chen, 2010; Magan, 1998).
57 Another main distinction between studies on foreign acc ent perception and studies on fluency perception is the comparative variables under investigation such as speech rate, mean length of utterance, phonation time ratio and the presence of disfluencies in the stimuli, which have been studied alongside syllabl e structure and resyllabification in studies on fluency (Ejzenberg, 2000; Kormos & Dnes, 2004; Riggenbach, 1991; Rossiter, 2009; Towell et al, 1996) while the role of segmental articulation and prosody have been evaluated in studies on foreign accent ( And erson Hsieh et al., 1992; Brennan & Brennan, 1981; Derwing & Munro, 1997; Magan, 1998 ). However, factors traditionally considered in studies on fluency such as pause duration and speech rate have been studied as non segmental parameters in studies on forei gn accent perception (Chen, 2010; MacKay et al., 2006; Munro & Derwing, 1998, 2001; Riney et al. 2005). Another distinction between studies on the perception of fluency and studies on the perception of foreign accent is the type of stimulus elicitation, wh ether narrative tasks or read aloud tasks Elicitation type is lar gely related to the previously discussed distinction of the comparative variables chosen for each type of study because elicitation methods are intended to include or exclude particular vari ables influencing speech perception. The current study is envisioned neither as strictly a study on fluency nor foreign accent perception given that the comparative factor of gender agreement is grammatical in nature and related to neither of the two field s Nevertheless, the speech elicitation method of the current study resembles that of foreign accent perception studies. This study utilizes the existing literature on both fluency perception and foreign accent perception as a knowledge base upon which eli citation and scaling methods as
58 well as hypotheses are based, namely, the use of read aloud speech and a four point scaling method, which will be discussed further in Chapter 3. Heritage Speakers Since the turn of the twenty first century, research on heri tage speakers of Spanish in the U.S. has been prolific and interdisciplinary (Beaudrie & Fairclough, 2012), with subfields ranging from pedagogical perspectives (Carreira, 2012; Potowski, 2005; Potowski, Jegerski, & Morgan Short, 2009; Roca & Colombi, 2003 ; Webb & Miller, 2000) and grammatical competence (Cuza & Frank, 2011; Mikulski, 2010; Montrul, 2002, 2004, 2006; Montrul, Foote, & Perpi an 2008a, 2008b; Montrul & Potowski, 2007; Zapata, Snchez, & Toribio, 2005) to sociolinguistic aspects including lang uage identity and language maintenance (Beaudrie, 2009; Bills, 2005; Carreira, 2000; Fishman, 2006; Martnez, 2009; Potowski, 2012; Valds, Fishman, Chvez, & Prez, 2006; Villa & Rivera Mills, 2009) among others. However the field of heritage language re search is nascent enough that there is to date, no universally accepted Fair c lough 2012). The be cross linguistic and applicable to variable contexts such as age of acquisition, generation, and proficiency level. The most widely used definition in the areas of research and education is that of Valds (2001) which states that a heritage language English language is spoken, who speaks or merely understands the heritage language, and is to some degree bilingual in proficiency requirement and has thus often been seen as too narrow (Beaudrie &
59 Fairclough, 2012). While v arious definitions have been proposed in heritage language litera ture, Beaudrie and Fairclough (2012) note d that most definitions have centered around two distinct elements: a personal or familiar connection to a particular group (Fishman, 2001) or a certain degree of proficiency in the language as in that of Valds (20 01 ). learners who learn their heritage language from scratch as adults are regular sec ond proficiency heritage learners] who from a linguistic standpoint resemble second language learners may have affective and intellectual need s that are Consequently, in their comprehensive anthology o f current research trends and the state of the field of Spanish as a heritage language in the U.S., Beaudrie and Fairclo ugh (Beaudrie & Fairclough, 2012; p.7; Fishman, 2001, p.81 ), arguing that this defin ition is inclusive and encompasses a diverse range of cultural profiles. Indeed, cultural and linguistic connection to Spanish to be included within the def inition of the home during childhood due to the language shift that has occurred in the region, as detailed previously in this chapter Yet this populati on still benefit s from participation in a heritage language program versus a second language program
60 Echoing the beliefs of Carreira (2004) is t he largest Spanish as a Heritage Language ( SHL ) program in the state of New Mexico, the Sabine Ulibarr Spanish as a Heritage Language program at the University of New Mexico The program implements inclusive definition of SHL learners in order to recognize the linguistic diversity found among which include beginning level students (Wilson, n.d.). In addition, the program states that many of their students see Spanish as an important part of their identity and acknowledges that a cultural connection to the language can serve as a powerful motivating factor for students dedicated to becoming more proficient in is that which is adopted for the current study It as the University of New Mexico (UNM) given the intermediate proficiency group of participants in the present study is composed of UNM students. In addition, the the low proficiency group of participants, whose characteristics are equivalent to those over hearers he two studies comparable. Although Polinsky and Kagan (2007) and other researchers ( e.g., Montrul 2012) have described low proficiency heritage speakers as regular second language learners with simply a different motivation for learning the language as st ated above, they admit that even low competence in
61 the heritage language that distinguish them from L2 speakers, prim arily in the area of phonology ( Au, Knightly, Jun, & Oh, 2002; Au & Romo, 1997; Godson, 2004; Oh, Jun, Knightly, & Au, 2003 ; Yeni Komshian, Flege, & Liu, 2000) These proposed advantages are outlined in further detail below. Heritage Language Phonology The competence of the lowest proficiency heritage speakers of various languages has been studied in order to determine any viable linguistic distinction with over hearers The description of these speakers centers on childhood exposure to and usage of the heritage language. Lipski (1993) stated that transitional bilinguals (TB) are typically yielded in situations of language shift and rarely communicate with wider groups or with each other in the heritage language. He used the example of TB Spanish speakers and described their use of Spanish as confined to conversation with a few relatives, typically quasi monolingual Spanish speakers of the by individuals known to be bilingual, TB speakers often respond wholly or partially in English (Lipski, 1993). Similarly, Polinsky and Kagan (2007) describe basilectal speakers as those who grew up with limited or early interrupted exposure to the home language and consist of over hearers (Au et al., 2002) who heard the home language in the background but ne ver responded in it and were not addressed in it consistently, as well as bilinguals who switched to the dominant language as early as preschool. Au et al. (2002) tested phonological production of voiceless stop consonants and voiced stop consonant lenitio n in 12 adult L2 learners and 11 low proficiency heritage speakers or
62 childhood over of Spanish. In addition, the researchers tested several aspects of morphosyntax including number and gender agreement in noun phrases. Results showed that the he ritage speakers scored closer to native speaker controls in measures of ph morphosyntax. In a follow up study, the researchers presented more robust data with additional morphosyntax and pronunciation assessments and presented measures to help rule out possible confounding prosodic factors such as speech rate, phrasing, and stress placement (Knightly et al 2003). They again found a pronunciation advantage for the low proficiency heritage speakers over the typical late L2 learners on phonetic analyse s (VOT and degree of lenition) and accent ratings (phoneme and story production) but no benefit in morphosyntax. Additionally, they found the pronunciation advantage did not seem attributable to prosodic factors. The same researchers found similar results with regard to low proficiency heritage speakers of Korean (Oh et al. 2003) With regard to phonemic perception, Oh et al. (2003) found that adults who regularly heard Korean during childhood performed better in their perception of Korean stops than those who had no exposure to Korean until college. In addition, low proficiency heritage speakers have been found to be able to perceive sentences even in noise (Au et al. 2008). In this study, low proficiency heritage speakers were compared proficiency heritage speakers with greater exposure to the heritage language during childhood), with typical adult L2 learners of Spanish, and with native Spanish speakers. All four groups were presented with five word sentences mixed digital ly with noise and were given four seconds to repeat out loud
63 exactly what they heard. The researchers found that low proficiency heritage speakers reliably outperformed the typical late L2 learners on this task and discussed the connection between mastery of phonological and higher level contextual information (e.g. semantic, grammatical) and the ability to decipher a sentence presented in noise which would suggest phonological mastery in low proficiency heritage speakers (Au et al 2008). Regarding highe r proficiency heritage speakers, t here is mixed evidence as to whether or not they perform better than typical adult learners and low proficiency heritage speakers on phonological perception tasks. In the aforementioned study, Au et al. (2008) found that w hile both the childhood speakers and the childhood over hearers reliably outperformed the typical adult L2 learners, the childhood speakers only marginally outperformed the childhood over hearers in perceiving sentences in noise. However in studying the pe rformance of Korean English bilinguals and adult Korean learners of English in detecting foreign accent in monosyllabic English stimuli, Park was able to detect a foreign accent f rom hearing stimuli consisting of simply the vowel / ( Park, 2009). This led him to propose a perceptual model for foreign accent perception among listeners w ith different L2 experiences in which listeners with limited L2 experience use L1 categories rather than L2 categories as the norm in foreign accent perception while listeners with enough L2 experience use L2 categories, leading to surprisingly better for eign accent perception by listeners with limited L2 experience (Park, 2009)
64 In summary, even low proficiency heritage speakers have been found to demonstrate greater phonological competence than traditional adult L2 learners, but below those of native mo nolinguals speakers in both perception and production. Higher proficiency heritage speakers have been found to slightly outperform low proficiency heritage speakers in phonological perception tasks, though this difference was not significant. Many of the s tudies discussed in this section have investigated the phonological competence of heritage speakers in comparison with their competence in morphosyntax, namely the morphosyntactic inflections of gender, number, and person. Their findings in this area will be discussed in the next section. Heritage Speakers and Gender Agreement Like phonology, morphosyntax is readily acquired by children but difficult to master by adult learners (Newport, 1990; Snow & Hoefnagel Hohle, 1978) and is therefore a good candidate for revealing any lasting befits of childhood language experience (Au et al., 2008). Previous studies have utilized native speaker knowledge of gender as a baseline from which to compare advanced, intermediate, and beginning L2 learners of Spanish (Keating 2009; Franceschina, 2003; 2005) and have shown that advanced L2 learners of Spanish show native like sensitivity to violations of gender agreement, while intermediate and beginning learners do not perform like native speakers on experimental tasks includ ing eye tracking (Keating, 2009) missing word tasks, cloze tests, grammaticality judgment tasks, and gender assignment checks (Franceschina 2005). In studies on the command of gender agreement in bilingual speakers, however, Montrul and Potowski (2007) fou nd that Spanish English bilingual children lag slightly behind monolingual Spanish speaking children in accuracy on gender agreement. Montrul, Foote, and Perpin (2008) found that gender agreement
65 is problematic for both L2 learners and heritage speakers. However even when both L2 learners and heritage speakers had incomplete knowledge of the language, the incidence of native speaker competence was higher in the heritage speaker group than in the L2 learner group (Montrul, Foote, & Perpian, 2008). Simila rly, Lipski (1993) found that low proficiency heritage or Spanish speakers we re aligned with second language learners with regard to While he found that errors of gender and number agreement we re quite frequent in productio n, Lipski did not assess the perceptive ability of these speakers. One may question why the lowest proficiency group sh ould be expected to have any knowledge on morphosyntactic inflections of gender given that they have such low proficiency in Spanish. How ever, there is evidence that receiving receptive training in a language could yield benefits in the perception of gender agreement. In a study on whether second language grammar can be learned through simply listening, De Jong (2005) provided 18 Dutch stud ents with receptive training only and 19 students with both receptive and productive training in Spanish. None of the participant s had had any prior exposure to Spanish, and no information about grammar was given to the learners in either group, therefore all grammar instruction was implicit. The target structure for this experiment was noun a dj ective gender agreement in a variety of production and perception tasks including a picture description task and a grammaticality judgment task. He found that partic ipants in both groups could to a large extent correctly identify correct and incorrect noun adjective gender agreement. In addition, he found that while the receptive training group showed faster processing times in comprehension tasks, they made a relativ ely large number of errors in production. De Jong concluded that the
66 receptive training had built a knowledge base in the learners that was available for comprehension, but it could not prevent errors in the production tasks. This finding suggests that per haps childhood over hearers of a language could benefit from having received receptive input only throughout childhood with little or no productive practice. T he findings of Au et al. (2002) however, are in contradiction with this suggestion They tested knowledge of phonology (as discussed in the previous section) and several aspects of morphosyntax, among them gender agreement among determiners, adjectives, and nouns in 12 adult L2 learners and 11 low proficiency heritage speakers of Spanish Participan ts completed oral elicited production and judgment tasks to assess their production of gender agreement and their performance in detect ing Results showed that the low proficiency heritage speakers were n om L2 learners in the production and perception tasks on marking gender agreement, and that both groups were outperformed by the native speaker control group in both tasks. These researchers concluded that there is no significan t morphosyntactic benefit to overhearing a language in childhood. In a follow up study, Au et al. (2008) co mpared the childhood over hearers group to a group of higher (as discussed in th e previous section) as well as typical adult L2 learners of Spanish and a native speaker control group in a narrative production task and a grammaticality gender agreement in noun phrases, they found that the native speakers reliably outperformed the other three groups in both the production and perception task.
67 However, the researchers found that on both tasks, childhood speakers were reliably outperformed by native speaker s yet the childhood speakers reliably outperformed the childhood over hearers and typical late L2 learners, whereas there was no reliable difference between the latter two groups (Au et al., 2008). They concluded that heritage speakers with greater experie nce speaking the heritage language during childhood seem better able to judge whether morphosyntactic markers, including gender markers, in that language are used correctly than heritage speakers with relatively no speaking experience during childhood. The se findings suggest that speaking a language during early childhood has lasting benefits beyond the domain phonology, even if the exposure to the heritage language was interrupted in later childhood years. In addition, speaking the heritage language during early childhood yields benefits beyond those of simply overhearing a language during childhood. These results have a substantial implication for receptive bilinguals, also called passive bilinguals, who are at one end of the bilingual range almost at the verge of culminating the language shift towards English monolingualism (Beaudrie, 2009). These speakers who tend to have receptive abilities in the language that allow them to comprehend oral and perhaps written language but have significant difficulties w hen producing the language (Myers Scotton, 2006). These speakers represent a highly important set of the Spanish speaking community in the Southwest United States in that many of the individuals that fit this description are actively involved in reclaiming their heritage language and reversing the language shift, as discussed in the previous section on NMS. What motivates these students and to what degree they are successful in part determines the future of Spanish in the Southwest (Villa & Rivera Mills, 20 09). However research on receptive heritage
68 language speakers is scant (Beaudrie & Ducar, 2005) despite predictions that this population will increase in the coming decades (Carreira, 2003). In summary, prior research suggests that heritage speaker compet ence in morphosyntax lie s along a continuum between typical adult L2 learners and native speakers of a language. The competence of low proficiency speakers at one end of the continuum has been tested to see if simply receiving exposure to the heritage la nguage can result in benefits where morphosyntax is concerned, particularly in the area of determiner/noun/adjective gender agreement. Although there is evidence that receiving receptive training in a language can yield benefits in recognizing errors in ge nder agreement, a study on the long term benefits of overhearing a language during childhood showed no benefits in the production or perception of gender agreement. However, measureable benefits were found in higher proficiency heritage speakers who actual ly produced the language regularly during childhood, even if the exposure to the heritage language ceased later in childhood. Summary T he linguistic competence of low proficiency heritage speakers, also called childhood over hearers or receptive bilinguals have been studied in comparison with L2 speakers, higher proficiency heritage speakers, and native speakers in order to determine whether there exists any benefit to simply overhearing a language in childhood as well as to gain better insight into the li nguistic abilities of individuals that lie along different points of the bilingualism spectrum The linguistic history of New Mexico, including the language shift that has occurred in the region, has yielded a high number of receptive bilinguals, making Ne w Mexico an ideal region to investigate the competencies of heritage speakers o f this type. However, it would be imprudent to do a
69 study on speakers of Spanish in New Mexico and not look at language attitudes toward the variety of NMS as a possible variabl e in assigning language ratings due to the prevalence of negative attitudes New Mexicans have traditionally held toward NMS. Nonetheless, the unique linguistic history of New Mexico make s this region a compelling area to study heritage speakers of various proficiency levels Research on the perception of non native speech has stemmed from two different sub fields: foreign accent perception and fluency perception. While the two fields generally investigate distinct speech variables that can indicate non nat ive speech, they have converged on the inquiry of temporal factors and linking, also called syllabic restructuring. Studies on both areas have shown the presence or absence of linking to be a significant factor in influencing listener ratings in both forei gn accent and fluency perception. Existing studies on linking as a signal of non native speech have utilized native speaker judges, whereas the current study seeks to examine the role of linking in heritage speaker perception processes. E mploying linking i n examining the performance of heritage speakers of varying proficiency levels could yield promising results. The field of heritage language development, in particular Spanish as a heritage language, is burgeoning and full of potential. As researchers sett le on a universal partial exposure to the heritage language during childhood and the impact on their potential for fully acquiring the language in adulthood continue t o emerge. Prior studies in the areas of heritage speaker competence hav e focused on their performance in phonological tasks in comparison with their performance on morphosyntactic tasks
70 namely their mastery of the morphosyntactic inflections of gender and number. While prior studies have investigated the perceptual competence of heritage speakers of various proficiency levels in the area of phonology in comparison with morphosyntax there are no data on the perceptual competence of these speakers when both variables are present simultaneously in the stimuli, in order to determine whether they rely more heavily on one variable over another. This is the gap that the present study seeks to fill. The inspiration for the current investigation stems largely from the findings of Au et al. (2008), who found that childhood speakers of Spanish reliably outperformed the childhood over hearers in the morphological tests of perception and production of gender agreement. With regard to phonology, they found that although the childhood speakers outperformed the childhood over hearers in both the perception and production tasks, the difference was slight and not statistically significant, suggesting a greater benefit for childhood over hearers in the area of phonology over m orphology. This and the remainder of the studies discuss ed in this chapter have provided the foundation for the current study, namely, the question of factors affecting heritage speaker judgments on speakers of the heritage language. In addition, the studi es discussed in this chapter have influenced the hypotheses that will be discussed in Chapter 3
71 CHAPTER 3 METHODS Research Questions and Hypotheses This dissertation examines listener and speaker effects in rating speech on a and in determining the language dominance of a given speaker. T hree different listener groups from New Mexico are examined : second generation heritage speakers of high Spanish proficiency, third generation heritage speakers of intermediate Spanish p roficiency, and third generation heritage speakers of low Spanish proficiency. The speech sample consisted of four New Mexican heritage speakers of high Spanish proficiency, each producing variants of sentences that resulted in four different conditions or combinations of the two linguistic factors under investigation: +/ connected speech (linking) and +/ errors in article noun gender agreement These two speaker produced factors compose the heart of the first two research questions, which are distinguishab le in that the first question investigates the of the speaker as the dependent variable whereas the second research question examines the ratings assigned to the stimuli (excellent, good, fair, or poor) as the indepen dent variable. In addition, the listener prompted effects of the perceived region of origin of the speaker and the listener attitudes toward different language varieties are examined as possible factors influencing language ratings, and drive the third and fourth research questions of this study, respectively. The research questions guiding this study and the corresponding hypotheses are detailed below:
72 R esearch Q uestion 1: How are linking and the presence or absence of errors in article noun gender agree ment related to the identification of a speaker as being Spanish dominant? T he pilot study upon which the current study is based examined the effect of various segmental and supra segmental factors on language ratings. The statistical analysis of the pil ot study found that only linking was highly correlated with the language ratings. In addition, multiple participants of the pilot study provided answers describing you n linking may be salient enough to describe and may play a significant role in the judgments of listeners. Another study that evaluated the phonological performance of heritage speakers in comparison with their morphosyntactic performance in both perception and production was that of Au et al. (2002). The researchers found both low and high proficiency heritage speakers to perform closer to L2 speakers than native Spanish speakers on gender agreement tasks, and closer to native speakers on phonological tasks. Based on these results, it is predicted that the participants of the present study will make judgments based on phonology rather than gender, thus it is predicted that stimuli within th perceived dominant language. Similarly it is predicted that In addition, it is predicted that there will be no difference between the outcome categori es of these two conditions regardless of the fact that one contains
73 errors in gender agreement and the other does not. Thus it is predicted that the independent variable of the presence of errors in gender agreement will have no effect on the outcome cate gory of the dependent variable. In other words, a speaker may be able to be perceived as Spanish dominant as long as their speech is fluid even if they produce gender agreement errors. These results are predicted to occur for participants of all proficienc y levels. R esearch Q uestion 2: How are linking and the presence or absence of errors in article noun gender agreement associated with language ratings? Listeners have been found to be able to identify foreign accent even in languages they do not speak, s uggesting the possibility of salient universal perceptual factors that come in to play in identifying foreign accent (Major, 2007) and part of this perception process may involve fluency of the speech Because prior research points to the salience of fluenc y in detecting non native speech, it is expected that listeners in the current study will be more mindful of the delivery of the speech than of errors in gender agreement in assigning language ratings. Therefore, i n the same vein as Hypothesis 1, it is exp ected that listeners will assign higher ratings on a four to sentences in the linked conditions than those not in the linked conditions even if the more fluid sent ences contain errors in gender agreement. Specifically, it is predicted that the condition This is expected because listeners may not notice the grammar of a sentence, even simple grammar errors, as long as it is spoken fluidly (Skehan, 2009). In addition, this outcome is expected among all listener proficiency level groups.
74 R esearch Q uestion 3: What is the nature of the relationship between langua ge ratings and the variety of Spanish believed to be heard on the speech sample? This research question investigates the relationship between the perceived region of origin of a given speaker and the language ratings assigned to that speaker Although th e speech sample consisted of four New Mexican heritage speakers, on this methodology will be discussed ahead in the chapter. As in other regions in which the code spoken, NMS has generally been a stigmatiz ed phenomenon in public language domains (Montes Acal, 2000; Valdez, 1997, 2000; Villa, 1996, 2004). In New Mexico, this can be evidenced by teaching methodology previously in use for language instruction at the university level, under which the original dialects brought by Spanish speaking or heritage students were eradicated in the classroom in favor of so called Enrquez & Hernndez Chavez 2003; Villa 2002). Villa (2004) notes that heritage students themselves or their teachers devalue their heritage language varieties and give them labels such as Spanglish mocho or slang which formed, and unsuitable for everyday communication, much less academic work. He adds that certain researchers involved in working with such students reinforce negative attitudes towards the
75 This negative attitude toward NMS by its own speakers, however, has been discussed in previous literature on NMS In their highly influential linguistic atlas of the Spanish of New Mexico and southern Colorado, Bills and Vigil (2008) refer to several at New Mexican speakers have toward standard Spanish note that ). The authors explain that the major nurturers of this myth are people from Mexico or other Spanish speaking countries who have been educated in Spanish and imbued with the strong prescriptivist tradition of the Real Academia Espaola (Royal Academy of Sp feelings about what is proper Spanish and find the local Spanish to be lacking, even to the point of being a corrupt and degenerate means of communication. Anglos who have acquired a good command of standard Spanish may sh are those attitudes for the same Based on this knowledge, it is hypothesized that listeners across all proficiency levels will assign lower ratings to stimuli that they identify as being spoken by a speaker of NMS R esearch Q uestion 4: How are language ratings associated with language attitudes towards NMS? This research question also investigates the relationship between the perceived region of origin of a given speaker and the language ratings assigned to him/her by listeners. How ever this question evaluates the role of the language attitudes toward NMS in assigning language ratings to speakers believed to be from New Mexico in comparison with listeners believed to be from the U.S. outside of New Mexico and those believe d to be from a Spanish speaking country other than the U.S.
