Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Articulatory settings of French and English monolingual and bilingual speakers Wilson, Ian Lewis 2006

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2006-199183.pdf [ 11.33MB ]
Metadata
JSON: 831-1.0092890.json
JSON-LD: 831-1.0092890-ld.json
RDF/XML (Pretty): 831-1.0092890-rdf.xml
RDF/JSON: 831-1.0092890-rdf.json
Turtle: 831-1.0092890-turtle.txt
N-Triples: 831-1.0092890-rdf-ntriples.txt
Original Record: 831-1.0092890-source.json
Full Text
831-1.0092890-fulltext.txt
Citation
831-1.0092890.ris

Full Text

A R T I C U L A T O R Y SETTINGS OF F R E N C H A N D E N G L I S H M O N O L I N G U A L A N D B I L I N G U A L SPEAKERS  by I A N LEWIS WILSON B.Math., University of Waterloo, 1988 M . A . (TESL/TEFL), University of Birmingham, 1998  A THESIS SUBMITTED IN P A R T I A L F U L F I L L M E N T OF THE REQUIREMENTS FOR T H E D E G R E E OF DOCTOR OF PHILOSOPHY  in  THE F A C U L T Y OF G R A D U A T E STUDIES  (Linguistics)  THE UNIVERSITY OF BRITISH C O L U M B I A April 2006  © Ian Lewis Wilson, 2006  Abstract This dissertation investigates articulatory setting (AS), a language's underlying or default posture of the articulators (i.e., the tongue, jaw, and lips). Inter-speech posture (ISP) of the articulators (the position of the articulators when they are motionless during inter-utterance pauses) is used as a measure of AS in Canadian English and Quebecois French. The dissertation reports two experiments using a combination of Optotrak and ultrasound imaging to test whether ISP is language specific in both monolingual and bilingual speakers, whether it is affected by phonetic context, and whether it is influenced by speech mode (monolingual or bilingual). Results of Experiment 1 show significant differences in ISP across the English and French monolingual groups, with English exhibiting a higher tongue tip, more protruded upper and lower lips, and narrower horizontal lip aperture. Results also show that for English speakers, the jaw ISP is somewhat influenced by phonetic context while the lip and tongue ISP are not. For French speakers, only certain lip components of ISP are influenced by phonetic context while the ISP of the tongue and jaw are not. Results of Experiment 2 show that upper and lower lip protrusion are greater for the English ISP than for the French ISP, in all bilinguals who were perceived as native speakers of both of their languages, but in none of the other bilinguals. Also, tongue tip height results mirrored those of the monolingual groups, for half of the bilinguals perceived as native speakers of both languages, but for no other bilinguals. Finally, results show that there is no unique bilingual-mode ISP, but instead one that is equivalent to the monolingual-mode ISP of a speaker's currently most-used language. This research empirically confirms centuries of non-instrumental evidence for the existence of A S , and thus supports calls for the teaching of AS to L2 learners. Additionally, the lack of phonetic carry-over effect on ISP is encouraging for studies that have used ISP as a measurement baseline. Finally, the fact that there is no unique ISP for bilingual speech mode suggests that differences between monolingual and bilingual modes do not hold at the phonetic level. I  ii  Table of Contents Abstract  ii  Table of Contents  iii  List of Tables  v  List of Figures  vi  Acknowledgements  vii  C H A P T E R I Introduction 1.1. Outline of the Dissertation 1.2. Articulatory Setting (AS) 1.2.1. Non-instrumental Views of AS 1.3. Measuring A S 1.3.1. A S and the Rest Position of the Tongue 1.4. Articulatory Setting in Bilingual Speakers 1.5. Purpose of this Research and Summary of Hypotheses  1 2 3 5 8 10 15 18  C H A P T E R II Method 2.1. Subjects 2.1.1. Criteria for Classifying Subjects as Monolingual or Bilingual 2.1.2. Monolingual subjects 2.1.3. Bilingual subjects 2.2. Apparatus 2.3. Procedure ; 2.3.1. Data Collection 2.3.1.1. Optotrak Setup 2.3.1.2. Ultrasound Setup 2.3.1.3. Preliminary Trials 2.3.1.4. Main Trials 2.3.2. Data Analysis  20 20 20 22 24 26 29 30 30 31 34 36 38  C H A P T E R III Experiment 1: A S in English and French Monolinguals 3.1. Results 3.1.1. Results: ISP Across Languages 3.1.2. Results: ISP Across Individuals Within a Language 3.1.3. Results: Carry-Over Effects of Phonetic Context on ISP 3.2. Discussion 3.2.1. Discussion Regarding Test of Hypothesis 1 3.2.2. Comparison of Results to Long-held Impressions of A S 3.2.3. Discussion Regarding Test of Hypothesis 2 3.3. Summary of Chapter III  57 57 57 60 62 69 69 73 75 79  iii  C H A P T E R IV Experiment 2: A S in English-French Bilinguals 4.1. Results 4.1.1. Results: English Versus French (Monolingual Mode) 4.1.2. Results: Bilingual Mode Versus Monolingual Mode 4.2. Discussion 4.2.1. Discussion Regarding Test of Hypothesis 3 4.2.2. Discussion Regarding Test of Hypothesis 4 4.3. Summary of Chapter IV  81 81 81 87 90 91 92 94  C H A P T E R V General Discussion and Conclusions 5.1. General Discussion 5.1.1. Implications of this Research 5.2. Limitations and Future Directions 5.3. Conclusions  95 95 97 99 103  References  105  Appendix I English advertisement for subjects Appendix II French advertisement for subjects Appendix III Detailed background on subjects Monolingual English subjects Monolingual French subjects Bilingual English-French subjects Appendix IV Definition of ratings in foreign accent rating task Appendix V Detailed results of foreign accent rating task Appendix VI English stimuli used Appendix VII French stimuli used Appendix VIII Sentences used in foreign accent rating task Appendix IX Bilingual-mode stimuli used Appendix X Geometrical formulas used in the M A T L A B code Appendix X I Means and standard deviations for all 24 subjects Tongue and jaw Lip height and protrusion Lip aperture and narrowing Appendix XII Detailed statistics for Table 3.2 Appendix XIII Questionnaire filled out by French subjects  114 115 116 116 117 118 119 120 121 124 127 128 129 132 133 134 135 136 148  iv  List of Tables  Table 2.1.  Summary of perceived language abilities of all bilingual subjects  Table 2.2.  Total rest frames available and number actually used (monolingual subjects).. 40  Table 2.3.  Total rest frames available and number actually used (bilingual subjects)  41  Table 2.4.  Total possible available ISPs for each pre-ISP word  44  Table 2.5.  Definitions of the broad phonetic contexts used in the analysis  45  Table 2.6.  Definitions of components of ISP used in statistical analyses  47  Table 2.7.  Contents of numeric output file of main M A T L A B m-file used  48  Table 2.8.  Mean distance from subject's nose bridge to alveolar ridge  55  Table 3.1.  Means and between-subject standard deviations of monolingual English and French groups for each component of ISP 58  Table 3.2.  Comparison using t tests (assuming unequal variances) of monolingual English and French group means by component of ISP  59  Comparison of English group versus French group means by component of ISP and by phonetic context  64  Significant differences (p < .05) between the ISP in French and English monolingual modes  85  For each bilingual who was perceived as native in both languages, a comparison of bilingual-mode ISP to each of the monolingual-mode ISPs ("MonoEng" and "MonoFre") for 4 components of ISP  89  Table 3.3.  Table 4.1.  Table 4.2.  v  25  List of Figures Figure 2.1.  Data collection setting  26  Figure 2.2.  Subject's view of the stimuli from the experiment chair  28  Figure 2.3.  Placement of ultrasound probe, head restraint, and Optotrak markers  29  Figure 2.4.  CT scan of upper vocal tract with ultrasound tongue image overlaid at various opacities  33  Figure 2.5.  Plexiglas "bite triangle" used in the second preliminary trial  34  Figure 2.6.  Ultrasound frame in M A T L A B with alveolar ridge visible  50  Figure 2.7.  Ultrasound frame in M A T L A B of an ISP to be analyzed  51  Figure 2.8.  Ultrasound frame in M A T L A B showing four measurement lines  53  Figure 3.1.  Box plots of monolingual subjects' distribution for the four components of ISP that were significantly different across languages 61 Number of components of ISP per subject where / tests showed a significant difference in any direction between the BackV context and the CoronalC context 66  Figure 3.2.  Figure 3.3;  Figure 4.1.  Figure 4.2.  Figure 4.3.  Number of subjects that showed a significant difference in the same direction between the BackV context and the CoronalC context by component of ISP  67  Box plot of distribution of tongue tip height values for all 9 bilingual subjects in monolingual mode for both English and French  82  Box plot of distribution of amount of jaw lowering for all 9 bilingual subjects in monolingual mode for both English and French  83  Box plot of distribution of lower lip protrusion values for all 9 bilingual subjects in monolingual mode for both English and French  83  vi  Acknowledgements I finally saw the light at the end of the tunnel but then realized that it was really a huge torch being held by my committee members who had come in to get me. First, I would like to thank my research supervisor, Bryan Gick, for six years of phenomenal support and mentoring. Without him, there is no way I would be where I am today. He has given me chances to publish and further my research experience throughout grad school, and his N S E R C and CFI grants have funded much of my research. M y other committee members, Eric Vatikiotis-Bateson and Stefka Marinova-Todd have also been very helpful, especially while Bryan was away on sabbatical. Thanks Eric for help with Optotrak and the methods used, but also for letting me figure out lots on my own. Thanks Stefka for help with statistics, L2 acquisition, and for assisting me in my search for French subjects. This dissertation would never have been completed were it not for Jason Chang's brilliant assistance with data collection and analysis. During experiments, I would ask him to do something and it would always be the case that he had anticipated my request and finished doing it 30 seconds before I asked. Jason was also instrumental in helping to keep up the pace of the research. Thanks to Fiona Campbell for getting the ball rolling helping to construct English stimuli, running experiments, and figuring out Optotrak in the early stages. Ritu Kumar very efficiently assisted with data preparation and analysis, always with a friendly smile on her face. A big thank you to Doug Pulleyblank and AnneMarie Comte for help with the creation of the French stimuli and for helping to find French and bilingual subjects. Thanks too to Rose-Marie Dechaine who helped translate my advertisement for subjects into French, and to Joe Stemberger and Fei X u , who provided me with many useful things to think about at my defence. I also wish to thank Maciej Mizerski, U B C Math Department, for help with formulas for coordinate system transformations and vector algebra, and Jean-Francois Plante, U B C Stats Department, for statistical consulting. Thanks to Fernanda Almeida and Alan Hannam of U B C Dentistry for answers to palate questions and to Elaine Orpe, who was very kind in giving up time to take CT scans of my head and help with D I C O M reader software. Heartfelt thanks to Shaffiq Rahemtulla for technical support in the lab over the years and to Edna Dhamaratne, our overworked graduate secretary for making life much easier. And thanks even to Marion Caldecott for incessantly asking me if I was finished. I so wanted to say "yes" and finally did! I also appreciate the questions and comments from various audiences at U B C , Ultrafest, A S A Vancouver, and the University of Pennsylvania. Thanks to all the U B C Linguistics grad students and profs over the years who have helped me in many ways! A very big thank you goes out to all my subjects, including my native listener judges, for their time and cooperation. I also couldn't have done it without my Mom and Dad's moral and financial support. Thanks Dad for the help with vector equations. Finally, the biggest thank you goes to my wife, Hiromi, and our three kids. You guys kept me sane though all of this and gave me space when I needed it most. It's been really painful often putting the dissertation before family commitments, and I know it has been even more painful for you. It's finally over! vii  CHAPTER I Introduction  If asked why different languages sound different, a layperson might answer that different languages use different sounds (i.e. they have different phonemes). A linguist would add that different languages use sounds differently (i.e. they have different phonologies). However, there is another factor that also plays a part in the sound of a language. As far back as 350 years ago (Wallis, 1653 / 1972) people sensed that when speaking a foreign language, one's articulators - the tongue, jaw, lips, etc. - seemed to have a whole different underlying or default posture than the one used for one's native language. A language's underlying articulatory posture is one part of what Honikman (1964) called articulatory setting (henceforth AS), and although it is something that has interested phoneticians for centuries, until very recently its existence had never been instrumentally verified. For reasons to be discussed in Section 1.3, there has been no support, in the form of direct articulatory measurements, for the existence of A S until recent work by Gick, Wilson, Koch, and Cook (2004). These authors looked at inter-speech posture (the position of the articulators when they are motionless during inter-utterance pauses; henceforth ISP) to investigate AS. The reasons for this connection between AS and ISP are made apparent in Section 1.3.1. The results of Gick et al. (2004) showed that the ISP for Quebecois French is significantly different from that for Canadian English . Their 1  study, however, examined only five speakers of each language and its methodology was constrained by the fact that the data they analyzed was based on existing x-ray movie films with limited spatial resolution and clarity, and they had no control over the linguistic stimuli or how they were presented to the subjects. The first major purpose of my research has been to partially replicate the study of Gick et al. (2004) using a greater number of speakers of French and English, and using an entirely different methodology  In this dissertation (as well as in Gick et al. (2004)), "Canadian English" refers to a general Canadian accent as would be heard from many television newscasters on national television, in the same way as "General American English" refers to a general accent in the U.S.A. Admittedly, in different Canadian' provinces there are slightly different accents and the term "Canadian English" is not meant to imply that there exists only one kind of English in Canada. See Section 5.2.3 for more on this issue. 1  1  that has enabled more measurement precision and has allowed for control over the phonetic context of the ISPs that were analyzed. In the course of this replication study, not only have I tested whether or not ISP is language specific, but I have also tested whether it is sensitive to carry-over effects of phonetic context. The measurement tools I have used to establish baseline data for English and French will allow new languages to be studied and systematically compared with these data in the future. The second major purpose of this study has been to extend previous research on AS and research on bilingual speech production by examining whether bilinguals who are perceived as native speakers of both of their languages have ISP differences that mirror monolingual group differences, and more broadly, how a bilingual's pronunciation proficiency relates to their ISP(s). Finally, I have also examined the effect of a bilingual's speaking mode (bilingual mode versus monolingual mode) on ISP.  1.1. Outline of the Dissertation  In the remainder of Chapter I of this dissertation, I describe AS and ISP in more detail, and I review a small subset of the literature dealing with both of these. Specifically, I review various researchers' non-instrumental views of A S , including some that have existed for more than a century, and I also review a study by Gick et al. (2004), the first study to quantitatively measure multiple articulatory components of AS. Shortcomings of that study are pointed out, setting the stage for Experiment 1 on how language and phonetic context affect ISP. I finish Chapter I by talking about measuring AS in bilinguals, thus motivating Experiment 2 on how perceived pronunciation proficiency and mode of speech production relate to ISP. In Chapter II, I present the method used in Experiments 1 and 2. Subjects and apparatus used, as well as the procedure followed are laid out in detail. Chapter III consists of the results and discussion of Experiment 1, an experiment designed to measure AS in monolingual speakers of Canadian English and monolingual speakers of Quebecois French. Through direct ultrasound measurements of .the tongue, and Optotrak measurements of the lips and jaw in ISP, it is tested whether or not the ISP 2  for Canadian English is significantly different from that of Quebecois French. It is also tested whether or not phonetic context has a carry-over effect on ISP. Chapter IV consists of the results and discussion of Experiment 2, a very similar experiment to Experiment 1, but instead using subjects who were fluent bilinguals in both Canadian English and Quebecois French. Data provided by these bilingual subjects is analyzed to test whether bilinguals who are perceived as native speakers of two languages have one or two ISPs. It is also tested whether ISP is sensitive to the mode of speech (bilingual mode versus monolingual mode) that the bilingual speaker is in. Finally, in Chapter V , there is a general discussion of the results of both experiments, including implications of the research for a wide variety of speech research applications. Then conclusions are drawn, limitations of this dissertation are considered, and future directions for research are given.  1.2. Articulatory Setting (AS) Much has been written about AS, and an attempt to provide a thorough historical review of it is not made here. For detailed historical surveys of A S , see Kelz (1971), Laver (1978), and Jenner (2001). One of the earliest references to the concept of AS was Wallis (1653/1972) cited by Van Buuren (1995, p. 136) as follows:  For instance, the English as it were push forward the whole of their pronunciation into the front of the mouth, speaking with a wide mouth cavity, so that their sounds are more distinct. The Germans, on the other hand, retract their pronunciation to the back of the mouth and the bottom of the throat. Van Buuren (1995, p. 136) also cited Sievers (1876, p. 47) as talking about "Operationsbasis" (basis of articulation), the correct tongue position for a different dialect, a position that is held even during the production of various sounds. Sweet (1890, p. 69) called the concept of AS the "organic basis" of a language:  3  Every language has certain tendencies which control its organic movements and positions, constituting its organic basis or the basis of articulation. A knowledge of the organic basis is a great help in acquiring the pronunciation of a language. As defined by Honikman (1964), A S actually includes more than simply the underlying posture of the articulators. She defined it as the "gross oral posture and mechanics" required for the "economic and fluent" production of the "established pronunciation of a language" [p. 73]. Honikman also divided AS into external setting (i.e. the lips and jaw) and internal setting (i.e. the tongue, velum, and larynx), but of the components of the internal A S , she focused primarily on the tongue, probably because it is difficult to know exactly what state the velum and larynx are in without proper measurement tools. Abercrombie (1967, pp. 92-93) made a distinction between aspects of AS that are within a speaker's control ("muscular tensions [...] which keep certain of the organs of speech adjusted in a way which is not their relaxed position of rest"), and those that are not (i.e. size and shape of an individual speaker's vocal tract). Lebrun (1970) took issue with the concept of tension and how it had been measured at the time, and he called for verification by reliable measurements. Presumably, these "muscular tensions" that Abercrombie referred to can be inferred indirectly from the position of the articulators. A n oversimplified but useful analogy would be inferring the tension of the marionette's strings by observing the position of the marionette. In this study, I looked only at the posture of the articulators, and not the mechanics (i.e. static, not dynamic properties). The reason for this choice was simply one of limiting the research to a manageable scale. Specifically, because I had available methods of measuring the tongue, lips and jaw positions (ultrasound and Optotrak), I limited my observations to these three articulators. I did not measure the velum or vocal folds here. This is not to say that these two areas are unimportant; they are simply beyond the scope of this research. There are other terms that AS is often known by, such as voice quality, voice setting, phonetic setting, and basis of articulation, but different researchers and lexicographers have defined these terms in different ways. In my research, since I only examine the tongue, lips and jaw, A S is the most appropriate term to use because terms such as voice quality seem to imply a focus on laryngeal settings. In the Applied Linguistics field, A S is usually referred to as voice quality (e.g., Esling & Wong, 1983). 4  In the language teaching and applied linguistics dictionary by Richards, Piatt, and Piatt (1992, p. 403), voice quality is equated with timbre and is seen as speaker-specific. Both Crystal's (2003) and Matthews' (1997) linguistics dictionaries have separate entries for AS and voice quality. Crystal (2003) defined A S as more of a language- or dialectspecific entity, and voice quality as a "person-identifying feature of speech" (Crystal, p. 496). Although Matthews (1997) defined voice quality as an individual quality, he also noted that AS "may identify the voice of an individual [,...] may also be characteristic of particular languages or accents [,...] and may carry affective meaning" (p. 26). In his dictionary of phonetics and phonology, Trask (1996, p. 34) defined articulatory setting as:  The overall tendency, on the part of an individual or of the speakers of a particular language, to maintain the organs of speech in some particular configuration throughout speech, as reflected in such factors as the height of the velum, the degree of lip-rounding and the tension of the tongue and lips. In my research, I was only concerned with the linguistic function of AS, specifically its function of characterizing a particular language, in this case Quebecois French versus Canadian English. This is the first of three functions of AS succinctly summarized by Esling & Wong (1983, p. 89):  Voice quality settings may function linguistically, to characterize the particular language or dialect or social group to which a speaker belongs; or they may function paralinguistically, to signal mood or emotion in conversational contexts; or they may also function extralinguistically to characterize or identify the individual speaker. 1.2.1. Non-instrumental Views of AS  The following non-instrumental views of English and French A S will be reviewed briefly in chronological order: Sweet (1890), Graff (1932), Heffner (1950), Honikman (1964), and Esling & Wong (1983). These are not the only non-instrumental views of AS that exist, but they are the ones most widely cited. Although these views are not quantitative, they are nevertheless important, as Laver (1978, p. 9) points out: "The  5  contribution of auditory judgment to the analysis of settings in voice quality is particularly important, since the clues that are available are slight and subtle". A number of these descriptions of AS are for Received Pronunciation (R.P.) of British English, as well as for Parisian French. These probably do not correspond to the AS for Canadian English and the AS for Quebecois French, so direct comparison must be done cautiously. Indeed, even within Canadian English and within Quebecois French there are many different dialects, each with a potentially different AS. In listing seven features of the AS for General American English, Esling and Wong (1983, p. 91) pointed out that "not all dialect groups will share the same features, and some dialect groups may even demonstrate opposite features, but settings that combine some if not all of these features are very common". Given this, I will still compare what has been said about English and French AS to the findings presented in this dissertation so that predictions are made explicit for future non-instrumental studies of Canadian English and Quebecois French. Sweet (1890, p. 72) stated that in English the tongue is flatter and lower, with the tongue blade hollowed and the tongue tip brought back from the teeth, while the lips are in a neutral position. In contrast, in French "the tongue is arched and raised and advanced as much as possible, and the lips articulate with energy". Where in English, the tongue is flat, in French it is narrow. , Graff (1932) described the tongue in (probably Parisian) French to be more forward than in British English, with the lips ready for frequent rounding and the tongue ready for tenser articulation in French. This account of the tongue in French being in a more forward position accords with Sweet's view. Heffner (1950) echoed Graffs and Sweet's description of the tongue being forward for French, and he added that it is also high and tense, although it is not clear whether it is the tongue body or the tongue tip that is high. He also stated that the tongue in British English is comparatively lower and more relaxed, and implied that the tongue is even more relaxed in American English than in British English. It is unclear though, how this notion of a "relaxed" tongue translates into a position or shape of the tongue, and, as mentioned in Section 1.2, Lebrun (1970) has convincingly questioned this notion of tenseness of the tongue.  6  Honikman (1964) also described the AS for Parisian French and for R.P. English, and, although Kelz (1971, p. 204) bluntly stated that "what Honikman said, Vietor and Sweet had said in similar words 80 or 90 years before", Honikman went into much more detail regarding these two languages and her account of the differences did not completely mirror Sweet's. Given the 74-year and 32-year differences between her account and Sweet's and Graffs accounts, respectively, it would not be surprising that differences in their accounts of the ASs are actual differences in the way sounds were produced at the time rather than simply differences of opinion. Honikman (1964, p. 78) stated about the French AS that it has the tongue tip "tethered to the lower front teeth", whereas English AS has a higher tongue tip because the sides of the tongue are tethered to the roof of the mouth and the molars. She said that the French jaw is open more often and perhaps open more widely than English due to the relatively high frequency of [a] in French as opposed to English. The English jaw is loosely closed, but not clenched. The English lips are neutral, whereas the French lips are rounded and vigorously active in spreading and rounding. She implied that the French tongue body would be higher because it is convex to the roof of the mouth, whereas the English tongue body is concave to the roof of the mouth and would therefore be lower. The paper by Esling and Wong (1983) appears to be the only account of the A S of any variety of North American English. Esling and Wong (p. 91) listed seven features of a General American English AS: spread lips, open jaw, palatalized tongue body position, retroflex articulation, nasal voice, lowered larynx, and creaky voice. This description of American English differs from the previous accounts of British English in a number of ways: In American English the lips are spread, the jaw open, and the tongue body "fronted and slightly raised" (p. 92) as opposed to the lips being neutral, the jaw loosely closed, and the tongue body lower in British English. Esling and Wong's description of "retroflex articulation" in American English applies to the tongue tip and implies that the tongue tip in American English should be more retracted than that of French. However in a study using MRI, Tiede, Boyce, Holland, and Choe (2004) showed that American English Ixl is actually articulated in a variety of ways ranging from a bunched articulation to a retroflex one. This means that many speakers of American English probably do not  7  necessarily have a retroflex A S and thus it cannot be assumed that the tongue tip is more retracted in English than in French. In summary, i f we assume that the Canadian-English A S is similar to the General American-English A S described by Esling and W o n g (1983), then the predictions for how it compares to the Quebecois-French A S depend on whether the Quebecois-French A S is more like the Parisian-French A S described by Honikman (1964) or that described by Sweet (1890) and Graff (1932). Whether the Quebecois-French A S is more like Honikman's description o f the Parisian French A S or Sweet's and G r a f f s descriptions o f the Parisian French A S , we would expect the lips to be relatively more spread in English than in French, and the tongue body to be equally high in English and i n French. A s for the tongue tip, we expect it to be higher in English i f the Quebecois-French A S is more like Honikman's description but lower in English i f the Quebecois-French A S is more like Sweet's description. Finally, we expect to see the j a w equally as open in French as in English, according to Honikman's description. In Section 3.2.2, the above noninstrumental views o f A S are compared to the results o f the present study.  1.3. Measuring AS  Although many people have described A S as being different for different languages and dialects, very few people have actually quantitatively measured aspects o f A S to be able to say conclusively how it is different for two languages. The biggest impediment to measuring a language's A S is ensuring that one is measuring only A S and not being influenced by the articulation o f the language's speech segments (see Laver's statements below). This problem was recognized over 55 years ago by Heffner (1950, p. 99): " N o method o f measurement has been devised which would permit the mathematical description o f a basis o f articulation." The problem still had not been solved 23 years later when O'Connor (1973, p. 289) called for future study o f "bases o f articulation" (i.e. A S ) and stated " W e know a good deal more about the detailed articulatory movements i n a language than we know about the general articulatory background on which they are superimposed." Even 22 years after that, the situation still had not changed as evidenced 8  by Collins and Mees' (1995, p. 422) statement that "At the moment, much of the description of AS features - including our own - is largely impressionistic." Some researchers have concluded that AS cannot be measured and the description of it must rely on cross-linguistic comparisons (Kelz, 1971). There have been a number of studies trying to characterize a given language in terms of its overall acoustic properties. If AS underlies speech, surely its effects must be audible in the speech signal. The most common method of measuring the overall acoustic properties of a language is to measure its long term average spectrum (henceforth LTAS), the average of many instantaneous spectra over a reasonably long speech sample. A number of L T A S studies have found a correlation between language spoken and L T A S for individual bilingual speakers, while other studies have failed to find any correlation (see Bruyninckx, Harmegnies, Llisterri, and Poch-Olive, 1994 for a brief summary). However, it is not necessarily the case that L T A S data should directly correlate with AS. Laver (2000, p. 40) pointed out that "all calculations of a long-term average (whether of articulatory position, auditory impression, or acoustic spectrum) based on all segments [...] will give obvious inaccuracies." The problem is that L T A S is a measure of the sounds of a language - i.e., it is directly affected by the phonetic context of the speech one is examining and there is no way to distinguish which aspects of the speech signal are based on A S , and which are a reflection of the frequency of specific articulations in the language's phonetic inventory. Laver (1978, p. 11) stated that "no articulatory setting normally applies to every single segment a speaker utters", and he called this property of speech segments segmental susceptibility. Laver (1980, p. 21) further added that "because the successive segments in the stream of continuous speech vary in their susceptibility to the effect of settings, a setting is audible only on an intermittent basis, and even when audible, varies in its prominence, depending on the susceptibility of the segment currently being uttered." Evidence supporting this comes from Harmegnies, Esling, and Delplancq (1989), who found that not all deliberate changes to voice quality have large effects on L T A S . Thus, although L T A S may provide a kind of spectral signature of a language, it is likely that L T A S does not accurately describe the underlying A S of that language.  