76 The diverse set of linguistic influences and characteristics that were described in Chapter 2 (e.g. phonological reduction processes, lexical archaisms anglicisms, and Native American lexical in fluences) have contributed to the unique nature of NMS and established it as a non standard variety. As with other non standard varieties of Spanish that incorporate code switching, NMS frequently carries with it a stigma that directly results in negative language attitudes toward the variety even by its own speakers ( Montes Acal, 2000; Valdez, 1997, 2000; Villa, 1996, 2004 ). Because of this, it is hypothesized that listeners that demonstrate less positive attitude s toward NMS will assign low ratings for s timulus sentences in which they believe the speakers to be from New Mexico, and will assign higher ratings to these same speakers in stimulus sentences in which they believe them to be from regions outside of New Mexico. In addition, it is hypothesized tha t these results will occur among all listener groups due to the fact that members of all listener groups are subject to the stigma associated with NMS by nature of their New Mexico origin. With these questions in mind, the following research design was cr eated to evaluate the role of linking in the identification of a speaker as being Spanish dominant and in language ratings. Research Design Participants The participants involved in this study are heritage speakers of Spanish from New Mexico. They have bee n classified according to their proficiency level and linguistic generation or distance from contact generation ancestors. The participants are further detailed below:
77 High proficiency, s econd generation heritage speakers (n=18) : In dividual s with at least one contact generation parent. These individuals have spoken both English and Spanish regularly since early childhood. Their self identified L1 and dominant language may be English, Spanish, or both. Participants in this group attained a score of 80% or a bove on an independent measure of proficiency which will be discussed ahead in this chapter This group was composed of nine males and nine females, and had an average age of 62. These participants were recruited for participation in this study via snowba ll sampling within various northern New Mexico communities and were not, at the time of their participation enrolled in a Spanish course nor had they been enrolled in a Spanish course within the past ten years. This group represents the closest entity to a base line or point of comparison to which all other NMS speakers would be compared. Strictly speaking, there is no monolingual baseline for NMS speakers given that NMS is inherently a bilingual variety. The ideal baseline for NMS would be members of the c ontact generation of NMS. However, given the linguistic history of this region, these speakers would be one biological generation above the current participant group and thus likely average around 85 years old a nearly defunct population at the time of th is research Intermediate proficiency, t hird generation heritage speakers (n=18) : I ndividual s with at least one contact generation grandparent. These individuals have been exposed to Spanish th r ough parents, grandparents, and/or other family members since birth (i.e. heritage speakers, may be passive bilinguals). Their self identified L1 and dominant language is English. Participants in this group scored between 60 79% on the independent measure of proficiency. This group was composed of twelve females
78 and six males, and had an average age of 21. At the time of their participation, these participants were currently enrolled in an intermediate level Spanish course designed for heritage speakers and were recruited for participation in this study during their S panish class The fact that t his is the only participant group that was enrolled in a Spanish course at the time of their participation could influence the results of these analyses given that this is the only group to have received recent formal education in Spanis h and thus this group may prove to be better prepared to recognize errors in gender agreement and rate accordingly. However this result is not expected, as it has been hypothesized that all participant groups will render similar rating tendencies Low proficiency, t hird generation heritage speakers (n=18) : I ndividual s with at least one contact generation grandparent. These individuals have been exposed to Spanish though parents, grandparents, and/or other family members since birth (i.e. heritage speakers, may be passive bilinguals). Their self identified L1 and dominant language is English. Participants in this group scored 59% or below on an independent definition o who heard the home language in the background but never responded in it and were not ad This group was composed of ten females and eight males, and had an average age of 33. These participants were recruited for participation in this study via snowball sampling within various northern New Mexico communiti es and were not, at the time of their participation, enrolled in a Spanish course nor had they been enrolled in a course within the past five years
79 In order to determine if langua ge attitudes play a role in performance on tasks on detect ing native/ non n ative speech, it is important to investigate an area that has a documented history of language attitudes towards different varieties of Spanish. This study examines in further detail the traditional assertion that speakers of NMS express negative attitudes toward their own variety (Bills, 1997), and therefore all members of the three listener groups were recruited from New Mexico. Given that Bills and Vigil (2008) have identified two different varieties of NMS (that of northern New Mexico and southern Color ado, and that of southern New Mexico), which differ in terms of social identity and lexicon (p. 6), all participants for this study were recruited from northern New Mexico in order to maintain homogeneity. A power analysis conducted prior to the experimen t revealed that a minimum of 16 participants per listener group were required to yield potentially significant results for a data set containing 60 items as in the current study. For good measure, 18 participants were recruited in each of the listener grou ps: high, intermediate, and low proficiency Participants in the high and low proficiency groups were recruited via snowball sampling and were remunerated $10 in cash for their participation 1 Participants in the intermediate proficiency group were recruit ed by the researcher during their class time and were asked to fill out a sign up sheet to meet with the researcher individually at a mutually agreed upon time outside of class; they were also remunerated $10 in cash for their participation. 1 Funded by the University of Florida Department of Spanish and Portuguese Studies 2011 Doctoral Student Summer Scholarship Award
80 All potential participants were asked to fill out a language background questionnaire (Appendix A), prior to their selection for participation in this study. In addition, in order to determine the proficiency level and to establish to which category each belonged, pote ntial participants were asked to complete an independent measure of proficiency (Appendix E) These materials are described below. Language background questionnaire Potential participants of the speaker sample and the three listener groups were asked to c omplete a language background questionnaire in order to ensure the most homogenous groups possible. The language background questionnaire was administered on paper in English to all potential participants, and elicited information relating to demographic d ata and sociolinguistic background such as gender, age, place of birth, birthplace of parents and grandparents, and possible residence in other languages spoken in the home during childhood, L1 of parents and grandparents, and language use, participants were asked to indicate how often they write, speak, listen to music, read, and wa tch television or movies in English and in Spanish, following a 7 point scale ranging from never (1) to every day, almost all day (7). Participants were also asked to select which language they felt was their strongest and were given the assessment in English and Spanish for each of the four language skills (reading, writing, comprehension an d speaking) as well as pronunciation ability, grammar ability, and overall ability. The scale ranged from minimal (1) to native like (4).
81 Independent measure of proficiency While participants filled out a language proficiency self assessment as part of th e language background questionnaire, we also wanted a less subjective measure of their linguistic competence To that end, potential participants completed a proficiency test (Appendix E) whose scores were used in conjunction with the language background questionnaire to place participants in to the appropriate listener group. Potential speakers to comprise the speech sample also completed the proficiency test in order to ensure homogeneity within the proficiency level of the speakers of the core sample. Th e proficiency test was a modified version of the Diploma de Espaol como Lengua Extranjera (DELE) proficiency test (http://nhlrc.ucla.edu/data/proficiency assessments example proficiency exams.asp), which has been widely used in recent SLA studies (Montrul Foote & Perpin, 2008; Montrul & Bowles, 2010; Montrul & Slabakova, 2003; Cuza & Frank, 2011; Prada Prez & Pascual y Cabo, 2011; Guijarro Fuentes & Marinis, 2011; Slabakova, Rothman, & Kempchinsky, 2011; White, Valenzuela, Kozlowska MacGregor & Leung, 2004; Montrul, 2004; Rothman & Iverson, 2008; among others). proficiency in Spanish and organize participants according to proficiency level. The task of measuring of se cond and third generation bilingual proficiency is a challenging one, made even more difficult in an area that has a documented history of a sense of language inferiority, as was discussed in Chapter 2 (e.g., Bills & Vigil, 2008). Therefore, administering a traditional proficiency test could have been inappropriate for this particular population. In order to remain sensitive to this delicate situation, the DELE was modified, as detailed below, to suit the population being studied.
82 Polinsky and Kagan (2007 ) propose lexical proficiency as an accurate diagnostic of overall language proficiency, and point to studies (e.g., Polinsky 1997, 2000, 2005, lexical items and that control of grammatical phenomena such as agreement, case marking, aspectual and temporal marking, and use of subordinate clauses (Polinsky & Kagan, 2007). They note that this correlation has been supported by studies investigating a variety heritage langu ages, (Lee, 2001; Godson, 2003), including Spanish (Montrul, 2006). With this in mind, only the multiple choice vocabulary section of the DELE was utilized in the proficiency test for the current study, in order to avoid complications associated with accep table variation within the NMS variety in the cloze test section of the DELE. Furthermore, the vocabulary section was reduced from 30 items to 20 in the interest of respecting the time commitment of the participants. Polinsky and Kagan (2007) argue that it is unrealistic to expect heritage speakers of one variety of Spanish to have rudimentary knowledge of other varieties, not to mention the standard language promoted by schooling, the media, and literature. Instead, the y propose using a baseline determined by the language or variety that the speaker was exposed to as a child. To this end, the vocabulary section of the DELE was modified in the present study specifically for speakers of NMS. After consulting with three second generation speakers of NMS, 26 le xical items on the multiple choice vocabulary section were replaced with items that the consultants agreed conformed to the lexicon of NMS and were likely to be understood by speakers of NMS with even limited fluency.
83 In adapting the scoring of the profici ency test from the 50 item test used in the studies cited above for the 20 item test used in the current study, scoring was converted to percentages, while the thresholds were maintained. Therefore, potential participants who achieved 80% 100% correct scor es were considered advanced proficiency speakers, those with 60% 79% correct were considered intermediate proficiency speakers, and those scoring 0 59% correct were considered low proficiency speakers. Tasks and Materials Audio sample The audio sample use d in this investigation was comprised of the elicited speech of four speakers, who read aloud pre written sentences. To the extent possible, the intent was to compose a sample of a homogenous speaker group to eliminate variables such as generation, profici ency level, and speaker age as factors in assigning language ratings. To that end, all potential members of the core speaker group were asked to fill out a language background questionnaire based 2 on that of de Prada Prez (2009; Appendix A) in which they provided information such as L1, frequency and nature of exposure to Spanish, and familial history of Spanish. In addition, in order to determine the proficiency level and to which category speakers belonged, potential participants were asked to complete t he same independent measure of proficiency that was completed by listener participants. In striving for homogeneity, all four speakers included in the core audio sample were second generation bilinguals, identifying bilingual parents and monolingual Spanis h speaking grandparents in their language background questionnaire. All 2 In addition to the questions asked by de Prada Prez (2009), the questionnaire used in the current study
84 speakers resided in the Albuquerque area and reported having parents that speak varieties of Mexican Spanish, thus, from outside the Northern New Mexico speech community. Due to the ge nerational language shift that has taken place in the region, second generation bilinguals from within the community would likely be older speakers (50+ years old). In an effort to avoid bias toward older speakers (i.e. listeners rating speakers as highly proficient simply because they sound older) younger speakers were solicited for the speech sample. The trade off, however, is that they have parents that speak a variety other than NMS. The effect of this choice is believed to be minimal, given that the pu rpose of the investigation was not for the listeners to truly identify the origin of the speakers in question but simply to identify the factors present (linking and / or gender agreement) when they believed the speaker was from New Mexico and other places All four speakers were between the ages of 20 29 and identified both English and Spanish as their (simultaneous) first and dominant languages. In addition, all four speakers of the core sample scored above 90% on the independent measure of proficiency. S peakers were recruited via snowball sampling and did not participate in the listening task In a review of factors that affect perceived degree of foreign accent, Piske, sign Fletcher, 1992; Olson & Samuels, 1973; Purcell & Suter, 1980; Snow & Hoefnagel Hhle, 1977; Suter, 1976). As such, gender was not among the factors considered, but was controlled for by including two male and two female speakers in the core sample.
85 In addition to the core sample, distracter sentences spoken by speakers from outside the core group were also included. T he distracter speakers were included in the sample to help avoid the overuse of stimuli spoken by the core group, thus reducing the possibility that the listeners would become familiar with the same four speakers and assign them the same ratings that they had assigned them previously due s imply to recognizin g the voice. Additionally the inclusion of speakers from outside the target detecting origin (and thus, in group inclusion/exclusion). Distracter sentences were provided by n ine different speakers from both within and outside of the northern New Mexican speech community. The distracter speaker group included: two second generation New Mexican heritage speakers 3 one third generation New Mexican heritage speaker, two second gen eration Floridian heritage speakers, two L2 speakers of Spanish and two native Spanish speakers (Panamanian and Colombian). Like the core speaker group, the distracter group included both male and females speakers, with four male speakers and five female speakers. Speakers included in the distracter group were of varying proficiency levels in order to provide variety within the overall speech sample. Table 3 1 illustrates the demographics of the speaker group and the make up of the speech sample 3 The two second generation New Mexican heritage speakers differ from those used in the core speaker group in age: the average age of the two second generation New Mexican heritage speakers was 62 years old whereas the average age of the core speaker group was 26 years old. In addition, one of the second generati on New Mexican heritage speakers in the distracter group was a low proficiency speaker whereas all other second generation New Mexican heritage speakers (in both the core and the distracter group) were high proficiency speakers of Spanish
86 Table 3 1. Distribution of speakers within audio sample. Speaker # Gender Linguistic generation Proficiency level Region of origin Total # stimulus sentences in sample Core 1 M 2 High New Mexico 10 (8 core + 2 distracter) Core 2 M 2 High New Mexico 10 (8 core + 2 distracter) Core 3 F 2 High New Mexico 10 (8 core + 2 distracter) Core 4 F 2 High New Mexico 10 (8 core + 2 distracter) Distracter 1 F 1 High Colombia 2 Distracter 2 F 2 Low New Mexico 2 Distracter 3 M 2 Intermed Florida 2 Distracter 4 M 1 High Panama 1 Distracter 5 F (L2) High Ohio 1 Distracter 6 M (L2) High Florida 1 Distracter 7 F 2 High New Mexico 8 Distracter 8 M 3 Low New Mexico 1 Distracter 9 F 2 Intermed Florida 2 This dissertation utilized elicited speech of read aloud sentences b y each of the four core speakers producing eight sentences. Each sentence contained one token of an [s] + [a] sequence across article noun word boundaries embedded within the target sentences in order to provide an opportunity for syllabic restructuring in the target phrase. For example, the phrase mis amigos + /a.mi.gos/ would be resyllabified as /mi.sa.mi.gos/. A ll stimuli were presented as an [s] + [a] sequence across article noun word boundaries (as opposed t o, for example, noun/verb word boundaries) to limit variability among the tokens. In addition, t he [s] + [a] sequence was chosen because of the ample possibilities it provided in creating stimuli : in the interest of utilizing simple lexical items, there we re a greater number high Additionally, t he vowel [a] was chosen as the initial segment for the nouns because this vowel has been shown in
87 previous research to be unaffected by lexical stress among speakers of Spanish in the southwest United States (Willis, 2005) thus eliminating one possible variable The full transcript of the read aloud elicitation task can be found in Appendix B. Each of the four core speakers read aloud six different versions o f 10 pre written sentences. These different versions included the following variations: Table 3 2 Sentence variants in audio sample including distracter variants Variant Gender agreement /s/ + /a/ realization Condition 1 Error [s.a] Disfluency with agr eement errors 2 No error [s.a] Disfluency without agreement errors 3 Error [sa] Fluency with agreement errors 4 No error [sa] Fluency without agreement errors 5 No error [za] 6 No error Variants 5 and 6 were produced in order to provide fur ther distracters in the event that listeners began to suspect the purpose of the experiment. Speakers were told to review the sentences beforehand to make sure they were familiar with all the words and were told to begin again from the beginning of the sen tence if they committed any errors while reading any given sentence. Speakers were told they could record each sentence as many times as they needed to until they felt comfortable with what they had produced. This was done in an effort to limit disfluencie s in the final speech sample. Participants were given a script to read from which indicated either a fluid sentence, evident by the lack of punctuation between the words of a given stimulus sentence or a sentence in which tokens of [s] + [a] were not link ed, evident by the use of a comma between the initial article and proceeding noun in the script Linking of [s] and [a] or the
88 lack thereof in each stimulus sentence was verified by the researcher in the spectrogram of each sentence in the acoustic analysi s program PRAAT ( Boersma & Weenink, 2010) manifested as either a continuous voiced waveform (linked token) or as a stop (unlinked token). Of the 240 sentences recorded by the four speakers, the core speech sample was pared to 32 item s by selecting two sen tence groups in which a speaker produced the first four 4 versions or conditions of each sentence with little or no disfluencies so as to avoid the effect that such disfluencies as hesitations, false starts, and fillers would have had in signaling non nativ e speech. In the event that a speaker produc ed more than two sentence groups that containe d no disfluencies, two sentence groups were selected at random to include in the sample and no sentence was presented in the sample as having been read by more than o ne speaker It was important to include all four conditions within each sentence group by any given speaker in order to determine whether linking or the presence of a gender agreement error had any effect on the ratings that a given speaker obtained, the a ssumption being that all other speech variables (such as speed, intonation, phonemic articulation) would remain equal when produced under the same conditions by the same speaker. Thus each of the four core speakers contributed eight core sentences to the f inal speech sample. In addition to the 32 core stimulus sentences, the final speech sample heard by the participants included 28 distracter sentences for a total of 60 items as illustrated in Table 3 1 Unlike the core speech sample, distracter sentences did not contain grammatical errors or intentional pauses in speech. Any variation in the distracter 4 The last two ve rsions, versions 5 and 6 shown in Table 3 2, were distracter versions and were only included in the final speech sample when necessary to space out the actual stimulus sentences.
89 sentences was due to variation in speaker proficiency and region of origin. Twenty of the 28 distracter sentences were true distracter sentences, spoken by nine different speakers outside of the core speaker group. Of these distract e r speakers, four speakers provided one distracter sentence, while four speakers provided two sentences One distracter speaker was initially slated to be a part of the core speake r sample, but was determined to have a very distinguishable voice in the pilot study and was omitted as a core speaker out of concern that listeners would recognize her voice and simply assign her the same ratings for all stimuli without regard to the vari ables distinguishing them. Although she was replaced by another core speaker for the current study, all eight of her sentences remained in the audio sample as distracter sentences. The remaining eight distracter sentences were spoken by members of the core speaker group, with each core speaker producing one of each of the distracter variants (variants 5 and 6 as detailed in Table 3 2 ). Sentence lengths were measured in the acoustic analysis program PRAAT prior to the creation of the speech sample to ensure that the two stimulus sentences of each type (+pause/ pause) produced by the same speaker differed by no more than 0.5 second s in terms of their overall duration. In two instances, the difference between two sentences produced by the same speaker exceeded 0.5 seco nds, and i n these cases, the excess pause time was manually removed (using PRAAT ) in order to maintain consistency across the stimuli produced by each speaker. The pauses that were intentionally produced by the speakers between the sentence initia l article and were also measured to ensure a consistent length of 50 milliseconds The majority of instances, pauses were
90 longer than 50 ms and i n eac h of these instances, the pause duration w as reduced to 50 ms in PRAAT prior to being inserted into the final speech sample. In one instance, the pause measured 48 ms; this stimulus sentence was included in the speech sample unedited. Additional measurements and analyses of the stimuli were compl eted post hoc in order to investigate whether segmental and temporal variables external to linking were present and could have influence d the results of the study. Recall that each speaker contributed four conditions of two different sentences for a total contribution of eight stimulus sentences per speaker in the final speech sample. The post hoc analysi s examined all four conditions of one randomly selected sentence produced by each speaker T hus 50% of the stimuli were examined to ensure consistency wit hin the stimuli created by each speaker T he following segments were analyzed in PRAAT in order to verify their production was comparable across all four sentence conditions for the same speaker: VOT in voiceless stops /p, t, k/; tonic vowel length; artic ulation of laterals /l/; lenition of voiced stops /b, d, g/; and articulation of rhotics / / vs. / / These segments were selected on the basis of feedback from participants of the pilot study (Trujillo, 2011) who indicated these to be factors that led the m to identify speakers as either native or non native speakers of Spanish based on the sounds or letters they said they notice d as either native or non native like It is noteworthy that many of these same segments have been shown to be responsible for th e presence of a notable foreign accent in L2 speech (Lord, 2005). In addition to verifying the variables were comparable within speakers, t he analysis of the production of these segments helped to
91 confirm that all stimuli was essentially native like with the exception of the variables under analysis. Means and s tandard deviations were calculated for the VOTs of each token of /p, t, k/ produced by each speaker as well as for the tonic vowel lengths, and first and second formants of /l/ tokens The quotient of the mean and standard deviation were calculated for each token of each speaker in order to provide a means of comparison for the standard deviations For each token quotients under 20% were deemed acceptable variation. It was found that the standard d eviations for all tokens of VOT did not exceed 3.08 ms with the highest deviation quotient being 16 %. A ll VOTs were 28.4 ms or shorter, which falls within the range of standard VOT values for native like Spanish 5 (Lisker & Abramson, 1964). F or length of t onic vowels t he greatest standard deviation across all speakers was 9.83 ms with the highest deviation quotient being 10 % First and second formants of the sentence initial laterals /l/ were measured and it was discovered that no standard deviation excee ded 67.59 Hz for F1 formants and 205.45 Hz for F2 formants produced by a given speaker resulting in deviation quotients of 16% and 14%, respectively Native like l enition of v oiced stops /b, d, g/ was determined in PRAAT by the absence of a closure in th e waveform and spectrogram ; alternately, the presence of a closure would have indicate d a stop in the airflow and thus a non native like occlusive sound (Lord, 2005) The presence of lenition was verified in each possible occurrence within each stimulus se ntence. Similarly, the articulation of the segment /r/ was verified in each possible instance as the native like alveolar tap [ ] ( versus the approximant [ ] ) 5 VOT values for native like Spanish range from 4 to 29 ms (Lisker & Abramson 1964).
92 as evidenced by a closure in the waveform. These measurements and analyses allow us to be confid ent that the stimuli created by an individual speaker were both native like and comparable to each other with the exception of the variables under investigation, thus minimizing external factors that may influence the results of this study. Speakers were r ecorded with a Shure SM10A microphone directly into the audio recording program Audacity on an Asus 1005 PEB Netbook. Audio recordings were stored as .mp3 files and embedded into a user friendly PowerPoint presentation. The speech sample was presented to l isteners over stereo headphones (Sennheiser HD 202) to limit variability among listeners in the quality of speakers through which they heard the speech sample. Stimulus sentences were randomized and organized onto the PowerPoint presentation within six sl ides or modules, each containing ten sentences. On each slide, audio icon next to each label, as in Figure 3 1. Each stimulus sentence was presented under a different sp eaker number, and sentences produced by the same speaker were presented a minimum of four sentences apart under different speaker numbers. Figure 3 1. Sample slide of PowerPoint presentation as presented to listeners
93 The modules were organized into thre e different orders, and approximately one third of the participants from each listener group listened to the speech sample under Order 1, one third listened to Order 2, and one third listened to Order 3; this was done in order to counterbalance any potenti al effects of practice or fatigue on the part of the listeners. Listeners were instructed to click on each audio icon within the PowerPoint presentation and listen to each stimulus sentence no more than twice, and then to fill out a rating sheet for each s entence. The rating sheet and other materials are described below. Rating s heet Prior to hearing the sample and completing the rating sheet, participants were given an alternate explanation of the task; they were told that the sample they were about to hea r was composed of speakers of varying proficiency levels who were given sentences with intentional grammar errors, and that the speakers were told to correct the error as they read the sentence out loud into the microphone. Participants were told that spea kers of lower proficiency levels were unable to locate and thus correct the grammar error, and therefore the speech sample would contain some sentences with grammar errors. They were told that speakers were given multiple opportunities to locate and correc t the error, thus explaining why they might hear particular speakers reading the same sentence multiple times throughout the sample. Participants were explicitly told that they could take into account the presence of grammar errors (or lack thereof) when a ssigning the language rating to each sentence. This was done not only justification in assigning ratings, but also to alert them to the fact that grammar errors might be present in the sample. Multiple participants in a pilot study indicated that
94 although they thought they heard errors in gender agreement in certain stimuli, they did not take these errors into account while rating the speech of the individual speakers because they figured the speaker had simply been reading a pre prepared sentence and were thus told to make the error (as they were). In addition, participants of the pilot study indicated that they recognized some of the speakers from previous stimulus se ntences and assigned them the same rating in striving for consistency, which they thought was the aim of the study, without taking into account the factors under investigation that differentiated the stimuli For this reason, participants of the present st udy were alerted to the possibility of grammar errors but were not informed of the exact nature of the error (gender agreement errors) nor were they told whether an error was present in each stimulus sentence This p roved to be beneficial for participants with lower Spanish proficiency who may have second guessed their overall knowledge of Spanish and would have simply assumed that errors in gender agreement were correct simply because they were being read by someone they believed to be a native speaker. T he rating sheet (Appendix C) asked participants The next it provided a space to provide further information if needed if they chose the latter. Particip ants were told that other could refer in general terms to than one answer in this category if needed.
9 5 The next item asked listeners to provide a global rating of the Spanish of the speaker of each stimulus sentence. Global ratings were elicited in this study in favor of more specific ratings ( such as for phonology and syntax ) in order to learn more about the relationship between the factors of li nking overall, general impression of the speech and speakers in the sample. Participants were not told what aspects of language to consider in this rating, but were given the options chosen rather than a numerical assessment in order to facilitate the rating process for the listeners. The four point categorical scale was used because it not only captures a qualitative distinction in terms of degree of overall language ability (Alba Salas, 2004), but it also allows for a straightforward binary comparison of positive and negative responses by collapsing the selection from four to two. Furth ermore, a four point scale eliminates the possibility that the listener will select a middle response out of indecisiveness or an attempt to remain neutral you notice in their optional, and were instructed to fill it out only if any words or sounds caught their attention while listening to the sentence. This section was included in the rating sheet in order to prov ide a space for any additional information participants wanted to share, and also to gauge the extent of metalinguistic awareness of the participants. Finally, listeners were asked to speculate on the place of birth of the speaker of each stimulus in the s ample in order to determine the Spanish variety they believed was being spoken. In determining place of birth, listeners were given the lead
96 of New Me While the reality is that Spanish speakers could come from hundreds of locations, options for this question were simplified to five categories so as to facilitate the choices for the listene rs and simplify the determination of in group affiliation or exclusion. The options of that New Mexican listeners were likely able to identify either because of proximity wi th speakers of that variety (e.g., Mexico) or because of well known linguistic characteristic s (e.g., i n ) Although many participants may not have been in contact with speakers of Castilian Spanish, this does not strictly imply that th ey were not familiar with salient characteristics of this variety. In studying accent recognition and language attitudes in South Africa, Smit (1996) asserted that a specific population group can, despite the fact that a specific accent might only be seldo m heard as an option in the as discussed in C hapter 1, in group affiliation vs. out of group distinction. For this reason, the five answer options for speaker region of origi n were collapsed in the analysis for Research Question speaking country Language a ttitudes assessment The present research also sought to investigate the relatio nship between the oral proficiency ratings and language attitudes of the listener group, in order to examine
97 whether language attitudes play a role in how heritage speakers perceive and rate the speech of others. In order to answer this question, it is imp ortant to consider previous works in this area. Traditionally, New Mexican speakers of Spanish have expressed negative attitudes toward their own variety of Spanish (Bills, 1997). However in order to further investigate this claim, the final section of th e language background questionnaire contained a language attitudes assessment (Appendix D) partially based 6 on that used by Bugel (2009). In an effort to avoid priming answers the participants thought the researcher was looking for, respondents were not ex plicitly told that this section of the questionnaire was focused on language attitudes. Instead, the top of the assessment In this assessment, questions asked about familiarity with different varieties of Spanish and any differences that they could perceive between other varieties of Spanish and the Spanish spoken in given space to elaborate on this choice. The assessment also asked participants the following questions: Which dialect(s) of Spanish do you think should be reported on the local Spanish news? Which dialect(s) of Spanish do you think should be taught in local Spanish classes? Which dialect(s) of Spanish would you like to speak? Other (Puerto Rican, Cuban, 6 Questions 2 and 4 are based on questions from Bugel (2009).
98 describe the Spanish of New Mexico and to describe, if familiar, other varieties of Spanish, in an attempt to elicit either positive or negativ e descriptions from the point of view of the participants. The analysis of answers provided on this survey helped to determine whether participants demonstrated positive or negative attitudes toward their own and other varieties of Spanish and whether or n ot there was a connection between these attitudes and the ratings they assigned to speakers based on what variety of Spanish they believed was being spoken. ended items on this survey are listed in Appendix G and will be d iscussed in further detail in Chapter 4. Data Analysis The following section discusses the methodology for analyzing the data and thus in evaluating the outcomes of the aforementioned research questions. Factors in assessing language dominance and accent r atings The first research question asks how linking and the presence or absence of errors in article noun gender agreement are related to the identification of a speaker as being Spanish dominant T o address this question, stimulus sentences were coded num erically from 1 4 according to variant or condition as demonstrated in T able 3 3 : Table 3 3. Sentence variants in audio sample Variant Gender agreement /s/ + /a/ realization Condition 1 Error [s.a] Disfluency with agreement errors 2 No error [s.a] D isfluency without agreement errors 3 Error [sa] Fluency with agreement errors 4 No error [sa] Fluency without agreement errors
99 E (Englis h), S (Spanish), or B (both) and constituted the dependent variable, thus both the independent and dependent variables were categorical/nominal Since the methodology used in this study resulted in multiple observations per listener, neither a chi square t est nor a log linear analysis could be used to determine whether any association existed between the independent and dependent variables. In their place a multinomial logistic regression model was used. The multinomial logistic regression model is an odds ratio measure which is appropriate for analyzing repeated measures for categorical data (Agresti, 2002). A multinomial logistic model that calculates odds ratios in order to examine the nature of the relationship between variables was chosen In general, t he model determines how the risk of the outcome falling into the comparison group compared to the risk of the outcome falling into the referent group changes with the variable in question (UCLA: Stat Consulting Group, n.d.). The nature of the multinomial l ogistic model is such that it requires one possible outcome, or dependent variable, to be designated as the baseline to which another outcome is compared. The three possible outcomes for this analysis were: 1) that a speaker would be identified as Spanish dominant; 2) that a speaker would be identified as English dominant; and 3) that a speaker would be identified as being equally proficient in both languages. For this research question, the outcome of Spanish ( S ) was chosen as the baseline. The multinomial logistic regression model first compared English ( E ) to the S baseline and then compared Both ( B ) to the S baseline. independent variable (condition) on the odds that liste ners would 1) rate a speaker as
100 English dominant ( E ) versus Spanish dominant ( S ) and 2) rate a speaker as equally proficient in both languages ( B ) versus Spanish dominant ( S ). A similar statistical analysis was used in answering the second research questio n which asked how linking and the presence of grammar errors are associated with language rating s. As stated above, the four stimulus conditions shown in Table 3 3 constituted the independent variable. The four categories of accent ratings were coded nume dependent variable As in Research Question 1, a chi square test could not be used to analyze the repeated response data. In its place, logistic regression appropriate for an alyzing repeated measures for categorical data was necessary (Agre sti, 2002). However since in this case the dependent variable was categorical/ordinal, an ordinal regression analysis based on the cumulative logit model was computed in order to determine the strength of the association between the independent and dependent variables. This ordinal regression analysis is a cumulative logit model that calculates the probability that a listener will assign a particular rating or lower given a set of independen t variables. In this case the independent variables in question are condition (as shown in T able 4 1) and participant group. In general, this model calculates ordered log odds (logit) regression coefficients, which can be interpreted as the expected level of change in the response variable for a one unit increase in the predictor in the ordered log odds scale while the other variables in the model are held constant (UCLA: Stat Consulting Group, n.d.). The model is based on the assumption that there is a lat ent continuous outcome variable and that the observed ordinal outcome arises from discretizing the underlying continuum into ordered groups, labeled as j
101 j coefficients, which are the thresholds that estimate these cutoff values. The threshold estimates represent the response variable in the ordered logistic regression and are the cutoff value between two differen t categories of this variable when values of the predictor are evaluated at zero (UCLA: Stat Consulting ession analysis gives inform ation on the odds that a listener will assign a particular rating or lower, for any given condition (Agresti, 2002) Correlation between region of origin perception and accent ratings The third research question asked about the nature of the relationship between language ratings with the variety of Spanish believed to be heard on the speech sample In addressing this question, answers to the following item on the rating sheet, s N (New Mexico), U (U.S. outside of New Mexico), while the three remaining categories (Mexico, Spain, and Other) were collapsed into one general category, O and constituted one of the independent variables ; the proficiency level of the listeners constitu ted the other independent variable As stated in the previous section, answers on the ratings section data set constituted the dependent variable. The dependent variab le was categorical /ordinal; therefore an ordinal regression analysis based on the cumulative logit model was used to calculate the probability that a listener will assign a particular rating or lower given a set of independent variables. The predictive na ture of the model is dependent on threshold values that organize the continuous outcome variable into ordered groups. The logit regression coefficients produced by the model as well as
102 their placement within the threshold values also produced by the mode l render a likely Correlation between accent rating and language attitudes The fourth research question asked h ow l anguage ratings are associated with positive/negative language attitudes towards NMS To answer this question, answers provided on the language attitudes assessment (Appendix D) were quantified categorically per question: each of the three dialect preferen ce questions, questions 3 5, garnered one point when the participant indicated a preference for NMS, even in cases where NMS was indicated as an answer in conjunction with other varieties, as how participants were allowed to select more than one answer. Qu this question if they expressed a positive answer in describing NMS. Only questions 3 6 were included in the analysis due to the binary nature of possibl e responses (positive vs. negative attitudes toward NMS). Questions 1, 2, and 7 were included in the Spanish; however this information was not included in the analys is. In tallying the four answers included in the analysis for each participant, each listener was assigned an attitudes value ranging from 0 4. The resulting language attitudes value (LAV) for each listener was one independent variable and the variety of S panish they believed they were listening to ( N, U or O ) was a second independent variable. Together, these two independent variables constituted the listener type. The listener type was then compared to the numerically coded language ratings (dependent va riable) for each stimulus sentence as assigned by each listener. Since the dependent variable was
103 categorical/ordinal, an ordinal regression analysis based on the cumulative logit model was computed in order to determine the strength of the association bet ween the independent and dependent variables. In this case, the ordinal regression analysis gave information on the odds that a listener will assign a particular rating or lower, for any given listener type (Agresti, 2002) Summary In this chapter the res earch questions that guide this study have been discussed, as well as the research design upon which this study is based. It has been hypothesized, based on the results of the pilot study and on previous research in related areas, that linking will have a stronger statistical association with the identification of a speaker as being Spanish dominant as well as with language ratings than will the presence of errors in gender agreement across all three listener groups of varying proficiency levels. In additio n, it has been hypothesized that listeners will assign lower ratings to stimuli that they identify as being spoken by a speaker of NMS and that there will be a moderate to high statistical association between listeners that have a negative attitude toward NMS and low ratings assigned to speakers that are believed to be from New Mexico. Chapter 4 provides a summary of the results that were obtained utilizing the aforementioned methods and data analyses, and a thorough discussion of the results will be presen ted in Chapter 5
104 CHAPTER 4 RESULTS In this chapter the results of the data analyses are presented. Quantitative data are presented in the following order: 1) the relationship between the factors in question, linking and the presence of gender agreement errors, and the identification of a speaker as being Spanish dominant; 2) relationship between the factors in question and the language ratings; 3) the relationship between language ratings with the variety of Spanish believed to be heard on the speech sam ple; and 4) the relationship between language ratings and language attitudes towards NMS. Next, a qualitative assessment of answers provided in the additional comments section of the language rating sheet and the attitudes assessment is presented, and the relation of these comments to the quantitative data are also discussed. The chapter concludes with a summary of the findings of this study while a thorough discussion of the implications of the results follow s in C hapter 5 Quantitative Results Factors I n volved in L anguage D ominance I dentification The first research question asked how linking and the presence or absence of errors in article noun gender agreement relate s to the identification of a speaker as being Spanish dominant. In order to answer this q uestion, a multinomial logistic regression analysis was used to calculate the odds of a particular outcome given the combination of independent variables present T he independent variable s in question are the participant group and the sentence condition as detailed in Table 4 1
105 Table 4 1. Sentence variants in audio sample Variant Gender agreement /s/ + /a/ realization Condition 1 Error [s.a] Disfluency with agreement errors 2 No error [s.a] Disfluency without agreement errors 3 Error [sa] Fluency wi th agreement errors 4 No error [sa] Fluency without agreement errors Recall that t he designation of a speaker as being Spanish dominant was chosen as the baseline due to the fact that Spanish is the language in which the task was conducted Thus the d ata presented below d isplays two sets of odds ratios; the first set compares provides the odds that a listener will assume a speaker is English dominant compared to a Spanish dominant baseline for a giv en condition for each listener group. Similarly the second set of odds ratios in Table 4 2 compares the provides the odds that a listener will assume a speaker has equal dominance in both languages compared t o a Spanish dominant baseline for a given condition for each listener group An odds ratio of >1 indicates that the risk of the outcome falling into the comparison group relative to the risk of the outcome falling into the referent group increases as the variable increases. In other words, the comparison outcome is more likely. In this case the comparison outcome is English, meaning that participants of a given listener group are more likely to identify the speaker of a given condition as an English domi nant speaker. Conversely, an odds ratio of <1 indicates that the risk of the outcome falling into the comparison group relative to the risk of the outcome falling into the referent group decreases as the variable increases. Therefore if the odds ratio is < 1, the outcome is more likely to be in the referent group, in this case Spanish. This
106 means that participants of a given listener group are more likely to identify the speaker of a given condition as a Spanish dominant speaker. An odds ratio of 1 would i ndicate that the independent variables have no effect on the dependent variable. The odds ratios can be found in the SPSS output parameter estimates table under the column labeled Exp(B) ( Appendix E ). Exp(B) is the change in the odds ratio associated with a one unit change in the predictor variable (Agresti, 2002) Table 4 2 illustrates the odds ratios computed for the comparison outcomes of English and versus the Spanish referent category and the likely outcome category for each participant grou p and condition based on the odds ratio values. Table 4 2. Odds ratios and likely outcome categories by condition and listener proficiency level Condition Participant group by proficiency level Odds ratio Exp(B) intercept vs. baseline Likely outcome category for dominant language, intercept vs. baseline Odds ratio Exp(B) intercept vs. baseline Likely outcome category for dominant language, intercept vs. baseline Disfluency with agre ement errors Low .534 Spanish .601 Spanish Intermediate 1.123 English 1.034 Both High 6.327 English 1.930 Both Disfluency without agreement errors Low 1.168 English .484 Spanish Intermediate .882 Spanish .670 Spanish High 2.993 English 1.734 B oth Fluency with agreement errors Low .246 Spanish .482 Spanish Intermediate .667 Spanish .887 Spanish High 3.024 English 1.393 Both Fluency without agreement errors Low 1.476 English 3.083 Both Intermediate 1.391 English 1.945 Both High .223 Spanish .506 Spanish
107 In T able 4 2 an odds ratio of .534 indicates that for there is a 47% reduction in the likelihood that low proficiency listeners will identify a speaker of this condition as being En glish dominant, meaning that they are more likely to identify speakers of this condition as Spanish dominant than English dominant On the other hand, an odds ratio of 1.123 indicates that for the there is a 12 .3 % increase in the likelihood that intermediate proficiency will identify speakers of this condition as English dominant, while high proficiency listeners are 6.327 times more likely to indicate English dominance for a speaker of this condition When the intercept is compared to the Spanish baseline, there is a 40% reduction in the likelihood that listeners will identify speakers as being equally proficient in both languages, making the outcome of Spanish more likely. However, the subsequent odd s ratios of 1.034 and 1.930 indicate that intermediate proficiency listeners show an increase of 3% while high proficiency n both languages over Spanish dominance. For the odds ratio of 1.168 indicates a 17% increase in the likelihood that low proficiency listeners will say that the speaker is English dominant. However ther e is a 12% reduction in the likelihood that intermediate listeners will assume English dominance among speakers of this condition as indicated by the odds ratio of .882 for this group Highly proficient listeners are 2.993 times more likely to say that th e speaker is English dominant than Spanish dominant for When the intercept is compared to the Spanish baseline low and intermediate proficiency listeners
108 demonstrate reductions of 52% and 33% r espectively in the likelihood of identifying speakers as being equally proficient in both languages while high proficiency listeners demonstrate a 73% increase in the likelihood of identifying speakers of this condition as being equally proficient. For th low proficiency listeners show a 75% reduction while intermediate listeners show a 33% reduction in the likelihood of assuming English dominance, making Spanish the more likely outcome for both groups. However for the high proficiency group, listeners are more than 3 times as likely to assume English dominance for speakers of this condition When the intercept is compared to the Spanish baseline a similar pattern emerges, with low and intermediate pro ficiency listeners demonstrating reductions of 52% and 11% respectively in the likelihood of selecting while high proficiency listeners show a 39% increase in the likelihood of selecting The reverse is tru e for with low and intermediate listeners showing increases of 48% and 39%, respectively, in the likelihood of identifying speakers as English dominant while high proficiency listeners showed a reduction of 78% in identifying listeners as English dominant, making Spanish the likely outcome category for this group. When the intercept is compared to the Spanish baseline, low proficiency listeners are 3.083 times as likely to select for speakers of this condition while intermediate proficiency listeners are nearly twice as likely, with an odds ratio of 1.945, to select for this condition High proficiency listeners, however, are more likely to identify speakers of thout
109 as Spanish dominant, with a reduction of 49% in the likelihood that they will select as the dominant language of this condition The results indicate that there is no condition that yielded a consistent likely outcome categor y across all listener groups. In both analyses, the intermediate and low proficiency groups displayed variation in their likely outcome categories that appears to be inconsistent with the expected outcomes given the variables present within each condition For example, the low proficiency group yielded a likely outcome category of proficient condition in this study. This result may not be surprising, however, as partici pants in this group, by definition, attained low proficiency scores on the independent measure of proficiency and therefore one possibility may be that these listeners were not aware of errors in gender agreement when present. Adding credence to this notio the only other condition to also contain errors in gender agreement, the condition agreement by this group. It may be tempting then, to infer that while the low proficiency pa rticipants did not take note of errors in gender agreement, perhaps they took note of the flow of the speech by way of the presence or lack of linking. The likely outcome based their choices on the presence of pauses in the sentences: the condition
110 low proficiency group. However, this assertion is refuted by the outcomes of the following childh ood over hearers in which they were shown to outperform typical late L2 learners in phonological perception tasks (Au et al., 2008). Another possible explanation for the apparent lack of consistency in the answers rendered by this group is that the errors in gender agreement are indeed perceived by these listeners, however the fact that they are presented in fluent, connected speech left them unsure of the correct form therefore they assumed the fluent sounding speaker to be producing the correct form. Thus the dominant language judgments rendered by this group may not necessarily be representative of their overall knowledge of gender. The intermediate proficiency group also displayed unexpected variation in their likely outcome categories given the variabl es under investigation, albeit with different outcomes than the low proficiency group for the first two conditions. For the condition which suggests that they had a greater likelihood than the low proficiency group of being able to recognize either lack of linking, lack of gender intermediate proficiency participants were
111 that they were at minimum able to discern that the least proficient condition was non native when both indicators of non native speech under investigation in this study were present. When either of the two indicat ors was present, listeners in this group seemed aware that at least one speech signal was indicating to them that the speaker was Spanish dominant, whether it be the presence of linking or gender agreement. The results of the intermediate proficiency group for the remaining two conditions concur with the results of the low proficiency group as described above. These results indicate no definite pattern in the likely outcomes of the low and intermediate proficiency listeners to suggest that they made their j udgments based on either linking or errors in gender agreement. The judgments of these two groups, however, were likely not entirely random. The standard deviations of the ratios that determine the likely outcome categories were low 1 suggesting similar int uitions within each group. This means that these groups likely did not make their judgments arbitrarily but rather based them on a factor or combination of factors other than linking and gender agreement. Only the high proficiency group produced hypothesi zed results by yielding a consistent in their judgments This group had a likely outcome category of either ken by a Spanish dominant speaker. This result indicates that the high proficiency participants were able to distinguish non native speech even when only one of the factors under investigation 1 Standard deviations for the low proficiency group were between .262 .523 with an average of .413. Standard deviations for the intermediate proficiency group were between .260 .478 wit h an average of .404.
112 that signals non native speech was present, and that only the m ost highly proficient participants were likely to consistently recognize these factors and make this distinction. In other words, high proficiency heritage listeners were able to recognize lack of gender agreement and utilize this factor as an indicator of non native speech, which is unsurprising given that this group by definition was highly proficient in Spanish. It is unsurprising that the high proficiency listeners would recognize gender agreement errors and rate accordingly, just as it is unsurprising that low proficiency listeners would not. The question then is where along the spectrum of ability to recognize non native speech intermediate proficiency listeners would be. Like the low proficiency group, the intermediate proficiency group was composed o f third generation heritage speakers. The key distinction between the two, other than the score they obtained on the proficiency measure, was the fact that members of the intermediate proficiency group were enrolled in Spanish classes at the time of their participation. The results appear to indicate that although this group had been given formal instruction on Spanish at the time of their participation, they were still not utilizing gender agreement as a factor in determining non native speech. Also intere sting is the result that the high proficiency group appeared to utilize errors in gender agreement as an indicator of non native speech even when the errors were embedded in fluid speech. This suggests that errors in gender agreement cannot be masked by fl uid speech when the interlocutor is a highly proficient, second generation heritage speaker. v
113 level groups. This fact demonstrates a level of consistency with regard to which conditions triggered a judgment that a speaker was Spanish dominant and reinforces the id ea that the judgments of the listeners, including the low and intermediate proficiency listeners, were not random but perhaps based on factors external to this investigation such as intonation, pitch, or segmental articulation. The only notable difference b their chosen outcome categories. Recall that the analysis predicted the odds that listeners would select the co over indicated a greater likelihood i n achieving this category. On the other hand, odds ratios fact that many odds rat b achieve their respective outcome categories when the comparative category was achieve the outcome of ove
114 likely to obtain the outcome of proficiency group was the only group to consistently yield higher likelihoods in achieving their respective analysis. There were two in stances in which the intermediate proficiency group yielded occurred for the low profic iency group. Overall however, the likelihoods were greater b demonstrating that participants were not as likely to achieve their respective outcome categories for each variable set when the comparative outcome was as when it Spanish distinction may have been easier for participants to make than the both Spanish distinction. This interpretation highlights the complexity of bilingualism and makes sense given the nuances that exist among speakers situated along the bilingualism spectrum. This result suggests that it is easier for listeners to determine that a speaker is English or Spanish dominant and residing c loser to the extremes of the spectrum than to determine that a speaker is equally proficient in both languages, residing closer to the middle of the spectrum. The odds ratio values also reinforce the fact that the low and intermediate groups rendered resul ts that were not as consistent with the variables under consideration as proficiency g in the respective
115 analyses, which is inconsistent with the presence of both linking and gender agreement within the condition. This indicates that not only was the low proficiency group likely to assume that the speakers of this most fluent condition were either English dominant or dominant in both languages rather than Spanish dominant, but this likelihood was higher than the likelihood that the intermediate and high proficiency groups would rende r in their respective outcomes. This result reinforces the assumptions that linking and gender agreement were not significant factors for the low and intermediate proficiency listeners and suggests that other factors instead were at play as mentioned above These the implications of these results will be discussed in further detail in Chapter 5 Factors I nvolved in Assigning L anguage R atings The second research question asked how linking and the presence or absence of errors in article noun gender agreemen t are associated with language ratings. In order to answer this question, an ordinal regression analysis was used. Table 4 3 illustrates the threshold values ( j ) that were rendered by the model for this analysis and serve as the cutoff between two outcome categories : Table 4 3. Estimate threshold values for predictor variables of condition and listener proficiency level. If estimate value is less than A nd mor e than The likely outcome category for language rating is 5.283 Poor 2.464 5.283 Fair 0.131 2.464 Good 0.131 Excellent is j = 5.283, which
116 regression coefficient below 5.283 indicates a likely language rating outcome category j = 2.464, which is the cutoff value bet threshold estimate of j = .131 indicates a likely language rating outcome category of The threshold estimates and the ordered log odds (logit) regression coefficients are found in the SPSS output parameter estimates table under the column labeled Estimate ( A ppendix E). Because the output of the model itself calculates the ex pected level of change in the response variable for a one unit increase in the predictor in the ordered log odds scale while the other variables in the model are held constant, the estimate coefficients provide the interpretation for one dependent variable at a time. However, the predictive nature of the model allows us to examine the logit regression coefficient as the independent variables change by combining estimate coefficients across independent variables utilizing the same threshold values. This is p ossible because these threshold values remain constant and do not depend on the value of the F igure 4 1 illustrates the formula used to calculate the final logit regression coefficients, which we re the values that were fitted to the threshold values and thus predict the likely outcome category for language rating: [Estimate coefficient for Group] + [Estimate coefficient for Sentence] + [Estimate coefficient for Group*Sentence] = fina l logit regre ssion coefficient Figure 4 1. Final logit regression coefficient calculation formula with interaction
117 For instance, the output parameter estimates provided by the model indicate that members of the low proficiency listener group are likely to provide a r indicated by the estimate value of .710 regardless of condition In order to determine the likely outcome rating by members of this group of different condition s, for example which carri es an estimate value of 1.845, the sum these two estimate coefficients can be applied to the same threshold values in order to provide a prediction of the response variable for this particular combination of independent variables ( for low proficiency listeners ). For this analysis, the sum of the two variables was added with the estimate coefficient calculated by the model for the interaction between the two variables, indicated in T able 4 4 by an asterisk, i n order to provide a more precise final logit regression coefficient. In this example, the estimate coefficient for the interaction between the low proficiency group and is 0.868, for a resulting final logit regression coefficient of 1.687, which falls within the Table 4 4 illustrates the final values of the logit regression coefficients as well as the likely outcome category for the lan guage rating, based on the aforementioned thresholds for each category. Table 4 4. Logit regression coefficients and likely outcome rating categories with condition and listener proficiency level predictor variables Condition Participant group by proficie ncy level Group estimate coefficient Sentence estimate coefficient Group* Sentence (interaction) estimate coefficient Final logit regression coefficient Likely outcome category for language rating Disfluency w/ agreement errors Low 0.710 1.845 0.868 1.687 Good Intermed. 0.430 1.845 0.440 1.835 Good High 0 1.845 0 1.845 Good
118 Table 4 4. Continued Condition Participant group by proficiency level Group estimate coefficient Sentence estimate coefficient Group* Sentence (interaction) estima te coefficient Final logit regression coefficient Likely outcome category for language rating Disfluency without agreement errors Low 0.710 0.621 0.359 1.69 Good Intermed. 0.430 0.621 0.037 1.421 Good High 0 0.621 0 0.621 Good Fluency with agreement errors Low 0.710 1.050 0.904 0.856 Good Intermed. 0.430 1.050 0.423 1.057 Good High 0 1.050 0 1.050 Good Fluency without agreement errors Low 0.710 0 0 0.710 Good Intermed. 0.430 0 0 0.430 Good High 0 0 0 0 Excellent Table 4 4 shows that the likely outcome category for all condition regardless of the proficiency level of the listener All logit regression coefficient values fall between .131 and 2.464, which are the threshold values required for a likely outcome T able 4 4 is the by the high proficiency listeners. The table shows estimate coefficient values of 0 f or Group, Sentence, and Group*Sentence. This is because the High Proficiency participant group is the referent group for the variable of proficiency level, and is the referent group for the variable of condi tion as selected by the statistical analysis program SPSS 2 As the referent groups, the default values of The threshold for a .131, therefore an estimate value of zero yields a likely 2 SPSS automatically selects as the referent the last category entered within a set of variables (UCLA: Stat Consulting Group, n.d.).