9  1.3.1. AS and the Rest Position of the Tongue If AS is an articulatory property that underlies speech, and yet to measure it properly we need to avoid contextual effects, then the challenge is to find a method of measuring it that eliminates the effect of context. Gick et al. (2004) proposed that the ISP is the most representative, least biased configuration at which to measure the position of the articulators in order to infer a language's AS. Because the ISP occurs between utterances, it is representative of speech in a way that absolute rest position (simply for respiration) is not. The position I take in the present research is the same as Gick et al. (2004) in assuming that one's ISP provides the best window for investigating one's AS. Sharpe (1970, p. 124) also equated AS with rest position, but it is assumed that she was talking about absolute rest position (for respiration) and not rest position when one is still in speech mode (i.e. the ISP): "these settings [AS] are learnt early in life, and are then usually the 'rest positions' taken up by the jaws, lips, tongue, etc., when not speaking." In contrast, Hartmann & Stork (1972, p. 19, my italics), in their dictionary definition of AS, explicitly stated that AS is different from an absolute rest position: "adjustments in the vocal tract, adopting a posture of the articulatory organs which is maintained by a speaker throughout the whole time he is talking, but which is different from the relaxed position." Abercrombie (1967, p. 91) stated that some of the features of voice quality are even present when we simply cough, or sigh, or clear the throat. Presumably though, he was referring to the innate components of A S , namely, a speaker's anatomical make up, and not the linguistic aspects of A S . A number of studies in the dental research literature have been conducted on the rest position of the tongue, with the intention of discovering a relationship between tongue rest position (but not pre-speech or between utterances) and malocclusion (e.g. Ballard, 1959; Eifert, 1960; Cookson, 1967). Ballard (1959), on the basis of x-ray data, reported two different types of rest positions for the tongue, The two rest positions he described are the habitual posture, the "true rest position" with the tongue tip making a seal with the incisors and lower lip, and the innate posture with the tongue backed and more arched accompanied by a lower mandible position. It is possible that Ballard's "true rest position" can be considered absolute rest position, when the person is relaxed and is 10  not about to engage in speech activity. However, neither of these are likely to be the ISP (speech rest position) because the subjects in Ballard's experiment were not speaking during the x-ray filming. In the speech research literature, Ohman (1967) was the first to write about a "basic speech posture", giving electromyographic (EMG) evidence of steady tonic activity in some facial muscles (e.g. the levator labii superioris) immediately prior to speech. Ohman (1967) proposed that the articulatory movements of speech are 2  superimposed on the basic speech posture, although he also found that the basic speech posture can be directly inhibited if it conflicts with some necessary articulatory posture (p. 43). Ohman did not mention the possibility that the basic speech posture could be language-specific, and Tatham (1997 (1969), p. 8) said that he (i.e. Tatham) "infer[s] that Ohman does not want, as [he] do[es] not, to make this notion of basic speech posture language-specific." Recently, Tatham (e-mail communication, Dec. 9, 2003) clarified the reasons for his opinion: He referred to "examples in the neuro-physiological literature about how tonic activity in the musculature 'gets ready' for activity" and proposed that "immediately prior to speaking the system sets itself up in a speech-ready state which is universal". He also proposed a "multi-layered modelling approach" for speech such that after the system is in its universal speech-ready state, "then the system re-sets in a language-specific speech-ready state which is overlaid on the previous one". This second layer that Tatham proposed is equivalent to the ISP that I am investigating in this study. Around the same time as Ohman's research, Perkell (1969, p. 41) used the term "speech posture" to describe the state of the vocal tract "at the beginning of a sentence." Through observations of nonsense-word x-ray speech data pronounced by speakers of American English, he determined that the larynx, the velum, and the tongue each have a speech posture that they take up when a speaker is preparing to speak. Specifically, for the larynx he (p. 41) stated that "at the beginning of a sentence the larynx seems to rise to a speech posture [...] from which smaller, context dependent fluctuations are made." For the velum, he (p. 52) described it as being "in an intermediate 'speech posture' position which is between rest and its next highest position: the one occurring during nasalized Note, however, that Bithell (1952, p. 58) used the term "Sprechbereitschaftslage", which is defined as "the position of readiness to speak before the organs required for the sound become active". This concept seems very similar to Ohman's basic speech posture. 2  11  speech". As for the tongue, Perkell (p. 65) stated that "the tongue shape or tonus is part of a speech posture and is basically the same for all vowels; the resulting semirigid tongue body is positioned in the vocal tract by extrinsic tongue muscles attached to its periphery." Like Ohman, Perkell did not mention whether he thought that the "speech posture" he had discovered is a general posture that is the same regardless of language, or whether language-specific speech postures are more likely. However, it is likely that what Perkell had found was the universal first layer that Tatham refers to above. Examples of other references where a distinction is made between pre-speech posture and absolute rest position include Daniloff & Moll (1968), Barry (1992), and Gick (2002). In motor control research, Brown & Rosenbaum (2002, p. 129) state that "Anticipating the perceptual consequences of one's own actions (feedforward) is a prerequisite for effective motor control." Thus it is reasonable to assume that in speech, preparatory vocal tract postures are used (see Schmidt and Lee, 1999, pp. 126-127 for a description of "preparatory postural reactions"), and that these postures are dependent on the phonetic context of the utterance one is about to produce. Indeed, this has been demonstrated for the jaw by Hamlet & Stone (1981), who found that the jaw's pre-speech posture (actually, ISP in their study, as the subjects arguably never left speech mode between stimuli) correlated with the position of the jaw required for sounds that appeared in the following utterance. In other words, anticipatory coarticulation affects not just the speech sounds one produces, but also the pre-speech posture adopted by one's articulators. If a speaker has an utterance-specific preparatory posture that is affected by the sounds to follow, then it follows that, more generally, the speaker could also have a language-specific  preparatory posture, in preparation to produce any of the given sounds  of a language (or possibly the most frequently occurring sounds). The existence of such a language-specific preparatory posture, i.e. a languagespecific ISP, was tested for and confirmed by Gick et al. (2004), who showed that not only was the ISP for Quebecois French different from that for Canadian English, but the accuracy of production of the ISP was as high as that for producing the speech sound [i], consistent with the view that the ISP is a speech target posture. Specifically, Gick et al. found the following differences between the Quebecois-French ISP and the CanadianEnglish ISP: The tongue tip (TT), tongue body (TB), and tongue root (TR) were all 12  farther away from the opposing vocal tract surface in the French group compared to the English group. The upper lip was significantly more protruded in English, but the lower lip was significantly more protruded in French. For both the jaw and the velum, there was no difference between the French ISP and the English ISP. Gick et al. did not measure the tongue dorsum, so it is not known whether this measurement was different across languages. Although the Gick et al. study was as accurate as possible under the circumstances, a number of problems exist, which warrant a replication of the study. First, their study examined only 5 speakers of each language and made crosslinguistic generalizations based on these 10 speakers. Obviously the more speakers used, the more accurate any generalizations will be. Second, the Gick et al. methodology is constrained by the fact that the data they analyzed were existing x-ray movie data (Munhall, Vatikiotis-Bateson, and Tohkura, 1994) with limited spatial resolution and clarity. X-ray films do not show slices of the vocal tract; they contain shadows of objects over other objects and edges are often difficult to define. Another methodological issue with the Gick et al. study is that because the data already existed, Gick et al. had no control over the phonetic content of the stimuli being used and how they were presented to the subjects. The stimuli in the original x-ray study were not designed to balance the phonetic context surrounding ISPs. In addition, they were presented to speakers as a list of sentences to be read, thus increasing the chances of anticipatory coarticulation effects on ISP, just as anticipatory coarticulation effects on the ISP of the jaw were found by Hamlet & Stone (1981). If these effects do exist and were not completely controlled for by Gick et al., it may be that the language-specific differences that they found were actually due to the phonetic context rather than language-specific properties of the ISP. Another potential factor that may have influenced the results of the Gick et al. study is the method of statistical analysis they employed. As the experimental unit for statistical comparison in their repeated measures study, they used the data obtained from each individual measurement token produced by each individual subject, and then used the jackknife procedure (i.e. verification that the means of every subset of N - l subjects was distributed in a similar way) to justify this choice. Although using data obtained from each individual measurement token as the experimental unit is a common practice among 13  speech researchers, M a x & Onghena (1999, pp. 265-266) point out that these types o f analyses are at risk o f having the assumption o f independent error effects violated. M a x & Onghena recommend using one value per measurement location per subject (i.e. the mean measurement value across all o f a given subject's productions in all trials) as the experimental unit. They state that "despite the agreement on this issue in the contemporary statistical literature, the potential for violations o f the assumption continues to occur rather frequently in studies addressing normal or disordered speech-languagehearing processes." (p. 266) If anything, the choice o f statistical method in Gick et al. (2004) would have resulted in a greater number o f significant differences being reported than should reasonably be expected. Thus, a replication o f the results o f Gick et al. (2004) was warranted, using a greater number o f speakers, a method o f data collection that allows for greater accuracy, stimuli presentation and design that balances phonetic context, and using the statistical method recommended by M a x & Onghena (1999). In order to replicate the first experiment o f Gick et al. (2004), the first hypothesis that was tested in the present research was that ISP is language dependent.  Hypothesis 1:  The inter-speech posture (ISP) for Canadian English is significantly different from the ISP for Quebecois French.  In my study, it is unlikely that utterance-specific anticipatory coarticulation effects were possible because the subjects could not see the next stimulus until they had had a chance to assume an ISP. Although anticipatory effects were unlikely, carry-over effects were impossible to eliminate while still being sure the subjects remained in speech mode. So, instead o f eliminating carry-over effects, the phonetic context o f the last syllables uttered was tightly controlled across languages. Hamlet & Stone (1981) tried to eliminate any carry-over effect by waiting "a few seconds" before manually presenting the next stimulus for the subject to read. Therefore, they assumed that the j a w went to some intermediate position (perhaps absolute rest position, but this is not made explicit) or simply drifted around before assuming the configuration o f the next pre-speech posture. Ohman (1967, p. 43) mentioned E M G "evidence" for "basic speech posture"  14  following  an utterance as i f it was common knowledge among E M G speech researchers.  Thus, he implied that one does not simply maintain the posture o f the last sound o f the previous utterance, but that one actively moves the articulators back to the basic speech posture. However, even i f Ohman was correct, the possibility o f carry-over effects o f phonetic context on ISP still exists and needed to be investigated. This led to the second hypothesis that was tested in these experiments.  Hypothesis 2:  Within a given speaker's speech in a given language, that speaker's ISP w i l l differ depending on the phonetic segment that precedes the ISP.  1.4. Articulatory Setting in Bilingual Speakers  It has long been realized that one o f the greatest benefits o f using bilinguals in phonetic research is that vocal tract morphology is automatically controlled for - in single-subject studies, no normalization o f measurements is necessary. Thus, it is tempting to think that a bilingual subject would be a good test for whether or not A S is a language-specific property. However, there are a number o f reasons why comparing each o f a bilingual's ISPs to the respective monolingual group's ISP is potentially imprudent. Grosjean (1989) convincingly argued that a bilingual is not equivalent to two monolinguals in one body, and researchers in bilingualism now generally agree that in the phonetic/phonological acquisition and retention o f more than one language, each language has an effect on the other (Paradis, 1996,2001). Birdsong (2005, p. 9) stated that "it is impossible for either the L I or the L 2 o f a bilingual to be identical in all respects to the language o f a monolingual". Elston-Guttler, Paulmann, and K o t z (2005, p. 1593) cited a number o f recent neurolinguistic studies that support a word-recognition system "that allows for parallel activation o f both languages where influence o f one language while processing in the other is likely." Although the authors were referring to studies o f speech perception, it stands to reason that "parallel activation" also applies to speech production. Given that it is unclear to what degree production o f one o f a  15  bilingual's languages affects production of the other, a direct comparison of bilinguals with monolinguals may be difficult to interpret. Just as quantitative studies of A S in monolingual speakers are very rare, so too are quantitative studies of AS in bilingual speakers. The only study that I am aware of that contains articulatory measurements of AS in bilinguals is that of Todaka (1993, 1995). In his study of four Japanese-English bilinguals (two men who were not perceived as native speakers of Japanese, and two women who were perceived as native speakers of both Japanese and English), Todaka (1993, 1995) used aerodynamic, electroglottograph (EGG) and a number of acoustic measures to assess the voice quality of his bilingual subjects in each of their two languages. He compared speech samples from each language, but as stated previously, the problem with this kind of approach is the same as the problem with using L T A S to infer a language's AS - namely the problem of interference from phonetic context. Todaka found that the two females had breathier voice in Japanese than in English and that all four subjects had higher fundamental frequency (fO) in Japanese than in English. However, because of a lack of language-specific properties of English and Japanese that could account for these results, Todaka admits that both these results must be due to sociolinguistic factors. One of these sociolinguistic factors could be Loveday's (1981) finding that as an indication of politeness, Japanese-speaking females use a higher pitch than English-speaking females, but it is unclear what other sociolinguistic factors Todaka is referring to. Todaka also found that the bilinguals' Japanese vowel space was smaller than their English vowel space, but this could simply be an indication of how each sound is articulated, and not a general setting that is different. In fact it is difficult to imagine an AS that would affect all vowels such that the vowel space is larger for one language than another. Harmegnies & Landercy (1985) did an L T A S study of 20 Dutch-French bilinguals from Belgium. Unfortunately, no information was given about the subjects' abilities in each language. Harmegnies & Landercy found that the L T A S differences were greater between speakers than between languages, and that the variability between languages "mainly rel[ied] on differences between the distribution of phonemes in the languages." (p. 72) Another bilingual L T A S study was done by Bruyninckx et al. (1994), who examined the speech of 24 Spanish-Catalan bilinguals, 12 of whom were Spanish16  dominant and 12 of whom were Catalan-dominant. In that study the between-language variability was higher than the within-language variability. Bruyninckx et al. ended their paper by calling for an articulatory study to be done to explain the L T A S results. Although a direct comparison of bilinguals with monolinguals may be difficult to interpret, comparing a given bilingual's ISP in one language to his/her ISP in another language could determine whether or not having the correct ISP for a language is an important component in native-like pronunciation, something that has never been empirically tested. For i f a bilingual speaker who is perceived as a native speaker of both languages does not have two distinct ISPs (one for each language), then it follows that having the correct ISP (and hence, the correct AS) is not a prerequisite for native-like pronunciation of a language. Not all bilinguals are perceived as native speakers of both of their languages, and in fact, because of first language attrition, some are perceived as native speakers of neither of their languages. There is a myriad of factors that influence L2 pronunciation proficiency: age of first exposure to the L2, language of the home, language of the community, frequency of exposure to the L2, amount of LI use, etc. It is possible that an additional factor is the ISP one uses in speaking an L2. It is reasonable to expect that bilinguals who are not perceived as native speakers of at least one of their languages might have only one ISP, or if they have different ISPs then the differences are not those that are most salient between the monolingual groups. The third hypothesis that was tested in these experiments is as follows:  Hypothesis 3:  A bilingual who is perceived as a native speaker of both languages has a different ISP for each language and will show the same types of crosslinguistic ISP differences that monolingual groups show; conversely, a bilingual who is perceived as not being a native speaker of at least one language will have fewer, if any, of the crosslinguistic ISP differences that monolingual groups show.  In addition to controlling for the phonetic context of the ISP (mentioned above), another factor, this one specific to bilingual studies, that should be controlled for is the 17  mode that the speakers are speaking in. Research on bilingual speakers has shown that the communicative setting (whether monolingual or bilingual) affects their speech production. Thus, in testing Hypothesis 3 above, only stimuli spoken in a "monolingual setting" were used (see Chapter II for more details). Grosjean (1998, p. 136) stated that bilinguals communicate in either monolingual mode or bilingual mode, where mode is defined as the "state of activation of the bilingual's languages and language processing mechanisms". This "state of activation" is difficult to define precisely, because it seems to refer to a cognitive measure. However, if part of this state of activation involves a readiness of the articulators to move into the needed configuration for a sound in either language, then it could be that bilinguals who are perceived as native speakers of two languages have only one intermediate ISP when in bilingual mode, especially in cases where they truly do not know what language they will use next. For these bilinguals, this one "bilingual-mode ISP" would be different from the "monolingual-mode ISP" of each of their two languages for ISP components where the two monolingual-mode ISPs differ. The fourth and final hypothesis that was tested in these experiments was as follows:  Hypothesis 4:  Bilingual speakers who are perceived as native speakers of each of their two languages have a unique bilingual-mode ISP that differs in all significant respects from both monolingual-mode ISPs (where "significant respects" are those respects in which differences obtain between the two monolingual modes).  1.5. Purpose of this Research and Summary of Hypotheses  In summary, one purpose of this research was to partially replicate Gick et al. (2004) using an improved methodology, thereby quantitatively determining what differences exist, if any, between the ISP (and thus, as it has been argued above, the AS) of Canadian English and that of Quebecois French. A group of monolingual speakers of each language was used to provide speech rest position data. In addition, the carry-over  18  effect of phonetic context on the ISP was examined within each speaker's speech and more generally within a given language-group's speech. Another purpose of the research was to test whether bilingual speakers of both of the above languages have one ISP (and hence, one AS) that is shared between their two languages, or instead show the same type of crosslinguistic ISP differences that monolingual groups show. This question was answered for bilinguals of various degrees of proficiency in their two languages. In addition, a comparison was made between the bilingual subjects' monolingual mode results and their bilingual mode results. Specifically, the following four hypotheses were tested:  Hypothesis 1:  The inter-speech posture (ISP) for Canadian English is significantly different from the ISP for Quebecois French.  Hypothesis 2:  Within a given speaker's speech in a given language, that speaker's ISP will differ depending on the phonetic segment that precedes the ISP.  Hypothesis 3:  A bilingual who is perceived as a native speaker of both languages has a different ISP for each language and will show the same types of crosslinguistic ISP differences that monolingual groups show; conversely, a bilingual who is perceived as not being a native speaker of at least one language will have fewer, if any, of the crosslinguistic ISP differences that monolingual groups show.  Hypothesis 4:  Bilingual speakers who are perceived as native speakers of each of their two languages have a unique bilingual-mode ISP that differs in all significant respects from both monolingual-mode ISPs (where "significant respects" are those respects in which differences obtain between the two monolingual modes).  19  C H A P T E R II Method  2.1. Subjects  A l l subjects who participated in this research either had responded to an advertisement for subjects or had been invited to respond through word of mouth. Both English and French advertisements for subjects were used (see Appendices I and II) and were placed at strategic locations in the city of Vancouver, including the University of British Columbia campus. Data for Experiments 1 and 2 was initially provided by 33 speakers, although 9 of these had to be excluded for various reasons (see below in Sections 2.1.2 and 2.1.3). Details about the 15 monolingual subjects and 9 bilingual subjects whose data was actually used in Experiments 1 and 2, respectively, can be found in Appendix III. None of the speakers who provided data for these experiments had noticeably missing teeth or an extreme overbite or underbite. This is important to note because the state of one's dentition can have an effect on one's tongue's absolute rest position. Kotsiomiti, Farmakis, and Kapari (2005) found that an abnormally retracted resting tongue position is much more likely in subjects who are partially or fully toothless. Note though that even in fully dentate subjects, they found that 12.3% had an abnormally retracted resting tongue position. A l l subjects were paid for taking part in the experiments, and none of them were aware of the purpose of the experiments. In addition, almost all subjects had no previous phonetic training.  2.1.1. Criteria for Classifying Subjects as Monolingual or Bilingual  As Experiment 1 used monolingual subjects and Experiment 2 used bilingual subjects, before a detailed description of the subjects is given, it is appropriate to define what is meant by "monolingual" and "bilingual" here. The literature contains a variety of disparate definitions for bilingualism, with the line between monolingual and bilingual 20  being drawn in many different places. A review of this research is not appropriate here, but see Baetens Beardsmore (1986) for an in-depth summary. In this research, pronunciation ability as perceived by native listeners, along with self-classification as bilingual or monolingual, were paramount in classifying the subjects. The speech of all subjects who classified themselves as bilingual, as well as the speech of a few who classified themselves as monolingual but were proficient enough in their L2 to be able to read the stimuli fluently, was judged by native-speaking listeners for degree of foreign accent. No subjects were classified as bilingual unless both of their languages received a rating of 3.0 or higher ("adequate" / "convenablement") out of 5, as judged by 10 monolingual native listeners - see below for a description of the foreign accent rating task, and see Appendix IV for the actual rating scale used. If a subject's L2 had an average rating of less than 3.0, then that subject was considered to be a monolingual for the purposes of this research. Using the native listener judgements to determine whether . someone was bilingual or not was appropriate for this research because part of the research is an investigation of how perceived pronunciation ability in a language relates to the AS that different groups use. The foreign accent rating task mentioned above was given to 10 native listeners of each of the two languages. These 20 listeners were paid for their participation. Ten monolingual French listeners (eight of whom were the monolingual subjects from Experiment 1) judged the French speech of all the bilinguals in a task in which they had to rate the bilinguals on a scale of 1 to 5. In addition, 10 monolingual.English speakers (none of whom were the monolingual subjects from Experiment 1) judged the English speech of the same bilinguals on a 5-point scale (see Appendix V for detailed results) . A 3  5-point scale has often been used in other studies where foreign accent is rated such as Bongaerts (1999), Marinova-Todd (2003), Dromey & Wheeler (2004), and Birdsong (to appear). Included as controls in the French speech samples that were judged were samples of two monolingual French speakers, as well as samples of two less-proficient L2 French In hindsight, it may have been better in this rating task to have used a 7- or 9-point Likert scale (see Jesney, 2004) where only the endpoints were given definitions and the rest of the points on the scale lacked descriptors. This would not have made a difference to whether subjects were perceived as native speakers or not (there would still be only one "native speaker" rating available), but in one or two cases it might have influenced whether a subject was considered monolingual (less than 3.0 out of 5, in the present study) or not. 3  21  speakers. The same was true of the English speech samples. Five sentences from each speaker were played in succession, with a six-second pause between speakers. The order in which the speakers were presented was the same as that listed in the tables in Appendix V . The sentences that were played were selected from the stimuli in Appendices VI and VII, and these sentences had been uttered in the experimental setting (i.e. with markers attached to their faces and ultrasound probes under their chins). Wherever possible, the same 5-sentence sample was chosen for each speaker, but if the speaker had stumbled over the words or not had time to finish the sentence, another sentence was chosen instead. The sentences chosen for each speaker's sample are given in Appendix-VIII. As mentioned above, if a subject's L2 was given an average rating of 3.0 or less, then that subject was deemed to be monolingual for the purposes of this research. Also, i f a subject was given an average rating of 4.2 or above in a language, he or she was deemed to be perceived as a native speaker of that language. A level of 4.2 was chosen because that ensured that at least two judges perceived that subject to be a native speaker of the language. A level below 4.0 was considered to be too low as it would explicitly signify, according to the definitions given in Appendix IV, that that speaker's average rating was as a "near native speaker". Note, however, that it would be possible for a speaker to have an average rating below 4.0, but still be perceived as a native speaker by some judges something that indeed occurred with Subject 23 in English. Her background is discussed in more depth in Section 4.2.1.  