119 outcome These results are confirmed by the raw data provided by the SPSS cross tabulation as illustrated in T able 4 5 Table 4 5. condition and listener proficiency level predictor variables Sentence r ating g roup c rosstabulation Count Group Rating Total Poor Fair Good Excel lent Low p roficiency Sentence variant 1 3 39 80 21 143 2 1 43 78 22 144 3 1 22 75 46 144 4 1 13 83 46 143 Total 6 117 316 135 574 Intermediate p roficiency Sentence variant 1 8 45 62 28 143 2 2 25 78 38 143 3 3 31 64 46 144 4 2 14 67 61 144 Total 15 115 271 173 574 High p roficiency Sentence variant 1 8 46 59 29 142 2 0 19 71 54 14 4 3 0 31 70 43 144 4 0 8 60 74 142 Total 8 104 260 200 572 T able 4 5 condition s for all three listener groups regardless of proficiency level It can be seen that low proficiency l atings. The ratings totals of the high proficiency listeners yielded a similar within
120 the responses for this condition condition or proficiency group These findings are in accordance with the results of the ordinal regression analysis described above. Eleven of the twelve likely outcome categories rendered by the model in this analysis resulted in a This result indicates that participants regardless of the presence of linking or errors in gender agreement. The notable exception to this f inding proficiency listeners were li kely to detect the presence of both fluency markers under investigation intermediate and low proficiency listeners were unable to distinguish between the conditions and the varia bles present therein. Upon first glance, it appears that the presence or absence of the variables in question within the four conditions studied has no bearing on the outcome category as o variation. It is important to remember, however, that the ordinal regression analysis is a cumulative model that calculates the likely outcome category or lower ; therefore no information is given in the output on ratings that did not fall into the majori ty category. A review of the raw data sheds light on this deficit of information. The raw data show that 316 tokens,
121 well over half of the 574 tokens assessed by the low proficiency group across all iate and low proficiency groups, and 260 for the high proficiency group. In viewing the raw data results for the individual the prevailing majority response category for proficiency group, thus supporting the results of the regression model. The raw data also help to evaluate the results against the backdrop of the original hypothesis, which predicted that listeners of all proficiency levels would assign higher ratings to speakers of sentences containing linking regardless of the presence or absence were assigned to the combined total of the two conditions that contained linking than for the combined total of the two conditions that did not contain linking for all participant groups. However when taking into account the variable of gender agreement, only the intermediate and high agreem proficiency group assigned an equal es assigned to sentences containing linking as a measure by which to evaluate the hypothesis did not yield supportive results. The low and high
122 containing fluency as to the stimuli conta ining disfluency, while the intermediate stimuli containing fluency. One possible explanation for the discord among the response category overall for all conditions, across all proficiency levels with the proficiency group and was therefore somewhat of a default r esponse for participants. listeners across all proficiency levels were likely to assign a rating regardless of the absence or presence of the variables under investigation as a default outcome out of a desire for respondents to be agreeable or sympathetic with the speakers in the sample. Indeed, the low, intermediate, and hig h proficiency groups s assigned by the high proficiency listeners were likely to detect the presence or absence of both fluency mar kers under investigation and assign ratings accordingly. response, a post hoc analysis of the ratings assigned to the distractor sentences spoken by low proficiency speake rs was conducted as a litmus test. The post hoc
123 analysis revealed that, for the low and intermediate proficiency group, as shown in T able 4 6 T able 4 6 Response category totals for low proficiency distractor sentences. Excellent Good Fair Poor Low proficiency 10 59 76 35 Intermediate proficiency 13 52 83 32 High proficiency 16 80 63 19 The post hoc analysis examined distractor sentences t hat were not coded for condition; therefore no information is given on whether the sentences include instances of linking or gender agreement. Because of this, the information in T able 4 6 should not be compared directly with the results of the primary res earch question and do not indicate that the variables under investigation have no effect on sentence ratings. This information does indicate, however, that listeners were indeed being discriminatory in Further information on the independent variables under investigation and their relationship to the likely outcome category for each group can be gleaned from the results of the regression analysis by investigating the logit regression values rendered by the cumulative logit model and the distance of these coefficients from the threshold values. Although eleven of the twelve values resulted in a likely outcome category of with some closer to the was 1.845 by the high proficiency grou
124 falls in line with the idea that high proficiency listeners were more likely than the other group to detect the presence or absence of both fluency markers under investigation and assign ratings accordingly. The next lowest value was 1.835 by the intermediate proficiency group proficiency group, but not as likely to do so as the high proficiency group. On the other end of the scale, the regression coefficient with the highest value that yielded a likely outcome of 0.430 by the intermediate 0.710 by the low proficiency group, showing that they were the least likely to assign a rating of the high proficiency group, the intermediate proficiency group was most likely to detect the absence or presence of both fluency markers and rate accordingly while low proficiency listeners were least likely to do so. These results will be analyzed further in Chapter 5 Language R atings and P erceived V ariety of Spanish The th ird research question asked about the nature of the relationship between language ratings with the variety of Spanish believed to be heard in the speech sample. In order to answer this question an ordinal regression analysis was used to calculate the proba bility that a listener will assign a particular rating or lower given a set of independent variables In this case one of the independent variables is the variety of Spanish the listeners believed they were listening to, as indicated by their answers to th
125 Spani sh proficiency level of the listeners. For this analysis, the following threshold values were yielded by the model: Table 4 7 Estimate threshold values for predictor variables o f perceived region of origin and listener proficiency level If estimate value is less than A nd more than The likely outcome category for language rating is 5.584 Poor 2.702 5.584 Fair 0.038 2.702 Good 0.038 Excellent Table 4 7 j = 5.584, which means that any logit regression coefficient below 5.584 indicates a likely language rating outcome cat 2.702, meaning that logit regression estimate less than 2.702 but higher than 5.584 indicates a likely language outcome category of is .038, therefore a logit regression estimate below .038 but above 2.702 has a likely .038 has a likely outcome category Table 4 8 illustrates the likely outcome category for language rating based on the resultant logit regression coefficients for each independent variable and their placement within the threshold estimates produced by the model as detailed in the previou s section. Because the threshold estimates that separate each outcome category are stationary regardless of the predictor variables in question, multiple variables may be
126 examined simultaneously by adding the estimate coefficients for each variable to be e xamined within one combination. In order to examine the effect of both the perceived region of origin and the proficiency level of the listeners as independent variables, the formula illustrated in Figure 4 1 was utilized to calculate the final logit regre ssion coefficients shown in Table 4 8 Table 4 8 Logit regression coefficients and likely outcome rating categories with perceived region of origin and listener proficiency level predictor variables Perceived region of origin Participant group by profici ency level Group estimate coefficient Region estimate coefficient Group* Region (interaction) Estimate coefficient Final logit regression coefficient (=SUM) Likely outcome category for language rating New Mexico Low .713 .964 .013 1.69 Good Inte rmed. .217 .964 .582 1.763 Good High 0 .964 0 .964 Good U.S. outside of New Mexico Low .713 2.654 .780 2.587 Good Intermed. .217 2.654 .254 2.617 Good High 0 2.654 0 2.654 Good Mexico Low .713 .460 .150 .103 Good Intermed .217 .460 .098 .341 Excellent High 0 .460 0 .460 Excellent Spain Low .713 .888 .504 .329 Good Intermed. .217 .888 .170 .841 Excellent High 0 .888 0 .888 Excellent Other Spanish speaking country Low .713 0 0 .713 Good Intermed. .2 17 0 0 .217 Good High 0 0 0 0 Excellent Table 4 8 illustrates the logit regression coefficients and the resulting likely outcome category for language rating. It can be seen that across all proficiency levels, all logit regression coefficient values corresponding to the perceived region of origin of
127 New Mexico and the U.S. outside of New Mexico are lower than .038 and higher than Although this is also true when the perceived region of orig in is Mexico for the low proficiency group, the logit regression coefficient for this region was .341 and .460 respectively for intermediate and high proficiency listeners. Both of these values exceed the threshold of .038 required interpretation: if a new listener of intermediate or high proficiency were to enter the study and identify a speaker as being from Mexico, they would likely assign the speaker T able 4 8 shows that the same is true when the perceived region of origin is Spain, where the logit regression coefficient for low proficiency listeners was coefficie nts for intermediate and high proficiency listeners is .841 and .888 respectively, intermediate proficiency listeners are .713 and .217 respectively, below the .038 sh speaking by the program as the reference category. Similarly, within the variable of proficiency as the reference category As the reference groups, the default values of the estimate coefficient for This means that the final logit regression coefficient for high proficiency listeners when the perceived region o
128 Spanish rating. The results of th is rating category across all proficiency levels when the perceived reg ion of origin was New Mexico and the U.S. outside of New Mexico. When the perceived region of origin was Mexico, Spain, or Other Spanish speaking country, the prevalent likely outcome rating was the likely outcome proficiency listeners for all three regions and by intermediate proficiency listeners for the final region only. This finding suggest s a bias on the part of intermediate and high proficiency listeners t oward varieties from Mexico, Spain, and other Spanish speaking countries since these were the only varieties likely proficiency group more reliably exemplifies this suggesti on given that they produced a speaking proficiency group was more consistent than the U.S. varieties of Spanish is not surprising given the fact that the high proficiency group was the only group to have produced results that were consistent with the variables under investigation in the previous analyses. There is an in teresting connection between these two results beyond the fact that they both demonstrate consistency on the part of the
129 high proficiency participants. In addition, the fact that the this group was consistent in U.S. varieties of Spanish and the fact they were the only group to have produced results that were consistent with the presence or absence of linking and gender agreement suggests a relationship between perception, sensitivity to phonolo gical and morphosyntactic characteristics, and language attitudes. Although the scope of this relationship is unclear, these results suggest that New Mexican listeners who are sensitive to phonological deviation and morphosyntactic errors are also more cri tical of NMS or, at the very least, not as likely to praise speakers that they believe to be from New Mexico or label their speech as believed to be from New Mexic o produced no more or no fewer errors in gender agreement or linking than speakers believed to be from other regions given that region of origin was largely in the mind of the listeners. The results of the present analysis are especially compelling given that the high proficiency listener group was the only group in this study composed of second they believed to be from Mexico, Spain, or other Spanish speaking countr ies while suggesting these participants hold a higher opinion toward varieties of Spanish that they believe not to be NMS. This conclusion exemplifies the assertion by Bills and Vigil generation bilinguals, the group with the highest average age of 63 year s, which most
130 consistently exemplified this belief in their results. This could be because, as detailed in C hapter 2, the linguistic history of this generation is complex and deeply personal, whereby members of this generation of New Mexicans were subject to overt language policies prohibiting Spanish use on school grounds and in other public places. These language policies contributed largely to the overall language shift that occurred in this region. It is conceivable that the relegation of Spanish to pri vate domains contributed not only to the loss of Spanish but also to an overall attitude of NMS as a deficient form of Spanish. Indeed, as early as 1917 language scholars in New Mexico were studying the language use and language shift in the region, and ov ertly or covertly passing ool children of the predominantly American cities and towns like Roswell, Albuquerque, East Las Vegas, etc. speak English as well as the English speaking people and speak very poor proficiency group was composed of members of the generation that was raised in a diglossic situation. Members of this generation were told by scholars (as in Espinosa above) and by speakers of other varieties of Spanish that as a result of their competency in English, their competency in Spanish is deficient, as well as their variety of Spanish which arose from the contact situation (Bills and Vigil, 2008). In this regard, it is reasonable that these participants would assign higher ratings to people they believed were speak ing varieties other than NMS. These results suggest that individuals raised within diglossic situations view the language or variety of lesser prestige in a more negative light than
131 they do the variety or language of greater prestige due to their upbringin g that devalued bilingualism. One potentially remarkable implication to these beliefs is that placing a positive value on bilingualism and on contact language varieties has important ramifications for how bilingual individuals view themselves, their commun ities, and their languages. However, given that at least some individuals in each participant group expressed negative comments toward NMS, further investigation on language attitudes in New Mexico is recommended. The original hypothesis predicted that li steners of all proficiency levels would assign lower ratings to stimuli that they identified as being spoken by a speaker of NMS than to stimuli they identified as being spoken by speakers of other varieties of Spanish. The results of the low proficiency g roup did not support this hypothesis, as this group region of origin. However the results of the intermediate and high proficiency groups support the hypothesis because t hese participants yielded likely outcome rating another Spanish speaking country but yielded a likely outcome rating category of s New Mexico or the U.S. outside of New Mexico. While this result does not necessarily suggest these listeners hold negative attitudes toward NMS, it does show that intermediate and high proficiency listeners in this region were likely to assign higher rat ings to varieties that they perceive not to be NMS. The implication of this finding is that these listeners hold a higher opinion of non NMS varieties than to NMS, likely due to the historical factors previously discussed. Further implications of these res ults will appear in Chapter 5
132 Language R atings and L anguage A ttitudes T owards NMS The fourth research question asked how language ratings are associated with language attitudes towards NMS. In order to answer this question an ordinal regression analysis was used, as in Research Question 2 and Research Question 3, due to the ordinal/categorical nature of the language ratings. Two of the independent variables in this analysis were examined in the three previous research questions: the perceived region of or igin of the speakers and the proficiency level of the listeners. At the heart of this research question is the investigation into whether listeners who express negative attitudes toward NMS are likely to assign low ratings to speakers that are believed to be from New Mexico, and/or higher ratings to speakers they believe to be from regions outside of New Mexico. The item on the rating sheet under investigation for this inq uiry for this question juxtaposes NMS with that which is not NMS, therefore for The third dependent variable examined in this analysis is language attitude, as expressed by the listeners on the language attitudes assessment (Appendix D) To examine this variable, answers provided on the language attitudes assessment were quantified in the following manner: each of the three dialect preference questions garnered one point when the participant indicated a preference for NMS. On the open
133 garnered another point if they expressed a positive 3 answer in describing NMS. In tallying the four answers, the resulting value of 0 4 was the resulting language attitudes value (LAV) assigned to each lis tener, a categorical/ordinal variable with zero representing those that did not express positive attitudes toward NMS, while listeners with an LAV of 4 expressed positive attitudes toward NMS. Table 4 9 illustrates the various LAVs rendered by each partici pant group. Table 4 9 Total n of LAVs rendered by proficiency level group Proficiency level group LAV n = High 0 5 1 4 2 4 3 3 4 2 Intermediate 0 7 1 3 2 3 3 3 4 2 Low 0 6 1 3 2 2 3 7 4 0 For each listener, three characterist ics are known: their proficiency level, their LAV, and the variety of Spanish they believed they were listening to for each token. The model itself calculates the estimate log odds (logit) regression coefficients for these variables individually. These coe fficients lie along a continuum and are discretized into 3 c ed a point in this tally. A full list of comments coded as 13.
134 ordered groups by the threshold estimates, which are the cutoff values between two different categories of this variable. In this analysis, the thresholds shown in Table 4 10 were rendered by the mod el: Table 4 10 Estimate threshold values for predictor variables of perceived region of origin, listener LAV, and listener proficiency level If estimate value is less than A nd more than The likely outcome category for language rating is 6.428 Poor 3.430 6.428 Fair 0.730 3.430 Good 0.730 Excellent The cumulative logit model ordinal regression analysis gives us information on the probability that, for a given combination of listener characteristics (proficiency level, LAV, and perceived va riety of Spanish), a listener will assign a particular rating or lower to a given speaker. The thresholds are static regardless of whether the independent variable changes, therefore the estimate coefficients for individual variables may be combined to cal culate a logit regression coefficient for any combination of variables, per the formula shown in Figure 4 2 : [Estimate coefficient for Group] + [Estimate coefficient for Region] + [Estimate coefficient for LAV] = final logit regress ion coefficient Figur e 4 2. Final logit regression coefficient calculation formula no interaction Table 4 1 1 lists the likely outcome categories for all possible combinations of the three independent variables in question based on the thresholds outlined in T able 4 9
135 Tabl e 4 1 1 Logit regression coefficients and likely outcome rating categories with perceived region of origin, listener LAV, and listener proficiency level predictor variables Perceived region of origin LAV Participant group by proficiency level Group estima te coefficient Region estimate coefficient LAV estimate coefficient Final logit regression coefficient (=SUM) Likely outcome category for language rating New Mexico 0 Low .451 1.571 .281 2.303 Good Intermed. .388 1.571 .281 2.24 Good High 0 1.571 .281 1.852 Good 1 Low .451 1.571 .742 2.764 Good Intermed. .388 1.571 .742 2.701 Good High 0 1.571 .742 2.313 Good 2 Low .451 1.571 .096 2.118 Good Intermed. .388 1.571 .096 2.055 Good High 0 1.571 .096 1.66 7 Good 3 Low .451 1.571 .126 2.148 Good Intermed. .388 1.571 .126 2.085 Good High 0 1.571 .126 1.697 Good 4 Low .451 1.571 0 2.022 Good Intermed. .388 1.571 0 1.959 Good High 0 1.571 0 1.571 Good U.S. outside of New M exico 0 Low .451 2.813 .281 3.545 Fair Intermed. .388 2.813 .281 3.482 Fair High 0 2.813 .281 3.094 Good 1 Low .451 2.813 .742 4.006 Fair Intermed. .388 2.813 .742 3.943 Fair High 0 2.813 .742 3.555 Fair 2 Low .451 2 .813 .096 3.36 Good Intermed. .388 2.813 .096 3.297 Good High 0 2.813 .096 2.909 Good 3 Low .451 2.813 .126 3.39 Good Intermed. .388 2.813 .126 3.327 Good High 0 2.813 .126 2.939 Good 4 Low .451 2.813 0 3.264 Good Intermed. .388 2.813 0 3.201 Good High 0 2.813 0 2.813 Good
136 Table 4 11. Continued Perceived region of origin LAV Participant group by proficiency level Group estimate coefficient Region estimate coefficient LAV estimate coefficient Final log it regression coefficient (=SUM) Likely outcome category for language rating Other Spanish speaking country 0 Low .451 0 .281 0.732 Good Intermed. .388 0 .281 0.669 Excellent High 0 0 .281 0.281 Excellent 1 Low .451 0 .742 1.193 Good Intermed. .388 0 .742 1.13 Good High 0 0 .742 0.742 Good 2 Low .451 0 .096 0.547 Excellent Intermed. .388 0 .096 0.484 Excellent High 0 0 .096 0.096 Excellent 3 Low .451 0 .126 0.577 Excellent Intermed. .388 0 .126 0.51 4 Excellent High 0 0 .126 0.126 Excellent 4 Low .451 0 0 0.451 Excellent Intermed. .388 0 0 0.388 Excellent High 0 0 0 0 Excellent Table 4 1 1 shows that all logit regression coefficient values corresponding to the perceived region of ori gin of New Mexico are lower than .730 and higher than 3.430, proficiency level and LAV of the listeners. In addition, all coefficient values corresponding to New Mexico are well within those thresholds, with a difference of no .730 ) and a difference of no less than 0.84 on the end of the threshold that is closest to a 3. 430) In contrast, when the perceived region of origin is the U.S. outside of New rating categories for listeners with low LAVs The logit regression coefficient for participants with an LAV
137 of 0 is 3.545 for low proficiency listeners and 3.482 for intermediate proficiency listeners, both of which lie within the thresholds for a likely outcome rating category of proficiency listeners with an LAV of 0 the logi t regression coefficient of 3.094 signifies a likely outcome perceived region of origin is the U.S. outside of New Mexico For this same perceived region of origin, listeners with an LAV of one yielded logit regression c oefficients within levels. These coefficients: 4.006 for low proficiency listeners, 3.943 for intermediate proficiency listeners, and 3.555 for high proficiency l isteners, are all closer to the 3.430 6.428 threshold that is the cutoff 2 3 and 4 all lie within the .730 and 3.430 th perceived region of origin is the U.S. outside of New Mexico, regardless of proficiency level. In addition the regression coefficients for these groups are well within the ere low and intermediate proficiency listeners with an LAV of 3 approach the ratings by the nearest margin. prevailin The only exception to the with an LAV of 0 and 1 those that did not express positive attitudes toward NMS. With a logit regression coefficient of 0.732, low proficiency listeners with an LAV of 0 have a likely outcome proficiency
138 listeners generated logit regression coefficients of 0.669 and 0.281 respectively, both exceeding the threshold for a likely outcome with an LAV of 1 speaking outc ome However, listeners of high proficiency generated a logit regression coefficient of 0.742, which narrowly approaches the regres sion coefficients for listeners with LAVs of 2 3 and 4 are above the .730 Spanish f nearest margin to the proficiency group with an LAV of 3 whose logit regression coefficient is 0.577 Table 4 10 shows logit regression coefficients of zero for all instances where the perceived 4 and all instances of high proficiency speakers This is because these categorie s were automatically designated as the reference categories by SPSS by virtue of the fact that they were the last categories entered for their respective variables. As the reference category, this category serves as the beginning of the slope of the line f or its respective Listeners were found to have a likely outcome category of perceived region of origin was New Mexico regardless of their LAV. This finding i s particularly interesting because it shows no effect of the independent variable of
139 language attitudes on the likely language rating. There are several possible explanations for this result, including a possible shift in attitudes, meaning the results rep resent an overall acceptability of NMS that is inconsistent with previous literature that states that New Mexicans traditionally hold negative attitudes toward NMS. However this possibility is less likely when we consider the comments provided by the parti cipants on the language attitudes assessment as examined in Chapter 4. The qualitative analysis of these comments will be further discussed ahead in this chapter. Another possible explanation for the lack of variability in the likely outcome rating categor served as the default language rating by respondents looking to be agreeable or sympathetic with the speakers in the sample. However the post hoc analysis of the ratings assigned to the distractor sentences spoken by low proficiency speakers as explained in the discussion for Research Question 2 revealed that for the low and intermediate demonstrating judiciousness o n the part of the listeners in assigning language ratings. rating category among some listeners when the perc related to the fact that all of the core speakers were indeed high proficiency speakers, whereas two of th e nine distracter speakers were of intermediate Spanish proficiency and two were of low Spanish proficiency. The listeners could have intuited that the target speakers were of greater proficiency than at least some of the distracter
140 speakers, on the basis of factors external to linking and gender agreement, and rated accordingly. One final possible explanation for the lack of variability in the likely outcome rating category when the perceived region of origin was New Mexico is that there were no linguistic markers of NMS in the speech sample. Stimulus sentences were pre written by the researcher to examine the specific linguistic features of linking and morphological gender agreement, but in focusing on these factors they necessarily excluded speech charact eristics such as code switching and lexical variation, which are two salient markers of NMS (Sanchez, 1983; Cobos, 2003). These characteristics were specifically excluded in order to control for their presence as a variable affecting the judgments of the l isteners. The result of this exclusion could be that listeners judged the switch, for example. The implication of this finding is t hat speakers of NMS will be judged by members of their speech community this variety such as code switching. Additionally, listeners may rely on these traditional markers in order to determine in group affiliation. This speculation however, would require further investigation such as on listener ratings on naturalistic speech utilizing tokens by the same speaker that include salient markers of NMS and others that exclude t hese markers. of origin was the U.S. outside of New Mexico, listeners that demonstrated negative attitudes toward NMS, those with LAVs of 0 and 1, produced a likely outcome rati ng of
141 proficiency groups for this perceived region of origin. It is interesting to note that the only variability in the likely outcome categories when the perceived region of origin was the U.S. outside of New Mexico was between speakers from other Spanish speaking countries. This result is consistent with the fact that Spanish is not the majority language of the U.S. as it is for other Spanish speaking note that the participants that yielded a lik believed to be from the U.S. outside of New Mexico were those that had an LAV of 0 and 1, those that demonstrated the least positive attitudes toward NMS. This is the same group of participants that garnered a li could be simply that participants that had an LAV o f 0 and 1 were simply more harsh critics of language and therefore in addition to not exhibiting positive attitudes toward NMS in the language attitudes questionnaire, judge language more strictly and are thus likely to provide lower ratings to speakers th an are participants with higher LAVs. This seems likely given that these individuals were likely to rate lower for two different perceived region of Spanish
142 implication of this finding is that the language attitudes value is indeed be related to the language ratings that listeners assign toward what they perceive to be different varieties of Spanish. However, the relationship between language attitudes and ratings is likely more evident for stimuli that contain salient markers of the different varieties under investigation. Further research in this area is recommended. It was hypothesized that listeners with low LAV scores would assign low ratings to speakers that were believed to be from New Mexico and that these listeners would assign higher ratings to speakers they believed to be from regions outside of New Mexico. In addition, it was hypothesized that these results would occur among participants of all proficiency levels. The only results that s upported this hypothesis were the likely outcome categories for intermediate and high proficiency participants Spanish gion of origin was New Mexico. However the results of these two groups only partially support the hypothesis, with regard to the prediction that listeners would assign higher ratings to speakers they believed to be from regions outside of New Mexico. The r esults of these two groups, intermediate and high proficiency participants with an LAV of 0, fail to support the hypotheses with regard to the prediction that listeners with low LAV scores would assign low ratings to speakers that were believed to be from New Mexico, since all participants were spoken by speakers from New Mexico. Further confounding the results in relation to the hypotheses is the fact that the results of these two groups, the intermediate and high proficiency participants with an LAV of 0, are reflected by participants of all
143 proficiency levels that demonstrated more positive attitudes toward NMS with LAVs of 2, 3 and 4: participants with high LAVs wer e also likely to assign higher ratings to speakers they perceived to be from another Spanish speaking country than they were to speakers they perceived to be from New Mexico or the U.S. outside of New Mexico. trated attitudes toward NMS as calculated by the LAV were not necessarily associated with language ratings, at least when linguistic markers of NMS were not present These results and their effects will be analyzed further in C hapter 5 Qualitative Result s A qualitative evaluation of the open ended questions on the rating sheet is included in this study in the interest of exploring other factors that may impact listener S ince it is qualitative in na ture, this analysis cannot be widely generalized across other populations of course, but it is included as supplementary to the quantitative data in order to provide further sents a summary of the answers provided to the open ended questions on the language attitudes assessment in order to provide a more holistic view of the attitudes held by the participants toward NMS. The implications of these answers will be presented in C hapter 5. Rating S heet Three of the five items on the rating sheet were multiple choice answer selections which elicited ratings, perceived dominant language of the speakers, and perceived region of origin of the speakers, as have been discussed in the pr evious sections. Also included on the rating sheet was a multiple choice item which asked
144 ved dominant language of the speakers. The only open ended item on the rating sheet assum es a level of metalinguistic awareness that may not be universal; therefore to avoid the possibility of participants struggling to provide an answer and the interest of respecting the time commitment of the participants, this item was designated as optiona l. Participants were told only to provide an answer for this section if something at them. As the question allowed participants to identify either words or sounds, some participants answered the question by list ing words in the stimulus sentence that appeared salient to them, while others provided a description of the sounds that were salient. For the present analysis, all answers provided in this section were classified into four categories. The first category c onsists of answers that referenced general grammar errors or specifically cited gender The second category consists of answers that refer enced the delivery of the speech such as speed, fluency, or pauses between pronunciation o f particular sounds or words in the stimuli or global accent or
145 above categories, Table 4 12 provides an abridged list of comments from the question on the rating A more complete list of responses t o this question is provided in Appendix F. Table 4 12 also includes the corresponding language rating, perceived region of origin, and perceived dominant language that pertain to each comment on the list as organized by pa rticipant group/proficiency level. Ten answers per participant group were chosen randomly and are included below in order to provide a brief overview of the nature of the comments provided by participants. Table 4 1 2. Sample of c omments provided on rating sheet Group by proficiency level Rating Perceived region of origin Perceived dominant language Comment High Good Other Both flow of speech, enunciation Good NM Both grammar, las abrigos, brazos, accent ok Excellent NM Span sounds like northern nm Spanish Excellent Other Span wrong grammar, great accent Excellent NM Both pronunciation/timing Fair U.S. Eng los/las Excellent U.S. Span speed of communication Fair U.S. Eng slow in speech Excellent Other Eng natural flow of speech Excell ent U.S. Both speech flows well and correct Intermediate Good NM Both heavy "h" sound common in NM Fair U.S. Both accent, los colores Good NM Eng slight accent Good Other Eng flautas emphasis on "au" Excellent Other Span Mexican accent Good Other Span llevan accent on "c" Fair NM Span good accent, but simple grammar mistake
146 Table 4 12. Continued Group by proficiency level Rating Perceived region of origin Perceived dominant language Comment Excellent Other Both avejas [sic] emphasi s on "j" Excellent Other Both speed of speech Excellent Other Eng very quick, good pronunciation Low Excellent Other Both the way he pronounced the "l's" Excellent NM Eng ease of pronunciation Excellent Other Span the rolling of the r's G ood Other Both outside NM Spanish Fair NM Span fluency Excellent Other Both spoke fast, flautas Good NM Span changed pitch, notes, through sentence Excellent Other Span B in brazos Good NM Both Sounds like NM Spanish Excellent Other Eng smoo th and fast Table 4 1 3 lists the percentage of comments under each category for each participant group, including the number of instances where participants declined to answer the question. In most cases participants provided only one comment per stimul us sentence (577 stimuli total), although in a few instances some participants provided two comments that were categorized under two separate categories. Table 4 1 3 Percentage of answers provided on rating sheet per category. Group by proficiency level G rammar Delivery ( s peed, linking etc.) Pronunciation of word or global accent Other High 58.20% 7.94% 28.04% 5.82% Intermediate 25.59% 12.35% 61.76% 0.29% Low 6.82% 28.03% 52.27% 12.88% Totals 32.59% 14.87% 52.53% 4.59%
147 In evaluating the totals with in the four answer category groups, Table 4 1 3 shows that the high proficiency group provided more answers that referenced grammatical errors and/or errors in gender agreement than the intermediate and low proficiency group s In addition, the high proficie ncy group provided more answers in this category than in any other answer category. It is interesting to note that with regard to answers that reference the delivery of the speech, the high proficiency group just 7.94% compared with 12.35 % for the intermediate proficiency group and 28.03% for the low proficiency group. The intermediate proficiency group provided more total answers than any other group; 61.76% of which fell under the category of pronunciation, 12.35% under the category of d elivery, and 25.59% under the category of grammar The low proficiency group provided 52.27% category, 28.03% 6.82% of their answers fell result may not be surprising that due to the fact that they scored the lowest on the proficiency test, as was discussed in Chapter 3 and it would therefore follow that they have limited knowledge or recognition of various grammatical aspects of Spanish. T he overall percentages of all three groups show the highest percentage this same trend is the intermediate pro ficiency group, w ho, as stated previously, provided a higher quantity of answers to the optional question than did the other two groups As detailed above, the high and low proficiency groups deviated from the global pattern. Figure 4 3 illustrates these t rends.