2.1.2. Monolingual  subjects  A l l monolingual subjects in this research had had at least some exposure to a second language - all had studied a foreign language in school, by choice and/or by law. However, all of the monolingual subjects considered themselves to be monolingual and had not been exposed to an L2 earlier than age 6. Of the eight monolingual French subjects, none had had formal schooling in English before age 10. A l l but one (Subject 14) lived in the province of Quebec at the time of the study, unless they had just moved to Vancouver within that week for a short homestay or temporary summer employment. Subject 14 had been living in Vancouver 22  for about one year, but had been using 60% French in her daily life as a nanny for a bilingual family. Before moving to Vancouver, she used 90% French in her daily life. A l l the monolingual French subjects had monolingual French parents. Of the seven monolingual English subjects, only two of them had studied French beyond high school, Subjects 2 and 5. After completing all their English trials, these two subjects were asked to read one French trial each. They received a French rating of 2.6 and 1.9, respectively, out of 5, and thus were classified as monolinguals. A l l the monolingual English subjects lived in Vancouver at the time of the study and all used nothing but English in their daily lives. The data for Experiment 1 was initially provided by 10 monolingual English speakers and 12 monolingual French speakers, although 3 of the English speakers and 4 of the French speakers had to be excluded, leaving 7 English speakers and 8 French speakers for analysis. There were a number of reasons for the omission of subjects. Of the three English subjects excluded, one was fluent in Cantonese, and thus was not monolingual, a second braced her tongue against her palate between most of the utterances, and a third subject was not cleanly shaven, resulting in a chin marker that would not stay taped on and an unclear ultrasound image. Of the four French subjects excluded, one had difficulty with the stimuli and when later questioned about it admitted that French was his second language and Arabic his first. A second French subject was excluded because she was not completely comfortable with the data collection procedure and only contributed less than half the amount of data as the other subjects. A third French subject was excluded because the ultrasound image of her tongue was not clear enough to be able to make reliable measurements. A fourth French subject was excluded because she was not comfortable with the stimuli as written and wanted to alter the form of the sentences. The mean age of all seven English subjects was 27. The mean age of all eight French subjects was 24. Since all subjects were adults and none had reached old age, their L1 was neither developing nor deteriorating, and therefore the difference in the two groups' mean ages was not considered an issue. Of the monolingual English subjects, four were female and three were male. As for the French, six were female and two were male. Since all data was scaled based on an anatomical measurement (see Section 2.3.2), 23  the slight gender mismatch between the two monolingual groups was not considered significant. Subject 2 had a fairly substantial amount of Quebecois-French schooling (outside of Quebec) from the age of 6, but her parents, siblings, and most of her friends are monolingual English. At the time of the study, she was attending university full time in English and living in Vancouver. At that time, 100% of any given week was completely in English for her and she had not spoken or listened to French for about 3 years. As mentioned above, because her French ability, as perceived by 10 native listeners of French, was 2.6 out of 5, she was classified as a monolingual speaker of English. Her English was given a rating of 5 by all 10 of the judges, the only English native speaker for which this happened. Thus, her English pronunciation was probably not influenced by her French schooling.  2.1.3. Bilingual  subjects  A l l subjects in Experiment 2 were bilingual in Canadian English and Quebecois French. Some of the subjects had knowledge of a third (or more) language, but had only used it (them) in the past to a minimal degree, or more than 10 years prior to participating in this experiment. A s mentioned in Section 2.1.1, all subjects who were classified as bilingual had been rated as 3.0 or above out of 5 in each of their languages. Data for Experiment 2 was initially provided by 11 bilingual speakers of English and French, although 2 of them had to be excluded, leaving a total of 9 subjects for analysis. O f the two subjects excluded, one was not cleanly shaven, resulting in a chin marker that would not stay taped on and an unclear ultrasound image. In addition, he had poor eyesight and had difficulty seeing the stimuli. The second subject who was excluded had a French rating of 2.6 out of 5, thus, by the criteria laid out in Section 2.1.1, she was considered to be monolingual. However, she was not included in the English monolingual group of Experiment 1 because the stimuli she read were the bilingual set of stimuli, not the full set of monolingual stimuli. The mean age of all bilingual subjects was 30. Seven subjects were female and two were male. A l l of the bilingual subjects admitted that they were comfortable 24  codeswitching (i.e. alternating between two or more languages during discourse). This was important because if a subject was not comfortable codeswitching, he or she may have found the bilingual-mode task (see Section 2.3.1) unnatural or overly difficult to do. A summary of the native listener judgements from Appendix V is given in Table 2.1 for all the bilingual subjects.  Table 2.1. Summary of perceived language abilities of all bilingual subjects (shaded cells indicate that the rating is high enough - 4.2 or above - that the subject was considered to be perceived as a native speaker) English  French  Perceived as native  rating  rating  speaker of...  21  4.9  4.6  Both  17  4.7  4.7  Both  22  4.3  4.9  Both  19  4.6  4.2  Both  18  3.9  4.9  French only  23  3.9  4.7  French only  20  3.8  4.4  French only  24  4.6  3.7  English only  16  3.3  3.7  Neither  Subject number  In Table 2.1, note that there was at least one subject in each of the four possible groups of bilinguals: Four subjects (21, 17, 22, and 19) were perceived to be native speakers of both languages, three subjects (18, 23, and 20) native speakers of French only, one subject (24) a native speaker of English only, and a final bilingual subject (16) a native speaker of neither English nor French. To aid in interpretation of the results throughout the remainder of the dissertation, the bilingual subjects will be presented in the order shown in Table 2.1 (i.e. grouped into four groups).  25  2.2. Apparatus The apparatus used to collect data in Experiments 1 and 2 can be seen in Figures 2.1,2.2, and 2.3.  Figure 2.1. Data collection setting  The main pieces of equipment for collecting data were an ultrasound monitor for viewing the movements of the tongue in real time, and an Optotrak (Northern Digital Inc.) 3020 optical tracking system for measuring the 3D positions of the lips, jaw, and head relative to the ultrasound probe. The ultrasound monitor used was an Aloka ProSound SSD-5000 with a UST-9118 endo-vaginal 180° electronic curved array probe. The probe is specified to have a variable frequency range of 3-9.0 M H z , and according to the Medicines and Healthcare products Regulatory Agency [MHRA] (2004), the mean  26  )  1 slice thickness width of the tissue viewed with this probe is approximately 3 mm. The Optotrak system used consists of a set of three single-axis C C D cameras, with 11-bit hardware resolution, that tracked the movements of 12 infrared-emitting diodes (markers). The Optotrak hardware was controlled using a Northern Digital software program, Collect (version 2.002), running on a PC (Micron Millennia X K U 333). Subjects were seated in "the experiment chair", a modified antique ophthalmic examination chair (American Optical Co., model 507-A) with a 2-cup rear headrest adjusted to contact the base of the skull just above the neck, and a forehead stabilizing head restraint ("head stabilizer") with two rubber pads which were positioned to be lightly touching the subject's forehead near the hairline. The ultrasound images seen on the ultrasound monitor were recorded onto digital video tape using a JVC SR-VS20 Mini DV/S-VHS V C R . Simultaneous audio for these ultrasound recordings was recorded using a Sennheiser M K H 416 P48 super-cardioid short shotgun condenser interference tube microphone. The microphone signal was fed into the V C R via a digital mixing console (Yamaha 01V). Stimuli were displayed to the subjects on an Apple PowerBook G4 17-inch laptop computer at a distance of about 2.5 metres, and at approximately eye level. The stimuli were presented as Microsoft PowerPoint (version 10.1.0) slides with the English stimuli displayed in Times 88 point font and the French stimuli displayed in Times New Roman 72 point font.  27  28  Figure 2 . 3 . Placement of ultrasound probe, head restraint, and Optotrak markers (numbered)  2.3. Procedure The data collection procedure used in this research was the same in Experiments 1 and 2 , except for which six blocks of stimuli were used and the order in which these blocks were presented. This difference in choice and order of stimuli blocks was simply due to the fact that the monolingual subjects of Experiment 1 were presented with stimuli in only one language, whereas the bilingual subjects of Experiment 2 were presented with stimuli in two languages. More details on stimuli presentation can be found in Section 2.3.1.4.  2 9  2.3.1. Data Collection  When a subject arrived for a data collection session, the procedure was as follows. First the subject was shown the equipment to be used, was told the procedure to be followed, and was given the opportunity to ask any questions. Then after signing ethics forms, the subject was seated in the experiment chair and the headrest, the head stabilizer, and the ultrasound probe were adjusted to the proper height. The subject was then moved to a more comfortable chair where the Optotrak markers were attached to his/her lips and jaw.  2.3.1.1. Optotrak Setup  The position of the 12 Optotrak markers can be seen in Figure 2.3. Markers 1 through 4 were all permanently attached to a pair of lensless glasses that were worn by each subject and it was assumed that these markers did not move relative to each other. Marker 3 was on the midsagittal plane and markers 2 and 4 were equidistant from it. Marker 3 was slightly higher and more protruded from the subject's face than markers 2 and 4. Marker 1 was situated on a rigid bamboo skewer that was mounted off the right arm of the glasses. Bamboo was used because it is strong enough to remain rigid but light enough not to put the glasses off balance. For all subjects, marker 1 was located to the right of, posterior to, and superior to the subject's right ear (see inset of Figure 2.3). Note that during the course of a trial, if it is assumed that the glasses do not move relative to the subject's head, then markers 1 through 4 defined a rigid body that included the subject's skull. This was important for being able to track the movement of the subject's skull (and thus the palate as well) during a trial. Markers 5 and 6 were attached to the ultrasound probe, 70 mm and 140 mm, respectively, from the tip of the probe (i.e. the end of the probe that made contact with the subject's skin). Marker 7 was mounted on a 1 cm cube of open cell foam that was taped under the chin using 3 M Micropore surgical tape. Markers 8 and 10 were placed at the right and left corners, respectively, of the subject's mouth, as close as possible to the 30  mouth opening without making it uncomfortable when closing the mouth. Marker 9 was placed as close as possible to the vermilion border of the upper lip on the midsagittal plane. Marker 11 was also placed on the midsagittal plane, but on the lower lip. Depending on how "pouty" the subject's lower lip was, it was sometimes necessary to place Marker 11 above the vermilion border in order for its light to be seen consistently by the Optotrak position sensor. Marker 12 was left in place on a wooden, hinged clapper between experiments. The clapper provided a sound that was used to synchronize the Optotrak data with the ultrasound data (see Section 2.3.1.4). A l l affixation of markers was done using double-sided clear tape that pulled off the skin easily, but usually not so easily as to come off during the course of an experiment. If a marker did come off during a trial, it was reattached before the following trial. However, it was serendipitously the case that data from all subjects for whom a marker did come off, were later excluded for other reasons (see Sections 2.1.2 and 2.1.3). The wires coming from the Optotrak markers were kept out of the way by taping them to the subjects' cheeks with surgical tape. Once all the Optotrak markers were in place on the subject, the subject was seated in the experiment chair and the Optotrak system parameters were set as follows: Marker frequency = 2600 Hz; Duty cycle = 25%; Strober voltage = 7V; and Dynamic duty cycle = On. A l l Optotrak data was collected at 90 Hz.  2.3.1.2. Ultrasound Setup  Throughout all preliminary and main trials involving ultrasound data collection, the forehead stabilizer and ultrasound probe were locked into position. Water-soluble ultrasound gel was applied to the head of the ultrasound transducer which was then placed against the subject's neck in the submental region. The probe was positioned so that a midsagittal image was being displayed with the tongue tip towards the right side of the screen. The probe angle was adjusted so that the image on the ultrasound monitor showed as wide a tongue region as possible, from the shadow of the hyoid bone on the left to the mandible shadow on the right. The exact angle of the probe was different for every subject, dependent on anatomy and posture. The average angle was 20° for the English subjects (with a range of 9° to 24°), 19° for the French subjects (with a range of 15° to 31  25°), and 18° for the bilingual subjects (with a range of 12° to 24°). As the probe was always placed so as to maximize the view of the tongue from the tongue root to the tongue tip while centring the tongue on the ultrasound monitor, the relatively similar average angle across groups indicates that vocal tract length was fairly consistent across the three groups. The probe angle was calculated from the absolute positions of Optotrak markers 5 and 6 on the ultrasound probe, and it is the number of degrees off of vertical that the tip was pointed away from the Optotrak cameras. To give the reader an idea of a typical ultrasound probe angle relative to the skull, Figure 2.4 shows a CT scan of the author's upper vocal tract with an ultrasound scan overlay of the author's tongue. Note that in Figure 2.4, the CT scan is angled at 23° relative to the ultrasound picture, an angle that is within the range of angles found in all three groups of subjects. Both the ultrasound image and the CT image were created during production of the sound [rj] in [on]. Since in the experiments the probe was at an angle off true vertical, each tongue image shown with ultrasound displays the tongue rotated clockwise. In Figure 2.4, where the tongue shapes are not identical, it is because the ultrasound image here was taken a number of days after the CT scan was made. The ultrasound probe that appears in the CT image was not the one used to make the ultrasound image on the right. Also, the probe in the CT image is not at an ideal angle relative to the tongue - it is simply serving as a landmark. It is interesting to note that the hyoid bone itself (not just its shadow) is clearly visible in the 100% ultrasound image, although this was certainly not always the case in the experiments. It is also clear in the 100% ultrasound image where the velum meets the tongue - the tongue line suddenly loses a lot of its brightness. But perhaps most interesting is that the tongue line that is seen in the 100% ultrasound image is actually the tongue line, all the way from the lower teeth to the hyoid shadow, even though the velum and epiglottis are pressed against the tongue (i.e., what is seen in the ultrasound is the tongue's surface, not the superior surface of the velum or the posterior surface of the epiglottis). This implies that the density of the tongue is different enough from the density of the epiglottis and the velum that the sound reflects at this border, confirming the reliability of the tongue surface images.  32  Figure 2.4. CT scan of upper vocal tract with ultrasound tongue image overlaid at various opacities 100% ultrasound overlay  70% ultrasound overlay  Both B-mode ("brightness modulation" mode) and M-mode ("motion" mode) ultrasound images were collected. B-mode shows a 2D section of the tongue, while M mode shows a continuous time series of a line of B-mode dots. Only the B-mode images were used as data in this dissertation. The "range" setting on the ultrasound, the total real distance represented in the window on the screen, was set at 10 cm for both the B-mode and M-mode displays. Although not relevant in this study, the M-mode sweep speed was set for 1.5 seconds per period. 33  2.3.1.3. Preliminary Trials  Three preliminary measurement trials were done prior to any of the main trials where the subject was asked to read sentences. The first preliminary trial was a 40-second "wag" trial, the purpose of which was to set baselines for movements of the head, lips, and jaw. The second preliminary trial was a 15-second "bite" trial using the 12 Optotrak markers described above in addition to 3 more markers that were placed on a triangular piece of Plexiglas (see Figure 2.5). The bite trial was a way of displaying, in the ultrasound image, an anatomical landmark whose position was known in an external frame of reference. The third and final preliminary trial was a 25-second "palate" trial that ensured that palate information was available for all subjects. Only the first of these three preliminary trials (i.e. the wag trial) was used in the data analysis for this study, but all three trials are described in more detail below.  Figure 2.5. Plexiglas "bite triangle" used in the second preliminary trial  In Preliminary Trial 1, the subject was asked to turn his/her head to the extreme right, left, up and down bringing the head back to a centre position and pausing between each direction. The subject cycled through this order twice. Although these head turns  34  could be used by the Optotrak system to calculate the centre of rotation of the head, they were not used in this study. After the head-turning task, the subject was asked to spread the corners of the lips as widely as possible, as i f saying an exaggerated [i]. This was followed by the subject protruding the lips as far as possible, as i f saying an exaggerated [u]. The subject was specifically asked to spread markers 8 and 10 as far to the sides as possible, and to protrude markers 9 and 11 out as far as possible. This was done twice each. This spreading and protruding of the lips enabled a baseline to be set for the extremes of lip movement of each subject. For the final few seconds of the wag trial, the subject was asked to relax, look straight ahead at the computer screen and keep the jaw and lips closed. The jaw here was not in a clenched position, but instead set a baseline for a maximally elevated rest position of the jaw. Unfortunately, for bilingual Subject 17, no wag trial was done, so the distance from the centre of the glasses to a maximally elevated jaw could not be calculated. Also, for bilingual Subject 23, although a wag trial was done, due to an oversight she was not wearing the head-tracking glasses during that trial. Thus, for these two subjects, jaw elevation data could not be properly analyzed. Next, in Preliminary Trial 2, the subject was.asked to bite down on the Plexiglas triangle seen in Figure 2.5. The subject bit the triangle with the incisors (not the molars) and held it motionless for about 10 seconds while forcing the blade of the tongue against the edge of the Plexiglas. The angle that was created as the tongue was bent around the Plexiglas was usually clearly visible in the ultrasound image, thus providing an anatomical landmark whose position was known in an external frame of reference. Although this data was not used for analysis in these experiments (because it was not collected for any of the monolingual English subjects), it also provided a way of determining both the distance from the central incisors to a point on the tongue, as well as the approximate occlusal plane of a given subject. Finally, in Preliminary Trial 3, the subject was asked to take a small amount of water into his/her mouth through a straw. While holding the water in his/her mouth, the subject then ran the tongue tip back and forth along the midsagittal line of the hard palate. The subject then swallowed the water and was usually asked to press the whole tongue against the hard palate for about 2-3 seconds. The palate trial ensured that a relatively good image of each subject's palate was available. Although this preliminary trial was 35  done with all subjects, it was only with the final few subjects that Optotrak data was collected simultaneously with this trial. Thus, this trial was not used in the data analysis, but the data that do exist show very clear swallowing images and may be useful in future work. The palate trial did not have to be used in this study because each subject swallowed at least once during each trial between sentences, and the alveolar ridge was visible at this time. See Epstein & Stone (2005) for issues related to imaging the palate with ultrasound.  2.3.1.4. Main Trials  After the three preliminary trials, the experiments then moved on to main trials involving the subjects reading a number of sentences aloud. Due to the fact that the English stimuli contained some nonsense or low frequency words for use in a different study (Campbell, 2004), all monolingual English subjects and all bilingual subjects were given a 15-sentence English practice trial. The practice trial was not deemed necessary for the monolingual French subjects because the French stimuli all contained standard vocabulary familiar to any French speaker. The French stimuli were chosen for a future study that needs many tense-lax minimal pairs in a carrier sentence. For a list of the English and French stimuli used, see Appendices VI and VII, respectively. While each monolingual subject in Experiment 1 read six blocks of monolingual data, each bilingual subject in Experiment 2 read two monolingual English blocks (Blocks 1 and 2 in Appendix VI), two monolingual French blocks (Blocks 1 and 2 in Appendix VII), and two blocks that each contained a mix of 15 English and 15 French sentences that were in a pseudo-randomized order (see Appendix IX). The order of presentation of the blocks was the same for all bilingual subjects: English Block 1, French Block 1, Mixed Block 1, English Block 2, French Block 2, Mixed Block 2. Before each block the subject was told what kind of block was to appear. Thus it was assumed that during the presentation of a monolingual block, the bilingual subject was in "monolingual mode" (Grosjean, 1998), prepared to read in only one language. However, during the presentation of a bilingual block, the bilingual subject was not aware of which language would be presented next. Thus, they were in "bilingual mode", ready to read in either 36  language with both languages fully activated at once (Grosjean, 1998). Regarding language mode, it should be noted that all of the bilingual subjects and many of the monolingual French subjects understood enough English to allow the experimenters to communicate in a mix of broken French and basic English. Thus, even though the French monolingual blocks contained only French, communication before and after the blocks was done mostly in English. This may have compromised the monolingual nature of the monolingual French mode. In future research, if a fluent French-speaking assistant were available to communicate with the subjects, it would help to create a monolingual French environment for French data collection. The duration of Optotrak data collection was 67 seconds for the practice trial, and 131 seconds for each real trial. Each of the real trials consisted of 30 sentences that were displayed one at a time to the subject. As mentioned in Section 2.2, stimuli were displayed as PowerPoint slides in a large font size. The PowerPoint "slide transition" for each 30-sentence trial was set so that each sentence slide was displayed for 3 seconds followed by a blank slide for 1 second. The final blank slide after the 30 sentence was th  accompanied by a distinct sound (a loon call), indicating to all that the trial was complete. As each sentence was displayed, the computer beeped, thus making a record on the ultrasound D V tape of when the subject saw what he or she was supposed to say next. It was assumed that before the beep, any preparatory vocal tract posture (see Schmidt and Lee (1999, pp. 126-127) for a description of "preparatory postural reactions") would be for the language or speech in general and not the task of articulating the first phoneme/syliable. Since the subject was not presented with a list of stimuli, there was no list effect to take into account. Also, since the first word of each sentence was sufficiently varied, there was no way that the subject could predict what articulation would be necessary next. This most probably eliminated any anticipatory coarticulation effects on the ISP. The clapper was used twice per trial - once at the beginning before the start of the stimuli presentation, and again at the end about 1-2 seconds after the subject had finished saying the final sentence. After the clapper was raised and released, the first minimum vertical position of marker 12 in the Optotrak data ("first", because in some cases there was a bounce) was taken to be synchronous with the start of the banging sound, on the 37  D V tape, of the clapper reaching a closed position. This allowed for synchronization of the Optotrak and ultrasound signals. At the beginning of each trial, the Optotrak data collection program, Collect (version 2.002), was initiated, then the clapper was dropped, and finally the PowerPoint slide show was started. While the subject was reading the stimuli, two experimenters (the author and an assistant) monitored a real-time display of the Optotrak data for missing markers and checked the real-time ultrasound display for any problems with the ultrasound data (e.g. screen going into sleep mode, or a fuzzy tongue line because the subject was gradually sliding sideways on the probe). The order of the trials was not randomized, but kept the same across subjects. For all monolingual subjects, the trials were collected in order from one to six. In a few cases where image quality was thought to be inferior, trial 1 was re-collected after trial 6 and this second version of trial 1 was used in the data analysis. In addition, the sentences within each trial were presented in the same order to every subject, but this order had been pseudo-randomized to ensure a balance of phonetic contexts throughout and across the trials. Since between-trial comparisons are not being made (all data from all trials are averaged together), the fact that the order of the trials is not randomized presumably affects every subject the same way.  2.3.2. Data Analysis  Since the data consisted of two types, Optotrak numeric data and ultrasound video data, the most efficient way to concurrently analyze both types of data was through a series of M A T L A B programs written by the author for this purpose. Before the data could be processed by the M A T L A B programs, though, it had to be pre-processed and narrowed down. As a first step, the D V tape of the ultrasound data was transferred to a manipulable file format (Adobe Premiere 6.0 movie files) by means of a Sony DCR-TRV900 digital video camera connected via a Fire Wire cable to an Apple PowerBook G4 laptop computer. The Premiere video capture settings were as follows: Compressor - DV-NTSC; Frame size: 720 x 480; Pixel aspect ratio: D l / D V NTSC (0.9); Frame rate: 29.97; Depth: 38  Millions, Quality: 100%. The Premiere audio capture settings were as follows: Rate: 32,000 Hz; Format: 16 bit stereo; Uncompressed; Interleave: 1 frame. In retrospect, using a pixel aspect ratio of 0.9 was not a good setting and it necessitated having an extra step resampling the data to a frame size of 656 x 480 with a pixel aspect ratio setting of "square pixels (1.0)" (See Aho (2004) for formulas showing why a frame size of 656 x 480 was chosen.) This resampling was necessary in order to ensure that measurements made later were on the same scale in all directions. The ultrasound movie files were then cropped so that the first frame in each file was the frame immediately after the clapper was first heard. Cropping the movie in this way made it easier later to determine which Optotrak frames corresponded to the ultrasound frames of interest (see below). Possible periods of rest to be used for analysis were found by playing back the ultrasound movie files and searching after every sentence for a period of at least 10 frames (i.e. 333 ms) of no tongue motion in the B-mode tongue shape and the M-mode lines. The reason for choosing a 10-frame period, as opposed to a longer or shorter period, was that a 10-frame period was the longest possible rest period such that the tongue was considered to be at rest in an average of about 50% of the intersentential pauses across all 24 subjects. If such a period of 10 frames of no tongue motion existed, then the centre frame of that period was chosen as a "possible rest frame" for analysis. For each subject, a list of all possible rest frames was constructed for all trials 4  that the subject completed. For each monolingual and bilingual subject, Tables 2.2 and 2.3, respectively, show the total number of times after a sentence (out of 180 possible sentences, unless otherwise stated) when that subject's tongue was at a complete stop for at least 10 frames of the ultrasound movie file.  4  "possible" because if it was not in one of the desired phonetic contexts, it was not used.  39  Table 2.2. Total rest frames available and number actually used (monolingual subjects) Subject  nglis  lolin  ca  5 w S  Fre nch  [onollingu  "c3  Total number of times  Total number of rest times in a  tongue at rest  required phonetic context  (out of 180)  (i.e. total used in this study)  1  116  61  2  94  51  3  131  74  4  101  63  5  76  46  6  71  51  7  103  59  8  n/a  45  9  n/a  68  10  65  22  11  47  22  12  n/a  37  13  122  58  14  n/a  56  15  n/a  57  40  Table 2.3. Total rest frames available and number actually used (bilingual subjects) Total number of rest times in a required  Perceived as native speaker  Subject  Total number of times  phonetic context  tongue at rest (out of 180,  (i.e. total used in this study)  unless otherwise stated)  of...  Monolingual mode  Bilingual mode  English  French  English  French  21  96  19  18  6  9  17  41 (out of 90)  0  6  13  5  22  111  21  19  7  10  19  154 (out of 210)  25  20  21  16  18  62  8  8  5  8  23  64 (out of 120)  12  9  7  5  20  121  23  19  14  8  Eng only  24  99  21  9  13  7  Neither  16  n/a  13  15  12  12  Both  Fre only  5  For example, in Table 2.2, it can be seen that Subject 1 brought her tongue to a complete stop after 116 of the 180 sentences - thus, a total 116 possible rest frames. The rightmost column shows the actual number of rest frames analyzed in Experiment 1 once frames outside the necessary phonetic contexts were eliminated. Thus, out of the 116 frames available for Subject 1,61 of these were in a desirable phonetic context - one that could be reasonably balanced across English and French (see below). Due to time constraints during data analysis, the total number of times the tongue was at rest was not investigated for 5 of the monolinguals and 1 of the bilinguals (appears as "n/a" in the table). The average number of frames used per monolingual English, monolingual French, and bilingual speaker was 58, 46, and 49, respectively. The total number of rest positions analyzed was 405 from monolingual English speakers, 365 from monolingual French speakers, and 443 from bilingual speakers, for a grand total of 1,213. In at least some of the cases, the speed of the subject's speech had a direct effect on the number of possible rest frames that could be used for analysis. For example, Subject 6 spoke noticeably slower than the other English subjects and also had the fewest 5  Due to accidental tape erasure, monolingual-mode English data is not available for Participant 17.  41  available rest frames to choose from. Subject 10 braced her tongue against her palate between at least 53 pairs of sentences and those tokens were not included in the analysis. Subject 11 often did not finish reading the complete sentence before it disappeared from the screen, so he ended up speaking during the time his tongue was expected to be at rest. Because there were only two blocks of bilingual-mode data and four blocks of monolingual-mode data and these were divided between two languages, there were not as many analyzable tokens in each language for the bilingual speakers as there were for the monolingual speakers. As mentioned above, only frames that were in certain phonetic contexts were used for data analysis. In order to test whether ISP is language specific, one first has to ensure that the phonetic contexts in which the ISPs appear across languages are balanced. To phrase this differently, if the phonetic context surrounding the ISP has an effect on the ISP itself, then in order to investigate whether there were other language-specific properties of the ISP, it would be necessary to control very tightly for context. In this study, since the sound following the ISP is not known to the subject until the next stimulus flashes on the screen, there could be no anticipatory effects, only carry-over effects. In this research, phonetic context was balanced by considering the IPA representation of the standard Canadian-English and Quebecois-French pronunciation of the final syllable of each sentence-final word, making the untested and probably naive assumption that, for example, an l\l in English is articulated the same way as an HI in French, and then approximately balancing the number of tokens of III across the two languages. For balancing the contexts, in order to have enough tokens to do a reliable statistical analysis, it was necessary to assume that a French nasalized vowel was equivalent (in terms of the articulatory configuration of the tongue, lips, and jaw) to its non-nasalized English counterpart. Admittedly, this is also probably a naive assumption as the velum is connected to the tongue body by means of the palatoglossus muscle. Since English contains sounds that are not found in French (and vice versa), out of the 180 sentences (6 blocks of 30) that each subject said, there were only a certain number that had final words whose final sound could be reasonably matched across English and French. In the six English monolingual blocks, there were 103 sentences out  42  of 180 that qualified as having appropriate final sounds to keep the phonetic context balanced across languages. In the six French monolingual blocks, there were 84 words out of 180 that had appropriate final sounds. Table 2.4 shows the final words from all of the English and French sentences for which the following rest position was eligible to be chosen (i.e., only if the tongue came to a complete stop for 10 frames during this time) for analysis in this study.  43  Table 2.4. Total possible available ISPs for each pre-ISP word Broad  Narrow  English  Narrow  Broad  context #  context #  word  total  total  1  1  2  Thai July day holiday  10  French word  ail  Narrow  Broad  total  total  3  plaie 11  26  musee  9  vallee  24  outils 3  January  5  nuit  12  radiographic 4 2 5 3  6 7  4  5  6  Sue through show scenario regatta again weekend  perdus  11  trou 22  11  9  auto chaudron  24  15  maison 5  5  assiette  18  8  class  5  9  lunch  6  10  spring  7  11  week  5  12  job  9  monsieur recettes  29  face roche sacoche camping  12  clinique grecques  9  champ etang  3  3  6 3  15  6 3 6 9  9  9  The words are grouped according to narrow, as well as broad phonetic context, but for analysis in Experiment 1, only the effects of broad phonetic context were examined. This was because there were not enough tokens of each narrow phonetic context to give sufficient statistical power. The broad phonetic contexts were as follows: 1, 2, and 3 were vowel-final contexts, 4 was a coronal-final context, 5 was a dorsal-final context, and 6 was [low vowel + labial] in English and [nasalized low vowel] in French. More 44  specifically, the sounds included in each broad context are listed in Table 2.5, and each broad context is given a name that will be used in Section 3.1.3.  Table 2.5. Definitions of the broad phonetic contexts used in the analysis Broad  Broad  Narrow  Included words with these  context name  context #  context #  final sounds  1  [ai]  2  [e(0]  3  [i]  4  [u]  5  [o(u)], [o]  6  [3]  7  [en], [end], [et]  8  [ass], [as]  9  [AntfJ, [oj]  10  N]  11  [ik], [ik], [ek]  12  [ab], [a]  FrontV  BackV  Schwa  CoronalC  DorsalC  LowV  1  2  3  4  5  6  A l l data were checked to make sure that what was actually said during the data collection matched the presented stimuli word for word. Any cases where the final word spoken was not the final word in the sentence presented to the subject were discarded. Finally, the rest frames were extracted from the video files and saved as .tiff image files. In order to determine correctly which Optotrak frame corresponded to a given ultrasound frame, it was necessary to search through the Optotrak data for marker 12 (the clapper marker) and find the lowest vertical position for the marker (lowest x-coordinate in the Optotrak coordinate system) after the clapper was dropped. In some cases the clapper bounced, resulting in marker 12's vertical position increasing slightly before dropping slightly again. In this case, the first minimum was taken (abbreviated "clprmin", 45  a M A T L A B variable). This frame where the marker first reached its minimum value was taken to correspond to the ultrasound video frame where the clapper noise was first heard. Because the Optotrak data was collected at 90 Hz, whereas the ultrasound data was at 29.97 Hz, a formula was used in the main M A T L A B program to calculate the Optotrak frames of interest based on the ultrasound frames of interest. The ultrasound frames of interest were simply multiplied by 90, divided by 29.97, and then the result was added to "clprmin" to get the Optotrak frames of interest. In each trial, the frame where the alveolar ridge was the most clearly visible was chosen and saved as a .tiff image file. These alveolar ridge files were later used in a M A T L A B program to define the (constant) location of the alveolar ridge with respect to the four glasses markers in each trial (i.e. the coordinates of the alveolar ridge in head space). This calculation of the position of the alveolar ridge in head space was accomplished by first using ultrasound data to calculate the location of the alveolar ridge relative to the probe in ultrasound image space, then using Optotrak data to calculate the location of the ultrasound probe with respect to the head. Knowing the alveolar ridge relative to the probe, and the probe relative to the head, gave us the position of the alveolar ridge relative to the head. Then knowing from the Optotrak data how the head moved about the probe during the course of a trial, we then knew how the alveolar ridge moved about the probe and we determined the coordinates of the alveolar ridge in all ultrasound frames of interest. See Appendix X for the formulas used in these geometrical calculations. Although one can analyze ultrasound data without correcting for head movement, especially if one tries to limit head movement during data collection (see Gick, Bird, and Wilson, 2005, for why and to what extent this is valid), correction for head movement is desirable for at least two reasons: When the head rotates about the probe, the tongue line in the ultrasound image also rotates and consequently one cannot be sure of where on the tongue one is measuring. Also, while the skull including the hard palate is moving, the mandible and tongue could remain motionless with respect to the probe, thus giving no indication on the ultrasound monitor of any actual change in shape of the vocal tract airspace.  46  As mentioned above, a series of M A T L A B programs ("m-files") were used for data organization and analysis. Certain functions contained in the M A T L A B Image Processing Toolbox were also used by the m-files. Optotrak data in its original floating point file format was converted into M A T L A B 3D matrixes by a program supplied by Mark Tiede (Haskins Laboratories / MIT). These 3D matrixes were used by the main mfile, which was written by the present author. This main m-file, in which measurements were made and calculations were performed, had as its input the rest position .tiff images, the alveolar ridge .tiff images, and a database of Optotrak numerical values from three other m-files. The articulator measurements that were relevant for this experiment and on which statistical analyses were performed are shown in Table 2.6. In this dissertation, these 12 measurement locations are hereafter referred to as the "components of ISP".  Table 2.6. Definitions of the components of ISP used in statistical analyses TTht  distance from the probe centre (a point exactly 1 cm below the surface of the probe on the midsagittal line and marked on the ultrasound) to the tongue tip  TBht  distance from the probe centre to the tongue body  TDht  distance from the probe centre to the tongue dorsum  TRrt  distance from the probe centre to the tongue root  JAW1  amount of jaw lowering from a maximally closed position  ULlo  upper lip height relative to the glasses  LLlo  lower lip height relative to the glasses  ULpr  upper lip protrusion - distance from the midsagittal upper lip marker to an imaginary plane constructed through the alveolar ridge and two end points of the glasses  LLpr  lower lip protrusion - same as upper lip, but using the lower lip marker  Lvap  vertical lip aperture  Lhap  horizontal lip aperture  Lnar  amount that horizontal lip aperture is narrowed from its maximally spread position  Coronal tongue shape was not measured, but it is admittedly an important factor to consider. It can be viewed with ultrasound, but with 2D ultrasound, it is not possible to  47  see both midsagittal and coronal views of the tongue simultaneously. A l l items generated in the .csv file output of the main m-file can be seen in the left column of Table 2.7. A l l 30 of these items were generated for each of the 770 rest frames analyzed for the monolingual speakers in Experiment 1 and each of the 443 rest frames analyzed for the bilingual speakers in Experiment 2.  Table 2.7. Contents of numeric output file of main M A T L A B m-file used Data output  Notes  1  Subject #  1-24  2  Language of previous sentence  l=English; 2=French  3  Mode (monolingual or bilingual)  l=Monolingual; 2=Bilingual  4  Trial #  1-6  5  Previous sentence #  1-30  6  Narrow phonetic context #  1-12  7  Broad phonetic context #  1-6  8  Ultrasound rest frame #  1-4000 (approx)  9  Optotrak rest frame #  1-12000 (approx)  10  Angle of ultrasound probe in degrees  See Tables 2.2 and 2.3  11  Distance in mm from probe centre to TT  Click on TT & program measures dist  12  Distance in mm from probe centre to TB  Click on TB & program measures dist  13  Distance in mm from probe centre to TD  Click on TD & program measures dist  14  Distance in mm from probe centre to TR  Click on TR & program measures dist  15  Distance in mm from probe to alveolar ridge  16  Distance in mm from TT to alveolar ridge  17  Distance in mm from bridge of nose to chin  18  Absolute min. in mm from nose to chin  19  Distance in mm from nose to alveolar ridge  20  Distance in mm from chin to alveolar ridge  Euclidean dist from marker 3 to marker 7 Used as the factor in normalization of data  48  21  Distance in mm from nose to upper lip  Euclidean dist from marker 3 to marker 9  22  Distance in mm from nose to lower lip  Euclidean dist from marker 3 to marker 11  23  Angle in degrees between tongue lines  one-third of angle between hyoid shadow & alveolar ridge  24  Maximum lip spread during WAG trial  Euclidean dist from marker 8 to marker 10  25  Horizontal lip aperture  Euclidean dist from marker 8 to marker 10  26  Difference between max lip spread & horizontal aperture  27  Vertical lip aperture  Euclidean dist from marker 9 to marker 11  28  Upper lip protrusion  Perpendicular dist from marker 9 to plane defined by markers 2, 4, and the alveolar ridge  29  Lower lip protrusion  Perpendicular dist from marker 11 to plane defined by markers 2, 4, and the alveolar ridge  30  Ultrasound image space coordinates of 5 points  Ptl=alveolar ridge; Pts2-5 are the 4 points on the tongue that were clicked on by user (TT, TB, TD, TR in that order)  The procedure that the M A T L A B m-files followed was first to prompt the user for the subject's 3-letter code name and a Trial number to analyze. After retrieving relevant data for the specified trial of that specified subject, the program then displayed the stored .tiff image of the frame where the alveolar ridge was visible (e.g., see Figure 2.6). The program prompted the user to click on the location of the alveolar ridge.  49  Figure 2.6. Ultrasound frame in M A T L A B with alveolar ridge visible. The three lines labelled A , B, and C, and cutting through the tongue image should be ignored here and in Figures 2.7 and 2.8. These M-mode lines, which correspond to the three horizontal data tracks (labelled "M-mode images") situated under the B-mode image, are not used in the present research.  UBC 1SRL _  •F  CPT  54Hz  B-mode image  3  ALOKA 04-AUD-04 09:44:40 9118 6.0 DUfl:100% MI = 0 . 3 9  alveolar ridge hyoid bone shadow RIO  fil  G b l CS  in  o d e  m S*<wr rtmmmm mmmm  „  ,... -  .-, -  -  .  g  e 2:TONGUE 90 DEG  After the user clicked on the alveolar ridge, the program retrieved the first stored rest position .tiff image, and placed two red marks on it, one at the probe centre (10 mm below the surface of the probe) and the other at the point where the previously-clicked-on alveolar ridge was now computed to be after corrections for head movement. Note that in order to register the ultrasound images in a physical space defined by the Optotrak, a simplifying assumption was made that the ultrasound images always showed the midsagittal plane. This allowed the 3D coordinates of the alveolar ridge to be mapped onto the 2D ultrasound image by simply ignoring the third coordinate (i.e. the one off the midsagittal plane). Although it is very likely that the ultrasound images were not always 50  showing the midsagittal plane, in a preliminary analysis of a subset of the data reported here, Gick et al. (2005, p. 512) showed that during ISP, the variation in head position in the direction perpendicular to the midsagittal plane was 1.86 mm, the smallest of the three possible translational movements. After zooming in on the tongue, the image was then displayed to the user (e.g., see Figure 2.7) and the user was prompted to click on the image enough above the "hyoid shadow" that a straight line drawn to the probe centre would intersect the tongue line. The hyoid shadow is the dark area to the lower-left of the tongue root, a shadow in the image caused by the absorption of the ultrasound waves by the hyoid bone. This can be seen most clearly in the ultrasound/CT scan overlay in Figure 2.4.  Figure 2.7. Ultrasound frame in M A T L A B of an ISP to be analyzed. The user is separately instructed to click above the hyoid shadow in the picture. File  Edit  View  Insert  Tools  Window  Help  computed position of alveolar ridge tongue root  tongue tip  RIO O b i CS  fij  51  After the user clicked above the hyoid shadow, a straight line was drawn through this point and the probe centre. A second straight line was drawn through the alveolar ridge and the probe centre. Finally, two more lines were drawn that trisected the angle between the first two lines (e.g., see Figure 2.8). The user was then prompted to click on the four points where each line intersected the surface of the tongue, and to do this in order from right to left (i.e. TT to TR). These measurement locations shall be called tongue tip (TT), tongue body (TB), tongue dorsum (TD), and tongue root (TR), and they correspond roughly to constrictions in the alveolar, palatal, uvular, and pharyngeal regions. Although the tongue line appears to be a thick white line, the actual surface of the tongue is the bottom edge of the white line, where it meets the black area. In the case of the tongue line not being visible, the user was prompted to click in a far corner of the image and such points were later eliminated from consideration in the analysis. After the user clicked on the four tongue points, the distance in mm from the probe centre to each point was calculated and saved.  52  Figure 2.8. Ultrasound frame in MATLAB showing four measurement lines  Before any statistical analyses were performed on the data, the data was normalized. Every speaker has a different sized vocal tract, and when comparing groups of speakers across languages, normalizing the articulatory measurements is likely to reduce some of the noise in the data. The method of normalization was the same for all subjects, monolinguals and bilinguals. For comparing an individual bilingual subject's French data to his/her own English data, obviously no scaling is necessary and any scaling that is done does not change the results (the same vocal tract produced both sets of data), but in order to compare across bilinguals or to compare the bilinguals to the monolinguals scaling is desirable. Although no perfect method of normalizing speech data from different speakers has been discovered yet, a number of methods have been used in other studies (see below). The method of normalization used in this dissertation  53  was to multiply each subject's data measurements by a factor that was calculated using the distance from each subject's nose bridge (as approximated by the centre glasses' marker) to the alveolar ridge (as seen in some ultrasound images). This is effectively an anatomical measure that approximately varies with some aspects of the size of the vocal tract. The multiplication factor for a given subject was the largest subject's distance (in this dissertation, that of Subject 6) divided by the given subject's own distance. Table 2.8 shows the mean distance from a given subject's nose bridge (as approximated by the centre glasses' marker) to the alveolar ridge (as seen in some ultrasound images and then calculated for each ISP). Table 2.8 also shows a ranking of the subjects from longest (1) to shortest (24). The mean of the mean distances for the seven English subjects was 72.19 mm, for the eight French subjects was 74.75 mm, and for the nine bilingual subjects was 73.68 mm.  54  Table 2.8. Mean distance from subject's nose bridge to alveolar ridge  English French English-French  Bilingual  Monolingual  Monolingual  Subject  Mean distance in mm (and standard deviation)  Rank (1 = longest; 24 =  from nose bridge to calculated alveolar ridge  shortest)  1  69.11 (5.69)  20  2  66.44 (1.87)  23  3  77.85 (2.76)  7  4  69.55 (2.12)  18  5  70.95 (2.90)  16  6  81.90 (2.11)  1  7  69.52 (1.52)  19  8  75.81 (1.23)  9  9  79.16 (3.67)  5  10  63.12 (4.79)  24  11  76.50 (2.28)  8  12  80.21 (1.98)  2  13  74.44 (1.33)  11  14  75.26 (3.35)  10  15  73.48 (1.67)  12  16  70.27 (1.75)  17  17  78.49 (1.55)  6  18  72.77 (2.09)  14  19  80.16 (0.82)  3  20  68.38 (5.04)  22  21  71.57 (2.12)  15  22  79.53 (1.26)  4  23  69.02 (2.85)  21  24  72.89 (4.79)  13  In Table 2.8, note that the standard deviations indicate that there is a reasonably high degree of variability in the distance from the nose bridge to the alveolar ridge. Ideally, this is a measurement that should not vary at all, assuming the glasses do not move relative to the skull. The standard deviation ranges from a low of 0.82 (Subject 19) to a high of 5.69 (Subject 1). The most probable reason for the high standard deviation in some subjects is trial-to-trial variation in the selected location of the alveolar ridge. This  55  variation would have been due to a lack of clarity in the ultrasound frames where the subjects were swallowing. It is possible that what looked like the alveolar ridge was actually not so in some trials. If anything, this extra noise would reduce the number of significant differences found across speakers and languages, and should not introduce artificial significant effects. Although it is intuitively apparent that tongue dimensions should vary with body size, just as across the animal kingdom, the size of the brain increases with body size (Seyfarth & Cheney, 2002), results have been mixed. Tongue measurements taken of 35 healthy Caucasian dental students by Oliver and Evans (1986) showed that the mean length, breadth, and thickness of the tongue is greater for males than for females. Note, however, that Chiang, Lee, Peng, and Lin (2003), who studied 20 Chinese medical students, found no significant difference between the 10 females' and the 10 males' mean tongue thicknesses (as measured with ultrasound from the mylohyoid muscle to the tongue body). In a three dimensional study of 25 Japanese female adults, Takada, Sakuda, Yoshida, and Kawamura (1980) showed a significant correlation between tongue volume and both the capacity of the oral cavity and the depth of the floor of the mouth, but not the height of the palatal vault. Thus, the anatomically-based method of normalization used in this dissertation is probably not perfect, but is probably an improvement over using nonnormalized data. A l l statistical analyses were performed using JMP 5.1 (SAS Institute Inc.) statistical analysis software.  56  CHAPTER III Experiment 1: AS in English and French Monolinguals  Experiment 1 was an investigation of whether or not ISP is language dependent and whether or not phonetic context has a carry-over effect on ISP. In this experiment, only monolingual speakers were used. Hypotheses 1 and 2 were tested, namely that the ISP for Canadian English is significantly different from the ISP for Quebecois French, and that within a given speaker's speech in a given language, that speaker's ISP differs depending on the phonetic segment that precedes the ISP.  3.1. Results  Results of Experiment 1 on monolingual subjects are now presented such that in Section 3.1.1 a comparison of languages' ISPs is presented, in Section 3.1.2 a comparison of individuals' ISPs is presented, and in Section 3.1.3 a comparison of phonetic context effects is presented. More specifically, in Section 3.1.1, English group means are compared to French group means, in order to test the hypothesis that the English ISP is different from the French ISP. Then in Section 3.1.2, box plots of the ISPs for individual subjects are presented for English and French, showing within- and between-subject variability. Finally, in Section 3.1.3, results are presented of a test of the hypothesis that phonetic context has a carry-over effect on ISP.  3.1.1. Results: ISP Across Languages For each measurement (e.g. tongue tip height, upper lip protrusion, etc.), group means and standard deviations were calculated for English and for French. Each group mean and standard deviation are the mean and standard deviation of the individual subject means for that measurement and that language. These individual subject means, as well as wtf/zw-subject standard deviations, can be found in Appendix XI. The English and French  57  group means and between-sub)QC\ standard deviations for each language and each measurement are given in Table 3.1.  Table 3.1. Means and between-subject standard deviations (in parentheses) of monolingual English and French groups for each component of ISP Component of ISP  English group  French group  mean (SD)  mean (SD)  TTht (tongue tip height)  63.88 mm (5.02)  58.35 mm (3.57)  TBht (tongue body height)  66.41 mm (4.55)  63.33 mm (5.46)  TDht (tongue dorsum height)  58.19 mm (8.83)  56.39 mm (5.77)  TRrn (tongue root retraction)  48.60 mm (9.59)  48.69 mm (6.65)  J A W l (jaw lowering)  6.36 mm (4.16)  6.53 mm (2.72)  ULlo (upper lip distance from  75.39 mm (2.49)  72.38 mm (5.07)  97.14 mm (3.41)  96.16 mm (6.90)  ULpr (upper lip protrusion)  31.80 mm (6.96)  23.60 mm (4.68)  LLpr (lower lip protrusion)  36.00 mm (6.90)  27.01 mm (5.15)  Lvap (vertical lip aperture)  22.13 mm (3.82)  23.89 mm (4.42)  Lhap (horizontal lip aperture)  61.96 mm (4.21)  60.81 mm (5.62)  Lnar (degree of lip narrowing from  14.11 mm (5.99)  7.38 mm (3.57)  bridge of nose) LLlo (lower lip distance from bridge of nose)  maximum spread)  To test whether the group means in Table 3.1 were significantly different across language groups, 12 / tests (assuming unequal variances) were done - one at each of the 12 components of ISP. These t tests compared the English group mean to the French group mean, where each group mean was the mean of the individual subject means from Appendix XI. Table 3.2 shows the results of these t tests. For more details about the statistics, including the exact degrees of freedom - reduced because of the more conservative assumption of unequal variances - see Appendix XII.  58  Table 3.2. Comparison using t tests (assuming unequal variances) of monolingual English and French group means by component of ISP. Note that because of the assumption of unequal variances, the actual degrees of freedom are fewer than the 13 reported here. See Appendix XII for the exact degrees of freedom. Component of ISP  Result  / Ratio  Prob > |*|  TTht (tongue tip height)  Eng significantly higher  /(13) = 2.43  p = .0340 *  TBht (tongue body height)  Eng tending higher  t(U) = 1.19  p = .2542  TDht (tongue dorsum height)  no difference  ^(13) = 0.46  p = .6560  TRrn (tongue root retraction)  no difference  f(13) = 0.02  p = .9848  JAW1 (jaw lowering)  no difference  ^(13) = 0.10  p = .9254  ULlo (upper lip distance from  Eng tending greater  t(13)=  1.49  p = .1656  bridge of nose)  (i.e. lower height)  LLlo (lower lip distance from  no difference  /(13) = 0.36  p = .7291  ULpr (upper lip protrusion)  Eng significantly greater  <13) = 2.64  p = .0242 *  LLpr (lower lip protrusion)  Eng significantly greater  ^(13) = 2.83  = .0163 *  Lvap (vertical lip aperture)  no difference  ^(13) = 0.83  p = .4217  Lhap (horizontal lip aperture)  no difference  ^(13) = 0.45  p = .6590  Lnar (degree of lip narrowing  Eng significantly greater  /(13) = 2.60  p = .0277 *  from maximum spread)  (i.e. more narrowing)  bridge of nose)  As can be seen in Table 3.2, significant differences between the English and French groups were found for tongue tip height (English higher than French), upper lip protrusion (English more protruded than French), lower lip protrusion (English more protruded than French), and degree of lip narrowing - the amount that the corners of the mouth are drawn in towards the midsagittal plane from a maximally spread position (English more narrowed than French).  59  3.1.2. Results: ISP Across Individuals  Within a  Language  In Figure 3.1, results in box plot format for individual subjects are presented for four components of ISP. These are the four components of ISP that were found to be significantly different across English and French groups immediately above in Section 3.1.1. In the box plots, the top, bottom, and line through the middle of each box correspond to the 75 percentile, 25 percentile, and 50 percentile (i.e. the median) th  th  th  respectively. The whiskers on the bottom and top extend from the 10 percentile and 90 th  th  percentile respectively. The box plots are provided here because they clearly indicate the amount of within-individual and across-individual variability in the measurements obtained.  60  Figure 3.1. Box plots of monolingual subjects' distribution for the four components of ISP that were significantly different across languages in Section 3.1.1. The dotted lines are the group means from Table 3.1. Male subjects are #3, #6, #7, #11, and #13. Upper lip protrusion (ULpr)  Tongue tip height (TTht) 8  0  1  2  3  _J  1  1  4 1  Participant number 5 6 7 8 9 10 11 12 13 14 1 1 I I 1 1 1 1 1 1  1  2  3  4  5  I ri H  I  I  I  I  15 L_ 80  75  75  70  70  50 45  Participant number 6 7 8 9 10 11 12 13 14 15  I  Ul  I  I I  I  I  I  I  40  • 35 E E  ;30  E  %60  30  -frt  5 25  60 •  CO  .c t  40  E 35 E  •£•65  = 20  25 '  4 — 4  20 =  r 15  55  60 45  15 •  50 English  50  10  French  participants  45  participants™  45 4 0  i—i—i—i—i—i—rh—i—i—i—i—i—i—r 1  2  3  4  5  6 7 8 9 10 11 12 13 14 Participant number  oi  4 0  15  1  Lower lip protrusion (LLpr) 1 60  !  2  3  I  I  4  I  Participant number 5 6 7 8 9 10 11 12 13 14  I  I  ]| I  I  i  I  I  I  I  EnglisR  French  participants  participants  i  i  i  2  3  4  i 5  i  m  i  10  i i i i  i  i  6 7 8 9 10 11 12 13 14 Participant number  t>  15  Degree of lip narrowing (Lnar) 15  I  60  25  1  2  3  4  5  I  I  I  I  !  Participant number 6 7 8 9 10 11 12 13 14  I  Ii I  I  I  I  i  I  15  I I  25  T 50 E E 40  50  T —  40  20  20-&-  I ! 15  ;30  30 •  : 20  20  £-  15  H  S 10 §>  ?10  10  I e  10 English  French  English  French  participants  participants  participants  participants  I 1  2  I I  3  4  I 5  I I 6 7 8 9 10 11 12 13 14 Participant number  -ho 15  o  I  I i3  2  4-44  5  M M  6 7 8 -9i —10r 11 12 13 14 Participant number  15  In Figure 3.1, one thing that is immediately noticeable is the fairly high degree of between-subject variability within a given language. All of the subjects in a given language do not tightly cluster about that language's mean. For each of the four components of ISP shown above, there is at least one subject per language that could 61  possibly be considered an outlier. Within-subject variability (shown by the length of each box and the distance the whiskers extend) also differs greatly across the 15 subjects. The most extreme example can be seen in the lower lip protrusion ISP values for the English subjects. Subject 1 has 50% of her lower lip protrusion data (the amount inside the box) within a range of approximately 10 mm, while Subject 7 has his corresponding data within a range of only 2 mm. Another thing to note in Figure 3.1 is that the within-subject variability of the degree of (horizontal) lip narrowing is much lower than the other three components of ISP. Most of the subjects in both languages have 80% of their data (from the top whisker to the bottom whisker) within a 1-2 mm range.  3.1.3. Results: Carry-Over Effects of Phonetic Context on ISP  The preceding phonetic context of the ISP was controlled for across languages, and thus the four crosslinguistic differences that were found in Section 3.1.1 are due to something other than phonetic context. Nevertheless, the question still remained whether or not the effect of language on ISP was also mirrored by a carry-over effect of phonetic context on ISP. As mentioned in Section 1.3.1, the existence of a carry-over effect has implications for theories of speech motor control, for studies that use the ISP as a reference point for making measurements, and for determining the best timing for stimuli presentation in future studies of ISP. Given the four crosslinguistic differences in ISP reported in Section 3.1.1, one would expect that if phonetic context had a very weak effect, or no effect, on ISP, then these four differences should also show up in the majority of phonetic contexts. For if the overall crosslinguistic differences from Table 3.2 only showed up in a few of the contexts, it would raise the concern that the differences were caused by the phonetic context. To investigate this issue, 24 t tests (assuming unequal variances) were conducted with the group-mean measurement of the ISP position of a given component of ISP in a given phonetic context as the dependent variable, and with language as the independent variable. There were 24 t tests because there were 6 phonetic contexts for each of the 4 components of ISP that had shown significant group differences in Section 3.1.1. Results 62  are listed in Table 3.3 and show that out of the 24 t tests, 20 were significant. Note that for each of the four components of ISP, a significant difference is present in almost every one of the six broad phonetic contexts. Thus, based on these group means in different phonetic contexts, it is unlikely that the crosslinguistic group differences that were presented in Table 3.2 were caused by phonetic context.  63  Table 3.3. Comparison o f English group versus French group means by component o f ISP and by phonetic context (phonetic contexts that did not have significant results at p < .05 are shaded; all others were significant across language groups)  Component  Direction of  Broad phonetic t Ratio  of ISP  difference  context number 1  r(13) = 2.49  £ = .0317  r(13)= 1.96  p = .0766  3  / ( l l ) = 2.52  £ = .0338  4  /(13) = 3.15  p = .0094  5  /(12) = 2.87  £ = .0153  6  r(12)=1.87  £ = .0863  1  r(13) = 2.63  p = .0256  2  <13) = 2.43  £ = .0352  3  /(12) = 1.70  ^=.1168  4  /(13) = 2.98  £ = .0139  5  /(12) = 2.70  £ = .0216  6  f(13) = 2.76  £ = .0174  1  /(13) = 2.77  £ = .0190  2  f(13) = 2.49  £ = .0318  3  r(12)= 1.70  £ = .1169  4  f(13) = 3.37  £ = .0064  5  r(12) = 2.78  £ = .0179  6  /(13) = 2.88  £ = .0132  1  /(13) = 2.68  £ = .0248  2  r(13) = 2.62  £ = .0263  3  / ( I I ) = 2.26  £ = .0454  4  r(13) = 2.59  £ = .0282  5  t(\2) = 2.91  £ = .0183  6  r(13) = 2.69  £ = .0239  •2. TTht (tongue tip  English higher  height)  ULpr (upper lip  English greater  protrusion)  LLpr (lower lip  English greater  protrusion)  Lnar (degree of lip narrowing from maximum spread)  English greater (i.e. more narrowed)  Prob > \t\  64  .  