148 Figure 4 3. Types of open ended answers provided on rating sheet Overall, participants provided more than twice as many comments referencing grammar as they did those referencing delivery, suggesting that gender agreement is more salient than disf luencies in speech. However, recall that participants were told that grammar errors may appear in the stimuli, but were not told which or how many stimulus sentences contained grammar errors nor were they informed of the nature of the grammar error. The hi gher number of comments referencing grammar was possibly a result of participants having been alerted to expect the possibility of an unknown type of grammar error whereas they were not necessarily told to expect changes in the delivery of the stimuli. Ano ther possibility is that hesitations in delivery are common even in native speech, but grammar errors are a fairly dependable indicator of non nativeness. The only group to provide more comments referencing delivery than grammar was the low proficiency gro up. In fact, 6.82% of the comments provided by this group referenced grammar. This suggests that, although they were informed that grammar errors may be present in the stimuli, the low proficiency participants were either unable to detect them or did not f ind them as salient as speech cues such as
149 pronunciation and delivery. This result is not surprising given that this group of participants scored the lowest on the independent measure of proficiency and embodies f a heritage language: those who heard the home language in the background. It is rational then, that participants in this group would rely more on the speech cues of delivery and pronunciation than on the recognition of grammar errors. However the results of Research Question 1 and Research Question 2 as previously discussed do not show that the low proficiency group relied on linking in assigning their language ratings. Instead, the comments discussed in this section support the conclusion that low profic iency heritage learners rely on other speech cues in producing judgments, presumably segmental cues as indicated by the comments referencing pronunciation. Also expected was the result that the high proficiency group provided the highest number of comment s referencing grammar than the other groups given that they have a greater mastery of grammar as demonstrated on the proficiency measure. It is also not surprising that this group provided twice as many comments referencing grammar as pronunciation, and si x times as many comments on grammar as on delivery. It is possible that because the high proficiency participants were alerted to expect grammar errors they were more conscious of their presence or absence than they were of linking. More information would be required to make this conclusion, though, such as stimulated recall questions on an exit interview. However, the fact remains that in research question 1, the high the dominant language of speakers only of sentences that contained both linking and no grammar errors; and in Research Question 2, this group yielded a likely outcome
150 proficient condition as well. These facts can lead us to infer tha t although the high proficiency participants more often identified grammar errors or the lack thereof to be the salient characteristics of the speech, they were also at least subconsciously aware of the presence or absence of linking, given that they did n ot rate sentences that did not feature this phenomenon as highly as they did stimuli in which it did occur. The implication of these findings is that high proficiency listeners possess a greater meta awareness of linking and the fluidity of the speech, eve n if they do not consciously acknowledge this awareness. The low and intermediate proficiency groups likely do not possess this meta awareness regarding delivery of the speech even though they cited it more than the high proficiency group did in the commen ts section. The results of Research Question 1 uphold this conclusion, in which participants in the low and intermediate proficiency groups yielded results highly inconsistent with the presence or absence of linking. Of particular interest to this study a re the comments made by participants on the rating sheet that reference the delivery or fluency of the speech or may otherwise reference linking What particular words or sounds did you Participants across all three proficiency level groups found a variety of ways to describe their perception of the occurrence or absence of linking in reference to both linked and unlinked stimuli, as demonstrated in the Table 4 14
151 Table 4 1 4 Rating sheet comments referencing delivery of speech Participant group by proficiency level Sentence variant /s/ + /a/ realization Comment High Variant 4 [sa] very good movement of words, abejas flow of speech over emphasized /slow* hesit ation* flow of speech, enunciation natural flow of speech hesitates* speaks fast Variant 3 [sa] speed of communication slow in speech* read not comfortable* Variant 2 [s.a] speed of speech reads but not well speech flows well a nd correct* Variant 1 [s.a] hesitates Intermediate Variant 4 [sa] spoken quickly and with confidence frequency, llevan very quick, good pronunciation very fluid Variant 3 [sa] fast pronunciation was great and spoken quickly he spoke s lowly* the words were spoken together really quickly words were separate in sentence and not quickly run together like native speaker* Variant 2 [s.a] proper pronunciation but slowly spoken words run together* pause between los and next wor d hesitant sounds uncomfortable sounds like they are familiar with speaking the language* staccato, not fluid words pronounced correctly but separated hesitant but good accent cantan kind of slow fluency
152 Table 4 14. Continued Pa rticipant group by proficiency level Sentence variant /s/ + /a/ realization Comment Variant 1 [s.a] spoken slowly with slight accent speech was choppy but proper accent pronounced right but spoken slowly good Spanish, fluid, but not used much you can tell in her accent Low Variant 4 [sa] ease of pronunciation trying at Spanish* smooth and fast speed Variant 3 [sa] confident pronunciation Variant 2 [s.a] words come to an abrupt stop enunciated each syllable hesitation Variant 1 [s.a] Reading not too familiar w. Spanish A little slower, seemed a little foreign to her letters blended together in their words* *Comments marked with an asterisk are inconsistent with the presence or absence of linking within the stimul us sentence In many cases, participants make reference to the speed or lack thereof in the speech of the sample. However, sentence lengths were measured in the acoustic analysis program PRAAT prior to the creation of the speech sample to ensure that spe ed would not be a variable in this study, thus limiting the variables in this study to linking within the coda of the sentence initial article and the proceeding noun and article noun gender agreement Many of the comments above do not mention speed of the speech but more specifically describe the occurrence of linking or lack thereof within the stimuli
153 absence of linking Some of these comments referencing what appears to be linking do not nece ssarily corroborate with the occurrence or lack of this phenomenon within the token in question. For example, a high proficiency listener commented on a token of the which has a /s/ + /a/ realization of [s.a] condition given that it was produced with an intentional break between the sentence initial article and noun. In total, eleven of the forty nine comments that referenc e the delivery of the speech in the T able 4 14 did not corroborate with the /s/ + /a/ realization as verified by the researcher in the spectrogram of each sentence in the acoustic analysis program PRAAT, manifested as either a continuous voiced waveform (l inked token) or as a stop (unlinked token). C omments that were inconsistent with the /s/ + /a/ realization of the token they reference are marked with an asterisk in Table 4 1 4 Thirty five percent of the comments by the high proficiency group were marked by the researcher as inaccurately describing the /s/ + /a/ realization of the token in question, compared to 16.6% for the intermediate proficiency group and 18% for the the low proficiency group. These numbers appear to suggest that the high proficiency g roup possesses less metalinguistic awareness than the other two groups based on the inaccuracy of their comments with regard t o the presence or absence of linking However this suggestion is in opposition with the assumption described above in which althou gh the high proficiency participants more often identified grammar as a salient cue in determining language dominance, they may have also been subconsciously aware of the presence
154 or absence of linking, as evidenced by the fact that the they group did not rate stimuli that lacked linking as highly as they did stimuli in which it did occur. In this case, it appears that their conscious, deliberate comments regarding the delivery of the speech did not reflect their subconscious judgments regarding the languag e dominance of the speakers in question. This finding calls into question the declarative vs. procedural knowledge of linking on the part of high proficiency heritage speakers. The high percentage of inaccurate comments regarding delivery / linking by the high proficiency group in relation to their accuracy in assigning ratings in accordance with this phenomenon suggests that the high proficiency participants have a greater procedural knowledge of linking than they do declarative knowledge. This means that they are able to perceive and utilize linking in making judgments but not necessarily able to put that knowledge into words. This finding lends credence to early work by Labov (1966) that found that even native speakers have difficulty accurately reporting on language use, as report tests and actual usage. An important methodological implication of this finding is that self reporting by participants should be interpreted judiciously and not utilized as the pr imary method of assessment in interpreting perception data. Instead, examining correlations with language ratings and specific language features, as in the current study, may provide more reliable results. Although the high proficiency group had the highe st inaccuracy rate in their comments regarding linking, it is important to remember that the comment section on the rating sheet was optional; therefore many of the participants who produced ratings that were congruent with the presence or absence of linki ng may simply not have
155 provided comments in this section. This would explain the disproportionateness of the accuracy of the comments made in this section in comparison with the accuracy of the ratings themselves. However, further research in this area f or example in an exit interview with participants to assess their awareness of these concepts, is necessary to make a solid assessment. Nonetheless, it is important to remember in viewing the comments above and in A ppendix F, many of which are in direct op position with one another, that all comments were made about a sample consisting of just four speakers, thus demonstrating the effect that the variables of linking and gender agreement have in the judgments of the listeners. The inconsistencies between the answers of at least some unreliable. The comments provided on the open ended question on rating sheet and their contribution to the overall study will be discussed in further detail in Chapter 5 Language A ttitudes A ssessment The language attitudes assessment completed by the participants contained four multiple choice questions and two op en ended questions. The first open ended question asked participants to describe the Spanish of New Mexico. Participants garnered one point toward their LAV if they expressed opinions toward NMS that were considered unequivocally positive in the judgment o f the researcher, for example awarded a point in this tally. The second op en ended question asked participants to describe, if possible, the Spanish of Mexico and/or other countries. Although answers to
156 this question were not applied toward the LAVs, these answers are included in this section in order to be compared with the ans wers describing NMS and to provide a more comprehensive view of the attitudes New Mexicans hold toward other varieties of Spanish. Appendix G contains a complete list of the answers provided for both questions along with information on whether the research er assigned a point for the comment toward the LAV. In viewing the comments in A ppendix G it is important to remember that all participants were New Mexican residents and bilingual or heritage speakers of the New Mexico variety of Spanish. The answers gi ven by the participants provide insight into the opinions that exist among New Mexicans regarding NMS and contextualize the LAV scores that were assigned to each of the participants based on these and other answers provided in the language attitudes assess ment. Many of these answers lend credence to the assertion by Bills (1997) that many speakers of NMS express negative attitudes toward their own variety. For example, one observation that appears often is the word ways in which participants note code often used as a negative descriptor ( Otheguy & Stern, 2011; Stavans, 2000), and this seems to be the case in the present s tudy as well, as these comments are sometimes
157 comments that appear multiple times in the descr (P#301, P#312, P#316, P#317, P#318 low; P#352, P#353, P# 358, P#370, P#372 e considered as carrying a negative connotation by the researcher and thus not awarded a point counting toward the overall LAV, which accrued points only on answers that were considered unequivocally positive by the researcher Yet of the three, only the w appeared just in three comments referencing other varieties of Spanish. Thus it can be concluded that participants provided a higher number of negative descriptions of NMS than they did of non NMS varieties as a singular group. Only 8% of the descriptions of NMS were judged by a panel 4 of judges to be unequivocally positive. These positive By comparison, 28% of the descriptions of other varieties were judged to be positive 5 This finding shows that participants provided a higher number of positive descriptions of other var ieties of Spanish than of NMS. This statement, however, is not mutually exclusive with the previous conclusion that participants provided a higher number of 4 The panel consisted of four judges including the primary researcher. Of the panel, three were linguistically trained an d the fourth judge was familiar with NMS as a third generation speaker. There was 89% inter rater reliability in which the judges agreed 3 1 or una nimously. Ten percent of the cases resulted in a 2 2 tie, in which case they were designated I n one occurrence (1%) the judgment was 2 1 1 in which case the participants discussed the token until a 3 1 majority was reached 5 Descriptions that were unanimously judged to be positive by the panel of judges are designated with an asterisk in Appendix G.
158 negative descriptions of NMS than of other varieties of Spanish, due to the possibility that partic ipants would provide neutral descriptions of either variety. D escriptions of NMS that were considered neutral were not awarded a point considered neutral provided a historical or educational perspective of NMS, such as (P# 211 high) Across all proficiency groups, eight answers in total provided this these eight answers were provided by the high proficiency group, which was composed of second generation bilinguals and the group with the highest average age of 63 years. As stated in Chapter 2 schoolchildren in New Mexico were punished for speaking Spanish on school grounds as late as t majority view of Spanish as the language of lesser prestige and thus played a large part in the diglossic situation of the region. It is reasonable then that participants belonging to this generation of New Mexica ns would provide answers in defense of NMS that push back against the language beliefs and policies that demonized Spanish use in the public domain. In their linguistic atlas of the Spanish of southern Colorado and Northern New Mexico, Bills and Vigil (200 8) discuss several prevalent myths that have shaped the attitudes that New Mexican speakers have toward NMS. Among these is the myth that NMS is the Spanish of sixteenth the romanticized image of New Mexico Spanish as proof of a
159 could be this same linguistic insecurity that inhibited participants from providing unequivocally positive descriptions of NMS on the language attitudes assessment. An alternative explanation could be the judgments themselves. Of the panel of four judges, two members of the panel each judged two different comments describing ve. In these four instances, the positive judgments were overruled by the other three judges who designated them as neutral. However these occurrences suggest that participants may have provided these comments in order to put NMS in a positive light within the context of an academic investigation, and thus according to the perspective of the participant these comments may have been considered positive. An exit interview with participants would be able to shed more light on the intended connotation of the co mments that were not unequivocally positive or negative. Many participants provided descriptions of NMS that are interesting to juxtapose with their comments toward other varieties. For example, one participant describes most striking example of opposing opinions of the two varieties comes from a participant and in discussing othe
160 made towards NMS in and of themselves appear to be neutral, except when juxtaposed against comment s mad e by the same participant toward other varieties. For example, two different high her participant uses both positive and negative descriptors to describe NMS : described by Bills and Vigil (2008) a nd detailed in Chapter 2 in which New Mexican speakers tend to view standard Spanish as good but non standard Spanish, specifically NMS, is viewed as bad or deficient. The answers provided by the participants in this section provide insight into the biase s and opinions that exist among New Mexicans toward NMS and contextualize the LAV scores that were assigned to each of the participants based on these and other answers provided in the language attitudes assessment. Many of these answers lend credence to t he assertion by Bills (1997) that speakers of NMS tend to express negative attitudes toward their own variety. These answers show that negative attitudes toward NMS still in fact do exist by New Mexicans of varying proficiency levels and across various gen erations. Further discussion of answers provided in this section will appear in Chapter 5 Summary The first research question determined the likely outcome category of dominant language identification given various combinations of two independent variable s: condition and proficiency level of the listeners. The four condition s in question consisted of combinations of gender agreement and connected speech Two analyses were
161 run, one comparing the likely outcome categories for language dominance of Englis h vs. Spanish and another comparing Both vs. Spanish In both analyses, no condition yielded a consistent likely outcome category across all listener groups. Arguably the least proficient of the four conditions in this study was gender agreement, connect ed speech Yet t he low proficiency listeners were more likely to designate speakers of this condition as Spanish dominant than English dominant. Only the high proficiency group produced expected results by yielding a likely outcome category of English fo r all condition s for all but arguably the most proficient of the four condition s, for which they had a likely outcome category of Spanish. Similar results were produced by the second analysis, with variat ion among the low proficiency and intermediate proficiency groups across all four condition s, but the high proficiency group yielded a likely outcome category of for all but the most proficient condition for which they had a likely outcome category of Spanish. The second research question determined the likely outcome category of language ratings given various combinations of the two independent variables of listener proficiency level and the presence or absence of gender agreement and linking a s manifested in the four condition s as discussed above The analysis found that condition s, regardless of the presence of linking or errors in gender agreement. The exception to this was the high proficiency group, which yielded a likely outcome The third research question determined the likely outcome category of language ratin g given various combinations of the two independent variables of proficiency level
162 of the listeners and the perceived region of origin of the speakers on the audio sample. rating category across all p roficiency levels when the perceived region of origin was New Mexico and the U.S. outside of New Mexico When the perceived region of origin was Mexico, Spain, or Other Spanish speaking country, the prevalent likely outcome rating category was proficiency listeners for all three regions and by intermediate proficiency listeners for the final region only. The fourth research question expanded on the prev ious question by introducing a third independent variable, that of the language attitudes value or LAV. As in the likely outcome rating category when the perceived region of origin was New Mexico. was the U.S. outside of New Mexico, listeners that demonstrated negative attitudes toward NMS, those with LAVs of 0 and 1, produced a likely outc ome across all proficiency levels save for the high proficiency group. In addition, when the perceived region of origin was Other Spanish speaking country, the only likely outcome ls for participants that demonstrated positive attitudes toward NMS, those with LAVs of 2, 3, and 4. However for listeners with LAVs of 0 or 1, the likely outcome rating category varies between A qualitative analysis of the optional open ended question on the rating sheet found more overall comments by intermediate proficiency listeners than the other two
163 groups. The majority of their comments referenced pronunciation, as did the majority of the comments by low proficiency speakers High proficiency listeners provided more comments referencing gender agreement than pronunciation or delivery and also provided more comments referencing gender agreement than any other group. The only group to provide more comments referencing delivery th an gender agreement was the low proficiency group. In viewing the comments referencing delivery, participants found several ways to describe the occurrence of linking demonstrating that at least some participants across all three proficiency levels posses s the linguistic awareness to describe this phenomenon and indicate it as a salient factor in their judgments of the listeners. The qualitative analysis on the open ended question on the language attitudes assessment revealed very few unequivocally positiv e descriptions of NMS across all three participant groups All results presented in this chapter and their implications will be discussed in further detail in C hapter 5.
164 C HAPTER 5 DISCUSSION AND CONCLUSIONS This chapter offers a discussion of the result s presented in Chapter 4 Answers to each of the research questions are presented, and a discussion of the contribution of the open ended answers provided by the participants is offered The limitations of the current study are examined and suggested direc tions of future research are discussed. The chapter conclud es by contextualizing the study within the realm of current research. Speaker Effects of Linking and Gender Agreement Research Question 1 linking and the presence or absence of errors in article noun gender agreement related to the identification of a speaker as being Spanish a multinomial logistic regression analysis compa red as the likely outcome c ategories for dominant language of the speakers of the sample as well as In this analysis only high proficiency listeners yielded outcome categories that were consistent with the factors under investigation, the presence of linking and gender agreement when determining language dominance of a speaker The results revealed that high proficiency listeners did recognize the presence or absence of both fluid speech and gender agreement and were likely to identify as Spanish dominant only s peakers of sentences that contained both factors. Low and intermediate proficiency listeners showed no definite pattern in their likely outcomes to indicate that they made their judgments on the language dominance of the speakers based on either linking or errors in gender agreement. These results do not corroborate the original hypothesis, which predicted that condition s that contain linking or fluent speech would result in a likely
165 hree listener groups. However it was also hypothesized that the presence of errors in gender agreement would have no effect on the outcome category of the dependent variable. The results for the low and intermediate proficiency groups support this aspect of the hypothesis, as there is no consistency between the presence of this independent variable and the likely outcome categories yielded by these two groups This means that listeners did not perceive speakers to be Spanish dominant as long as their speec h was presented in fluid, connected speech, even if the speech contained errors in gender agreement, as predicted One very important possible explanation that the results of the low and intermediate proficiency groups deviated so drastically from the hy pothesis could be that at least some listeners in these groups understand that connected speech is not obligatory for native speakers as is gender agreement, and these listeners made judgments accordingly. Another possible explanation for the apparent lack of consistency in the answers rendered by low and intermediate proficiency listeners is that the errors in gender agreement are indeed perceived by these listeners, however the fact that they are presented in fluent, connected speech left them unsure of the correct form therefore they assumed the fluent sounding speaker to be producing the correct form. Thus, the dominant language judgments rendered by this group may not necessarily be representative of their overall knowledge of gender. Furthermore, t he low standard deviations within the intermediate and low proficiency groups imply that the judgments of the two groups were not random but based on a factor or combination of factors extraneous to the variables currently under
166 investigation. This is a nota ble result that merits further research into what those variables are in order to fully understand the processes by which low and intermediate proficiency heritage speakers of Spanish distinguish native and non native speech and whether or not these proces ses include speaker effects other than linking and gender agreement such as intonation and vowel measurements or listener effects such as proficiency level or age. b were similar with the exception of the outcome category of for the low proficiency group. This shows that all groups were fairly consistent in their judgments of whether the speakers of each condition were Spanish do minant rega rdless of the comparison outcome, lending credence to the idea that low and intermediate However, b Spanish disti nction may have been easier for participants to make than the Both Spanish distinction The implication of this finding is that that it is easier for listeners to determine that a speaker resides closer to the extreme s of the bilingualism spectrum than to determine that a speaker is equally proficient in both languages, residing closer to the middle of the spectrum. This belief highlights the complexity of bilingualism in general, and demonstrates why it is problematic to attempt to define heritage speakers and individuals who were raised in a bilingual environment as pertaining strictly to either end of the bilingualism spectrum. Research Question 2 linking and the presence or absence of errors in article noun ge
167 A n ordina l regression analysis / cumulative logit model was used to assess the likelihood that participants would assign a particular rating or lower to a stimulus sentence based on the condition. The resul ts showed that low and intermediate speakers of all conditions regardless of the absence or presence of linking and gender agreement. This finding does not corroborate the original hypothesis, which predicted that listeners across all proficiency levels would assign higher ratings to sentences that and the condition t this would occur regardless of whether the sentence contained errors in gender agreement. The high proficiency group was that contain both linking and gender agreem ent, suggesting that high proficiency listeners were able to detect the presence or absence of these variables and assign ratings accordingly. Examinations of the raw data and the values of the regression coefficients rendered by the statistical analysis support the idea that high proficiency listeners were more likely than intermediate and low proficiency listeners to assign ratings in accordance with the absence or presence of the variables in question In a similar vein to the results of Research Questi on 1, these results imply that a speaker must employ both linking proficient heritage speakers of Spanish. Intermediate and low proficiency heritage speakers, however, appear to base language ratings on factors external to the scope of this study.