These results in Table 3.3 above were obtained by comparing group means in different phonetic contexts. A more reliable indication of whether phonetic context has carry-over effects on ISP can be obtained by analyzing each subject's data individually. Hypothesis 2 stated that "Within a given speaker's speech in a given language, that speaker's ISP will differ depending on the phonetic segment that precedes the ISP." In order to statistically test whether phonetic context affects a following ISP, two contexts that intuitively seem quite different and also happen to have the most tokens available were compared - the BackV context (back rounded vowels) and the CoronalC context 6  (coronal obstruents). With 15 subjects and 12 components of ISP per subject, the total number of t tests performed was 180. The dependent variable was the individual mean measurement of the ISP position for a given subject-component pairing, and the independent variable was the phonetic context that preceded the ISP. A summary of the results of these 180 t tests (assuming unequal variance) appears in Figures 3.2 and 3.3. In the interest of clarity of presentation of the data, t Ratios and probabilities are not reported. A significance level of .05 was used.  As a comparison of this one pair of phonetic contexts (i.e. BackV versus CoronalC) involved 180 / tests, a comparison of more pairs of contexts (with probably more similar articulator positions) was not done.  6  65  Figure 3.2. Number of components of ISP (out of 12), per subject, where t tests (assuming unequal variance) showed a significant difference (p < .05) in any direction between the BackV context and the CoronalC context  Lips (ULIo, LLIo, ULpr, LLpr, Lvap, Lhap, Lnar) g]  Jaw(JAWI)  •  Tongue (TTht, TBht, TDht, TRm)  In Figure 3.2, note that all 15 subjects have at least one component of ISP (i.e. one of TTht, TBht, TDht,..., Lnar) in which the ISP measurements are significantly different across the two phonetic contexts. The highest number of components of ISP where a significant difference between contexts is found is five, and this is true for both Subjects 6 and 12. However, this is still fewer than half of the 12 total components of ISP. Also note that all of the French subjects have at least two lip components of ISP that are influenced by phonetic context. On the contrary, two of the English subjects have no context effects on the lips, and another two English subjects have only one lip component of ISP that is influenced by phonetic context. As for the tongue, all but one of the English subjects have at least one component of ISP that is influenced by phonetic context. For the French subjects, this is true for only three of them.  66  Figure 3.3. Number of subjects (out of 7 English and 8 French) that showed a significant difference, in the same direction, between the BackV context and the CoronalC context by component of ISP English subjects  •jijll  French subjects  8 tongue components of S P v-  lip compbnentsl of ISA  jaw ISP  6  O-Ul  TTht TBht TDht TRrn JAWI ULIo  I  LLIo ULpr LLpr Lvap Lhap Lnar  ISP component  In Figure 3.3, note that for the monolingual English subjects (solid shading), there is only one component of ISP where at least three of the seven subjects have the same significant effect of context on ISP, namely the degree of lowering of the jaw. The three subjects here all had the jaw more open in the BackV context than in the CoronalC context, i.e. more open after high and mid, back rounded vowels than after coronal obstruents. For the monolingual French subjects, there are four components of ISP where at least three of the eight subjects have the same significant effect of context on ISP: the height of the lower lip (LLIo), vertical lip aperture (Lvap), horizontal lip aperture (Lhap), and degree of lip narrowness (Lnar). For five of the eight subjects, the lower lip was lower after coronal obstruents than after back rounded vowels. For the same five subjects,  67  the vertical aperture of the lips was greater after coronal obstruents than after back rounded vowels. Three of the French subjects had a significantly greater horizontal lip aperture and less narrowing of the lips after the rounded vowels than after the coronal obstruents. At a significance level of .05, it can be expected that 9 out of 1801 tests would give false positives, i.e. 9 of the 180 / tests would indicate significant differences when in fact these results were not significant. Since 46 out of 180 t tests were significant in Figure 3.2, and 41 out of 180 t tests were significant in Figure 3.3, it is very likely that a great majority of the differences obtained were indeed significant. However, since it is impossible to determine which of the significant differences were false positives, the results in Figure 3.2, where no strong trend was found, must be observed cautiously. As for the results in Figure 3.3, where a stronger trend was found, the data were analyzed further by comparing the directionality of all differences (both significant and nonsignificant). The prediction, if there were no phonetic carry-over effect, was that for any given component of ISP, half the subjects in a given language should have a greater value in the BackV context and the other half should have a greater value in the CoronalC context. For English monolinguals, the directionality of differences showed a trend for three components: 6 out of 7 subjects had TTht higher in the CoronalC context, 6 out of 7 subjects had TRrn greater in the CoronalC context, and 5 out of 7 subjects had ULpr greater in the BackV context. Since there is no difference in directionality of the jaw results, it is possible that this is a false positive in Figure 3.3. However, For French monolinguals, the directionality of differences showed a strong pattern. For all 12 components of ISP, at least 6 out of 8 French monolinguals showed the same directionality. Specifically, TTht, TBht, TDht, ULlo, ULpr, LLpr, Lhap were all greater for 6 or more out of 8 speakers in the BackV context, while TRrn, JAW1, LLlo, Lvap, and Lspr were all greater for 6 or more out of 8 speakers in the CoronalC context. These directionality results for the French monolinguals support the claim that the significant differences seen in Figure 3.3 are not simply false positives.  68  3.2. Discussion  In Experiment 1, ISP was measured in seven monolingual speakers o f Canadian English and eight monolingual speakers o f Quebecois French i n order to test Hypotheses 1 and 2. Recall that Hypothesis 1 stated that the ISP for Canadian English is significantly different from the ISP for Quebecois French, and Hypothesis 2 stated that within a given speaker's speech in a given language, that speaker'sTSP w i l l differ depending on the phonetic segment that precedes the ISP. The results just presented in Section 3.1 partially support both hypotheses.  3.2.1. Discussion Regarding  Test of Hypothesis 1  In a test o f Hypothesis 1, the results in Table 3.2 show that the ISP for monolingual English speakers is significantly different from the ISP for monolingual French speakers in the following ways: For English speakers, the tongue tip is higher and both lips are more protruded, and the corners o f the mouth are drawn farther away from a maximum spread position than for the French speakers. These significant differences match those o f G i c k et al. (2004) for the tongue tip height and the upper lip protrusion, but they are opposite those o f G i c k et al. for the lower lip protrusion. Note that the lip protrusion results are also contrary to expectation based on the non-instrumental accounts o f Honikman (1964) and others (see Section 1.2.1). Since G i c k et al. were not able to measure lip aperture with the x-ray data they used, no comparison o f lip aperture or degree o f spreading can be made. G i c k et al. found that the tongue body was higher for English speakers. Table 3.2 shows that in the present study, although the English tongue body tended to be higher, there was no significant difference between the English and French speakers (p = .2542). A l s o , results from G i c k et al. showed that the tongue root was more retracted for English speakers. The present results show absolutely no difference in tongue root position between English and French speakers (p = .9848). Finally, neither the results from G i c k et al. nor the present results show any difference in jaw height between the English group and the French group.  69  Thus, out of the six possible comparisons that can be made between the present study and that of Gick et al., three show the same results: the same significant differences for tongue tip height and upper lip protrusion, and the same lack of significant difference for jaw height. Of the other three comparisons that do not completely agree, two were found to be significant by Gick et al. but do not differ significantly in the present results namely, tongue body height and tongue root retraction. Although tongue body height was not found to be significantly different between English and French in this study, the tendency was for English to be higher, the same direction as the Gick et al. results. A n explanation for these two differences in results between the Gick et al. study and the present one may be the more stringent statistical method employed in the present study. As mentioned in Section 1.3.1, the choice in Gick et al. of using each individual token as the experimental unit for statistical comparison makes it more likely that statistically significant differences will be found. The third comparison between the Gick et al. study and the present experiment that does not agree (i.e. lower lip protrusion) is found in both studies to be significantly different across languages, but in opposite directions. One reason for this may have to do with the effect of phonetic context on the position of the lower lip. While the definition of an ISP in Gick et al. was a minimum length of 3 ultrasound frames (i.e. about 100 ms), the minimum ISP length in this study is 10 frames (i.e. about 333 ms). Thus although the articulators may have appeared to be at rest in the Gick et al. study, it is possible that there was simply not enough time for the articulators to return to a rest position, especially given the fact that the subjects could already see the next sentence and could continue reading when ready. It should also be pointed out that one or more of the differences between the Gick et al. results and the results in Section 3.1.1 could be due to (1) the many years difference between when the x-ray data and the present data were collected, and (2) the different distribution of where the CanadianEnglish speakers in each study originated from. Although they did not measure it, Gick et al. posited that the tongue dorsum could be higher for French than for English because the other three tongue measurements all indicated that French speakers' tongues have a smaller midsagittal area than English speakers' tongues. The results of the present study do not support this view - no difference was found between the English and the French tongue dorsum height. Thus it 70  is more likely that Gick et al.'s (p. 226) other explanation is true, namely that there could be more lateral expansion of the tongue for French speakers, and that due to the fact that the tongue is a muscular hydrostat (Kier & Smith, 1985), this lateral expansion causes a reduction in the total midsagittal area of the tongue. This explanation agrees with Honikman's (1964) assertion that the English tongue tip is "tapered" whereas the French tongue tip is "untapered". In addition to the tongue dorsum measurement, five other measurements were made in this study that were not made in the Gick et al. study: upper and lower lip height (measured as the distance from the bridge of the nose), vertical lip aperture, horizontal lip aperture, and the distance that the horizontal lip aperture differed from a maximally spread position (i.e. "degree of lip narrowing"). Neither the lip height nor the lip aperture were different across languages, but the degree of lip narrowing was significantly greater for English, meaning the lips were closer to a maximally spread position for French. As increased lip spreading naturally decreases the amount of lip protrusion, this difference is consistent with the above findings that both lips were more protruded for English speakers. Given the higher frequency of rounded segment types in the phonemic inventory of French, it is perhaps surprising that French had a more spread-lip ISP than English did. It is customary to think of rounding as involving lip protrusion. However, "rounding" in Quebecois French actually could primarily involve a decrease in vertical lip aperture and spreading the lips could cause this decrease. This type of rounding is what Heffner (1950, p. 98) referred to as "vertical lip rounding". Heffner stated that lip protrusion is "much less frequently found with vertical lip rounding" than with horizontal lip rounding. However, i f vertical lip aperture were a salient component of the ISP of the lips, then we would expect to see a cross-linguistic difference in this component (i.e. "Lvap"), but we did not. Although the type frequency of rounded segments in the phonemic inventory of French is high compared to English, it is possible that the token frequency of rounded segments in French is comparable to or even lower than that of English. In that case, the results showing French having a more spread-lip ISP than English would not be surprising. Future work relating AS to token frequencies could shed light on this issue.  71  Another result that at first seems surprising is the fact that the English group had a higher tongue tip during ISP than the French group did. This seems surprising given the fact that coronal consonants in French have a dental place of articulation, more anterior than English coronals, which have an alveolar place of articulation. However, considering what was actually measured by TTht, at least one reasonable explanation presents itself. The measurement denoted by TTht was the distance from the centre of the ultrasound probe to the surface of the tongue and this was measured along a line that intersected the alveolar ridge. Thus, if the tongue tip were anterior to the alveolar ridge (as it is in the case of a French coronal), then TTht would actually be measuring the height of the tongue in a location posterior to the tip (i.e. the tongue blade). During ISP, if the anterior part of the tongue were in an optimal position for articulating a coronal sound (which it may or may not be), the tongue would be higher for English than for French along the line running through the alveolar ridge. Turning from the group comparisons and examining the individual results in Figure 3.1, it is apparent that in each language the results do not cluster tightly around the group mean. Thus, although significant differences between group means were found for tongue tip height (TTht), upper lip protrusion (ULpr), lower lip protrusion (LLpr), and degree of lip narrowing (Lnar), there is a considerable amount of allowable variation in ISP across native speakers of a given language. However, it is almost always the case that the mean of a given measurement for a given speaker is closer to that speaker's group's mean than the other group's mean is. There are four exceptions to this: In the case of TTht, two of the seven English monolingual subjects had lower TTht means than the French group mean, and one of the eight French monolingual subjects had a higher TTht mean than the English group mean. In the case of Lnar, one of the English monolingual subjects had a lower Lnar than the French group mean. No answer is forthcoming as to why Subject 1 had such a great degree of lip protrusion. Since lip protrusion was measured as a distance from an anatomically determined plane, it is possible that Subject 1 had much thicker lips than the rest of the subjects and that because the normalization was done based on a face-length measurement (mid-glasses marker to alveolar ridge), lip thickness was not corrected for. However, variation in anatomical size and proportion, even after normalization, could 72  certainly be the reason for the above anomalies in the data. In the comparison of group means done in Section 3.1.1, it is assumed that these individual anomalies are averaged out, and thus not a concern. It is worth pointing out that even i f Subject 1 is eliminated from consideration in the / test calculations for ULpr and LLpr, the crosslinguistic group differences are still significant (p = .0300 for ULpr and p = .0188 for LLpr). When interpreting the unique TTht results for Subject 5 in Figure 3.1, it should be pointed out that although she is monolingual English, her language background is quite different from the other monolingual English subjects. Her parents speak Frisian and Dutch together, although the home language of her childhood was English. Also, she lived in Ontario for the first 11 years of her life and then lived in the United States for 10 years until the age of about 21, and this may have influenced her ISP tongue tip setting. Note in Appendix X I that her other tongue measurements are not out of the ordinary - it is simply her tongue tip position that is significantly higher than the other English subjects' tongue tip positions. It is interesting to note that Subject 7, who has the highest TTht value of the remaining six English subjects, grew up in Ontario like Subject 5, both further east in Canada than the other five English subjects. Due to the limited number of subjects, no conclusions can be made about differences between the ISP of speakers from Ontario and speakers from further west, but with more subjects from various parts of the country, differences in ISP within the broad category of "Canadian English" could be pinpointed. It is worth noting that if Subject 5 is eliminated from consideration when doing t tests on the TTht data, thep value changes from .0340 to .0528, pushing the difference to insignificance at/? < .05. However, if along with Subject 5, Subject 11 (the most extreme outlier on the French side) is also eliminated, then the crosslinguistic difference in tongue tip height remains significant and the p value actually gets stronger at £ = .0177.  3.2.2. Comparison of Results to Long-held Impressions of AS Recall from Section 1.2.1 that although there exist many published impressions of A S , none of them specifically describe Canadian English or Quebecois French. Thus, any comparisons of the present results to long-held impressions of A S should be taken with a 73  grain of salt until further non-instrumental work has been done on Canadian English and Quebecois French. Sweet's (1890) description of the tongue tip and tongue body being lower in RPEnglish than in Parisian-French is opposite to the results of Section 3.1.1, where it was found that the tongue tip is higher for Canadian English than for Quebecois French. No difference in tongue body height was found, although the tendency in the present data is in the opposite direction to Sweet's impressions. Graff (1932) and Heffner (1950) both described similar differences between RP-English and Parisian-French as Sweet did, at least as far as the tongue tip is concerned, and thus their descriptions also differ from the results of Section 3.1.1, but again for different dialects. Heffner (1950) did say that the tongue is even lower for American-English than it is for British-English (and thus for Parisian-French), and this would make his impressions of the AS for the AmericanEnglish tongue very different from the results of this dissertation for the CanadianEnglish tongue. As for the lips, although Sweet (1890) stated that they are in neutral position for English, he did not describe their position for French, only that they "articulate with energy". Graff (1932) described the lips as being ready for frequent rounding, but it is unclear what this means as far as their ISP is concerned. Honikman's (1964) detailed description of RP-English versus Parisian-French differs somewhat from the studies described above. As stated in Section 1.2.1, Honikman's impression was that the French tongue tip is lower, tongue body is higher, lips rounded, and jaw more open than in English. The results in Section 3.1.1 of this dissertation support only her claim of the French tongue tip being lower. The results are opposite for the tongue body height (English tends to be higher, though not significantly), and there is no significant difference for jaw height. As for the lips, it is usually the case that when one talks about lip rounding, lip protrusion is implicitly assumed. However, Zerling (1992) shows that frontal lip shape for French and English vowels is much more complex than first imagined, and that [+round] vowels can actually involve flat lips in an effort to reduce the cross-sectional area of the lip opening. Thus, depending on what Honikman meant by "lips rounded" for French, these results may agree with her impressions i f she meant "flat", or disagree with her impressions i f she meant "protruded". 74  The only similarity between Esling and Wong's (1983) description of AmericanEnglish AS and the Canadian-English results here is that of a palatalized tongue body position. The results here indicate that the English tongue body tends to be higher than the French tongue body, although it cannot be said with certainty that the tongue body is "palatalized". Overall, it can be seen that the results of this experiment do not match the longheld impressions of English and French AS by phoneticians. While these results may be taken as an indication that a language's ISP may not accurately reflect its AS in its entirety, this difference is not surprising given the fact that none of the impressions in the previous literature were of the dialects under investigation in this dissertation. In future work, when non-instrumental studies are done on the AS of Canadian English and Quebecois French, they will provide an effective measure as to how close the ISP mirrors the A S .  3.2.3. Discussion Regarding  Test of Hypothesis  2  The carry-over effect of phonetic context on ISP was examined systematically. First a detailed analysis was carried out on the four components of ISP in which a crosslinguistic difference was found (i.e. TTht, ULpr, LLpr, and Lnar). It was expected that if the cross-linguistic differences only showed up in a few of the phonetic contexts, the differences might have in fact been caused by the context instead of being language specific. However, Table 3.3 shows that for each of the four significant components of ISP, at least four of the six broad phonetic contexts showed cross-linguistic differences four contexts for TTht, five for ULpr, five for LLpr, and all six contexts for Lnar. Thus, based on these group means in different phonetic contexts, it is unlikely that the crosslinguistic group differences presented in Section 3.1.1 were caused by phonetic context. The four t tests that were not significant in Table 3.3 all have plausible explanations. The two broad phonetic contexts in which TTht is not significantly different across languages (namely, the BackV context and the LowV context) are both back vowel contexts. In such contexts, the tongue tip presumably remains low and out of the way 75  while a constriction is being made with the tongue dorsum and/or tongue root in the posterior part of the oral tract, and since it is not active in the articulation there should not be a significant difference across languages. In addition, the Schwa context is the one context in which ULpr and LLpr do not show significant crosslinguistic differences. Perhaps here, if French schwa is produced "with noticeably rounded lips" (Price, 1991, p. 77), and i f lip rounding entails lip protrusion, then the underlying AS of greater lip protrusion for English (from the results in Table 3.2) is similar to the schwa's demand for greater lip protrusion for French, and this eliminates any significant difference between the French ISP and the English ISP following schwa. In the test of Hypothesis 2, which proposes that phonetic context has a carry-over effect on a given individual's ISP, Figures 3.2 and 3.3 show mixed results. Hypothesis 2 was partially supported in that carry-over effects of phonetic context do exist, but not for the majority of components of ISP. Figure 3.3 shows that for 7 of the 12 components of ISP (i.e. for all 4 tongue components of ISP, both lip protrusion components of ISP, and the height of the upper lip), there is no clear pattern of a measurement being significantly greater after one context than the other - at most only two subjects show similar significant differences for a given language. Although there was no systematic phonetic context effect on the position of the tongue, there was an effect on the position of the lips, but curiously only for French speakers and not for lip protrusion. Five of the eight French speakers had LLIo greater after the CoronalC context than after the BackV context, meaning that the lower lip was at a lower height (relative to the bridge of the nose) after coronal consonants than after back, rounded vowels. This makes sense as the rounding of the vowels involved a reduction of vertical lip aperture via raising of the lower lip. The same five speakers had a significantly greater vertical lip aperture after the CoronalC context compared to the BackV context. Interestingly this same pattern (LLIo and Lvap greater after the CoronalC context) only happened with one of the seven English speakers, and the opposite happened for LLIo with two of the other six English speakers. It is possible that this is an indication of a tighter degree of constriction in the French back, rounded vowels than the English ones, something that Zerling (1992) also reported based on his own EuropeanFrench data and the American-English data of Fromkin (1964) and Linker (1982). 76  In addition to LLIo and Lvap being context dependent for French speakers, Lhap and Lnar (horizontal lip aperture and degree of lip narrowing from a fully spread position) also showed identical contextual effects for three of the eight French speakers. In these three speakers, Lhap was greater after the BackV context than after the CoronalC context, and Lnar was greater after the CoronalC context than after the BackV context. Taken together with the fact that only 1 of the 8 French subjects showed context-related differences for upper lip and lower lip protrusion, this seems to indicate that either Quebecois-French back, rounded vowels are produced with lip spreading as opposed to lip protrusion, or, if they are produced with lip protrusion, then overcompensatory lip spreading occurs when returning to ISP after the rounded vowel. Subject 10 had a significant difference in the opposite direction though (i.e. significantly less lip spreading after back rounded vowels than after coronal obstruents). Due to this subject's age difference with the rest of the subjects (in her 50's as opposed to her twenties or teens), it is possibly an age related difference with older people producing rounding by protruding, while younger people produce rounding by spreading their lips and decreasing the vertical aperture. It should also be noted that Subject 10 remarked about her own pronunciation that some Quebecois have heard her speaking and have asked her where (outside Quebec) she is from. There is a part-whole problem (Barry, 1983; Munhall, 1985; Benoit, 1986) that should be mentioned when analyzing the above results for the French speakers' lower lip height (LLIo), vertical lip aperture (Lvap), horizontal lip aperture (Lhap), and degree of lip narrowing (Lnar). Since LLIo probably accounts for much of the variation of Lvap, it is likely that i f one were to remove the cross-context difference in LLIo from that of Lvap, the cross-context differences in Lvap would no longer be significant. The same can be said for Lhap and Lnar. This concern was not addressed here, but is left for future research. As for jaw height, three out of the seven English speakers had a lower jaw after the BackV context than after the CoronalC context. This is completely logical, given that the vocal tract is more open (i.e. the jaw is lower) when producing vowels than when producing coronal consonants. However, the opposite was true for two of the seven French speakers for whom data was available. Due to the apparently narrower labial 77  constriction for back, rounded vowels in French (see above), perhaps the jaw is raised to allow the lower lip to make a constriction more easily. A speaker-specific factor that may have had an effect on the apparent degree.of influence of phonetic context on ISP is the speed at which each subject spoke. In Figure 3.2, the English subject with the greatest number of differences in ISP between the two phonetic contexts (Subject 6) is also one of two subjects who seemed to speak the slowest. He sometimes barely had time to finish one sentence before the next one was automatically presented to him to be read. In French, Subject 8 spoke noticeably faster than the other subjects and she had one of the smallest number of differences in ISP between the two phonetic contexts. Thus, it seems like speed of speech may influence the degree of context effects on ISP. However, the other English subject who spoke noticeably slowly (Subject 3) had the smallest number of differences in ISP between the two phonetic contexts. So, while it might be possible that for Subjects 6 and 8, speed of speech caused the phonetic context to have a greater effect on ISP than for most other subjects, it was certainly not the case for Subject 3. The fact that ISP was found to be sensitive to carry-over effects of phonetic context is not surprising. Given that Hamlet and Stone (1981) found that jaw ISP is sensitive to anticipatory effects of phonetic context, it is possible that carry-over effects would also exist, and that they would exist for the tongue and lips as well. What is surprising is that no carry-over context effects were found for the tongue - only for the jaw and lips. It is perhaps also surprising that the phonetic effects on the jaw and lip ISP 7  were not the same across language groups. Since the decision on which frames to analyze as ISP frames was based solely on the lack of movement of the tongue, it is possible that the lips and jaw had not come to rest yet even though the tongue had. If this were the case, it might explain why the lips showed more phonetic context effects than the tongue in French, and why the jaw showed more than the tongue in English - the tongue and jaw had not yet moved as far away from the configuration they had been in for the sound preceding the ISP. However, that still does not explain why the lips did not show a similar effect of phonetic context in English, or the jaw a similar effect in French. In future Note, however, that the results at the end of Section 3.1.3 show directionality differences across the two phonetic contexts for all components of the tongue's ISP for French speakers and for the tongue tip and tongue root for English speakers. 7  78  studies it may be best to have articulator-specific criteria for what is counted as ISP. Thus, tongue measurements could be made when the tongue stops moving, and likewise for the lips and jaw. It may turn out that there is a time when all three are motionless, but this is an empirical question that is left for future research.  3.3. Summary of Chapter III  In Chapter III, the results were presented of Experiment 1, an investigation of whether or not ISP is language dependent (in a balanced phonetic context), and whether or not phonetic context has a carry-over effect on ISP (within a given monolingual speaker's speech). Hypotheses 1 and 2 were tested, namely that the ISP for Canadian English is significantly different from the ISP for Quebecois French, and that within a given speaker's speech in a given language, that speaker's ISP differs depending on the phonetic segment that precedes the ISP. Results support Hypothesis 1, but only for four components of ISP. For the tongue tip, the mean ISP for the monolingual English group was higher than that for the monolingual French group. This matches the findings of Gick et al. (2004) but is contrary to all of the existing non-instrumental evidence on the AS of RP English versus Parisian French. For upper and lower lip protrusion, again the mean ISP for the monolingual English group was higher (i.e. more protruded) than that for the monolingual French group. This is in accordance with Gick et al.'s findings for the upper lip, but not for the lower lip. For the degree of lip narrowing compared to a fully spread position, once again the mean ISP for the monolingual English group was higher (i.e. the corners of the mouth drawn in more in English from a fully spread position) than that for the monolingual French group. Results in Chapter III also support Hypothesis 2, but only for some components of ISP, and those components are different in each language. Specifically, in English, only the jaw's ISP is influenced by phonetic context, but this is only in the speech of three of seven speakers. For the English speakers, no systematic effects of phonetic context were found for the ISP of the tongue or lips. In French, four components of ISP are influenced 79  by phonetic context: the height of the lower lip and the vertical lip aperture for five speakers, and the horizontal lip aperture and the degree of lip narrowing for three speakers. For the French speakers, no systematic effects of phonetic context were found for the ISP of the tongue or jaw.  80  CHAPTER IV Experiment 2: AS in English-French Bilinguals  In Experiment 2, Hypotheses 3 and 4 were tested to broadly determine how a bilingual's pronunciation proficiency relates to his or her ISP(s) and how speaking mode affects a bilingual's ISP. Hypothesis 3 stated that a bilingual who is perceived as a native speaker of both languages has a different ISP for each language and will show the same types of crosslinguistic ISP differences that monolingual groups show; conversely, a bilingual who is perceived as not being a native speaker of at least one language will have fewer, if any, of the crosslinguistic ISP differences that monolingual groups show. Hypothesis 4 stated that bilingual speakers who are perceived as native speakers of each of their two languages have a unique bilingual-mode ISP that differs in all significant respects from both monolingual-mode ISPs (where "significant respects" are those respects in which differences obtain between the two monolingual modes).  4.1. Results  Results of Experiment 2 on bilingual subjects are now presented. In Section 4.1.1, the ISP for English monolingual mode is compared to the ISP for French monolingual mode. This comparison is done for all bilingual subjects in a test of Hypothesis 3. In Section 4.1.2, within the subset of bilinguals who were perceived as native speakers of both languages, ISP for bilingual mode is compared to ISP for monolingual mode in a test of Hypothesis 4.  4.1.1. Results: English Versus French (Monolingual Mode)  Each of Figures 4.1 to 4.3 shows a box plot containing the distribution of values at one of the components of ISP for all nine bilingual subjects in monolingual mode. Specifically, Figure 4.1 shows a box plot of values for tongue tip height; Figure 4.2 shows amount of jaw lowering, and Figure 4.3 shows lower lip protrusion. For each subject, the  81  distribution of values is plotted for both monolingual English mode (e.g. " 2 I E " for Subject 21) and monolingual French mode (e.g. "2IF" for Subject 21). As with the box plots in Figure 3.1, the top, bottom, and line through the middle of each box correspond to the 75 percentile, 25 percentile, and 50 percentile (i.e. the median) respectively. The th  th  th  whiskers on the bottom and top extend from the 10 percentile and 90 percentile th  th  respectively. To allow for comparison, the dotted lines in each box plot show the group means for the French and English monolingual subjects. Figures 4.1 to 4.3 are given to show the individual variation within the bilingual subjects for one tongue, one jaw, and one lip measurement, and to show how their data compare to the monolingual group means. Results of statistical analyses of this data are presented following these figures.  Figure 4.1. Box plot of distribution of tongue tip height values for all 9 bilingual subjects in monolingual mode for both English and French Participant number 21E 21F 17E 17F 22E 22F 19E 19F 18E 18F 23E 23F 20E 20F 24E 24F 16E 16F 80  . 1  1  1  I  I  1  i  1  '11  -I  1  -  S ?!  T  t^y English monolingual T  B  -&60  1  t  |  $*•  M  participants' mean  T  65  ..[. w  N  & 9  55  Native in both Eng & Fre  Native in Native in;  Native in Fre only  Eng only  45  ' 1  1  1  1  1  1  I  60 55 French monolingual  50  40  80  T • 75  75 70  1—  I  i  i  i  i  i  i  i  i  neither '.  i  45  i ' 40  21E 21F 17E 17F 22E 22F 19E 19F 18E 18F 23E 23F 20E 20F 24E 24F 16E 16F Participant number  82  50  participants' mean  Figure 4.2. Box plot of distribution of amount ofjaw lowering for all 9 bilingual subjects in monolingual mode for both English and French Participant number 21E 21F 17E 17F 22E 22F 19E 19F 18E 18F 23E 23F 20E 20F 24E 24F 16E 16F 20  15  10 'J  . French monolingual participants' mean  5 * English monolingual participants' mean 9  Native in Native in  Native in Fre only  Native in both Eng & Fre  Eng only  neither  21E 21F 17E 17F 22E 22F 19E 19F 18E 18F 23E 23F 20E 20F 24E 24F 16E 16F Participant number  Figure 4.3. Box plot of distribution of lower lip protrusion values for all 9 bilingual subjects in monolingual mode for both English and French Participant number 21E 21F 17E 17F 22E 22F 19E 19F 18E 18F 23E 23F 20E 20F 24E 24F 16E 16F 45  45  40  40 English monolingual  f 35  35  3 30  30  *»  8 a.  •I 25  T  a 25  T 20  20  T  15 Native in both Eng & Fre  Native in Fre only  Native in Native in Eng only  15  neither  10  10 21E 21F 17E 17F 22E 22F 19E 19F 18E 18F 23E 23F 20E 20F 24E 24F 16E 16F Participant number  83  participants' mean  French monolingual participants' mean  Two things from Figures 4.1 to 4.3 are especially noteworthy. First, it is evident that there was no clear pattern as to how the monolingual groups' mean ISPs (as indicated by the dotted lines) compared to the ISPs of the different types of bilinguals. Second, the results for Subject 16, perceived as a native speaker of neither English nor French, were anomalous both from the perspective of her tongue tip height measurements and her lower lip protrusion variability. These points will be discussed further in Section 4.2. The results in Figures 4.1 to 4.3 above show each bilingual subject's data distribution for three components of ISP, but they do not directly test Hypothesis 3. To test this hypothesis, t tests (assuming unequal variances) were carried out for each component of ISP to compare the monolingual-mode English means to the monolingualmode French means for each bilingual subject. The dependent variable was the individual mean measurement of the ISP position of a given articulator for a given subject. The independent variable was the language of the monolingual-mode stimuli set (i.e. English versus French). Results of these 96 t tests (12 components of ISP for 8 subjects monolingual-mode English data was not available for Subject 17) can be seen in Table 4.1 below. Grey shading indicates individual differences in these bilingual subjects that were the same as the four differences found between the monolingual groups in Section 3.1.1. A striped border surrounding a cell indicates significant differences that were opposite to the monolingual-group differences in Section 3.1.1.  84  Table 4.1. Significant differences (p < .05) between the ISP in French (F) and English (E) monolingual modes. The symbol "-" indicates no significant difference and "n/a" indicates data not available. Cell shading indicates identical results to monolingual-group differences, while striped cell borders indicate opposite results from monolingual-group differences. JAW  LIPS  Subj  TTht  TBht  TDht  TRrn  JAW1  ULlo  LLlo  ULpr  LLpr  Lvap  Lhap  Lnar  21  E>F  -  -  F>E  E>F  -  E>F  E>F  E>F  E>F  -  -  E>F  E>F  E>F  E>F  -  17  (n/a)  22  -  -  -  -  F>E  n/a vmmmmm  -  E>F  E>F  -  F>E  -  F>E  18  -  E>F  E>F  -  F>E  F>E  F>E  23  E>F  E>F  -  -  -  -  -  E>F  E>F  -  -  -  F>E  -  -  -  -  -  -  -  -  -  -  -  24  -  -  -  F>E  E>F  -  -  -  -  -  F>E  E>F  16  -  -  F>E  F>E  E>F  F>E  E>F  -  -  E>F  E>F  Perceived as Fre only  19  Perceived as Eng only  -  Perceived as Neither  Perceived as Both  TONGUE  20  i  2  E>F i  F>E  F>E  E>F 1  i  iA F>E 5i -  E>F  i F>E t  -  I  F>E  Thus, for example, Subject 21 showed a significant difference between the English ISP and the French ISP for 7 of the 12 components of ISP. Three of these seven differences, namely for TTht, ULpr, and LLpr, were the same differences that were found between the French monolingual group and the English monolingual group in Section 3.1.1. The remaining four of the seven differences, namely for TRrn, JAW1, LLlo, and Lvap, are components of ISP that had shown no significant differences between the monolingual groups. In Table 4.1, it is readily apparent that for all eight subjects for whom data was available, there was at least one component of ISP where the English ISP was different from the French ISP. Subject 19 had the greatest number of differences - 9 out of 12. 85  !  Subject 20 had the smallest number of differences - only 1 out of 12. A l l 3 subjects who were perceived as native speakers of both languages had 5 or more differences out of 12. In addition, all 3 subjects had greater upper and lower lip protrusion for the English ISP compared to the French ISP, the same difference that was found between the English and French monolingual groups in Section 3.1.1. Only one of the other five subjects had greater upper and lower lip protrusion for the English ISP, Subject 23, who could possibly be classified as a native speaker of both languages (and was indeed perceived to be a native speaker by two native listeners - see discussion in Section 4.2.1 below). In Table 4.1, also note that three of the four subjects (18, 20, and 16) who had significant differences that were in the opposite direction to the monolingual group differences in Section 3.1.1 were all bilinguals who were either perceived as being native speakers of only one language or of none. These three speakers not only had significant differences that were opposite to the monolingual group differences, but they also had no significant differences that were in the same direction to monolingual group differences. The fourth subject who had a significant difference in the opposite direction to the monolingual group differences was Subject 19, perceived as a native speaker of both English and French. He had greater horizontal narrowing of the lips in French than in English. As for Subject 16, who was perceived to be a native speaker of neither English nor French, note that she had eight significant differences between the English ISP and the French ISP, more differences than two of the bilinguals perceived as native speakers of both languages had. None of these eight are in the same direction as the monolingual group differences though, and one is in the opposite direction, as mentioned above. The other seven are all at components of ISP where no significant differences were found across monolingual groups in Section 3.1.1. However, out of these seven, three (i.e. TDht, ULlo, and Lvap) are significant in the opposite direction to tendencies, but not significant differences, in the monolingual group data.  86  4.1.2. Results: Bilingual Mode Versus Monolingual Mode  In order to test Hypothesis 4, ISP during bilingual mode was compared to ISP during monolingual mode for the subjects who were perceived as native speakers of both English and French (i.e. Subjects 21, 22, and 19), as well as for Subject 23 who was perceived to be a native speaker by two listeners. Recall that in bilingual mode the subjects had no way of knowing what language the next sentence to read would be in, whereas in monolingual mode the language of the stimuli was kept constant and the subject was aware of this. Hypothesis 4 proposed that bilingual speakers who are perceived as native speakers of each of their two languages have a unique bilingual-mode ISP that differs in all significant respects from both monolingual-mode ISPs (where "significant respects" are those respects in which differences obtain between the two monolingual modes). The results in Section 4.1.1 indicate that "significant respects" here means ULpr and LLpr for all four subjects, as well as TTht for Subjects 21 and 23, and Lnar for Subject 19. Prior to conducting a test of Hypothesis 4, it was necessary to contend with the problem, mentioned in Section 2.3.2 and illustrated in Table 2.5, of having a small number of tokens available for each mode-language pairing (i.e. bilingual-mode English, bilingual-mode French, monolingual-mode English, monolingual-mode French). With such a small number of tokens per subject in each of these four categories, it was likely that some statistical differences between modes might not have been discovered. Thus, in order to prepare the data such that statistical power was increased, it was desirable to combine categories where possible. Since monolingual-mode English ISP was shown to be significantly different from monolingual-mode French ISP in Experiment Section 4.1.1, these two monolingual-mode categories could not be combined. However, the two bilingual-mode categories (i.e. bilingual-mode English and bilingual-mode French) had not been compared to see whether they could be combined in order to increase statistical power for the test of Hypothesis 4 across modes. Thus, bilingual-mode English ISP was compared to bilingual-mode French ISP by performing 15 t tests (assuming unequal variances) where the dependent variable was the individual mean measurement of the ISP position of a given ISP component (i.e. TTht, ULpr, LLpr, or Lnar) for a given subject  87  (i.e. Subjects 21, 22, 19, and 23). The independent variable was the language (English or French) of the sentence preceding the ISP. Results of the t tests indicated no significant differences at a level ofp < .05. Thus, the two categories of bilingual-mode data were combined for each subject and then a test of Hypothesis 4 was carried out. The combined bilingual-mode dataset for each subject was compared to that subject's set of monolingual English data and the set of monolingual French data, in turn. Table 4.2 shows a summary of the outcome of 30 t tests (assuming unequal variances) comparing the combined ISP for bilingual mode with the ISP for each language's monolingual mode. There were 30 tests because there were 2 monolingual-mode data sets (one English and one French) for each of the 1.6 subject-measurement pairings (4 subjects X 4 components of ISP), minus 2 because Lnar data (in both languages) was unavailable for Subject 22. The dependent variable for these t tests was the mean ISP position of a given component of ISP for a given speaker for a given monolingual-mode language. The independent variable was the stimuli-presentation mode (monolingual versus bilingual).  88  Table 4.2. For each bilingual who was perceived as native in both languages, a comparison of bilingual-mode ISP ("Bil") to each of the monolingual-mode ISPs ("MonoEng" and "MonoFre") for 4 components of ISP. [ "=" means no significant difference between the two; "<" and ">" indicate significant differences (p < .05); "n/a" means data not available ] Subject  21  22  19  23  Tongue tip height  Upper lip protrusion  Lower lip protrusion  Horizontal lip narrowing  (TTht)  (ULpr)  (LLpr)  (Lnar)  MonoEng = Bil  MonoEng > Bil  MonoEng > Bil  MonoEng = Bil  MonoFre < Bil  MonoFre = Bil  MonoFre = Bil  MonoFre < Bil  MonoEng = Bil  MonoEng > Bil  MonoEng > Bil  MonoFre = Bil  MonoFre = Bil  MonoFre = Bil  MonoEng = Bil  MonoEng = Bil  MonoEng = Bil  MonoEng < Bil  MonoFre = Bil  MonoFre < Bil  MonoFre < Bil  MonoFre = Bil  MonoEng = Bil  MonoEng = Bil  MonoEng = Bil  MonoEng > Bil  MonoFre < Bil  MonoFre < Bil  MonoFre < Bil  MonoFre = Bil  n/a  In Table 4.2, note that for every pairing of a subject and a component of ISP, the bilingual-mode ISP was always equivalent to at least one of the monolingual-mode ISPs. It was never different from both language's monolingual-mode ISPs. For the TTht measurement, note that Subjects 21 and 23 showed identical results, with the bilingual-mode ISP being equivalent to the monolingual-mode English ISP but greater than (i.e. having a higher tongue tip than) the monolingual-mode French ISP. Subjects 22 and 19 showed identical results to each other, with no significant differences between the bilingual mode and either of the monolingual modes. These two subjects had one ISP for the tongue tip, whether they were speaking French, English, or in bilingual mode. For lip protrusion, note that the upper lip's results match the lower lip's results for all four subjects. Subjects 21 and 22 pattern together in having a bilingual-mode ISP equivalent to the monolingual-mode French ISP, and Subjects 19 and 23 pattern together in having a bilingual-mode ISP equivalent to the monolingual-mode English ISP.  89  4.2. Discussion  In Experiment 2, ISP was measured in nine bilingual speakers of Canadian English and Quebecois French in order to test Hypotheses 3 and 4. The results in Section 4.1 support Hypothesis 3, showing that bilinguals who are perceived as native in two languages have two ISPs and that the differences between these two ISPs mirror the differences between monolingual groups' ISPs. Results also show that bilinguals who are not perceived as native in at least one language do not have the same significant differences in ISPs. The results in Section 4.1 do not support Hypothesis 4, showing that bilinguals perceived as native in two languages do not have a unique bilingual-mode ISP that differs from both monolingual-mode ISPs in significant respects. As mentioned above, from Figures 4.1 to 4.3, it is evident that there was no clear pattern as to how the monolingual groups' mean ISPs compared to the ISPs of the different types of bilinguals. For example, it is not necessarily the case that bilinguals perceived as native speakers of both languages have ISPs that fall between the two monolingual group means. Nor is it necessarily the case that bilinguals who are only perceived as native speakers of French or English have ISPs that are closer to the French or English monolingual groups, respectively. In addition, Figures 4.1 to 4.3 illustrate the anomalous results for Subject 16, perceived as a native speaker of neither English nor French. In Figure 4.1, the ISP for the tongue tip in both her languages is very different (much higher) than the tongue tip for all the other subjects. Although not shown in Section 4.1 in figures, the same is true for the other three of her tongue components of ISP (i.e. TBht, TDht, and TRrn). Her tongue's ISP is closer to the opposing vocal tract surface for all four of these tongue components of ISP. In Figure 4.3^ for all subjects except Subject 16, lower lip protrusion is more variable in French than in English. Although not illustrated with a figure, the same is true for upper lip protrusion - it is more variable in French than in English for all subjects except Subject 16. It is possible that these factors could contribute to her not being perceived as a native speaker of either of her languages.  90  4.2.1. Discussion  Regarding  Test of Hypothesis  3  In Table 4.1, note that all three bilingual subjects who were perceived as native in two languages, and for whom data was available, showed a greater upper and lower lip protrusion in English than in French. These match two of the four significant differences between the two monolingual groups. Subject 23 also shows these differences and, although she was not perceived to be a native speaker of English, the native listener judgements of her English speech were very mixed. She was perceived to be a native speaker by 2 out of 10 judges, a near-native speaker by 5 out of 10 judges, and adequate, but not near-native by 3 out of 10 judges (see Appendix V). Since reasons for the judgements were not collected, it was impossible to determine if there was an anomalous reason for them such as a slightly different intonation pattern or one sound that was slightly mispronounced, but it is clear from her language background (see Appendix III) that she was exposed to a balance of both French and English from an early age at home. Thus, her results were also included with the results of the bilinguals perceived as native speakers of both languages in Section 4.1 and they will be included in the discussion here. Hypothesis 3 stated that a bilingual who is perceived as a native speaker of both languages has a different ISP for each language and will show the same types of crosslinguistic ISP differences that monolingual groups show; conversely, a bilingual who is perceived as not being a native speaker of at least one language will have fewer, i f any, of the crosslinguistic ISP differences that monolingual groups show. The results in Table 4.1 fully support this hypothesis. A l l subjects who were perceived to be native speakers of both of their languages had at least two, and at most three of the four crosslinguistic differences that the monolingual groups showed in Experiment 1. Of the other four subjects who were perceived not to be a native speaker of one or both of their languages, only one subject (24) showed a difference that matched the monolingual group differences found in Experiment 1. That difference was for the ISP of Lnar, the degree of lip narrowing from a fully spread position. In addition, three of the four subjects who were perceived to be non-native in at least one language had at least one crosslinguistic difference that was opposite to the monolingual group differences. This only happened  91  once with one of the bilinguals who was perceived as native in both languages, and this was for the ISP of Lnar. A l l subjects except Subject 22 showed at least one and at most two differences across languages in tongue components of ISP. However, the type and direction of the differences was not systematic, especially within each of the four groups of bilinguals. Overall, the one common thread between all bilinguals who were perceived as native speakers of both of their languages was a greater upper and lower lip protrusion for the English ISP than for the French ISP. It appears that in bilinguals who are native in both languages, this component of ISP must mirror crosslinguistic differences in monolingual speakers. As two of the four bilinguals perceived as native in both languages had a higher tongue tip in English, just like the monolinguals in Experiment 1, it seems that this is a salient measure of proficiency, but one of secondary importance compared to lip protrusion. Subject 20, perceived as a native speaker of French only, had a higher tongue tip in French, the opposite of the monolinguals in Experiment 1. Although it is not possible to determine with certainty the reason she was not perceived as a native English speaker, it is possibly due to her opposite tongue tip setting, but certainly also is due to her lack of a difference in lip protrusion across languages. For lip protrusion, this same reason can be used with all the bilinguals who were not perceived to be native in one or more of their languages. As for Lnar, the degree of lip narrowing from a fully spread position, results were very mixed and thus it can be concluded that this component of ISP is probably not an important factor in whether a bilingual is perceived as a native speaker of a given language or not.  4.2.2. Discussion Regarding  Test of Hypothesis 4  Hypothesis 4 was tested and the results presented in Section 4.1.2. This hypothesis was not supported by the data. In Table 4.2, for every cell in the table (i.e. every subjectcomponent pairing), bilingual-mode ISP was the same as at least one of the monolingualmode ISPs. It was never different from both language's monolingual-mode ISPs. This is a clear refutation of Hypothesis 4 because Hypothesis 4 predicts that there should exist a  92  unique bilingual-mode ISP for TTht for Subjects 21 and 23, for ULpr and LLpr for all four subjects, and for Lnar for Subject 19. As mentioned in Section 4.1.2, Subjects 21 and 22 pattern together in having a bilingual-mode ISP equivalent to the monolingual-mode French ISP, and Subjects 19 and 23 pattern together in having a bilingual-mode ISP equivalent to the monolingual-mode English ISP. It is interesting to note that for all four of these subjects, the ISP for lip protrusion for bilingual mode resembled the monolingual-mode language that was dominant in the subject's daily use at the time of the experiment. For subjects 21 and 22, French was the dominantly used language. Subject 21 was living in Vancouver but used exclusively French at home with her children and husband. She only used English at her part-time job and with some friends outside of the home. Subject 22 was living in Montreal and was working in French. For Subjects 19 and 23, English was the dominantly used language. Both subjects were living in Vancouver at the time of the experiment. Subject 19 judged his typical week to be 90% English - school and work were both entirely in English. The only chances he had to speak French were with his parents and siblings. Subject 23 judged her typical week to be 70% English. Thus, for the most salient components of ISP (ULpr and LLpr), all four subjects have bilingual-mode ISPs that are equivalent to the ISP of their dominantly used language at the time of the experiment. Perhaps this was because the bilingual-mode task was more complex than the monolingual-mode task, and so these subjects simply chose the ISP they were most habituated to using at the time. Notice that there was no relationship between these subjects' L I and the language their bilingual mode resembled. The L I of Subject 21 was English, and that of Subjects 22, 19, and 23 was French. Lnar showed mixed results, but note that its importance for the bilinguals is negligible. In Table 4.1, it was clear that the bilinguals followed the monolingual group differences in their lip protrusion and for some subjects in their tongue tip height, but not in their degree of lip narrowing. Thus, it is not surprising to see no clear pattern emerging from the Lnar results for these bilinguals.  93  4.3. S u m m a r y of C h a p t e r I V  In Chapter IV, the results were presented of Experiment 2, an investigation of (1) whether or not a bilingual's ISP is language dependent (in a balanced phonetic context in monolingual-mode speech), and (2) whether or not a bilingual speaker's ISP is mode dependent (i.e. different for monolingual-mode speech versus bilingual-mode speech). Hypotheses 3 and 4 were tested, and results strongly support Hypothesis 3, but do not support Hypothesis 4. Specifically, upper and lower lip protrusion were greater for the English ISP than for the French ISP, in all bilinguals who were perceived as native speakers of both of their languages (and Subject 23), but in none of the other bilinguals. Tongue tip height was also a salient component of ISP for bilinguals perceived as native in both languages, but it was of secondary importance compared to lip protrusion. As for the degree of lip narrowing from a fully spread position, this component of ISP was concluded not to be an important factor related to a bilingual's proficiency in Canadian English and Quebecois French. As for bilingual mode versus monolingual mode, Hypothesis 4 was not supported, meaning that bilingual-mode ISP was not uniquely different from both monolingual-mode ISPs. However, an interesting finding was that for the most salient ISP components (ULpr and LLpr), all four subjects have bilingual-mode ISPs that are equivalent to the ISP of their dominantly used language at the time of the experiment.  94  CHAPTER V General Discussion and Conclusions  5.1. G e n e r a l Discussion  Results of Experiment 1 showed that, for monolinguals, ISP is sensitive to the language being spoken, as well as being sensitive to the phonetic context preceding the ISP. The four components of ISP that show language-specific differences are tongue tip height, upper and lower lip protrusion, and the degree of horizontal lip narrowing compared to a fully spread position. The components of ISP that show sensitivity to phonetic context are the amount of jaw lowering in English (in three of seven subjects), the height of the lower lip and the vertical lip aperture in French (in five of eight speakers), and the horizontal lip aperture and the degree of lip narrowing, also in French (in three of eight speakers). Results of Experiment 2 showed that, for bilinguals, ISP is once again sensitive to the language being spoken, and also that it is sensitive to speaking mode, in that bilingual-mode ISP is identical to the monolingual-mode ISP of the speaker's dominanfly used language at the time. For bilinguals who are perceived as native speakers of both of their languages, their two ISPs differ from each other in three of the four ways that the monolingual groups differed in Experiment 1. In addition to confirming Hypothesis 3, these within-speaker bilingual results validate the method and normalization used across speakers in Experiment 1. Mirroring the two monolingual groups, upper and lower lip protrusion for all 4 of these bilinguals were greater in English than in French. However, none of the bilinguals who had an accent in one or both languages showed this lip protrusion difference between languages. Tongue tip height differences between languages also had a tendency to mirror the two monolingual groups, with the tongue tip being higher in English ISP than in French ISP for two of the four bilinguals perceived as native in both languages. Again, none of the bilinguals who had an accent in one or both languages showed this tongue tip height difference between languages, and in fact one showed a difference in the opposite direction. The degree of horizontal lip narrowing compared to a fully spread position is not a salient difference across languages in order to 95  be perceived as a native speaker of both languages - none of the four bilinguals perceived as a native speaker of both languages had this difference, and in fact one had the opposite difference from the monolingual groups. In Experiment 1, it was shown that upper and lower lip protrusion are the components of ISP that are least affected by phonetic context. It is interesting to note that these are the two strongest differences in ISP across monolingual language groups and they are also the two differences that are shared in common among every bilingual speaker perceived to be native in both languages. As mentioned in Chapter 1, Laver (1980) suggested that a language's AS can be overridden by the requirements of a particular sound segment in a language. Since upper and lower lip protrusion are least affected by phonetic context and the most salient differences across Canadian English and Quebecois French, it appears that lip protrusion is a component of AS that is less apt to be overridden by the articulatory requirements of individual sound segments. On the other hand, based on the effects of context on ISP observed in Experiment 1, degree of lip narrowing from a maximally spread position is more apt to be overridden by the demands of individual segments, at least in French. With the tongue only coming to rest in about half of the pauses between sentences (55% for English subjects, 43% for French subjects, and 57% for bilingual subjects), one may argue that the ISP is not a target configuration, contrary to what Gick et al. (2004) found for English and French. However, there are at least two factors to be considered before jumping to this conclusion. First, many of the tokens where the tongue failed to reach a speech rest position did so because of non-linguistic events such as swallowing. Second, because of the speed and automaticity of the stimuli presentation style in Experiments 1 and 2, it was often the case that a subject did not finish saying a previous sentence before the next one was presented (leaving no time to put the system into a rest posture). The consequences of this second factor were also seen when a speaker made a speech error that he or she decided to correct, leaving no time to pause between sentences. In an initial pilot study, a 2-second pause was used but it was found that some subjects had too much time and were bracing their tongues against their palates, possibly indicative of an absolute rest position, or a swallow, instead of a speech rest position. A 1-second pause was chosen for the present study to reduce this type of occurrence. If, 96  contrary to the findings in the second experiment of Gick et al. (2004), ISP does not behave like a speech target, the phonetic context should have a very noticeable effect on ISP as there would be no reason to move one's articulators between stimuli, other than simply to relax one's muscles. Since, especially in English, phonetic context does not have a clear effect on ISP (except for a weak effect on jaw posture), this is consistent with the view that ISP behaves like a speech target in English. To further test this question, in future