168 Qualitative Analysis Chapter 4 presented a qualitative analysis of the answers provided by participants to the open particular words or sounds did you notice in their ac cent or grammar? Recall that t he majority of the comments by intermediate and low proficiency participants referenced speaker pronunciation, while high proficiency participants provided more comments referencing grammar than pronunciation or delivery and also provided more comments referencing grammar than any other group. This finding supports the results of Research Question 1 which implied that intermediate and low proficiency listeners utilized factors other than linking and gender agreement in making language judgments, while high proficiency listeners made language judgments in accordance with the factors linking and gender agreement. Although the high proficiency group provided fewer comments referencing linking than any other group 40% of those co mments inaccurately described the occurrence of linking within the token in question. Taking into account the fact that the high proficiency group was likely to say a speaker was Spanish dominant only if they demonstrated gender agreement as well as linked speech, these findings suggest that high proficiency listeners possess greater procedural knowledge or sub conscious awareness of linking than they do explicit, declarative knowledge. T hese results demonstrate how difficult it is for listeners to accurate ly express the speech cues that they tune into in assessing the speech of others It was noted that listeners demonstrated difficulty in accurately expressing the speech cues that they tuned into in assessing the speech of others. Adding another layer to the difficulty in defining heritage speakers is the speculation that the apparent lack of consistency in the answers rendered by this group may not necessarily be
169 representative of their overall knowledge of gender but rather a demonstration of the lingui stic insecurity experienced by listeners when presented with what they indeed perceive to be errors in gender agreement embedded with in fluent, connected speech Indeed, the documented feelings of linguistic insecurity felt by speakers of this region (Bil ls & Vigil, 2008) may have left participants feeling inadequate and unable to judge the speech of a speaker who is able to produce fluid speech or even non fluid speech. Implications The contribution of these findings to the field of heritage language rese arch is that they provide further evidence that heritage speakers are not a homogenous group that can or should be easily defined. As speakers that reside within the gray areas of the bilingualism spectrum rather than at either extreme, the many facets of these speakers must be explored not only for pedagogical purposes to create a better s but also f or the purpose s of linguistic science and learning more about the nature of bilingu alism. These facets include linguistic experience, proficiency level, linguistic generation, All of these factors contribute to the diversity of the population and li kely will have an implication on any linguistic study involving these speakers. Thus, future studies involving heritage speakers should take into consideration these fa ctors in selecting participants and interpreting results. Listener Effect s of Language A ttitudes and Perceived Region of Origin Research Question 3 The third research question inquired about the nature of the relationship between language ratings and the variety of Spanish believed to be heard on the speech sample
170 An ordinal regression analy outcome rating category across all proficiency levels when the perceived region of origin was New Mexico or the U.S. outside of New Mexico. When the perceiv ed region of origin was Mexico or Spain, intermediate and hig h proficiency listeners yielded a likely outcome rating category while low proficiency listeners maintained a likely outcome speaking proficiency liste intermediate and low proficiency listeners maintained a like ly outcome category of T he results of the low proficiency listener group failed to uphold the hypothesis that listeners would as sign lower ratings to stimuli that they identified as being spoken by a speaker of NMS than to speakers of other varieties. While both the intermediate and high proficiency groups did uphold this hypothesis, the high proficiency group more consistently did possible Spanish speaking region of origin options, while the intermediate proficiency results s uggest a bias on the part of intermediate and high proficiency listeners toward varieties from Mexico, Spain, and other Spanish speaking countries since these were The finding that the high proficiency group was more consistent in demonstrating this bias is in line with their overall performance in other areas of this study in which this group provided consistent results throughout various analyses The fact that it was the hig h proficiency group that demonstrated a bias toward non NMS varieties was consistent with the
171 personal linguistic history of this group, which was composed of second generation New Mexican bilinguals. This is a group that has been documented to have been t old by people outside their speech community that NMS is deficient by virtue of the fact that it is a contact variety. This result hints at the ramifications of growing up in a society that places little value on contact language varieties and on bilingual ism in general, as this group did throughout the twentieth century. While these results are particularly fascinating, they alone do not necessarily suggest these listeners hold negative attitudes toward NMS because they do not make reference to the explic it attitudes of the participants but rather reflect their implicit judgments towards the speakers In order to more accurately investigate the language attitudes of the participants, the upcoming research question will evaluate the effect of the independen language ratings. Research Question 4 The fourth research question investigated the nature of the relationship between language ratings and the language attitudes of the participants tow ards NMS. Recall that the language attitudes of the participants were quantified by their LAV on a scale of 0 4, whereby participants with a higher LAV demonstrated more positive attitudes toward NMS on the language attitudes questionnaire. The analysis fo und that when the perceived region of origin was New Mexico regardless of the attitudes expressed toward NMS on the attitudes assessment as calculated by their LAV. In contrast, variability was found in the likely outcome rating categories of listeners with varying LAVs when the perceived region of origin was the U.S. outside of New Mexico
172 h proficiency participants who did not demonstrate n the perceived region of origin was another Spanish it was New Mexico. This was the only aspect of the results that upheld the hypotheses, which predicted that participants with low LAVs would assign lower ratings to spea kers believed to be from New Mexico than to speakers believed to be from another Spanish speaking country. There appears to be an association between language attitudes expressed toward NMS and the likely rating assigned to speech perceived to be spoken b y someone from the U.S. and from other Spanish speaking countries, but not from New Mexico. The lack of association between the LAVs of the participants and the ratings they assigned to speakers they believed to be from New Mexico could be due to the omiss ion of linguistic markers of NMS in the stimuli. This result alludes to the importance of these markers in triggering the attitudes that participants demonstrated on the language attitudes assessment. Future studies on the association between language atti tudes and language ratings should focus on the presence and absence of linguistic variables that are markers of NMS such as code switching and the use of specific lexical items that are unique to NMS and investigate any possible variability in language rat ings that are assigned to stimuli containing and omitting these variables. Qualitative Analysis The qualitative analysis of the answers provided on the language attitudes assessment illuminate s the biases that some participants demonstrated against NMS
173 an d contextualize s the LAVs. This analysis revealed that the majority of participants at best failed to provide a positive description of NMS and at worst provided a negative description of NMS. Thus while not mutually exclusive, participants provided more n egative descriptions of NMS than they did of non NMS varieties as a singular group. Additionally, participants provided more positive descriptions of non NMS varieties as a singular group than of NMS. Thus like the quantitative data, the qualitative data i n this section show that participants demonstrated higher opinions toward other varieties of Spanish than toward NMS. While these results do not necessarily show that New Mexicans invariably express negative attitudes toward NMS, they do show that New Mexi cans of varying proficiency levels and across various generations display neutral or negative opinions of NMS and more frequently express higher opinions of other varieties of Spanish. This conclusion tangentially reflects the assertion by Bills and Vigil investiga tion confirm the answers to Research Question 3 and Research Question 4 which demonstrate that New Mexicans do indeed demonstrate more negative attitudes toward NMS than they do toward other varieties. This finding is particularly compelling given that previous documentation on negative attitudes by New Mexicans toward their own variety of Spanish has been primarily anecdotal. The intermediate and high proficiency p articipant groups provided more neutral answers than positive answers regarding NMS and provided a higher number of
174 positive comments in reference to other varieties than to NMS. These facts are in agreement with the answers to Research Question 3 and Rese arch Question 4 in which from Mexico, Spain, or other Spanish speaking countries while assigning a rating of believed to be from New Mexico. This finding suggests that while these participants do not necessarily hold a negative opinion of NMS, they express higher opinions toward varieties of Spanish than to NMS. Although the neutral and negative answers provided by the low proficiency group als o outnumber the positive answers they provided regarding NMS, this qualitative result does not reflect the answers to Research Question 3 for this listener group, in which they resulted in a igin. When viewed in the light of the results for Research Question 3, the comments of the low proficiency group regarding NMS support the previously drawn conclusion that listeners who demonstrate less believed to be speakers of this variety due to the fact that traditional markers of NMS such as code switching were absent from the stimuli. The descriptions of NMS and of other varieties of Spanish provided by participants on the lan guage attitudes assessment are in ac cordance with the answers to Research Question 3 and Research Question 4 in which participants overwhelmingly yielded a likely outcome speaker was New Mexico an when the perceived region of origin of the speaker was another Spanish speaking country. In conjunction with the results of this qualitative investigation of the language
175 attitudes assessment, th ese results confirm suggest that participants believe that NMS Implications The contribution of these results to the field of foreign accent perception is the finding that although listener judgment data rendering un biased results may be difficult due to the listener effects that may influence their judgments. Existing languag e attitudes among the listeners were shown to have an effect on language ratings. This result is notable because other studies on listener effects on language ratings and on foreign accent detection have focused on factors ngth of residence in the U.S. (Flege, 1992) that may affect ability to detect foreign accent, yet no known study has examined the effect of existing language attitudes toward different varieties of a given language Negative language attitudes may affect h ow listeners assess the speech of speakers believed to be speakers of those and other varieties of the language and affect the results. The implication of this finding suggests that in future studies on foreign accent perception where language attitudes ar e not a desired variable, participants should be vetted prior to their participation to ensure neutral language attitudes toward different varieties of the language in question, in order to limit one possible variable that may affect the results of the stu dy. The contribution of this study to the body of knowledge on NMS is substantial T he current investigation is one of the only empirical works that have evaluated the language attitudes of New Mexican speakers toward NMS. This study was conducted in a reg ion where both bilingualism and the local non standard variety of Spanish were
176 once devalued and have since experienced a resurgence of support and efforts to reverse the language shift to English. Despite these efforts, the stigma toward NMS persists amo ng speakers in this area, even among third generation speakers of NMS as demonstrated in their language attitudes assessments The results encountered in this study support the anecdotal belief that New Mexicans display negative attitudes toward their own variety of Spanish The implication of this finding is that bilingualism and non standard varieties should be valued and respected both outside of and within the communities that speak these varieties, otherwise it may take several generations to overcome the stigma created by a society that does not value these varieties. The results of the analyses discussed in this chapter have answered the research questions that guided the study; however they have also brought to light new questions and directions for future research. These ideas are discussed below. Limitations and Directions for Future Research As with any empirical work, the current study is inhibited by a small number of limitations. Some of these limitations are due to adjustments made to the meth ods used in order to accommodate the population of participants currently under investigation, while others are related to the structure of the study itself and the materials used to assess participants. These limitations and suggestions on how to overcome them in future research are discussed below. One limit ation of this study is that participants were signaled beforehand of the possibility that grammar errors may be present in some of the stimuli, as has been previously mentioned in an effort to avoid t he possibility that listeners would assign speakers the same rating regardless of the variables present in the stimuli simply This limitation could be remedied in
177 future studies by having a higher number of spea kers each reading a lower number of stimulus sentences. In the current study, a total of four core speakers each contributed eight stimulus sentences to the speech sample. Future studies could have eight speakers each reading four sentences (one of each co ndition) and maintain the sample size of the current study. This would reduce the repetition of the speakers appearing in the sample and the likelihood that participants would recognize th e voices of individual speakers Another limitation of this study is related to the analysis of the language attitudes of the participants. Even with the open ended answers provided on the language attitudes assessment, it was difficult in many cases to ascertain the intended implication of the participant. An example of t his is the comments providing a historical background made by participants of the high proficiency group. Future studies could remedy this limitation by providing a more explicit language attitudes assessment to elicit more specific answers regarding the a ttitudes of participants toward NMS and other varieties of Spanish. Similarly, brief exit interviews could allow participants to more explicitly clarify their attitudes and ratings through stimulated recall protocols. An exit interview, though time consumi ng, would provide information on the motivations behind participants assigning particular ratings to certain speakers, as well as general tendencies toward language judgments. Future studies that investigate language attitudes in this or any region might b enefit from focusing solely on language attitudes without looking at linguistic factors. As we have seen, when linguistic and attitudinal factors are combined, it becomes difficult to accurately gauge the language attitudes of participants when expected
178 li nguistic markers of NMS are not present in the stimuli. This study has shown that research on existing language attitudes in New Mexico has the potential to produce compelling results and merits further study In spite of these limitations, the results ob tained here reveal that high proficiency listeners stand apart from intermediate and low proficiency bilinguals in their performance on detecting both gender disagreement and fluid speech and to assign ordingly. In addition, this study has shown that participants were likely to assign higher ratings to speaker s that they believe to be from another Spanish speaking country than to speakers they believe to be from New Mexico or the U.S. outside of New Mexi co a conclusion that was confirmed by many of the comments made on the open ended questions of the language attitudes assessment. Summary This investigation sought to learn more about heritage language processes and performance with regard to speech perce ption, and to learn more about the listener and speaker effects on dominant language perception and language ratings. The results obtained herein have important implications for the fields of foreign language perception, heritage language development, and studies on NMS. This study adds to the body of knowledge on foreign language perception by providing information on the role of fluid speech and morphosyntactic errors in B oth connected speech and gender agreemen t were shown to be necessary in order for a speaker to receive high ratings from highly proficient heritage listeners and to be perceived as Spanish dominant by these listeners. These results ran contrary to the hypothesis that listeners would
179 perceive a s peaker to be Spanish dominant as long as her/his speech was presented in fluid, connected sentences, even if those sentences contained errors in gender agreement. The implication of this finding for the fields of second language acquisition and foreign lan guage teaching is that listeners who wish to be perceived by a high proficiency interlocutor as Spanish Spanish should work on presenting speech that is both fluid and accurate in terms of gender agreement. In addition, we have seen that not all listeners rate speakers and perceive them to be Spanish dominant based on the presence or absence of fluid speech. The impact of this finding on existing knowledge of foreign language perception is that factors indi cating non native speech that are salient to some listener groups may not be salient for other listeners. This means that factors that are found to be significant predictors of accent ratings for one listener group should also be tested on various listener groups of varying proficiency levels and linguistic experiences. Existing language attitudes among the listeners were also shown to have an effect on language ratings. This result is notable because other studies on listener effects on language ratings an d on foreign accent detection length of residence in the U.S. (Flege, 1992). Although prior research has investigated the effect of extrinsic factors such as listening context and word frequency on lis to accent markers (Levi, Winters, & Pisoni, 2007) no known studies have viewed listener language attitudes as a variable in detecting non native speech and assigning language ratings. Furthermore, this study adds a new dimension to resear ch on language attitudes. Most prior studies on language attitudes have utilized the matched
180 guise technique to elicit language attitudes from listeners after having heard a speech sample. The present study has shown that existing negative language attitud es toward one or various varieties may be a predictor of low language ratings of individuals believed to be speakers of those and even of other varieties. Future studies should evaluate the language attitudes of potential participants toward the language i n question in order to limit one possible undesired variable. The results of this study also have important implications for understanding heritage language processes and development It was revealed that heritage listeners of varying proficiency levels an d generations likely base their judgments on different speech variables in assigning language ratings and in determining language dominance of a speaker. These results may provide further evidence that the generalization that heritage speakers possess nati ve like phonological competence does not necessarily apply to speakers of all generations and proficiency levels at least where performance in detect ing language dominance based on the variables under investigation of this study is concerned Although more research would be needed to conclude that ability to judge accent is equivalent to individual phonological ability we have nonetheless seen differences among the perceptive performances of speakers of varying linguistic backgrounds The findings encounte red in this study add to e xisting research that indicates that not all heritage speakers are homogenous and highlights the importance both instructing heritage learners and placing them in heritage language programs in order to serve them in a manner that meets their specific needs and not the needs of a (Carreira, 2007; 2012). In addition, the results of this study
181 indicate that if the aim of heritage speaker development is to gain proficiency in their heritage language and strengthen their ethnic identity, speakers should strive to perfect both the fluidity of their speech as well as accuracy in gender agreement, as mentioned above, in order t speech community. The contribution of this study to the body of knowledge on NMS is also significant. The results of this study allude to the importance of traditional linguistic markers of t his variety such as code switching in determining in group affiliation and in Most notably, the current investigation is one of the only empirical works that have evaluat ed the language attitudes of New Mexican speakers toward NMS. This study has shown that New Mexican bilinguals of varying proficiency levels express negative or neutral opinions of NMS while expressing more positive opinions of other varieties of Spanish. The results encountered in this study support the anecdotal belief that New Mexicans display negative attitudes toward their own variety of Spanish, as discussed in Chapter 2. This finding is particularly compelling and suggests this area merits further st udy. Future studies on language attitudes toward NMS and toward other contact language varieties should be familiar with the methodology used in the current study and heed recommendations presented here within for future research in this area.
1 82 APPENDIX A LANGUAGE BACKGROUND QUESTIONNAIRE Language Background questionnaire Participant number: ______________ Personal History Date of birth: ____________________________ Place of birth: (City, State, Country) _______________ ______ _____ How long have yo u lived in ____________________ ? ____ __________________ Have you ever lived in any other Spanish speaking places? (F or how long?): ____________________________________________________________________ Language History 1. W hat was your first language? ___ __________________________________ 2. Overall, which language do you think is your strongest? ________ A. English ________ B. Spanish ________ C. I think I am equally fluent in both ________ D. It depends on the situation 3. Family birth history New M exico (Where?) U.S. outside of New Mexico (Where?) Outside of the U.S. (Where?) Where was your father born? Where was your mother born? Where was your grandmother on born?
183 Where was your grandfather on your ? Where was your grandmother on born? Where was your grandfather on your 4. Please mark the answer that best corresponds to your family language history with an X. Mostly English Mostly Spanish Both equally De pends on the situation / talking to What language does your father speak? What language does your mother speak? What language does (did) your maternal grandmother speak? What language does (did) your maternal grandfather speak? What language does (did) your paternal grandmother speak? What language does (did) your paternal grandfather speak? 5. How much Spanish was spoken in your family when you were a child? Did this ever change? Please explain.
184 Language Use 6. Using the scale below please indicate how often you: In English In Spanish Write Speak Listen to music Read (newspapers, magazines, books or websites) Watch television or movies 1. Never 2. Every few years 3. One or two times a year 4. One or t wo times a month 5. A few times a week 6. Every day, sporadically throughout the day 7. Every day, almost all day 7) On a scale of 1 4 please estimate your linguistic ability in English and in Spanish. 1 = minimal ability 4 = native speaker abili t y In Englis h In Spanish Ability to SPEAK Ability to UNDERSTAND Ability to WRITE Ability to READ PRONUNCIATION ability GRAMMAR ability Overall ability 8 ) Are you satisfied with your abilities in Spanish or would you like to improve them? In what way?
185 APPENDIX B ELICITATION TASK Please read the following sentences out loud before we begin recording so that you can get familiar with each of the words Then read each sentence into the microphone If you mess up, go ahead and start from the beginning of the sentence You may record each sentence as many times as you need to until you feel comfortable with it. Group A Group B Los amigos ven las pelculas. Las amigos ven las pelculas. Los aviones llevan las maletas. Las aviones llevan las maletas. Los adultos pagan las cuentas. Las adultos pagan las cuentas. Los altos cubren los bajos. Las altos cubren los bajos. Los abuelos compran las vitaminas. Las abuelos compran las vitaminas. Las abejas cantan las canciones. Los abe jas cantan las canciones. Los abrigos cubren los brazos. Las abrigos cubren los brazos. Los andinos tocan las flautas. Las andinos tocan las flautas. Los aos llevan los dolores. Las aos llevan los dolores. Los anuncios venden los productos. Las anuncios venden los productos.
186 ELICITATION TASK Group C Group D Los, amigos ven las pelculas. Las, amigos ven las pelculas. Los, aviones llevan las maletas. Las, aviones llevan las maletas. Los, adultos pagan las cuentas. Las, adultos pagan las cuentas. Los, altos cubren los bajos. Las, altos cubren los bajos. Los, abuelos compran las vitaminas. Las, abuelos compran las vitaminas. Las, abejas cantan las canciones. Los, abejas cantan las canciones. Los, abrigos cubren los brazos. L as, abrigos cubren los brazos. Los, andinos tocan las flautas. Las, andinos tocan las flautas. Los, aos llevan los dolores. Las, aos llevan los dolores. Los, anuncios venden los productos. Las, anuncios venden los productos.
187 ELICITATION TASK G roup E Loz amigoz ven laz pelculaz. Loz avionez llevan laz maletaz. Loz adultoz pagan laz cuentaz. Loz altoz cubren loz bajoz. Loz abueloz compran laz vitaminaz. Laz abejaz cantan laz cancionez. Loz abrigoz cubren loz brazoz. Lo z andinoz tocan laz flautaz. Loz aoz llevan loz dolorez. Loz anuncioz venden loz productoz. Group F Lo(th) amigo(th) ven la(th) pelcula(th). Lo(th) avione(th) llevan la(th) maleta(th). Lo(th) adulto(th) pagan la(th) cuenta(th). Lo(th) alto(th) c ubren la(th) bajo(th). Lo(th) abuelo(th) compran la(th) vitamina(th). Lo(th) abrigo(th) cubren la(th) brazo(th). Lo(th) andino(th) tocan la(th) flauta(th). La(th) abeja(th) cantan la(th) cancione(th). Lo(th) ao(th) llevan la(th) dolore(th). Lo(th) anunci o(th) venden la(th) producto(th).
188 APPENDIX C RATING SHEET Participant # _____________ Module Order # _________________ Speaker # T his person probably communicates mostly in: What helped you reach this conclusion? Spanish was : What p articular words or sounds did you notice in their accent or grammar? This speaker is probably from : 1 English Spanish Both Their Grammar Their pronunciation Other_________ _________________ Excellent Good Fair Poor New Mexico The U.S. outside of N.M. Mexico Spain A nother Spanish speaking country 2 English Sp anish Both Their Grammar Their pronunciation Other_________ _________________ Excellent Good Fair Poor New Mexico The U.S. outside of N.M. Mexico Spain A nother Spanish speaking country
189 APPENDIX D LANGUAGE ATTITUDES ASSESSMENT Please fill out this survey to the best of your ability 1. Have you ever heard Spanish spoken by speakers outside of New Mexico (in person or on television, radio, movies)? Do you know where they were from? ___________________________________ _________________________________ 2. Have you ever noticed any differences between the Spanish spoken in New Mexico and the Spanish heard on TV or the Spanish of other places? _______ pronunciation/accent _______ words _______ other differenc es _______ ___________ ______________________ _______ None noticed 3. Which dialect (s) of Spanish do you think should be reported on the local Spanish news? _______ Mexican Spanish _______ New Mexican Spanish _______ Castilian Spanish (from Spain) _______ Other (Puerto Rican, Cuban, South Am erican, etc.) ______ _______ _______ Not sure 4. Which dialect (s) of Spanish do you think should be taught in local Spanish classes? _______ Mexican Spanish _______ New Mexican Spanish _______ Castilian Spanish (from Spain ) _______ Other (Puerto Rican, Cuban, South American, etc.) ____ _________ _______ Not sure 5. Which dialect (s) of Spanish would you like to speak? _______ Mexican Spanish _______ New Mexican Spanish _______ Castilian Spanish (from Spain) _______ Oth er (Puerto Rican, Cuban, Sout h American, etc.) ___ __________ _______ Not sure
190 6. How would you describe the Spanish spoken in New Mexico? 7. If familiar, how would you describe the Spanish of Mexico and/or other countries? Thank you for partic ipating in this study!
191 APPENDIX E PARAMETER ESTIMATES TABLES FROM SPSS FOR MULTINOMIAL LOGISTIC MODEL (RQ1) Parameter Estimates Language a B Std. Error Wald df Sig. Exp(B) 95% Confidence Interval for Exp(B) Lower Bound Upper Bound English Intercept 1.528 .260 34.558 1 .000 [Sentence=1.00] 1.845 .328 31.568 1 .000 6.327 3.324 12.041 [Sentence=2.00] 1.096 .335 10.715 1 .001 2.993 1.553 5.771 [Sentence=3.00] 1.106 .330 11.232 1 .001 3.024 1.583 5.775 [Sentence=4.00] 0 b . 0 . . [Group=1.00] .389 .387 1.008 1 .315 1.476 .691 3.153 [Group=2.00] .330 .369 .799 1 .371 1.391 .675 2.866 [Group=3.00] 0 b . 0 . . [Sentence=1.00] [Group=1.00] .627 .492 1.624 1 .202 .534 .203 1.401 [Sentence=1.00] [Group=2.00] .116 478 .059 1 .808 1.123 .440 2.867 [Sentence=1.00] [Group=3.00] 0 b . 0 . . [Sentence=2.00] [Group=1.00] .156 .490 .101 1 .751 1.168 .447 3.050 [Sentence=2.00] [Group=2.00] .126 .476 .070 1 .791 .882 .347 2.242 [Sentence=2.00] [Group=3. 00] 0 b . 0 . . [Sentence=3.00] [Group=1.00] 1.402 .523 7.174 1 .007 .246 .088 .687 [Sentence=3.00] [Group=2.00] .406 .479 .717 1 .397 .667 .261 1.704
192 [Sentence=3.00] [Group=3.00] 0 b . 0 . . [Sentence=4.00] [Group=1.00] 0 b . 0 . . [Sentence=4.00] [Group=2.00] 0 b . 0 . . [Sentence=4.00] [Group=3.00] 0 b . 0 . . Both Intercept .681 .189 12.940 1 .000 [Sentence=1.00] .658 .288 5.216 1 .022 1.930 1.098 3.394 [Sentence=2.00] .550 .271 4.123 1 .042 1. 734 1.019 2.948 [Sentence=3.00] .331 .275 1.455 1 .228 1.393 .813 2.387 [Sentence=4.00] 0 b . 0 . . [Group=1.00] 1.126 .262 18.456 1 .000 3.083 1.845 5.153 [Group=2.00] .665 .260 6.520 1 .011 1.945 1.167 3.241 [Group=3.00] 0 b . 0 . . [Sentence=1.00] [Group=1.00] .509 .397 1.640 1 .200 .601 .276 1.310 [Sentence=1.00] [Group=2.00] .033 .411 .007 1 .935 1.034 .462 2.314 [Sentence=1.00] [Group=3.00] 0 b . 0 . . [Sentence=2.00] [Group=1.00] .725 .385 3.548 1 .060 .484 .228 1.030 [Sentence=2.00] [Group=2.00] .401 .379 1.117 1 .291 .670 .319 1.408 [Sentence=2.00] [Group=3.00] 0 b . 0 . . [Sentence=3.00] [Group=1.00] .730 .373 3.821 1 .051 .482 .232 1.002 [Sentence=3.00] [Group=2.00] .120 .378 .101 1 .751 .887 .422 1.862 [Sentence=3.00] [Group=3.00] 0 b . 0 . .
193 [Sentence=4.00] [Group=1.00] 0 b . 0 . . [Sentence=4.00] [Group=2.00] 0 b . 0 . . [Sentence=4.00] [Group=3.00] 0 b . 0 . . a. The reference category is: Sp anish. b. This parameter is set to zero because it is redundant.
194 APPENDIX F RATING SHEET ANSWERS Appendix F provides an abridged list of comments from the question on the ccent origin, and perceived dominant language that pertain to each comment on the list. Out of the 1731 or so stimulus sentences analyzed (577 tokens for each of the three participant groups), all instances where participants declined to answer the optional question were omitted from the appendix. Comments that were duplicated or otherwise provided more than once were also omitted from the abridged list of comments in the a ppendix, in order to provide an overview of the nature of the comments without being excessively repetitive In these cases, the comment contains an asterisk to denote that the comment appeared more than once for the particular participant group in which i t is included. The corresponding ratings for those comments were omitted as well due to the fact that each instance of the comments in question likely corresponded to more than one rating category The final items omitted from the appendix were comments th at listed one singular word in the stimulus sentence, the pronunciation of which appeared salient to the participant. This was done in order to trim the extensive list of comments and provide example of the variety of comments without focusing on particula r words in the stimulus sent ences that appeared salient to the participants.
195 Group by proficiency level Rating Perceived region of origin Perceived dominant language Comment High grammatical error* las/los* Good U.S. Eng academic [sou nding] Excellent Other Eng accent, grammar Excellent NM Eng very good movement of words, abejas Excellent U.S. Both flow of speech Good Other Both good pronunciation, bad grammar Good NM Both over emphasized /slow hesitates* X U.S. Eng hesitation accent* intonation* Good Other Both flow of speech, enunciation Good NM Both grammar, las abrigos, brazos, accent ok Excellent NM Span sounds like northern nm Spanish Good NM Eng has heard Spanish, does not know it Exce llent Other Span wrong grammar, great accent Excellent NM Both pronunciation/timing Fair U.S. Eng poor grammar Good xx Span speed of speech Good NM Eng reads well but not first language Excellent Other Eng could be first language Excellent Othe r Span good Excellent U.S. Span speed of communication Good NM Eng method of pronunciation Fair U.S. Eng slow in speech Good NM Both accent, grammar, las adultos, cuentas Fair U.S. Eng read not comfortable Fair U.S. Span incorrect grammar Exc ellent Other Eng very good Excellent Other Eng natural flow of speech Good U.S. Eng reads but not well Excellent U.S. Both speech flows well and correct Good Other Span accent dolores Good Other Both speaks fast Excellent Other Span accent flautas Good U.S. Span grammatical error Excellent NM Span pronunciation
196 Excellent NM Span enunciates well Good U.S. Both grammar, las abuelos, accent Fair U.S. Span poor grammar Intermediate Poor NM Eng fair accent Good NM Both heavy "h" soun d common in NM Good Other Span different accent las/los* Good Other Eng good pronunciation Fair U.S. Span pause btwn "los" and next word Good NM Both las flautas, long vowels Poor U.S. Both hesitant Good NM Span paused at end, r Excel lent Other Eng words run together r* Excellent NM Eng good Spanish, fluid, but not used much, you can tell in her accent Good Other Both good pronunciation and grammar Fair U.S. Both accent, los colores Good NM Eng slight accent Good Othe r Eng flautas, emphasis on "au" Excellent Other Span Mexican accent Good U.S. Both sounds uncomfortable Fair NM Both emphasis on "los" Fair U.S. Eng stresses/accents on "los" Good Other Span llevan, accent on "c" Good Other Span confident Fai r NM Span good accent, but simple grammar mistake Excellent Other Both avejas, emphasis on "j" f luency Excellent Other Both speed of speech Excellent Other Eng very quick, good pronuciation Good Other Span los brazos, hard "s" = z Good N M Eng the last word in the phrase was very Mexican sounding Excellent NM Both sounds like they are familiar with speaking the language Good NM Eng speech was choppy but proper accent Good U.S. Both grammar is good but accent is not native Excellent Other Eng upward inflection, abejas and canciones Fair NM Span sounded hesitant
197 Excellent Other Span slightly accented in English Fair U.S. Both "pagan", hesitant Good Other Both proper pronunciation but slowly spoken Good U.S. Both los aos d ropped the s Good Other Both very articulate Good NM Both accent from NM Excellent Other Span productos "u" Good Other Both good grammar Fair NM Span las flautas long melodic tone Excellent NM Span the words were spoken together really quickl y very fluid Excellent Other Span speed Good Other Eng fast Excellent NM Both pronunciation was great and spoken quickly Excellent NM Eng had a slight English accent Excellent Other Both U's, anuncios, productos Fair NM Both he spoke s lowly Good U.S. Span a, cuentas Good U.S. Eng words were separate in sentence and not quickly run together like native speaker Good NM Both pagan las cuentas, slowed down at end Good NM Span spoken slowly with slight accent Good U.S. Both compran roll of r Excellent Other Span spoken quickly and with confidence Good NM Both words pronounced correctly but separated staccato not fluid Excellent NM Both good grammar Good Other Span frequency, llevan Good U.S. Span pronounced right but spoken slowly Good Other Span emphasis on first word slight English accent Poor NM Eng hesitant but good accent Fair NM Eng cantan, kind of slow Low Excellent Other Both the way he pronounced the "l's" studied Spanish* Excel lent Other Span y sound Excellent NM Eng ease of pronunciation Excellent Other Span the rolling of the r's
198 Fair NM Both could be New Mexican Excellent Other Span structured Good Other Both Outside NM Spanish enunciated each syllable* Go od NM Span trying at Spanish Fair NM Span slowly spoken the "s" sound* accent* Fair NM Both accentuated vowels Good U.S. Eng A little slower. Seemed a little foreign to her. Fair NM Span fluency Excellent Other Both spoke fast, fl autas Good NM Span changed pitch, notes, through sentence b/v sound* Good Other Span Unfamiliar to NM Spanish Excellent Other Span B in brazos Excellent Other Span clear accent Good U.S. Both words come to an abrupt stop Good Other Span confident pronunciation Good NM Both Sounds like NM Spanish Excellent Other Eng smooth and fast Excellent U.S. Span letters blended together in their words Fair NM Span Reading not too familiar w. Spanish Good Other Span the letters rolled toget her sing songy* trying at Spanish* Excellent Other Eng letter "e" strong Fair U.S. Span hesitation s low Fair U.S. Both speed of speech Good Other Both the "z" sounds Excellent Other Span flautas with an o Excellent Othe r Span spoke quickly Excellent Other Span grammar Excellent Other Span just the speed which he read Excellent Other Span smooth word transition Excellent Other Both aos, llevan v pronounced as a b speech* Good Other Span the pronunciatio n of hard consonants the pronunciation of a's* Excellent Other Eng Fluent, easily read.
199 Excellent Other Both steady flow Fair NM Both Didn't feel comfortable reading but knew what she was saying. Excellent NM Both changes in pitch s peed* Good Other Span Unfamiliar New Mexican Spanish Fair NM Both choppy Excellent Other Span smooth, blended words Good Other Span accent, wording Excellent Other Span fast! Good NM Both pronunciation Excellent NM Span pitch changes, smooth
200 APPENDIX G COMMENTS PROVIDED ON LANGUAGE ATTITUDES ASSESSMENT Although only the second question was designated as optional, five participants declined to answer the question describing NMS. One participant declined to answer either question and was ther efore omitted from the following table. Points were added toward the LAV only for descriptions of NMS that were unanimously judged by the panel to be positive. Descriptions marked with an asterisk were unanimously judged by the panel as positive descriptor s. Participant group by proficiency level Participant number Points toward LAV Description of NMS Description of other varieties Low 301 0 Sing songy [sic] slang terms Also has slang, but more accurate to the original language? 302 0 Eng lish and Spanish, it uses a little bit of both Pronounced better spoken faster 303 0 Some English thrown into the mix Spoken faster? Not sure 304 0 L ot of colloquialisms I didn't understand 305 0 Does not always use the right words and other words are made up for it 306 0 Somewhat broken or "Gringo" Mexican = fast, Spain = elegant 307 0 More Castilian in northern NM 308 1 Beautiful Beautiful 310 0 Broken, unique Pure and proper 311 0 Talk fast and use of different words 312 0 Mish mash of English, Spanish and slang that probably would not be understandable in other regions. Less structured, more conversational 313 0 Broken, mixed in with English words 314 0 Poor Beautiful
201 315 0 Very different from more modern speaking countries More modern 316 0 Mixed sentences w both Spanish and English words. Also a lot of slang, not very formal More formal and more traditional 317 0 Broken, slang, choppy More fluid, beautiful 318 0 Slang of Mexican Spanish Very different from Mexican Spanish Intermed. 352 0 Slower, not so much slang Laid back and varied, like English 353 0 Slang, not appropriate in Mexico City for example More relaxed than in some places where spoken fast 354 0 Certain accent, certain word choice Different accents/slang/conjugations 355 0 English/American twang. Spanish is a very pretty want to learn Castilian. NMS has l ittle or no intere st to me 357 0 More casual than other places Very formal 358 0 Combination of a lot of other dialects and accents. Lot of slang More proper 359 0 Thickness of different accents 361 1 Unique. It is what I am most comfortable with Mexican = c lose to New Mexican Spanish, Castilian = foreign 362 0 Old, native & English influences Very fast, blended differently 363 0 English based, relaxed, not as structured More structured, different idiomatic expressions 364 0 Spanglish, sometimes impro per More proper 365 0 Different with wordings Different words & accents 367 0 Americanized 368 1 C lassic, elegant, melodic, consistent, clear pronunciations Fast and full of slang terms. Consistently changing. Pronunciation is mashed between wo rds. 369 0 Mixed, Spanglish More formal sounding 370 0 Slang 371 0 Poor not widely spoken Different words, pronunciation 372 0 mixture of slang and Spanglish, for the most part fluent and grammatically More fluent, grammatically correct
202 correct High 205 0 Mix from other dialects and some English More Castilian 206 0 3 Dialects differing from north, central and southern NM 207 0 Very traditional, close to Spain. Varies throughout the state Not sure 208 0 Different, a lot of the words are individually made up 209 0 Faster, some words are different 210 0 Unique, archaic, some words here are from old Spanish as we were isolated [linguistically, geographically] especially Taos, Que sta, Peasco Rapid, flowery and flowing, I like to hear it 211 0 Old Spain Spanish (400 years old) Evolutionized [sic] 212 0 Spanglish Spoken perfectly 213 0 Spanglish Good Spanish 214 0 Very old Hard to understand sometimes 216 0 Old styl e Spanish, archaic Up to date, evolving rapidly 218 0 Lots of slang, northern NM is different from the rest of the state Words differ 219 1 Sweet* Like old English, but of course in Spanish 221 0 Weak, decreased identity Supported among family / communities 222 0 It varies Speed and intonation is very different 223 0 Sometimes archaic, sometimes too much "Spanglish" Beautiful, but too fast 224 0 North country Castilian Spanish 225 0 Fast, different pronunciation, grammar
203 LIS T OF REFERENCES Agresti, A. (2002). Categorical data analysis New York, NY: Wiley Interscience. Alba Salas, J. (2004). Voice onset time and foreign accent detection: Are L2 learners better than bilinguals? Revista Alicantina de Estudios Ingleses 17 9 30. Altenberg, E. (2005). The judgment, perception, and production of consonant clusters in a second language. International Review of Applied Linguistics 43 53 80. Amengual, M. (2012). Interlingual influence in bilingual speech: Cognate status effect in a continuum of bilingualism. Bilingualism: Language and Cognition, 15 (3), 517 530. Anderson Hsieh, J., Johnson, R., & Koehler, K. (1992). The relationship between native speaker judgments of nonnative pronunciation and deviance in segmentals, prosod y, and syllable structure. Language Learning, 42 (4) 529 555. Annotated SPSS output: Multinomial logistic regression. (n.d.). UCLA: Statistical Consulting Group. Retrieved from http://www.ats.ucla.edu/stat/spss/out put/mlogit.htm Annotated SPSS output: Ordered logistic regression. (n.d.). UCLA: Statistical Consulting Group. Retrieved from http://www.ats.ucla.edu/stat/spss/out put/ologit.htm Asher, J., & Garca, R. (1969). The optimal age to learn a foreign language. The Modern Language Journal, 53 3 34 341. Au, T.K., Knightly, L.M., Jun, S. & Oh, J.S. (2002). Overhearing a language during childhood. Psychological Science, 13 (3), 238 243. Au, T.K., Oh, J.S., Knightly, L.M., Jun, S. & Romo, L.F. (2008). Salvaging a childhood language. Journal of M emory and Language, 58 998 1011. Au, T.K., & Romo, L. (1997). Does childhood language experience help adult learners? In. H.C. Chen (Ed.), The cognitive processing of Chinese and related Asian languages (pp.417 441). Beijing: Chinese University Press. Beaudrie, S.M. (2009). Spanish receptive bilinguals: Understanding the cultural and linguistic profile of learners from three different generations. Spanish in Context, 6 (1), 85 104. Beaudre, S.M. & Ducar, C. (2005). Beginning level university heritage programs: Creating a space for all heritage language learners. Heritage Language Journal, 3 1 26.
204 Beaudrie, S.M., & Fairclough, M. (2012). Introduction: Spanish as a heritage language in the United States. In S.M. Beaudrie & M. Fairclough (Eds.), Spa nish as a heritage language in the United States: The state of the field (pp. 1 17). Georgetown University Press: Washington, DC. Berman, R., & Slobin, D. (1994). Relating events in narrative: A crosslinguistic developmental study Hillsdale, NJ: Lawren ce Erlbaum Associates. Becker, L.A. (1999a). C rosstabs : Measures for nominal data Retrieved from http://www.uccs.edu/~faculty/lbecker/SPSS/ctabs1.htm#10. Symmetric Measures of Association Becker, L.A. (1999b). Crosstabs: Measures for ordinal data. Retr ieved from http://www.uccs.edu/~faculty/lbecker/SPSS/ctabs2.htm#5. Symmetric Measures for Ordinal Data Bernal Enriquez, Y., & Hernndez Chvez, E. (2003). La enseanza del espaol en Nuevo Mxico: Revitalizacin o erradicacin de la variedad Chicana. In A. Roca & M. C. Colombi (Eds.), Mi lengua: Spanish as a heritage language in the United States, research and practice (pp. 96 119). Washington, DC: Georgetown University Press. Bills, G.D. (1997). New Mexican Spanish: Demise of the earliest European var iety in the United States. American Speech, 72 (2), 154 171. Bills, G.D., Hernndez Chvez, A., & Hudson, A. (1993). The geography of language shift: Distance from the Mexican border and Spanish language claiming in the Southwestern United States. Intern ational Journal of the Sociologoy of Language 114 9 27. Bills, G.D., Hudson, A., & Hernndez Chvez, E. (2000). Spanish home language use and English proficiency as differential measures of language maintenance and shift. Southwest Journal of Linguistic s, 19 (1), 11 27. Bills, G.D. & Vigil, N.A. (1999). The Spanish language of New Mexico and southern Colorado: A linguistic atlas. Journal of Sociolinguistics, 16 (2), 292 94. Bills, G.D. & Vigil, N.A. (2008). The Spanish language of New Mexico and south ern Colorado: A linguistic atlas Albuquerque: University of New Mexico Press. Boersma, P. & Weenink, D. (2010). Praat: doing phonetics by computer (Version 5.2.01) [Software]. Available from http://www.praat.org/ Brennan, E., & Brennan, J. (1981). Ac cent scaling and language attitudes: Reactions to Mexican American English speech. Language & Speech, 24, 207 221.
205 Buell Hill, R. (2001). Production and perception of authentic and feigned Spanish accents rida, Gainesville, FL. Caldas, S. & Caron Caldas, S. (1999). Language immersion and cultural identity: Conflicting influences and values. Language, Culture, and Curriculum, 12 (1), 42 58. Carreira, M. (2000). Validating and promoting Spanish in the Unit ed States: Lessons from linguistic science. Bilingual Research Journal, 24 (4), 423 442. Carreira, M. (2003). Profiles of SNS students in the twenty first century: Pedagogical implications of the changing demographics and social status of U.S. Hispanics. In A. Roca & M.C. Colombi (Eds.), Mi lengua: Spanish as a heritage language in the United States (pp. 51 77). Washington, DC: Georgetown University Press. Carreira, M. (2004). Seeking explanatory adequacy: A dual approach to understanding the term Herit age Language Learner. Heritage Language Journal, 2 (1). Retrieved from http://heritagelanguages.org/Journal.aspx Carreira, M. (2007). Teaching Spanish to native speakers in mixed ability language classrooms. In K. Potowski & R. Cameron (Eds.), Spanish in Contact: Policy, Social and Linguistic Inquiries. Washington, DC: Georgetown University Press. Carreira, M. (2012). Meeting the needs of heritage language learners: Approaches, strategies, research. In S.M. Beaudrie & M. Fairclough (Eds.), Spanish as a h eritage language in the United States: The state of the field (pp. 223 240). Washington, DC: Georgetown University Press. Chambers, F. (1997). What do we mean by fluency? System, 25 535 544. Chen, H.C. (2010). Second language timing patterns and their e ffects on native Concentric: Studies in Linguistics, 36 (2), 183 212. Chen, Q. (1997). Toward a sequential approach for tonal error analysis. Journal of the Chinese Language Teachers Association, 32 (1), 21 39. Clegg, J. H., & Wal termire, M. (2009). Gender assignment to English origin nouns in the Spanish of the southwestern United States. Southwest Journal of Linguistics, 28 (1), 1 17. Cobos, R. (2003). A dictionary of New Mexico & southern Colorado Spanish Santa Fe, NM: Museum of New Mexico Press. Cunningham Andersson, U. (1993). Stigmatized pronunciations in nonnative Swedish. PERILUS, 18 81 105.
206 Cuza, A. & Frank, J. ( 2011). Transfer effects at the syntax semantics interface: The case of double que questions in heritage S panish. The Heritage Language Journal, 8 (1). Retrieved from http://heritagelanguages.org/Journal.aspx Dvila, A., Bohara, A., & Saenz, R. (1993). Accent penalties and the earnings of Mexican Americans. Social Science Quarterly, 74 902 915. Derwing, T.M. & Munro, M.J. (1997). Accent, intelligibility, and comprehensibility: Evidence from four L1s. Studies in Second Language Acquisition, 19 1 16. Derwing, T.M. & Munro, M.J. (1998). The effects of speaking rate on listener evaluations of native and fore ign accented speech. Language Learning, 48 (2), 159 182. Derwing, T.M. & Munro, M.J. (2009). Putting accent in its place: Rethinking obstacles to communication. Language Teaching, 42 (4), 476 490. Derwing, T. M., Munro, M. J., & Thomson, R. I. (2008). A longitudinal study of ESL learners' fluency and comprehensibility development. Applied Linguistics, 29 (3), 359 380. Derwing, T.M., Rossiter, M.J., & Munro, M.J. (2002). Teaching native speakers to listen to foreign accented speech. Journal of Multiling ual and Multicultural Development, 23 245 259. Derwing, T.M., Rossiter, M.J., Munro, M.J., & Thomson, R.I. (2004). Second language fluency: Judgments on different tasks. Language Learning, 54 (4), 655 679. Douglas, G.J. (1999). Habl e mos E spaol en casa (We speak Spanish at home). The New Mexico Association for Bilingual Education. Retrieved from http://www.nmabe.net/nmabe_publications/we_speak_spanish.html Durn Urrea, E. (2011). Bilingual usage and linguistics attitudes in a community in northern Ne w Mexico. 40th Annual Meeting of the Linguistic Association of the Southwest (LASSO) [Conference presentation]. Retrieved from http://www.academia.edu/1702713/Bilingual_Usage_and_Linguistic_Attitudes_in _a_Community_in_Northern_New_Mexico Ejzenberg, R. (2 sociolinguistic metaphor. In: H. Riggenbach (Ed.), pp. 287 314). Ann Arbor, MI: University of Michigan Press. Elliott, R. E. (1995) Field independence/dependence, hemispheric special ization, and attitude in relation to pronunciation accuracy in Spanish as a foreign language. The Modern Language Journal, 79 356 371.
207 Ellis, R. (2008). The study of second language acquisition (2 nd ed.). Oxford, NY: Oxford University Press. Espinosa, A. (1975). Speech mixture in New Mexico: The influence of the English language on New Mexican Spanish. In E. Hernndez Chvez, A.D. Cohen, & A.F. Beltramo (Eds.), El lenguaje de los Chicanos (pp. 99 114). Arlington, VA: Center for Applied Linguistics. ( Original work published 1917 in H. M. Stephens & Bolton (Eds.), The Pacific Ocean in History pp. 408 428. New York: The Macmillan Co.) Face, T. (2006). Intervocalic rhotic pronunciation by adult learners of Spanish as a second language. In C.A. Klee & T L. Face, (Eds.), Selected proceedings of the 7th conference on the acquisition of Spanish and Portuguese as first and second languages (pp.47 58). Somerville, MA: Cascadilla Proceedings Project. Fernndez, M. (1999). Patterns in gender agreement in the speech of second language learners. In J. Gutirrez Rexach & F. Martnez Gil (Eds.), Advances in Hispanic linguistics (pp. 3 15). Sommerville, MA: Cascadilla Press. Fishman, J.A. (1964). Language maintenance and language shift as a field of inquiry: A de finition of the field and suggestions for its further development. Linguistics, 2 (9), 32 70. Fishman, J.A. (2001). 300 Plus years of heritage language education in the United States. In J.K. Peyton, D.A. Ranard, & S. McGinnis (Eds.), Heritage languages in America: Preserving a national resource (pp.81 97). Washington, DC: Delta Systems; and McHenry, IL: Center for Applied Linguistics. Fishman, J. (2006). Acquisition, maintenance and recovery of heritage languages. In G. Valds, J. Fishman, R. Chvez, & W. Prez (Eds.), Developing minority language resources: The case of Spanish in California (pp. 12 22). Clevedon, UK: Multilingual Matters. Flege, J.E. (1981). The phonological basis of foreign accent. TESOL Quarterly, 15 443 455. Flege, J.E. (1984). The detection of French accent by American listeners. Journal of the Acoustical Society of America, 76 (3), 692 707. Flege, J.E. (1987). The production of "new" and "similar" phones in a foreign language: Evidence for the effect of equivalence classifica tion. Journal of Phonetics, 15 47 65. Flege, J. E. (1988a). Factors affecting degree of perceived foreign accent in English sentences. Journal of the Acoustical Society of America, 84 70 79.
208 Flege, J. E. (1988b). The production and perception of foreig n language speech sounds. In: H. Winitz (Ed.), Human Communication and its Disorders: A Review. Volume I (pp 224 401). Norwood, NJ: Ablex Publishers. Flege, J.E. & Davidian, R.D. (1984). Transfer and developmental processes in adult foreign language sp eech production. Applied Psycholinguistics, 5 323 347. Flege, J.E. & Eefting, W. (1987). Production and perception of English stops by native Spanish speakers. Journal of Phonetics, 15 67 83. Flege, J. E. & Fletcher, K. L. (1992) Talker and listene r effects on degree of perceived foreign accent, Journal of the Acoustical Society of America, 91 370 389. Flege, J. E., Munro, M. J., & MacKay, I. R. A. (1995). Factors affecting strength of perceived foreign accent in a second language. Journal of th e Acoustical Society of America, 97 (5) 3125 3134. Fox, R.A., Flege, J.E., & Munro, M.J. (1995). The perception of English and Spanish vowels by native English and Spanish listeners: A multidimensional scaling analysis. Journal of the Acoustical Society of America, 97 (4) 2540 2551. Franceschina, F. (2003). The nature of grammatical representations in mature L2 grammars: The case of Spanish grammatical gender. (Doctoral dissertation). Retrieved from ProQuest dissertation and theses (85581748; 200402760 ). Franceschina, F. (2005). Fossilized second language grammars: The acquisition of grammatical gender Amsterdam: John Benjamins. Gabriele, A., Fiorentino, R., & Alemn Ba n, J. (2013). Examining second language development using event related potenti als: A cross sectional study on the processing of gender and number agreement. Linguistic Approaches to Bilingualism, 3 (2), 213 232. Galindo, D.L. (1995). Language attitudes toward Spanish and English varieties: A Chicano perspective. Hispanic Journal of Behavioral Sciences, 17 (1), 77 9 9. Garca, O. & Otheguy, R. (1988). The language situation of Cuban Americans. In S. McKay & S. Wong (Eds.), Language diversity: Problem or resource? (pp. 166 192). New York, NY: Harper & Row. Gatbonton, E., Trofimovich and L2 pronunciation accuracy: A sociolinguistic investigation. TESOL Quarterly, 39 (3), 489 511.
209 Godson, L. (2004). Vowel production in the speech of Western Armenian heritage speakers. Heri tage Language Journal, 2 1 26. Godson, Linda. ( 2003 ) Phonetics of language attrition: vowel production and articulatory setting in the speech of Western Armenian heritage speakers (Doctora l dissertation). Retrieved from ProQuest dissertation and thes es (85593258; 200410512). Gonzales Berry. (2000). Which language will our children speak? The Spanish language and public education policy in New Mexico, 1890 1930. In E. Gonzales Berry & D.R. Maciel (Eds.), The contested homeland: A Chicano history of New Mexico (pp. 167 189). Albuquerque: University of New Mexico Press. Grosjean, F. (1989). Neurolinguists, beware! The bilingual is not two monolinguals in one person. Brain and Language, 36 3 15. Guijarro Fuentes, P., & Marinis, T. (2011). Voicing l anguage dominance: The acquisition of interfaces by English/Spanish bilingual adolescents. In K. Potowski and J. Rothman (Eds.) Bilingual youth: Spanish in English speaking societies (pp. 227 248). Amsterdam: John Benjamins. Hammerly, H. (1991). Fluency and accuracy: Toward balance in language teaching and learning Clevedon, Avon, England: Multilingual Matters. Hancin Bhatt, B. & Bhatt, R.M. (1997). Optimal L2 syllables: Interactions of transfer and developmental effects. Studies in Second Language A cquisition, 19 331 378. Hannum, T. (1978). Attitudes of bilingual students toward Spanish. Hispania, 61 (1), 90 94. Hawkins, R., & Franceschina, F. (2004). Explaining the acquisition and non acquisition of determiner noun gender concord in French and Sp anish. In P. Prvost & J. Paradis (Eds.), The acquisition of French in different contexts (pp. 175 205). Amsterdam: John Benjamins. Hernndez Chavez, E. (1994). Language policy in the United States: A history of cultural genocide. In T. Skutnabb Kangas, R. Phillipson & M. Rannut (Eds.), Linguistic human rights: Overcoming linguistic discrimination (pp. 141 158). New York, NY: Moulton de Gruyter. Hernndez Chvez, E., Cohen, A.D. & Beltramo, A.F. (1975). El lenguaje de los Chicanos: Regional and social characteristics used by Mexican Americans Arlington, VA: Center for Applied Linguistics.