studies, velocity profiles of the articulators can be examined to determine whether or not movement into ISP is systematically similar in velocity to movement into a speech target. In Experiment 1, the components of ISP with the greatest crosslinguistic similarities were the position of the tongue root (p = .9848 across the two monolingual groups) and the jaw (p = .9254 across the two monolingual groups). Note that these two articulators somewhat determine the position of some of the other articulators. Specifically, the tongue is resting on the jaw and hence jaw height will have a strong effect on tongue height. Also, because of the hydrostatic nature of the tongue, the degree of tongue root retraction can have a great effect on the height of the tongue body. Perhaps then, the jaw and tongue root are grossly positioned (and English and French have similar gross positions for these) and then the finer adjustments are made by the rest of the components of the tongue and the lips. Note that since jaw height was not significantly different across the two languages in Experiment 1, the difference found in tongue tip height had nothing to do with the jaw.  5.1.1. Implications  of this Research  One important implication of these results is for the field of L2 acquisition, especially pronunciation teaching and learning. In the last 50 years, the methods and status of pronunciation teaching have fluctuated greatly (see Morley, 1991, and CelceMurcia et al., 1996, for thorough reviews), but recently there have been an increasing number of calls for the inclusion of AS in second language teaching curricula (Brown, 1995; Celce-Murcia et a l , 1996; Collins & Mees, 1995, 2003; Esling, 1987; Esling & Wong, 1983; Jenkins, 1998; Jones & Evans, 1995; Kerr, 2000; Mompean-Gonzalez, 97  2003; Pennington, 1996; Pennington & Richards, 1986; Rich, 2003; Thornbury, 1993). These calls and the methods that are used to teach AS exist in spite of the fact that there has been no empirical evidence for language-specific ASs. Studies of L T A S have demonstrated similarities and differences in the acoustics of two different languages, but as mentioned previously, L T A S does not necessarily directly relate to A S , and this acoustic information provided by L T A S is often very difficult if not impossible to map onto articulatory parameters for L2 learners. The results of this dissertation have shown that AS is indeed language specific, and have shown exactly where the relevant differences in AS occur between Canadian English and Quebecois French. These results, along with those of Gick et al. (2004), provide much-needed quantitative evidence to support the teaching of AS. Another implication of the present results is for studies that have used the ISP as a baseline from which to compare and measure components of the postures of various speech sounds. Examples of this type of study include Adler-Bock (2004), who used the default rest position for correcting her before-treatment versus after-treatment ultrasound images of the tongue, McDowell (2004), who also used the rest position across her ultrasound data as a reference for comparing tongue shapes, and Oh (2004), who used the average ISP in her ultrasound images of the tongue and made measurements relative to that one position. The fact that in Experiment 1 none of the components of the tongue ISP were systematically influenced by phonetic context is encouraging for the above studies, as none of them had systematically controlled for phonetic context around the ISP. This research also has implications for the claims of some researchers who equate a language's schwa with that language's AS. This position has been taken despite the fact that Gick (2002) has shown that, contrary to traditional belief, American English schwa has an articulatory target - retraction of the tongue root relative to pre-speech posture. For example, Kuhnert & Fougeron (2004) state that "Broadly speaking, the neutral vowel schwa can be considered as a kind of homebase to which the tongue returns frequently in the course of speech. As such, it can be considered as an indicator of the overall articulatory setting of a language." The fact that European-French schwa is phonetically rounded (Price, 1991) and impressions of Parisian-French AS described in Section 1.2.1 have the lips rounded supports Kuhnert & Fougeron's position. More support comes from 98  Barnes and Kavitskaya (2002), who found significant visible lip rounding remaining on "deleted" inaudible schwas spoken by one speaker of (presumably European) French. If Kuhnert & Fougeron's claim is true, then the results of Experiments 1 and 2 make specific predictions for the differences in articulation of Canadian-English schwa versus Quebecois-French schwa: Canadian-English schwa should have greater lip protrusion and a higher tongue tip than Quebecois-French schwa and should not have the retracted tongue root position Gick (2002) found for American English. A test of these predictions is left for future research. The results from Chapter IV showing that there is no unique ISP for bilingual speech mode (i.e. one that is different from each monolingual-mode ISP) suggest that differences between monolingual mode and bilingual mode (Grosjean, 1998) do not hold at the phonetic level. It is possible that bilingual mode is the norm for a bilingual's AS, and hence their bilingual-mode ISP defaults to the same ISP as the dominant language of their present life. For a bilingual, speaking in the non-dominant monolingual mode is not the norm and perhaps the ISP for this mode must be actively set. Evidence showing that bilinguals reset the phonetic parameters of their languages, depending on the conversational setting and on their proficiency in each language, has come from various studies, e.g., see Flege, Schirru, and MacKay (2003), Sancier and Fowler (1997), and Watson (1990, 1991).  5.2. L i m i t a t i o n s and F u t u r e Directions  There are a number of limitations to be considered in this study, such as the relative accuracy of the measurement systems used, the method of normalizing the data, the criteria for distinguishing between a bilingual and a monolingual, and other issues. These limitations, as well as directions for future studies are now considered below. One limitation of the present study was the relative degrees of accuracy of the Optotrak and ultrasound measurement systems. The spatial resolution of Optotrak is much higher than that of ultrasound. Very small differences (1 mm or less) in lip position can be detected with Optotrak, but these same differences in tongue position may be 99  missed with ultrasound. This difference in spatial resolution may have been a factor in (1) the finding of more crosslinguistic differences in lip ISP than in tongue ISP, and (2) the finding of many more phonetic context effects on the lips than on the tongue. As for the accuracy of the ultrasound system, there have been conflicting reports. Beasley, Stefansic, Herline, Guttierez, and Galloway (1999, p. 132), in an Optotrak calibration study of an ultrasound probe, show that "it is possible to track and (sic) ultrasound probe in space, with errors on the order of 1.0 mm" and that "it is possible to register ultrasound images with physical space, with average target errors on the order of 3.0 mm." On the other hand, Schreiner, Galloway, Lewis, Bass, and Muratore (1998, p. 640) had results that showed ultrasound to be much more accurate: "the ultrasonic system differed from the [Optotrak] pointing system by a mean of 0.5 mm with a 95% confidence interval of +/0.1 mm when localizing the same point in space." As cited by Kaburagi & Honda (1994, p. 2270): "Honda (1985) showed that the tracing error of the tongue contours from the [Bmode] ultrasonic images was 1.1 mm on average by comparison with x-ray pictures taken at the same time as the ultrasonic scanning." It should be pointed out though that x-ray images are not true slices of the vocal tract. They contain shadows that potentially make small measurements unreliable. A major challenge of any phonetic study where group means are calculated based on results from a number of different speakers is how to normalize the data. Honda, Maeda, Hashi, Dembowski, and Westbury (1996, p. 784) point out that the shape of "the space within which articulation takes place [...] is not the same among individuals or races". Although this study has serendipitously controlled for race, individual differences in vocal tract morphology undoubtedly added noise to the results, possibly masking some differences between the A S of each language. However, the normalizing of the data that was done attempted to minimize across-subject differences. In future studies, an MRI or CT scan of each subject's vocal tract could be taken and used for scaling purposes. However, because of the lack of agreement on what anatomical distances are best to use for scaling data, even that method may not provide the best answer. Another limitation of the present research has been the ability to make a systematic statement about the relationship between ISP and the various proficiencies of bilinguals who are not perceived as native in both languages. Because of the multitude of 100  factors that influence the success of a bilingual's acquisition of more than one language (Marinova-Todd, 2003), it is very difficult to control for the background of the subjects such that all factors are balanced. In the present study, because of the difficulty of finding, in Vancouver, relatively balanced bilinguals who are perceived as native speakers of both English and French, sacrifices had to be made as to the degree of similarity between bilinguals in this study. For degree of jaw lowering, the chin marker is attached to the skin on the chin and not to the mandible directly, and it is highly likely that the skin, and therefore the marker, can move (stretch) while the mandible remains stationary. However, this is probably the case mostly when the lower lip is raised to make a labial constriction during speech. It is expected that during ISP, assuming the lips are not closed, the chin marker would represent a reasonably accurate indication of the position of the mandible. Zerling (1992, p. 3) stated that "behind the apparent simplicity of the binary phonological feature [+/round] there lies a complex pattern of activity for the lips". There are a number of measurements one could make to calculate lip aperture. The most salient measurement is probably the area of the opening between the lips, but this is impossible to measure using Optotrak, and even with video it is difficult to determine the correct coronal cross-section at which to measure this area. Using Optotrak, the best possible measurements to make are the width of the opening (very roughly approximated in this case by the distance between the markers at each corner of the mouth) and the height of the opening (very roughly approximated here by the distance between the upper lip marker and the lower lip marker). Takano, Honda, and Dang (2002) give evidence suggesting that lateral tongue shape is much different for different vowels. As the present study has only focused on midsagittal tongue shape, there is the potential that other crosslinguistic differences in ISP may exist in the coronal view of the tongue - something that is, in fact, predicted by the results of Section 3.1.1 showing greater English tongue measures for all components of the tongue ISP. Crosslinguistic differences in coronal tongue shape would most certainly interact with crosslinguistic differences in midsagittal tongue shape. A multivariate analysis to see which components of ISP correlate could be done in future work. Since the tongue is a hydrostat, it would not be surprising to discover a causal relationship among 101  different components of ISP, both within the midsagittal plane and between the coronal and midsagittal planes. Another possible limitation of the present study is the fact that the English stimuli all contained some low frequency and/or nonsense words, whereas none of the French stimuli contained such types of words. Thus, the task was technically not exactly the same across languages, and was probably more complex in English. A n attempt was made to compensate for this by allowing a practice trial in English, but it is not known whether there was still a task effect present in the results. Another difference between the English stimuli and the French stimuli is the length of the sentences. As can be seen in Appendix VI, the English sentences range in length from 6 to 11 syllables. However, the French sentences, which can be seen in Appendix VII, range in length from 9 to 16 syllables. With the average sentence to read being longer in French, if the English and French speakers read at the same rate (syllables per second), then there was less time between Q  sentences to return to ISP in French , and if the lips have more inertia than the tongue, then they will lag behind in the return to the ISP. This may be one reason why phonetic context played a greater role for the French speakers, at least for the lips. A difficult question for any study that attempts to define certain phonetic properties of a group of speakers of a language is how narrowly to define that language. I have used the terms "Canadian English" and "Quebecois French" in this dissertation, but there certainly could be great variation in ISP among the English dialects present in Canada, and the French dialects present in Quebec. Each of these dialects could actually have its own ISP, and so the results of this study necessarily contain added noise from the variability of ISPs within Canadian English speakers and Quebecois French speakers. In future instrumental studies, tighter control of the origin of each group of speakers might help to reduce such noise and show dialectal differences in ISP. Such studies would synergize with non-instrumental data from a wider range of dialects and languages than have been studied to date. Although carry-over effects of phonetic context were analyzed in Chapter 3, anticipatory effects were assumed not to exist because of the method of stimuli However, with the complex codas and diphthongs in English, it would not be surprising if the syllable rate in French were higher than in English.  8  102  presentation. However, it would have been prudent to do an analysis of anticipatory effects of phonetic context because if anticipatory effects were found, it would call into question the carry-over coarticulation results. In future studies, the origin of AS and its possible relation to the frequency of occurrence of a language's sounds should be investigated. Gick et al. (2004, p. 222) pointed out that it is possible that language-specific ASs are "specified parts of a language's inventory (and hence learned from other speakers) or functionally derived properties of speech motor production". In the latter case, if AS is functional, it probably arises out of motor efficiency requirements and is directly related to the frequency of the sounds of a language. This idea was suggested as long ago as Wilkins (1668, p. 381), cited in Laver (1978, p. 3): "Another different mode of Pronunciation betwixt several Nations may be in regard of strength and distinctness of pronouncing, which will specially appear in those kind of Letters which do most abound in a Language." Laver (2000, p. 39) pointed out that AS could be "an emergent property of segmental performance". This is exactly what Honikman (1964, p. 76) proposed when she stated that "the internal articulatory setting of a language is determined, to a great extent, by the most frequently occurring sounds and sound combinations in that language." Honikman (1964) predicted that based on the greater frequency of [a] in French, the jaw AS would be lower in French than in English. This prediction was not upheld by the results of the present research, which showed no difference in jaw ISP between English and French.  5.3. Conclusions  This research has shown that articulatory setting (AS), observed through the window of inter-speech posture (ISP) of the articulators, is significantly different between Canadian English and Quebecois French, both across monolingual groups and within individual bilingual speakers. The components of ISP that differ across these languages between monolingual groups are upper and lower lip protrusion, tongue tip height, and the degree to which the corners of the mouth are drawn towards the midsagittal plane from a maximally-spread position. In Canadian English, the upper and lower lips are 103  significantly more protruded, the tongue tip is higher, and the corners of the mouth are drawn farther toward the midsagittal plane. It was also shown that ISP is significantly affected by carry-over coarticulation of phonetic context, but in different ways in different languages and for different speakers. In English, only the ISP of the jaw is systematically affected by phonetic context, and this in only three of seven subjects. In French, only the ISP of the lips is affected by phonetic context - specifically the height of the lower lip and the vertical lip aperture for five of eight speakers, and the horizontal lip aperture and the degree of lip narrowing for three speakers. Within individual bilingual speakers who are perceived to be native speakers of both Canadian English and Quebecois French, all speakers show the same upper and lower lip protrusion differences (i.e. English more protruded than French) as the monolingual groups, and half of the speakers show the same tongue tip differences (i.e. English higher than French) as the monolingual groups. These are the only relevant crosslinguistic differences between ISPs for bilinguals who are perceived as native in both languages. Finally, it was shown that bilinguals who are perceived as native speakers of both Canadian English and Quebecois French do not have a unique ISP for bilingual speech mode (i.e. when the bilingual is ready to speak in either language). Instead, the ISP for each of these speakers in bilingual speech mode is equivalent to the monolingualmode ISP of that speaker's dominantly-used language. Thus, in summary, this research shows that ISP (and hence AS) is language specific between monolingual subjects and within bilingual subjects. It is also phonetic context specific, but is not mode specific.  104  References Abercrombie, D. (1967). Elements of general phonetics. Edinburgh: Edinburgh University Press. Adler-Bock, M . (2004). Visual feedback from ultrasound in remediation ofpersistent errors: Case studies of two adolescents. M.Sc. thesis, University of British  M  Columbia. Aho, J. (2004). A quick guide to digital video resolution and aspect ratio conversions. Manuscript downloaded on May 1, 2005 from <http://www.uwasa.fi/~f76998/video/conversion>. Baetens Beardsmore, H . (1986). Bilingualism: Multilingual Matters Ltd.  Basic Principles (2 ed.). Clevedon, U.K.: nd  Ballard, C. F. (1959). Some observations on variations of tongue posture as seen in lateral skull radiographs and their significance. Report of the Congress - European Orthodontic Society, 35, 69-74.  Barnes, J., & Kavitskaya, D. (2002). Phonetic analogy and schwa deletion in French. Proceedings  of the Berkeley Linguistic Society, 27, 39-50.  Barry, W. (1983). Note on interarticulator phasing as an index of temporal regularity in speech. Journal of Experimental  Psychology: Human Perception  and  Performance, 9, 826-828. Barry, W. (1992). Comments on Chapter 2 (Browman & Goldstein). In G. J. Docherty & D. R. Ladd (Eds.), Papers in laboratory phonology II: Gesture, segment, prosody  (pp. 65-67). Cambridge, U.K.: Cambridge University Press. Beasley, R. A., Stefansic, J. D., Herline, A. J., Guttierez, L., & Galloway, R. L . Jr. (1999). Registration of ultrasound images. Proceedings of SPIE (The International for Optical Engineering), 3658, 125-132.  Society  Benoit, C. (1986). Note on the use of correlations in speech timing. Journal of the Acoustical Society of America, 80, 1846-1849.  Birdsong, D. (2005). Nativelikeness and non-nativelikeness in L 2 A research. Manuscript downloaded on November 1, 2005 from <http://ccwf.cc.utexas.edu/ ~birdsong/publications/native_non_native.pdf>.  105  Birdsong, D. (to appear). Nativelike pronunciation among late learners of French as a second language. In O.-S. Bohn & M . Munro (Eds.), Second language speech learning: The role of language experience in speech perception  and  production.  John Benjamins. Bithell, J. (1952). German pronunciation Kelz(1971,p. 203).]  and phonology. London: Methuen. [Cited in  Bongaerts, T. (1999). Ultimate attainment in L2 pronunciation: The case of very advanced late L2 learners. In D. Birdsong (Ed.), Second language acquisition and the critical period hypothesis (pp. 133-159). Mahwah, NJ: Erlbaum. Browman, C. P., & Goldstein, L. (1992). "Targetless" schwa: an articulatory analysis. In G. J. Docherty & D. R. Ladd (Eds.), Papers in laboratory phonology II: Gesture,  segment, prosody (pp. 26-56). Cambridge, U.K.: Cambridge University Press. Brown, A . (1995). Minimal pairs: Minimal importance? ELT Journal, 49, 169-175. Brown, L. E., & Rosenbaum, D. A. (2002). Motor control: Models. In L. Nadel (Ed.), Encyclopedia of Cognitive Science (pp. 127-133). London: Macmillan. Bruyninckx, M . , Harmegnies, B., Llisterri, J., & Poch-Olive, D. (1994). Languageinduced voice quality variability in bilinguals. Journal of Phonetics, 22, 19-31. Buuren, L. Van. (1995). Postura: Clear and dark consonants, etcetera. In J. W. Lewis, (Ed.), Studies in General and English Phonetics: Essays in Honour of Professor J.  D. O'Connor (pp. 130-142). New York: Routledge. Campbell, F. M . (2004). The gestural organization of North American English M: A  study of timing and magnitude. M . A . Thesis, University of British Columbia. Celce-Murcia, M . , Brinton, D. M . , & Goodwin, J. M . (1996). Teaching pronunciation:  A  reference for teachers of English to speakers of other languages. Cambridge,  U.K.: Cambridge University Press. Chiang, Y . - C , Lee, F.-P., Peng, C.-L., & Lin, C.-T. (2003). Measurement of tongue movement during vowels production with computer-assisted B-mode and M-mode ultrasonography. Otolaryngology  - Head and Neck Surgery, 128, 805-814.  Collins, B., & Mees, I. M . (1995). Approaches to articulatory setting in foreign-language teaching. In J. W. Lewis, (Ed.), Studies in General and English Phonetics:  Essays  in Honour of Professor J. D. O 'Connor (pp. 415-424). New York: Routledge. Collins, B . , & Mees, I. M . (2003). Practical phonetics and phonology:  for students. London: Routledge.  106  A resource book  Cookson, A . M . (1967). Tongue resting position: A method for its measurement and correlation to skeletal and occlusal patterns. The Dental Practitioner and Dental Record, 18, 115-120. Cote, M . - H . , & Morrison, G. S. (2004). Experimental evidence and the nature of the schwa/zero alternation in French. Poster presented at the Ninth Conference on Laboratory Phonology (LabPhon 9), Urbana, IL. Crystal, D. (2003). A dictionary of linguistics and phonetics (5 ed.). Oxford: Blackwell Publishing. th  Daniloff, R., & Moll, K. (1968). Coarticulation of lip rounding. Journal of Speech and Hearing Research, 11, 707-721. Dromey, C , & Wheeler, C. (2004). First and second language articulatory kinematics in  bilingual speakers. Poster presented at the American Speech-Language-Hearing Association's (ASHA) Annual Convention, Philadelphia, PA. Eifert, D. F. (1960). A roentgenographic cephalometric study of the tongue. American Journal of Orthodontics, 46, 226-227. (Abstract of paper presented before the Research Section of the American Association of Orthodontists, Detroit, Michigan, May 3-7, 1959.) Elston-Giittler, K. E., Paulmann, S., & Kotz, S. A. (2005). Who's in control? Proficiency and L I influence on L2 processing. Journal of Cognitive Neuroscience,  17, 1593-  1610. Epstein, M . A., & Stone, M . (2005). The tongue stops here: Ultrasound imaging of the palate. Journal of the Acoustical Society of America, 118, 2128-2131.  Esling, J. H. (1987). Methodology for voice setting awareness in language classes. Revue de Phonetique Appliquee, 85, 449-473.  Esling, J. H., & Wong, R. F. (1983). Voice quality settings and the teaching of pronunciation. TESOL Quarterly, 17, 89-95. Fitch, W.T., & Giedd, J. (1999). Morphology and development of the human vocal tract: a study using magnetic resonance imaging. Journal of the Acoustical Society of America, 106, 1511-1522. Flege, J. E., Schirru, C , & MacKay, I. R. A. (2003). Interaction between the native and second language phonetic subsystems. Speech Communication, 40, 467-491. Fromkin, V . A. (1964). Lips positions in American English vowels. Language and Speech, 7, 215-225.  107  Gick, B. (2002). A n X-ray investigation of pharyngeal constriction in American English schwa. Phonetica, 59, 38-48. Gick, B., Wilson, I., Koch, K., & Cook, C. (2004). Language-specific articulatory settings: Evidence from inter-utterance rest position. Phonetica, 61, 220-233. Gick, B., Bird, S., & Wilson, I. (2005). Techniques for field application of lingual ultrasound imaging. Clinical Linguistics & Phonetics, 19, 503-514. Graff, W. L . (1932). Language and languages: An introduction to linguistics.  New  York/London: D. Appleton & Co. Grosjean, F. (1989). Neurolinguists, beware! The bilingual is not two monolinguals in one person. Brain and Language, 36, 3-15.  Grosjean, F. (1998). Studying bilinguals: Methodological and conceptual issues. Bilingualism:  Language and Cognition, 1, 131-149.  Hamlet, S. L., & Stone, M . L . (1981). Pre-speech posturing of the mandible in relation to jaw activity during speech. Journal of Phonetics, 9, 425-436. Harmegnies, B., Esling, J. H., & Delplancq, V . (1989). Quantitative study of the effects of setting changes on the L T A S . In J. P. Tubach & J. J. Mariani (Eds.), European conference on speech communication  and technology (Vol. 2, pp. 139-142).  Edinburgh: CEP Consultants. Harmegnies, B., & Landercy, A . (1985). Language features in the long-term average spectrum. Revue de Phonetique Appliquee,  73-75, 69-80.  Hartmann, R. R. K., & Stork, F. C. (1972). Dictionary of language and linguistics. New York: John Wiley & Sons. Heffner, R - M . S. (1950). General phonetics. Madison, W l : The University of Wisconsin Press. Honda, M . (1985). Dynamical ultrasonic measurement system of tongue shape. Trans. Committee Speech Res., Acoustical Society of Japan, pp. 84-105. Honda, K., Maeda, S., Hashi, M . , Dembowski, J. S., & Westbury, J. R. (1996). Human palate and related structures: Their articulatory consequences. Proceedings of the International  Conference on Spoken Language Processing  (ICSLP-1996),  pp.  784-787. Honikman, B . (1964). Articulatory settings. In D. Abercrombie, D. B . Fry, P. A . D. MacCarthy, N . C. Scott, & J. L . M . Trim (Eds.), In Honour of Daniel Jones (pp. 73-84). London: Longman. 108  Jenkins, J. (1998). Which pronunciation norms and models for English as an international language? ELTJournal, 52, 119-126. Jenner, B . (2001). 'Articulatory setting': genealogies of an idea. Linguistica,  Historiographia  XXVIII, 121-141.  Jesney, K . (2004). The use of global foreign accent rating in studies of L2 acquisition. Calgary, A B : University of Calgary Language Research Centre Reports. Downloaded on January 31, 2006 from <http://fis.ucalgary.ca/LRC/doc/GAReport_L2.pdf>. Jones, R. H., & Evans, S. (1995). Teaching pronunciation through voice quality. ELT Journal, 49, 244-251. Kaburagi, T., & Honda, M . (1994). An ultrasonic method for monitoring tongue shape and the position of a fixed point on the tongue surface. Journal of the Acoustical Society of America, 95, 2268-2270.  Kelz, H. P. (1971). Articulatory basis and second language teaching. Phonetica, 24, 193211. Kerr, J. (2000). Articulatory setting and voice production: Issues in accent modification. Prospect, 15, 4-15.  Kier, W. M . , & Smith, K . K . (1985). Tongues, tentacles and trunks: the biomechanics of movement in muscular-hydrostats. Zoological Journal of the Linnean Society, 83,  307-324. Kotsiomiti, E., Farmakis, N . , & Kapari, D. (2005). Factors related to the resting tongue position among partially and completely edentulous subjects. Journal of Oral Rehabilitation, 32, 397-402. Kuhnert, B., & Fougeron, C. (2004). Schwa-productions and hesitations by native French speakers during English L2 performance. Abstract of poster presented at the 9 th  Conference on Laboratory Phonology. Urbana, IL. Laver, J. (1978). The concept of articulatory settings: an historical survey. Historiographia  Linguistica,  V, 1-14.  Laver, J. (1980). The phonetic description of voice quality. Cambridge, U.K.: Cambridge University Press. Laver, J. (2000). Phonetic evaluation of voice quality. In R. D. Kent & M . J. Ball (Eds.), Voice quality measurement (pp. 37-48). San Diego: Singular.  109  Lebrun, Y . (1970). On tension. In J. Dierickx & Y. Lebrun (Eds.), Linguistique contemporaine:  Hommage a Eric Buyssens (pp. 115-124). Bruxelles: Editions de  lTnstitut de Sociologie de l'Universite Libre de Bruxelles. Linker, W. (1982). Articulatory and acoustic correlates of labial activity in vowels: A cross-linguistic survey. UCLA Working Papers in Phonetics, 56, 1-1-34.  Loveday, L. (1981). Pitch, politeness and sexual role: A n exploratory investigation into the pitch correlates of English and Japanese politeness formulae. Language and Speech, 24, 71-89. Marinova-Todd, S. (2003). Comprehensive analysis of ultimate attainment in adult  second language acquisition. Ph.D. Dissertation, Harvard University. Matthews, P. H . (1997). The concise Oxford dictionary of linguistics. Oxford, U . K . :  Oxford University Press. Max, L., & Onghena, P. (1999). Some issues in the statistical analysis of completely randomized and repeated measures designs for speech, language, and hearing research. Journal of Speech, Language, and Hearing Research, 42, 261-270. McDowell, R. E. (2004). Retraction in Montana Salish lateral consonants: An ultrasonic  study. M A Thesis, University of British Columbia. Medicines and Healthcare products Regulatory Agency (MHRA). (2004). Evaluation Report MHRA 03107. Freely downloaded on May 7, 2005 from: <http://devices.mhra.gov.uk/mda/mdawebsitev2.nsf/710be55fc5a340d785256b58 00547efa/54241deb3354eaal80256el90051e536/$FILE/MHRA%2003107.pdf>. Mompean-Gonzalez, J. A . (2003). Pedagogical tools for teaching articulatory  setting.  Poster presented at the 15 International Congress of Phonetic Sciences (ICPhS 15), Barcelona. th  Morley, J. (1991). The pronunciation component in teaching English to speakers of other languages. TESOL Quarterly, 25, 481-520. Munhall, K. G. (1985). A n examination of intra-articulator relative timing. Journal of the Acoustical Society of America, 78, 1548-1553.  Munhall, K. G., Vatikiotis-Bateson, E., & Tohkura, Y . (1994). X-ray film database for speech research, ATR technical report TR-H-116. Kyoto: A T R Human  Information Processing Research Laboratories. Nilsson, A. (2001). Artefacts in sonography and Doppler. European Radiology, 11, 13081315.  110  O'Connor, J. D. (1973). Phonetics. Harmondsworth: Penguin. Oh, S. (2004). Articulatory characteristics  of English /I/ in speech development. Ph.D.  Dissertation, University of British Columbia. Ohman, S. E. G. (1967). Peripheral Motor Commands in Labial Articulation. Speech Transmission Laboratory - Quarterly Progress and Status Report (Royal Institute of Technology (KTH), Stockholm), 4/1967, 30-63.  Oliver, R. G., & Evans, S. P. (1986). Tongue size, oral cavity size and speech. The Angle Orthodontist, 56, 234-243. Paradis, J. (1996). Prosodic development and differentiation in bilingual first language acquisition. In A. Stringfellow, D. Cahana-Amitay, E. Hughes, & A . Zukowski (Eds.), Proceedings of the 20th Annual Boston University Conference on  Language Development (pp. 528-539). Somerville, M A : Cascadilla Press. Paradis, J. (2001). Do bilingual two-year-olds have separate phonological systems? International Journal of Bilingualism,  5, 19-38.  Pennington, M . C. (1996). Phonology in English language teaching: An  international  approach. London: Longman. Pennington, M . C , & Richards, J. C. (1986). Pronunciation revisited. TESOL 20, 207-225.  Quarterly,  Perkell, J. S. (1969). Physiology of speech production: Results and implications of a quantitative cineradiographic study. Cambridge, M A : MIT Press.  Price, G. (1991). An introduction to French pronunciation.  Oxford: Blackwell.  