210 Hernndez, P. F. (1984). Teoras psicosociolingsticas y su aplicacin a la adquisicin del espaol como lengua maternal Madrid: Siglo XXI de Espaa Editores. Hieke, A. E. (1984). Linking as a marker of fluent speech. Language and Speech, 27 343 354. Hieke, A. E. (1985). A componential approach to oral fluency evaluation. Modern Language Journal, 69 ,135 142. Housen, A., & Kuiken, F. (2009). Complexity, acc uracy, and fluency in second language acquisition. Applied Linguistics, 30 (4), 461 473. Hudson Edwards, A. & Bills, G.D. (1982). Intergenerational language shift in an Albuquerque barrio. In J. Amastae & L. Elas Olivares (Eds.), Spanish in the Unite d States: Sociolinguistic aspects (pp. 135 153). Cambridge: Cambridge University Press. James, C. (1998). Errors in language learning and use: Exploring error analysis London: Longman. Keating, G. D. (2009). Sensitivity to violations of gender agreemen t in native and nonnative Spanish: An eye movement investigation. Language Learning, 59 (3), 503 535. Kingston, J. (2002). Keeping and losing contrasts. Proceedings of the 28th Annual Meeting of the Berkeley Linguistics Society, 155 176. Knightly, L. M ., Jun, S. A., Oh, J. S., & Au, T. K. (2003). Production benefits of childhood overhearing. Journal of the Acoustical Society of America, 114 465 474. Kormos, J., & Denes, M. (2004). Exploring measures and perceptions of fluency in the speech of second language learners. System, 32 (2), 145 164. Labov, W. (1966). The Social Stratification of English in New York City. Washington, DC: Center for Applied Linguistics. Labov, W. (1972). Sociolinguistic Patterns. Philadelphia, PA: University of Pennsylvania Press. Larsen Freeman, D. (2006). The emergence of complexity, fluency, and accuracy in the oral and written production of five Chinese learners of English. Applied Linguistics, 27 (4), 590 619.
211 Lee, D. J. (2001). Recent trends in foreign language tea ching in the United States: The role of heritage learners. In J. Ree (Ed.), The Korean language in America: Volume 6. Papers from the Annual Conference and Teacher Training Workshop on the Teaching of Korean Language, Culture, and Literature. (pp.203 211) Tallahassee, FL: American Association of Teachers of Korean. Leikin, M., Ibrahim, R., Eviatar, Z., & Sapir, S. (2009). Listening with an accent: Speech perception in a second language by late bilinguals. Journal of Psycholinguistic Research, 38 447 457. Levi, S.V., Winters, S.J., Pisoni, D.B. (2007). Speaker independent factors affecting the perception of a foreign accent in a second language. Journal of the Acoustical Society of America, 121 (4), 2327 2338. Lippi Green, R. (1997). English with an accent: Language, ideology, and discrimination in the United States New York, NY: Routledge. Lipski, J. (1993). Creoloid phenomena in the Spanish of transitional bilinguals. In A. Roca & J. Lipski (Eds.), Spanish in the United States (pp. 155 173). Ber lin: Mouton. Lisker, L. & Abramson, A. (1964). A cross language study of voicing in initial stops: Acoustical measurements Word, 20 384 422. Llurda, E. (2000). Effects of intelligibility and speaking rate on judgments of non native lities. International Review of Applied Linguistics in Language Teaching, 38 (3 4), 289 299. Long, M. (1990). Maturational constraints on language development. Studies in Second Language Acquisition, 12 251 285. Lord, G. (2005). (How) Can we teach forei gn language pronunciation? The effects of a phonetics class on second language pronunciation. Hispania 88 (3), 557 567. Lynch, A. (2012). Key concepts for theorizing Spanish as a heritage language. In S.M. Beaudrie & M. Fairclough (Eds.), Spanish as a he ritage language in the United States: The state of the field (pp. 79 97). Washington, DC: Georgetown University Press. Maciel, D.R., & Pea, J.J. (2000). La reconquista: The Chicano movement in New Mexico. In E. Gonzales Berry & D.R. Maciel, (Eds.), The contested homeland: A Chicano history of New Mexico (pp. 269 301). Albuquerque: University of New Mexico Press.
212 Mackay, I.R.A., Flege, J.E., & Imai, S. (2006). Evaluating the effects of chronological age and sentence duration on degree of perceived for eign accent. Applied Psycholinguistics, 27 (2), 157 183. Magan, H. (1998). The perception of foreign accented speech. Journal of Phonetics, 26 381 400. Major, R.C. (1986). Paragoge and degree of foreign accent in Brazilian English. Second Language Rese arch, 2 53 71. Major, R.C. (2007). Identifying a foreign accent in an unfamiliar language. Studies in Second Language Acquisition, 29, 539 556. Maloof, V., Rubin, D., & Miller, A.N. (2006). Cultural competence and identity in cross cultural adaptation : The role of a Vietnamese heritage language school. International Journal of Bilingual Education and Bilingualism, 9 (2), 255 273. Martinez, G. A. (1998). Vowel harmony and the process of sound change in early Mexican American Spanish. The Bilingual Rev iew/La Revista Bilingue, 23 (1), 3 10. Martnez, G.A. (2009). Hacia una sociolingstica de la esperanza: El mantenimiento inter generacional del espaol y el desarrollo de comunidades hispanohablantes en el sudoeste de los Estados Unidos. Spanish in Cont ext, 6 (1), 127 137. McCarthy, C. (2008). Morphological variability in the comprehension of agreement: An argument for representation over computation. Second Language Research, 24 (4), 459 486. Mikulski, A.M. (2010). Receptive volitional subjunctive abil ities in heritage and traditional foreign language learners of Spanish. Modern Language Journal, 94 (2), 217 233. Montes Acal, C. (2000). Attitudes towards oral and written codeswitching in Spanish English bilingual youths. In A. Roca (Ed.), Research on Spanish in the U.S. (pp 218 227). Somerville, MA: Cascadilla Press. Montrul, S. (2002). Incomplete acquisition and attrition of Spanish tense/aspect distinctions in adult bilinguals. Bilingualism: Language and Cognition, 5 39 68. Montrul, S. (2004 a) Psycholinguistic evidence for split intransitivity in Spanish L2. Applied Psycholinguistics, 25 (2), 239 267. Montrul, S. (2004b). Subject and object expression in Spanish heritage speakers: A case of morpho syntactic convergence. Bilingualism: Languag e and Cognition, 7 1 18.
213 Montrul, S. (2004c). The acquisition of Spanish: Morphosyntactic development in monolingual and bilingual L1 acquisition and adult L2 acquisition Amsterdam: John Benjamins. Montrul, S. (2006 a ). Incomplete acquisition as a featu re of L2 and bilingual grammars. In R. Slabakova, S.A. Montrul, & P. Prvost (Eds.), Inquiries in linguistic development: In honor of Lydia White (pp. 335 360). Amsterdam: John Benjamins. Montrul, S. (2006b). On the bilingual competence of Spanish herita ge speakers: Syntax, lexical semantics and processing. International Journal of Bilingualism, 10 37 69. Montrul, S. (2010). Current issues in heritage language acquisition. Annual Review of Applied Linguistics, 30 3 23. Montrul, S. (2012) The gramm atical competence of Spanish heritage speakers. In S.M. Beaudrie & M. Fairclough (Eds.), Spanish as a heritage language in the United States: The state of the field Washington, DC: Georgetown University Press. Montrul, S., & Bowles, M. (2010). Is gramma r instruction beneficial for heritage language learners? Dative case marking in Spanish. The Heritage Language Journal, 7 (1), 47 63. Montrul, S., Foote, R., & Perpi an S. (2008a). Gender agreement in adult second language learners and Spanish heritage speakers: The effects of age and context of acquisition. Language Learning, 58 (3), 503 553. Montrul, S., Foote, R., & Perpin, S. (2008b). Knowledge of Wh movement in Spanish L2 learners and heritage speakers. In J. B. de Garavito & E. Valenzuela (Eds. ), Proceedings of the 10th Hispanics Linguistics Symposium Somerville, MA: Cascadilla Proceedings Project. Montrul, S., & Potowski, K. (2007). Command of gender agreement in school age Spanish English bilingual children. International Journal of Bilingu alism, 11 (3), 301 328. Montrul, S., & Slabakova, R. (2003). Competence similarities between native and near native speakers: An investigation of the preterite/imperfect contrast in Spanish. Studies in Second Language Acquisition, 25 (3), 351 398. Munro, M. J. (1995). Nonsegmental factors in foreign accent. Studies in Second Language Acquisition, 17 17 34.
214 Munro, M.J., 2003. A primer on accent discrimination in the Canadian context. TESL Canada Journal, 20 (2), 38 51. Munro, M. J., & Derwing, T. M. ( 1998). The effects of speaking rate on listener evaluations of native and foreign accented speech. Language Learning, 48 159 182. Munro, M. J. & Derwing, T. M. (1995) Foreign accent, comprehensibility and intelligibility in the speech of second languag e learners. Language Learning, 45 73 97. Munro, M. J., Derwing, T. M., & Morton, S. L. (2006). The mutual intelligibility of foreign accents. Studies in Second Language Acquisition, 28 111 131. Munro, M. J., Derwing, T. M., & Burgess, C. S. (2010). De tection of nonnative speaker status from content masked speech. Speech Communication, 52 (7 8), 626 637. Munro, M. J., & Derwing, T. M. (2001). Modeling perceptions of the accentedness and comprehensibility of L2 speech: The role of speaking rate. Studi es in Second Language Acquisition, 23 (4), 451 468. Myers Scotton, C. (2006). Multiple voices: An introduction to bilingualism Malden, MA: Blackwell Publishing. Neufeld, G.G. (1980). On the adult's ability to acquire phonology. TESOL Quarterly, 14 (3), 285 298. Newport, E. L. (1990). Maturational constraints on language learning. Cognitive Science, 14 11 28. Norris, J. M., & Ortega, L. (2009). Towards an organic approach to investigating CAF in instructed SLA: The case of complexity. Applied Linguis tics, 30 (4), 555 578. Norris, J.M. & Ortega, L. (2003). Defining and measuring SLA. In C. Doughty & M. Long (Eds.), The handbook of second language acquisition Malden, MA: Blackwell Publishing. IBM SPSS Statistics 19 statistica l procedures companion Upper Saddle River, NJ: Prentice Hall. Oh, J., Jun, S., Knightly, L., & Au, T. (2003). Holding on to childhood language memory. Cognition, 86 53 64. Olson, L. L., & Samuels, S. J. (1973) The relationship between age and accura cy of foreign language pronunciation. Journal of Educational Research, 66 263 268.
215 Ortiz, L.I. (1975). A sociolinguistic study of language maintenance in the northern New Mexico community of Arroyo Seco (Unpublished doctoral dissertation). University of New Mexico, Albuquerque. Otheguy, R., & Stern, N. (2011). On so called Spanglish. International Journal of Bilingualism, 15 (1), 85 100. Park, H. (2009). Phonological information and linguistic experience in foreign accent detection Dissertation Ab stracts International, A: The Humanities and Social Sciences, 4707 4707. Pease Alvarz, L. (1993). Moving in and out of bilingualism: Investigating native language maintenance and shift in Mexican descent children. National Center for Research on Cultur al Diversity and Second Language Learning, Research Report, 6 Washington, DC: Center for Applied Linguistics. Retrieved from http://www.ncela.gwu.edu/files/rcd/BE019083/RR6_Moving_In_and_Out.pdf Pease Alvarz, L. (2002). Moving beyond linear trajectories of language shift and bilingual language socialization. Hispanic Journal of Behavioral Sciences, 24 (2), 114 137. Pealosa, F. (1980). Chicano sociolinguistics: A brief introduction Rowley, MA: Newberry House Publishers, Inc. Prez Pereira, M. (1991). The acquisition of gender: What Spanish children tell us. Journal of Child Language, 18 571 590. Phinney, J., Romero, I., Nava, M., & Huang, D. (2001). The role of language, parents, and peers in ethnic identity among adolescents in immigrant families. Journal of Youth and Adolescence, 30 (2), 135 53. Piske, T., MacKay, I., & Flege, J. (2001). Factors affecting degree of foreign accent in an L2: A review. Journal of Phonetics, 29 191 215. Piske, T., Flege, J.E., MacKay, I.R., & Meador, D. (2002). Th e Production of English vowels by fluent early and late Italian English bilinguals. Phonetica, 59 49 71. Polinsky, M. (1997). American Russian: Language loss meets language acquisition. In W. Brown, E. Dornisch, N. Kondrashova, & D. Zec (Eds.), Formal approaches to Slavic linguistics (pp 370 407). Ann Arbor, MI: Michigan Slavic Publications. Polinsky, M. (2000). The composite linguistic profile of speakers of Russian in the US. In O. Kagan & B. Rifkin (Eds.), The learning and teaching of Slavic langua ges and cultures (pp. 437 65). Bloomington, IN: Slavica.
216 Polinsky, M. (2005). Word class distinctions in an incomplete grammar. In D.D. Ravid, H. B Z Shildrkrodt, & R.A. Berman (Eds.), Perspectives on language and language development (pp. 419 436). Dor drecht, The Netherlands: Kluwer Academic. Polinsky, M. (2006). Incomplete acquisition: American Russian. Journal of Slavic Linguistics, 14 161 219. Polinsky, M. (2007). Russian gender under incomplete acquisition. The Heritage Language Journal, 6 (1). Retrieved from http://www.heritagelanguages.org/ classroom. Language and Linguistics Compass, 1 (5), 368 395. Polio, C. (1997). Measures of linguistic accuracy in second lan guage writing research. Language Learning, 47 101 143. Potowski, K. (2004). Spanish language shift in Chicago. So uthwest Journal of Linguistics, 23 (1), 87 116. Potowski, K. (2005). Fundamentos de la enseanza del espaol a los hablantes nativos en lo s Estados Unidos Madrid, Spain: Arco/Libros. Potowski, K. (2012). Identity and heritage learners: Moving beyond essentializations. In S.M. Beaudrie & M. Fairclough (Eds.), Spanish as a heritage language in the United States: The state of the field (pp. 179 199). Washington, DC: Georgetown University Press. Potowski, K., Jegerski, J., & Morgan Short, K. (2009). The effects of instruction on linguistic development in Spanish heritage language speakers. Language Learning, 59 537 579. Potowski, K., & Ma tts, J. (2008). Interethnic language and identity: MexiRicans in Chicago. Journal of Language, Identity, and Education, 7 (2), 137 160. Prada Prez, A. de & Pascual y Cabo, D. (2011). Invariable gusta in the Spanish of Heritage Speakers in the US. In J. Herschensohn & D. Tanner (Eds.), Proceedings of the 11th Generative Approaches to Second Language Acquisition Conference (GASLA) (pp. 110 120). Somerville, MA: Cascadilla Proceedings Project. Prada Perez, A. de. (2009). Subject expression in Minorcan Spa nish: Consequences of contact with Catalan. (Doctoral dissertation). Retrieved from ProQuest dissertation and theses. (304981820)
217 Purcell, E. T., & Suter, R. W. (1980). Predictors of pronunciation accuracy: A reexamination. Language Learning, 30 271 2 87. Ramrez Verdugo, D. (2002). Non native interlanguage intonation systems: A study based on a computerized corpus of Spanish learners of English. International Computer Archive of Modern and Medieval English Journal, 26 115 132. Raupach, M. (1980). Temporal variables in first and second language speech production. In H. W. Dechert & M. Raupach (Eds.), Temporal variables in speech: Studies in honour of Frieda Goldman Eisler (pp. 263 270). The Hague: Mouton. Riazantseva, A. (2001). Second language p roficiency and pausing: A study of Russian speakers of English. Studies in Second Language Acquisition, 23 497 526. Riggenbach, H. (1991). Towards an understanding of fluency: A microanalysis of nonnative speaker conversation. Discourse Processes, 14 423 441. Riney, T.J., Takagi, N., & Inutsuka, K. (2005). Phonetic parameters and perceptual judgments of accent in English by American and Japanese listeners. TESOL Quarterly, 39 (3), 441 466. Rivera Mills, S. (2001). Acculturation and communicative nee d: Language shift in an ethnically diverse Hispanic community. Southwest Journal of Linguistics, 20 211 223. Roca, A., & Colombi, M.C. (Eds.). (2003). Mi lengua: Spanish as a heritage language in the United States Washington, DC: Georgetown University Press. Rosado, L.A. (2005). The language of Cervantes: Alive and well in Texas Implications for bilingual education programs. Hispania, 88 (4), 834 847. Rossiter, M.J. (2009). Perceptions of L2 fluency by native and non native speakers of English. The Canadian Modern Language Review, 65 (3), 395 412. Rothman, J. & Iverson, M. (2008). Poverty of the stimulus and L2 epistemology: Considering L2 knowledge of aspectual phrasal semantics. Language Acquisition: A Journal of Developmental Linguistics, 15 (4) 270 314. Ryan, E.B., & Carranza, M.A. (1977). Ingroup and outgroup reactions to Mexican American language varieties. In H. Giles (Ed.), Language, ethnicity and intergroup relations (pp. 59 82). London: Academic Press. Ryan, E.B., Carranza, M.A., & Mo ffie, R.W. (1977). Reactions toward varying degrees of accentedness in the speech of Spanish English bilinguals. Language and Speech, 20 267 273.
218 Sakuragi, T. (2011). The construct validity of the measures of complexity, accuracy, and fluency: Analyzin g the speaking performance of learners of Japanese. JALT Journal, 33 (2), 157 173. Snchez, R. (1983). Chicano discourse: Socio historic perspectives Houston, TX: Arte Pblico Press. Sanz, I., & Villa, D.J. (2011). The genesis of traditional New Mexica n Spanish: The emergence of a unique dialect in the Americas. Studies in Hispanic & Lusophone Linguistics, 4 (2), 417 442. Scales, J., Wennerstrom, A., Richard, D., & Wu, S.H. (2006). Language Learners' Perceptions of Accent. TESOL Quarterly, 40 (4), 715 738. Scheaffer, R.L. (1999). Categorical Data Analysis. NCSSM Statistics Leadership Institute. Retrieved from http://courses.ncssm.edu/math/Stat_Inst/PDFS/ Categorical%20Data%20Analysis.pdf Schmid, C. (2001). The politics of language: Conflict, identit y and cultural pluralism in comparative perspective New York, NY: Oxford University Press. authenticating discourse. Journal of Sociolinguistics, 11 194 220. Sheppard, C. (2004). The measurement of second language production: The validity of fluency, accuracy and complexity. ICU Language Research Bulletin, 19 139 156. Simes, A.R.M. (1996). Phonetics in second language acquisition: An acoustic study of fluency in adu lt learners of Spanish. Hispania, 79 (1), 87 95. Skehan, P. (1989). Individual differences in second language learning London : Edward Arnold. Skehan, P. (1998). A cognitive approach to language learning Oxford, NY: Oxford University Press. Skehan, P. (2009). Modeling second language performance: Integrating complexity, accuracy, fluency, and lexis. Applied Linguistics, 30 (4), 510 532. Skehan, P., & Foster, P. (1999). The influence of task structure and processing conditions on narrative retellings. Language Learning, 49 93 120. Slabakova, R., Rothman, J., & Kempchinsky, P. (2011). Gradient competence at the syntax discourse interface. EUROSLA Yearbook, 11 218 243.
219 Smit, U. (1996). Who speaks like that? Accent recognition and language attitudes South African Journal of Linguistics, 14 (3), 100 108. Snow, C. E., & Hoefnagel Hhle, M. (1977). Age differences in the pronunciation of foreign sounds. Language & Speech, 20 357 365. Snow, C., & Hoefnagel Hohle, M. (1978). The critical period for l anguage acquisition: Evidence from second language learning. Child Development, 49 1114 1128. Sol, Y.R. (1990). Bilingualism: Stable or transitional? The case of Spanish in the United States. International Journal of the Sociology of Language, 84 35 80. Southwood, M.H., & Flege, J.E. (1999). Scaling foreign accent: Direct magnitude estimation versus interval scaling. Clinical Linguistics & Phonetics, 13 (5), 335 349. Stavans, I. (2000). The gravitas of Spanglish. Chronicle of Higher Education, 13 B7. Suter, R. W. (1976). Predictors of pronunciation accuracy in second language learning. Language Learning, 26 233 253. Tajima, K., Port, R., & Dalby, J. (1997). Effects of temporal correction on intelligibility of foreign accented English. Journal of Phonetics, 25 1 24. Torres Cacoullos, R., & Aaron, J. E. (2003). Determiner variation with English origin nouns in New Mexican Spanish: Borrowing bare forms. University of Pennsylvania Working Papers in Linguistics, 9 (2), 159 172. Towell, R. (2007) Complexity, accuracy and fluency in second language acquisition research. In S. Van Daele, A. Housen, M. Pierrard, F. Kuiken & I. Vedder (Eds.), Complexity, accuracy and fluency in second language use, learning and teaching (pp.260 273). Brussels, Belgi um: Koninklijke Vlaamse Academie van Belgie voor Wetenschappen en Kunsten Towell, R., Hawkins, R., & Bazergui, N. (1996). The development of fluency in advanced learners of French. Applied Linguistics, 17 84 119. Trujillo, J.A. (1997). Archaism and in novation: A diachronic perspective on New Mexico Spanish, 1684 1893. (Doctoral dissertation). Retrieved from ProQuest dissertation and theses. (304373686). Trujillo, V.J. (2011). Evaluating listener ability in foreign accent ratings tasks. [Presentatio n]. Paper presented at the University of Florida Graduate Student Council Annual Interdisciplinary Research Conference Gainesville, FL.
220 Ueyama, M. (2000). Prosodic transfer: An acoustic study of L2 English vs. L2 Japanese (Doctoral dissertation). Retr ieved from ProQuest dissertation and theses. (304582382). Valds, G. (1997). The teaching of Spanish to bilingual Spanish speaking students: Outstanding issues and unanswered questions. In M.C. Colombi & F.X. Alarcn (Eds.), La enseanza del espaol a hi spanohablantes (pp. 8 44). Boston, MA: Houghton Mifflin Company. Valds, G. (2000). Introduction. In G. Valds (Ed.), Spanish for Native Speakers, Volume 1. AATSP Professional Development Series Handbook for Teachers K 16 New York, NY: Harcourt College Publishers. Valds, G. (2001). Heritage language students: Profiles and possibilities. In J.K. Peyton, D.A. Ranard, & S. McGinnis. (Eds.), Heritage languages in America: Preserving a national resource (pp.37 77). Washington, DC: Delta Systems; and McHen ry, IL: Center for Applied Linguistics. Valds, G., Fishman, J., Chvez, R., & Prez, W. (2006). Developing minority language resources: The case of Spanish in California Clevedon, UK: Multilingual Matters. Van Daele, S., Housen, A., Kuiken, F., Pierra rd, M., & Vedder, I. (Eds). (2007). Complexity, Accuracy and Fluency in Second Language Use, Learning and Teaching. Brussels, Belgium: Koninklijke Vlaamse Academie van Belgie voor Wetenschappen en Kunsten Vanderplank, R. (1993). Pacing and spacing as pr edictors of difficulty in speaking and understanding English English Language Teaching Journal, 47 117 125. Varonis, E.M., & Gass, S., (1982). The comprehensibility of non native speech. Studies in Second Language Acquisition, 4 114 136. Villa, D.J. Spanish speakers in the U.S. Foreign Language Annals, 29 (2), 191 200. Villa, D. J. (2002). The sanitizing of U.S. Spanish in academia. Foreign Language Annals, 35 (2), 222 230 Villa, D.J. (2004). No nos dejaremos: Writing in Spanish as an act of resistance. In M. Hall Kells, V. Balester, & V. Villanueva (Eds.), Latino/a Discourses on Language, Identity & Literacy Education (pp. 85 95). Portsmouth, NH: Heinemann. Villa, D.J. & Rivera Mills, S.V. (2009a). An integrated multi generational model for language maintenance and shift: The case of Spanish in the Southwest. Spanish in Context, 6 (1), 26 42.
221 Villa, D.J. & Rivera Mills, S.V. (2009b). Spanish maintenance and loss in t he U.S. southwest: History in the making. Spanish in Context, 6 (1), 1 5. Wah Lai, E.Y. (2008). Prosody and prosodic transfer in foreign language acquisition: Cantonese and Japanese Mnchen: LINCOM Europa. Webb, J.B., & Miller, B.L. (Eds.). (2000) Teac hing heritage language learners: Voices from the Classroom Yonkers, NY: American Council on the Teaching of Foreign Languages. H. Riggenbach (Ed.), (pp. 102 127). Ann Arbor, MI: University of Michigan Press. White, L., Valenzuela, E., Kozlowska MacGregor, M., & Leung, Y. K. (200 4). Gender and number agreement in nonnative Spanish. Applied Psycholinguistics, 25 105 133. Wiley, T.G. (2001). On defining heritage language and their speakers. In J.K. Peyton, D.A. Ranard, & S. McGinnis (Eds.), Heritage languages in America: Preserv ing a national resource (pp. 29 36). Washington, DC: Delta Systems; and McHenry, IL: Center for Applied Linguistics. Willis, E. (2005). An initial examination of southwest Spanish vowels. Southwest Journal of Linguistics,24 (1 2), 185 198. Wilson, D.V. (n.d.). The Sabine Ulibarr Spanish as a heritage language program. Retrieved from http://spanport.unm.edu/undergraduate/language programs/spanish heritage program.html Yeni Komshian, G.H., Flege, J.E., & Liu, S. (2000). Pronunciation proficiency in the first and second languages of Korean English bilinguals. Bilingualism: Language and Cognition, 3 131 149. Zampini, M. (1994). The role of native language transfer and task formality in the acquisition of Spanish spirantization. Hispania, 77 (3), 470 48 1. Zapata, G. C., Sanchez, L., & Toribio, A. J. (2005). Contact and contracting Spanish. International Journal of Bilingualism, 9 (3 4), 377 395. Zentella, A.C. (1997). Growing Up Bilingual Malden, MA: Blackwell. Zuengler,J. (1988). Identity markers an d L2 pronunciation. Studies in Second Language Acquisition, 10 33 49.
222 BIOGRAPHICAL SKETCH Valerie J. Trujillo is a native New Mexican and heritage speaker of Spanish. She j ournalism and Spanish from the University of N ew s outhwest s tudies from the University of New Mexico She received her Ph.D. in Spanish with a concentration in Hispanic linguistics from the University of Florida in the summer of 2013. Her research focuses on heritage language maintenance and acquisition, second language acquisition, and heritage speaker phonology and syntax.