Rich, S. (2003). Introducing voice-setting phonology, part 3. Manuscript downloaded on December 1, 2003 from <http://www.developingteachers.com/articles_tchtraining/ voiceset3_sarn.htm>. Richards, J. C , Piatt, J., & Piatt, H. (1992). Dictionary of language teaching & applied linguistics. Harlow, U.K.: Longman. Sancier, M . L., & Fowler, C. A . (1997). Gestural drift in a bilingual speaker of Brazilian Portuguese and English. Journal of Phonetics, 25, 421-436. Schmidt, R. A . , & Lee, T. D. (1999). Motor control and learning: A behavioral emphasis  (3 ed.). Champaign, IL: Human Kinetics. rd  Schreiner, S., Galloway, R. L . Jr., Lewis, J. T., Bass, W. A., & Muratore, D. M . (1998). A n ultrasonic approach to localization of fiducial markers for interactive, image111  guided neurosurgery - Part II: Implementation and automation. IEEE on Biomedical Engineering,  Transactions  45, 631-641.  Seyfarth, R. M . , & Cheney, D. L . (2002). What are big brains for? Proceedings of the National Academy of Sciences, 99, 4141-4142.  Sharpe, M . C. (1970). Voice quality: A suggested framework for description and some observations. In S. A . Wurm & D. C. Laycock (Eds.), Pacific Linguistic Studies in Honour of Arthur Capell (pp. 115-134). Canberra: Linguistic Circle of Canberra. Sievers, E. (1876). Grundziige der Lautphysiologie.  Leipzig: Breitkopf & Hartel.  Sweet, H . (1890). A primer of phonetics. Oxford: Clarendon Press. Takada, K., Sakuda, M . , Yoshida, K., & Kawamura, Y . (1980). Relations between tongue volume and capacity of the oral cavity proper. Journal of Dental Research, 59, 2026-2031. Takano, S., Honda, K., & Dang, J. (2002). Geometrical analysis of the tongue muscles based on MRI and functional modeling of the tongue. Journal of the Acoustical Society of America, 112(5), Part 2, 2418.  Tatham, M . (1997). Retrieved August 16, 2003 from <http://vsrww.essex.ac.uk/speech/ archive/control/control.html>. (Reproduced from Occasional Papers No. 3 (1969), Language Centre, University of Essex.) Thornbury, S. (1993). Having a good jaw: Voice-setting phonology. ELT Journal, 47, 126-131. Tiede, M . K., Boyce, S. E., Holland, C. K., & Choe, K . A . (2004). A new taxonomy of American English /r/ using MRI and ultrasound. Poster presented at the 147  th  Meeting of the Acoustical Society of America, New York. Todaka, Y . (1993). A cross-language study of voice quality: Bilingual Japanese and American English speakers. Ph.D. dissertation, U C L A .  Todaka, Y . (1995). A preliminary study of voice quality differences between Japanese and American English: Some pedagogical suggestions. JALT Journal, 17, 261268. Trask, R. L . (1996). A dictionary ofphonetics and phonology. London: Routledge. Traunmuller, H. (1984). Articulatory and perceptual factors controlling the age- and sexconditioned variability in formant frequencies of vowels. Speech Communication, 3, 49-61.  112  Wallis, J. (1653/1972). Grammatica linguae anglicanae. Edited/translated by J. A . Kemp, 1972. London: Longman. Watson, I. (1990). Acquiring the voicing contrast in French: A comparative study of monolingual and bilingual children. In J. N . Green & W. Ayres-Bennett (Eds.), Variation and change in French: Essays presented to Rebecca Posner on the  occasion of her sixtieth birthday (pp. 37-60). London: Routledge. Watson, I. (1991). Phonological processing in two languages. In E. Bialystok (Ed.), Language processing  in bilingual children (pp. 25-48). Cambridge, U . K . :  Cambridge University Press. Wilkins, J. (1668/1967). An essay towards a real character, and a  philosophical  language. London: J. Martyn. (Reprinted by Scholar Press, Menston 1967.) Zerling, J. P. (1992). Frontal lip shape for French and English vowels. Journal of Phonetics, 20, 3-14.  113  Appendix III Detailed background on subjects Monolingual English subjects Subject 1  Gender Female  Age 28  Place of Origin Winnipeg, M B  2  Female  25  Kelowna, B C (about 250 km east of Vancouver)  3  Male  25  Vancouver, B C  4 5  Female Female  22 36  6  Male  27  Vancouver, B C Southern O N (ages 0-11); various U.S. states (ages 12-20); Fraser Valley, B C (ages 21-36) Richmond, B C (part of Greater Vancouver)  7  Male  24  Sudbury, ON  116  Notes "dabbled" in Hebrew, French, Japanese & Spanish, but not fluent in any of them; had been living in Vancouver for over a year. French Immersion from Grades 1-7 (all day French, but in Grades 4-7 she had 1 hr per day in English); high school - half French, half English; parents & siblings monolingual English. studied French in high school; very slow, deliberate speech; slight facial twitch, but didn't seem to affect lips; was not so comfortable during data collection. studied French in high school. studied French from Grade 7 to college; parents spoke Frisian & Dutch together, but the home language was always English. studied a little bit of Japanese in high school. His mother is Japanese, but she apparently speaks English with no accent. had been living in Vancouver for 1 year at time of data collection.  M o n o l i n g u a l F r e n c h subjects Subject 8  Gender F  Age 18  Place of origin St-Jean-Chrysostome, QC (across the St. Lawrence R. south of Quebec City)  9  F  21  10  F  51  Sherbrooke, QC (equidistant - about 150 km - east of Montreal & south of Quebec City) Montreal, QC  11  M  18  Levis, QC (across the St. Lawrence R. south of Quebec City)  12  F  19  St-Adolphe d'Howard, QC (about 70 km northwest of Montreal)  13  M  18  14  F  22  St. Henri de Levis, QC (across the St. Lawrence R. south of Quebec City) Mont Tremblant (about 100 km northwest of Montreal)  15  F  22  Jonquiere, QC (near Chicoutimi, about 175 km north of Quebec City)  117  Notes ended every stimuli sentence with a rising intonation. Noticeably thin lips. Seemed to slur words at times - possibly due to fast speech. spoke loudly and clearly. Wore the experiment glasses over her own glasses. spoke some English, but with difficulty & heavy accent. Started learning in high school. Very thin lips - especially upper lip. Wore experiment glasses over her own. spoke some English, but with a heavy accent. Had been in Vancouver for 1 week. Said his jaw was a bit sore from speaking English this week and that his friend (Subject 13) felt the same way! Trials 1 & 2 were done with only the experiment glasses. Trials 3-6 were done with the experiment glasses over top of her own glasses (her eyes were getting tired). spoke English with a heavy accent. Had been in Vancouver for 1 week. Basic English vocabulary missing. Studied English in college (compulsory) and spoke it with a noticeable but not too heavy accent. Has been living in Vancouver for 1 year but 60% of her time in French. Before coming to Vancouver, 90% of her time in Quebec was spent in French. clear voice.  Bilingual English-French subjects Subject 16  Age& gender 33 - F  Places lived  Notes  St. Jerome, QC (ages 0-22); Vancouver, BC (ages 22-33)  17  23-F  Ottawa, ON (ages 0-23); Vancouver, BC (last 1.5 mths)  Ll=Fre; Age of exposure to Eng = 12; Parents both monolingual Fre; Typical week now 90% Eng, 10% Fre. Ll=Eng; Age of exposure to Fre = 5; Only Fre from kindergarten to 3 year university; Spoke to Mom in both Eng and Fre; Typical week now 95% Eng, 5% Fre; In Ottawa, typical week was 65% Eng, 35% Fre. L1 =Swiss German - spoke it with parents; Age of exposure to Fre = 3 (spoke Fre with sister); Age of exposure to Eng = 5 (but she said her Eng was not so good until university in Ottawa.) Typical week now 80% Eng, 20% Fre. Ll=Fre; Age of exposure to Eng = 9; Typical week now 90% Eng, 10% Fre; Monolingual Fre parents. Has always spoken to siblings in Fre. Not particularly used to reading in Fre. Ll=Fre; Age of exposure to Eng = 3; Typical week now 60% Fre, 40% Eng; attended Eng preschool, then Fre kindergarten & above (but 1 hr/wk Eng lesson & Eng with friends on her street) until Eng full-time in university. Ll=Eng; Age of exposure to Fre = 3; Mother's Ll=Fre; Father's Ll=Italian, but not spoken at home; Fre only from ages 11-22 (home, school, peers); Eng only from ages 24-34; From 34-37, Fre only at home with husband & kids, Eng only at work. (Typical week now: "Home days" = 66% Fre; "Work days" = 66% Eng.) Ll=Fre; Age of exposure to Eng = 13, but exposed to Eng TV at age 6; Both parents & brother monoling Fre. Ll=Fre; Age of exposure to Eng < 5; spoke Fre with mother, Eng with father, but parents spoke Eng to each other; schooled in Fre only (including CGEP); Typical week now 70% Eng, 30% Fre; Phonetically trained, speech-language pathologist. Ll=both Eng & Slovene; Age of exposure to Fre = 7; At age 4.5, Eng was dominant and parents switched to Slovene-only at home. From ages 4.5 to 7.5, Slovene was dominant. Eng schooling in Montreal, but worked in Fre; 100% Fre for 16yrs in Quebec City; Typical week now 30% Fre, 70% Eng. Hardly any Slovene spoken since age 8. rd  18  23-F  Quebec City, QC (ages 0-4); Ottawa, ON (ages 5-23)  19  19 - M  Greenwood, NS (ages 0-4); Montreal, QC (ages 5-9); Vancouver, BC (ages 10-19)  20  25-F  Montreal, QC (ages 0-25, except for 1 year in Japan teaching Eng)  21  37- F  ON (ages 0-10); Quebec City, QC (ages 11-22); Vancouver, BC (ages 24-37)  22  23-M  Montreal, QC (ages 0-23)  23  38-F  Montreal, QC; Vancouver, BC  24  46- F  Australia (ages 0-6); Montreal, QC (ages 7-28); Quebec City, QC (ages 29-45); Vancouver, BC (age 46)  118  Appendix IV  Definition of ratings in foreign accent rating task  Native French listeners were given the following guidelines and asked to rate speakers on the following scale. Although not shown here, there was space for Speakers A to N.  5  =  Le Francais est sa langue maternelle.  4  =  Le Francais est son deuxieme langage etudie, mais il est tres bien maitrise.  3  =  Le Francais est son deuxieme langage etudie, et elle/il le parle convenablement.  2  =  Le Francais est son deuxieme langage etudie, mais s'exprime avec difficulte.  1  =  Le Francais est son deuxieme langage etudie, mais franchement mediocre.  la personne A: la personne B: la personne C:  Native English listeners were given the following guidelines and asked to rate speakers on the following scale. Although not shown here, there was space for Speakers A to M. 5  -  English is her/his mother tongue.  4  =  English is her/his 2 language, but s/he has mastered it very well.  3  =  English is her/his 2 language, and s/he speaks it adequately.  2  =  English is her/his 2 language, and s/he speaks it with difficulty.  1  =  English is her/his 2 language, and s/he speaks it very poorly.  nd  nd  nd  nd  Person A: Person B: Person C:  119  Appendix V Detailed results of foreign accent rating task Table 3.2. French native listener judgements of subjects' French utterances on a scale from 1 to 5 where only 5 signifies a native speaker Ral.er number 2 (mono E) 16 17 18  other (mono F) E  19 5 (mono E) 20  o  other (mono F)  -§  21  other (E-dom) 22 23 24  avg.  8  9  2 3 4 4 5 2 1 4 4 5 2 5 4 3 3.4  11  12  13  14  15  A  B  avg.  2 3 4 5 4 4 2 4 5 4 3 4 4 3  10 2 3 4 5 5 4 2 5 4 5 3 5 5 -  3 4 5 5 5 4 2 4 4 5 2 5 5 -  3 4 5 5 5 4 2 4 5 4 2 5 5 4  3 4 5 5 5 4 2 5 5 5 3 5 5  -  3 4 5 5 5 5 2 5 5 5 3 5 5 4  2 4 5 5 5 5 1 4 5 4 2 5 5 4  3 4 5 5 5 5 2 5 4 5 3 5 4 4  3 4 5 5 5 5 3 4 5 4 3 5 5 4  2.6 3.7 4.7 4.9 4.9 4.2 1.9 4.4 4.6 4.6 2.6 4.9 4.7 3.7  3.6  4.0  4.1  4.1  4.3  4.4  4.0  4.2  4.3  4.0  Table 3.3. English native listener judgements of subjects' English utterances on a scale from 1 to 5 where only 5 signifies a native speaker later identi iication  tu  B O <D  2 (mono E) 16 17 18 19 5 (mono E) 6 (mono E) 20 '21  other (E-dom) 22 23 24 avg.  PB  CX  K.K  RW  MC  RK  KL  BS  DQ  HM  avg.  5 4 5 4 4 5 5 4 5 5 5 4  5 3 4 4 5 5 4 3 5 4 4 3  5 3 5 4 5 5 5 4 5 5 5 4  5 3  -  -  4 5 5 5 4 5 5 5 4 4  5 4 5 3 5 5 5 4 5 4 2 4 5  5 2 5 4 4 5 4 3 4 5 3 3 4  5 3 5 4  -  5 4 5 4 5 5 5 4 5 5 5 5 5  4 5  5 4 5 4 5 5 5 4 5 5 5 5 5  5 3 4 4 4 5 5 4 5 4 5 3 4  5.0 3.3 4.7 3.9 4.6 4.9 4.8 3.8 4.9 4.7 4.3 3.9 4.6  4.6  4.1  4.6  4.8  4.5  4.3  3.9  4.4  4.8  4.2  4.4  4*  4* means that the rater wrote " 4 / 5 " or " 5 / 4 " for this speaker.  120  4* 4 5 4 5 5  4*  Appendix VI English stimuli used Block 1  Block 2  Gavin said "hear ee" each spring. Dave said "hear hee" each June. Vance said "hair hay" each time again. Danny said "hoar owe" each minute. Joanna said "who roo" each holiday. Edwin said "hair A" each time again. Whitney said "hoar hoe" each day with Sue. Casey said "hear ee" each scenario. Mike said "hee wee" each fall. Seth said "hear ee" each week. Ann said "har awe" each turn in Thai. Betty said "har haw" each April. Heidi said "hee ree' each week. Judy said "hear ee" each April. Joyce said "hay ray" each holiday. Jamie said "hay ray" each month. Ben said "hee wee" each month. Tanya said "hoe roe" each job. Otto said "har haw" each lunch. Cindy said "har haw" each show. Jenny said "hee ree" each scenario. Becky said "har haw" each autumn. Mason said "hoar hoe" each term. Matt said "hee wee" each time through. Chuck said "haw yaw" each job. Nate said "hair A" each show. Jon said "hoar owe" each semester. Nick said "hee ree" each time through. Sandy said "who roo" each fall. Stan said "har awe" each time again.  Anne said "who're hoo" each minute. Sam said "hoar hoe" each day. Cathy said "hair A" each day. Hannah said "haw raw" each July. Evan said "hee wee" each summer. Simon said "haw yaw" each month. Fabian said "who roo" each evening with joy Tina said "hay ray" each spring. Jacob said "who're oo" each time through. Dan said "who're hoo" each April. Denise said "hee ree" each sailing regatta. Matt said "who're oo" each weekend. Joey said "who're hoo" each show. Hank said "hee wee" each minute. Maggie said "har awe" each semester. Ken said "who roo" each show. Tony said "hair A" each January. Ben said "hoar hoe" each day. Tom said "who're hoo" each job. Joseph said "har haw" each semester. Noah said "haw raw" each sailing regatta. Eva said "who're oo" each January. Jody said "who roo" each autumn. Bonnie said "har haw" each term. Dean said "who're oo" each holiday. Ivan said "hair hay" each evening with joy. Guy said "hair hay" each turn in Thai. Donna said "hoar owe" each day with Sue. Jim said "hair hay" each turn in Thai. Steve said "har awe" each month.  ?  121  Block 3  Block 4  Gina said "hay ray" each scenario. Toby said "hoar owe" each week. Joan said "hear ee" each scenario. Ned said "har awe" each year. Mitch said "who're hoo" each year. Kevin said "hoe roe" each weekend. Suzie said "hay ray" each minute. Joe said "hear ee" each minute. Anna said "hay ray" each evening with joy. Vicky said "hear hee" each April. Jan said "hair hay" each time through. Vanessa said "hay ray" each time again. Diane said "hear ee" each June. Tommy said "who're oo" each show. Dustin said "har awe" each fall. Hank said "hee wee" each month. Janice said "haw raw" each job. Beth said "hee ree" each time through. Vickie said "hoe roe" each time through. Kate said "hoe roe" each spring. Mike said "hoe roe" each evening with joy. Beth said "haw yaw" each job. Zack said "hear ee" each spring. Josie said "har awe" each semester. Scott said "hair hay" each weekend. Wendy said "hair hay" each lunch. Wanda said "who're hoo" each time again. Kenny said "haw raw" each January. Pat said "hear hee" each sailing regatta. Katie said "haw yaw" each summer.  Ethan said "hair A" each job. John said "har haw" each weekend. Jean said "hoar hoe" each week. Tom said "haw yaw" each fall. Justin said "hear hee" each holiday. Wayne said "hear hee" each time again. Jane said "har haw" each year. Teddy said "haw raw" each time again. Jessie said "hay ray" each autumn. Zane said "har haw" each June. Pete said "haw raw" each term. Simon said "haw yaw" each autumn. June said "hee ree" each scenario. Evan said "hoe roe" each summer. Bob said "hair hay" each class. Tammy said "hear hee" each afternoon. Jasmine said "haw raw" each autumn. Ike said "haw raw" each turn in Thai. Ted said "hoe roe" each year. Tim said "hear hee" each job. Dennis said "hair A" each sailing regatta. Jody said "hoar hoe" each weekend. Ken said "haw yaw" each time again. Don said "who're oo" each fall. Jason said "hoar hoe" each semester. Max said "hoar owe" each July. Vince said "haw raw" each month. Keith said "who roo" each afternoon. Jed said "hair A" each term. Mindy said "hair A" each day.  122  Block 5  Block 6  Peggy said "hair A " each afternoon. Hugh said "hair hay" each month. Josh said "hear hee" each minute. Patty said "who roo" each time again. Missy said "who're hoo" each lunch. Fay said "who're oo" each June. Evan said "hee wee" each summer. Bob said "hee wee" each autumn. Chuck said "haw yaw" each job. James said "who're oo" each fall. Jackie said "har haw" each class. Kim said "hear hee" each January. Steven said "who roo" each day. Todd said "hear ee" each afternoon. Janet said "hee ree" each sailing regatta. Joanne said "har awe" each class. Ian said "hoe roe" each July. Shane said "who're oo" each time again. Nancy said "hoar owe" each summer. Jay said "hee ree" each lunch. Dana said "hee ree" each week. Mandy said "hear hee" each holiday. Megan said "haw raw" each June. Hank said "hoar hoe" each day with Sue. Nathan said "har awe" each April. Finn said "hear ee" each month. Kay said "hoar hoe" each summer. Jimmy said "who're oo" each class. Susan said "who roo" each January. Sean said "hoar hoe" each afternoon.  Missy said "who're hoo" each lunch. Gavin said "hear ee" each spring. Katie said "hoar owe" each lunch. Anne said "who're hoo" each minute. Jessica said "hoar owe" each fall. Gina said "hay ray" each scenario. Shawn said "hee ree" each evening with joy. Ethan said "hair A " each job. Dominic said "who're hoo" each time again. Peggy said "hair A " each afternoon. Stacy said "har awe" each year. Dave said "hear hee" each June. Vance said "hair hay" each time again. Danny said "hoar owe" each minute. Joanna said "who roo" each holiday. Jake said "hoar owe" each July. Bobby said "who're hoo" each autumn. Diana said "who're hoo" each day with Sue. Scott said "haw yaw" each month. Sam said "haw yaw" each time again. Gina said "hee wee" each spring. Katie said "hee wee" each fall. Doug said "hay ray" each term. Chuck said "hoe roe" each class. Dawn said "hoar owe" each July. Eddie said "hair hay" each spring. Simon said "who roo" each turn in Thai. Monty said "hair A " each summer. Debbie said "hay ray" each day with Sue. Gene said "hoe roe" each summer.  123  Appendix VII French stimuli used There are 6 blocks, each consisting of 30 sentences and lasting 131 seconds. • • •  There are 60 names randomized throughout Blocks 1&2, Blocks 3&4, and Blocks 5&6 There are 60 predicates randomized throughout Blocks 1&2, Blocks 3&4, and Blocks 5&6, Within the 60 predicates, there are 20 keywords with i/I, 20 with y/Y, and 20 with uAJ. Within each of these groups of 20, there are 9 keywords with the tense vowel and 11 keywords with the lax vowel.  Block 1 Vincent, il a vu une ville dans la vallde. Alexandre, il a vu une nuque dans le dessin. Catherine, elle a vu une loupe dans le pupitre. Frederic, il a vu un loup dans la foret. Veronique, elle a vu une russe dans le ballet. Simon, il a vu un loup dans la foret. Arianne, elle a vu une pousse dans le jardin. Gabrielle, elle a fait un vide dans la maison. Sophie, elle a vu une pipe dans la bouche du monsieur. Juliette, elle a vu un cube dans la boite a jouets. Philippe, il a vu du riz dans le chaudron. Thomas, il a mis de la vie dans la maison. Audrey, elle a vu une ruche dans le champ. Gabriel, il a vu le cul de la chevre. Alicia, elle a vu un pouce dans la radiographie. Claudia, elle a vu une rue dans la vieille ville. Jerome, il a vu une pie dans l'arbre. Marie Eve, elle a vu un lit dans la clinique. Etienne, il a vu une cuve dans la salle de bain. Nicolas, il a vu un tube dans le trou. Zachary, il a vu un nu dans le musee. Chlo6, elle a vu un bouc dans le champ. Samuel, il a vu une pie dans l'arbre. Benoit, il a vu une lime dans la sacoche. Xavier, il a vu un tutu dans le vestiaire. Sebastien, il a vu une tuque dans le linge sale. Annabelle, elle a vu un lys dans l'etang. Valerie, elle a vu un pou dans les poils du rat. Amelie, elle a vu de la boue dans Pentree de la maison. Alexandra, elle a vu une frite dans l'assiette.  Block 2 Raphael, il a vu la lune dans le ciel. Daphnee, elle a vu une soupe dans le livre de recettes. Beatrice, elle a vu du pus dans la plaie. Marc-Olivier, il a vu une trousse dans les objets perdus. Rosalie, elle a vu un jus dans le frigidaire. Louis, il a vu une troupe dans la parade. Isabelle, elle a vu un trou dans Pail. FeJix, il a vu une jupe dans la vitrine. Hugo, il a entendu un cri dans la nuit. Maxime, il a vu une coupe dans le salon de coiffure. Emilie, elle a vu un trou dans la roche. Antoine, il a vu une puce dans la perruque. Elizabeth, elle a vu une lutte dans le bar. Justine, elle a vu une luge dans la neige. Julien, il a vu une cruche dans les mines grecques. Alexia, elle a vu un grand cru dans la cave a vin. Guillaume, il a vu une niche dans la cour. Christophe, il a vu un cou dans le dessin. Adam, il a vu un saoul dans le bar. M6gan, elle a entendu un cri dans la nuit. Florence, elle a vu une crique dans la foret. Tristan, il a vu un nid dans l'arbre. Olivier, il a vu un eric dans la valise de l'auto. M^lodie, elle a vu un site dans le terrain de camping. Cddric, il a recu un coup dans la face. Genevieve, elle a vu une niche dans la cour. Francois, il a vu une trousse dans les objets perdus. Eric, il a vu une scie dans la boite a outils. Camille, elle a vu une coupe dans le salon de coiffure. Mathieu, il a vu une troupe dans la parade.  124  Block 3 Alexia, elle a vu un loup dans la foret. Louis, il a vu de la boue dans Pentree de la maison. Eric, il a vu un tube dans le trou. Beatrice, elle a vu une pie dans l'arbre. Olivier, il a vu une pipe dans la bouche du monsieur. Camille, elle a mis de la vie dans la maison. Mathieu, il a vu un lys dans I'dtang. S^bastien, il a vu une frite dans Passiette. Justine, elle a fait un vide dans la maison. Christophe, il a vu une cuve dans la salle de bain. Sophie, elle a vu une ville dans la vallee. Felix, il a vu un loup dans la foret. Audrey, elle a vu un bouc dans le champ. Alicia, elle a vu un pouce dans la radiographic Gabriel, il a vu une russe dans le ballet. Frederic, il a vu une nuque dans le dessin. Tristan, il a vu un lit dans la clinique. BenoTt, il a vu un nu dans le musee. Juliette, elle a vu une tuque dans le linge sale. Veronique, elle a vu une lime dans la sacoche. Emilie, elle a vu le cul de la chevre. Florence, elle a vu un tutu dans le vestiaire. Nicolas, il a vu une pie dans l'arbre. Marie Eve, elle a vu une ruche dans le champ. Annabelle, elle a vu un cube dans la boite a jouets. Maxime, il a vu du riz dans le chaudron. Xavier, il a vu un pou dans les poils du rat. Genevieve, elle a vu une loupe dans le pupitre. Chlod, elle a vu une rue dans la vieille ville. Simon, il a vu une pousse dans le jardin.  Block 4 Zachary, il a vu une troupe dans la parade. Thomas, il a vu une lutte dans le bar. Elizabeth, elle a vu un trou dans la roche. Adam, il a vu un trou dans Pail. Amelie, elle a entendu un cri dans la nuit. Hugo, il a vu un cou dans le dessin. Philippe, il a vu un site dans le terrain de camping. Samuel, il a vu une soupe dans le livre de recettes. Vincent, il a vu un nid dans l'arbre. Etienne, il a vu une trousse dans les objets perdus. Daphn6e, elle a vu la lune dans le ciel. Guillaume, il a vu une crique dans la foret. Megan, elle a entendu un cri dans la nuit. Rosalie, elle a vu une trousse dans les objets perdus. Arianne, elle a vu un jus dans le frigidaire. Raphael, il a vu une jupe dans la vitrine. Isabelle, elle a vu un grand cru dans la cave a vin. Antoine, il a vu un saoul dans le bar. C6dric, il a vu une luge dans la neige. Gabrielle, elle a vu une puce dans la perruque. Fran?ois, il a vu un eric dans la valise de Pauto. Jerome, il a vu du pus dans la plaie. Claudia, elle a vu une coupe dans le salon de coiffure. Valerie, elle a vu une niche dans la cour. Catherine, elle a vu une coupe dans le salon de coiffure. Alexandra, elle a recu un coup dans la face. Alexandre, il a vu une scie dans la boite a outils. Marc-Olivier, il a vu une cruche dans les ruines grecques. Julien, il a vu une troupe dans la parade. Mdlodie, elle a vu une niche dans la cour.  125  Block 5 Tristan, il a vu un nu dans le musee. Louis, il a vu une loupe dans le pupitre. Audrey, elle a vu une pousse dans le jardin. Marie Eve, elle a vu un pou dans les poils du rat. Francois, il a vu un lit dans la clinique. Benoit, il a mis de la vie dans la maison. Justine, elle a vu un bouc dans le champ. Samuel, il a vu un loup dans la foret. Gabriel, il a vu une lime dans la sacoche. Amelie, elle a vu une cuve dans la salle de bain. Alexandre, il a vu le cul de la chevre. Maxime, il a vu une pie dans I'arbre. Arianne, elle a vu un pouce dans la radiographic Marc-Olivier, il a vu une ville dans la vallee. B6atrice, elle a vu une pie dans I'arbre. Chloe, elle a vu une rue dans la vieille ville. Nicolas, il a vu de la boue dans Pentr6e de la maison. Isabelle, elle a vu une tuque dans le linge sale. Sebastien, il a vu une ruche dans le champ. Cedric, il a fait un vide dans la maison. Genevieve, elle a vu un tube dans le trou. Camille, elle a vu une frite dans Passiette. Florence, elle a vu une pipe dans la bouche du monsieur. Alexandra, elle a vu un lys dans Peteng. Christophe, il a vu du riz dans le chaudron. Hugo, il a vu un tutu dans le vestiaire. Elizabeth, elle a vu un cube dans la boite k jouets. Thomas, il a vu un loup dans la foret. Frederic, il a vu une nuque dans le dessin. Rosalie, elle a vu une russe dans le ballet.  Block 6 Melodie, elle a vu une niche dans la cour. Felix, il a vu un grand cru dans la cave a vin. Antoine, il a vu un saoul dans le bar. Alicia, elle a vu une cruche dans les ruines grecques. Julien, il a vu une jupe dans la vitrine. Guillaume, il a entendu un cri dans la nuit. Valene, elle a vu une scie dans la boite a outils. Catherine, elle a vu une troupe dans la parade. Veronique, elle a vu du pus dans la plaie. Jerome, il a vu un nid dans I'arbre. Claudia, elle a vu une trousse dans les objets perdus. Gabrielle, elle a vu une crique dans la foret. Simon, il a vu une coupe dans le salon de coiffure. Emilie, elle a vu un trou dans la roche. Juliette, elle a vu un trou dans Fail. M6gan, elle a vu la lune dans le ciel. Olivier, il a vu un cou dans le dessin. Eric, il a vu une lutte dans le bar. Mathieu, il a vu une coupe dans le salon de coiffure. Raphael, il a vu une niche dans la cour. Annabelle, elle a entendu un cri dans la nuit. Vincent, il a vu une troupe dans la parade. Zachary, il a vu une trousse dans les objets perdus. Xavier, il a vu un jus dans le frigidaire. Daphnee, elle a recu un coup dans la face. Etienne, il a vu une soupe dans le livre de recettes. Philippe, il a vu une luge dans la neige. Sophie, elle a vu une puce dans la perruque. Alexia, elle a vu un eric dans la valise de I'auto. Adam, il a vu un site dans le terrain de camping.  126  Appendix VIII Sentences used in foreign accent rating task English:  English sentences used for each subject 2 (mono E)  16 17 18 ; x> 19 S 5 (mono E) c 6 (mono E) -t-» TP 20 21 OO other (E-dom) 22 23 24  Stimuli block used  First word of sentences used (see Appendix VI for full sentences)  Block 2 Block 2 Block 1 Block 2 Block 2 Block 2 Block 2 Block 2 Block 2 Block 2 Block 2 Block 2 Block 2  Hank, Maggie, Ken, Tony, Ben Hank, Maggie, Tony, Ben, Tom Dave, Vance, Danny, Joanna, Edwin Hank, Maggie, Ken, Tony, Ben Hank, Maggie, Ken, Tony, Ben Hank, Maggie, Ken, Tony, Ben Hank, Maggie, Ken, Tony, Ben Hank, Tony, Ben, Joseph, Dean Hank, Maggie, Ken, Tony, Ben Hank, Maggie, Ken, Tony, Ben Hank, Tony, Ben, Tom, Joseph Hank, Maggie, Ken, Tony, Ben Hank, Maggie, Ken, Tony, Ben  French:  French sentences used for each subject  Subject number  2 (mono E)  16 17 18 other (mono F) 19 5 (mono E)  20 Other (mono F) 21 other (E-dom) 22 23 24  Stimuli block used  First word of sentences used (see Appendix VII for full sentences)  Block 1 Block 1 Block 1 Block 1 Block 1 Block 1 Block 1 Block 1 Block 1 Block 1 Block 1 Block 1 Block 2 Block 1  Chloe, Samuel, Benoit, Xavier, Sebastien Chloe, Samuel, Benoit, Xavier, Sebastien Zachary, Chloe, Samuel, Benoit, Xavier Chloe, Samuel, Benoit, Xavier, Sebastien Chloe, Samuel, Benoit, Xavier, Sebastien Samuel, Benoit, Xavier, Sebastien, Annabelle Chloe, Samuel, Benoit, Xavier, Sebastien Chloe, Samuel, Benoit, Xavier, Sebastien Samuel, Benoit, Xavier, Sebastien, Annabelle Chloe, Samuel, Benoit, Xavier, Sebastien Nicolas, Zachary, Chloe, Samuel, Benoit Chloe, Samuel, Benoit, Xavier, Sebastien Raphael, Daphnee, Beatrice, Marc-Olivier, Rosalie Samuel, Benoit, Xavier, Annabelle, Valerie  127  Appendix IX Bilingual-mode stimuli used Block 1 contains the first 15 sentences from Fr. Block 1 & the first 15 from Eng. Block 1. B l o c k 2 contains the last 15 sentences from Fr. Block 1 & the last 15 from Eng. Block 1. The order o f language that comes next is different between blocks and was decided randomly.  Block 1  Block 2  Gavin said "hear ee" each spring. Vincent, il a vu une ville dans la vallee. Alexandre, il a vu une nuque dans le dessin. Dave said "hear hee" each June. Vance said "hair hay" each time again. Danny said "hoar owe" each minute. Catherine, elle a vu une loupe dans le pupitre. Joanna said "who roo" each holiday. Edwin said "hair A" each time again. Whitney said "hoar hoe" each day with Sue. Fred6ric, il a vu un loup dans la foret. Casey said "hear ee" each scenario. Veronique, elle a vu une russe dans le ballet. Simon, il a vu un loup dans la foret. Arianne, elle a vu une pousse dans le jardin. Gabrielle, elle a fait un vide dans la maison. Mike said "hee wee" each fall. Seth said "hear ee" each week. Sophie, elle a vu une pipe dans la bouche du monsieur. Juliette, elle a vu un cube dans la boite a jouets. Ann said "har awe" each turn in Thai. Betty said "har haw" each April. Philippe, il a vu du riz dans le chaudron. Heidi said "hee ree" each week. Judy said "hear ee" each April. Thomas, il a mis de la vie dans la maison. Audrey, elle a vu une ruche dans le champ. Gabriel, il a vu le cul de la chevre. Alicia, elle a vu un pouce dans la radiographic Joyce said "hay ray" each holiday.  Jamie said "hay ray" each month. Claudia, elle a vu une rue dans la vieille ville. Ben said "hee wee" each month. Tanya said "hoe roe" each job. Jerome, il a vu une pie dans I'arbre. Otto said "har haw" each lunch. Marie Eve, elle a vu un lit dans la clinique. £tienne, il a vu une cuve dans la salle de bain. Cindy said "har haw" each show. Jenny said "hee ree" each scenario. Becky said "har haw" each autumn. Nicolas, il a vu un tube dans le trou. Zachary, il a vu un nu dans le musee. Mason said "hoar hoe" each term. Matt said "hee wee" each time through. Chloe, elle a vu un bouc dans le champ. Samuel, il a vu une pie dans I'arbre. Benoit, il a vu une lime dans la sacoche. Chuck said "haw yaw" each job. Xavier, il a vu un tutu dans le vestiaire. Nate said "hair A" each show. S6bastien, il a vu une tuque dans le linge sale. Annabelle, elle a vu un lys dans Petang. Jon said "hoar owe" each semester. Valerie, elle a vu un pou dans les poils du rat. Nick said "hee ree" each time through. Sandy said "who roo" each fall. AmeJie, elle a vu de la boue dans Pentree de la maison. Stan said "har awe" each time again. Alexandra, elle a vu une frite dans l'assiette.  128  Appendix X Geometrical formulas used in the MATLAB code LINGUISTIC PROJECT M A C I E J MIZERSKI F O R IAN W I L S O N  1. B O N E POINT MOVES WITH THE HEAD  Let mi. rn2 and m be the vectors in the lab coordinates of 3 markers at any time t where take mj be the marker behind the head (these are functions of t). Define 3  i » l ( t ) : = m (t) -  mi(t)  it (t) : = m (t) -  rni(t)  2  2  3  V3(t)  := vi(t) X v (t) 2  This will be the basis for the new coordinate system centered at mi: the coordinate system of the head. For better results choose m and m to be the coordinates of the extreme points on the glasses. Recall the vector product formula: ( 0 1 , 0 2 , 0 3 ) x (cj, co, c ) = (0203 - 0 3 C 2 , a 3 C i — a i C 3 . ajca — 0201) We need to express b(t) — mi(t) in the coordinates U ) ( t ) , i>2.(t) and V2(t): (1) b(r) -m,(t) = civi(t) + C2V (t) + C3V (t) and so 2  3  3  2  (2)  b(t) =mi(t)  + civi(t) +  3  c v (t)+C3V3(t). 2  2  where c i , c and C3 are constants that do not change with time as b(t) is a fixed point on the head. To find these let M(t) be the matrix whose columns are v\(t), v (t) and v (t): 2  2  3  w(t)-(«i(t)h(t)hW)  then ^ cj j .  b(t)- (t)=M(t)mi  Hence we express c C2 and c by inverting M(t): lf  3  ^  ca j  -M^t)-*  .(b(t)- (t)) mi  At some time t = 0, measure the coordinates of the bone 6(0) in the coordinates of the lab, so we can find (3)  ( ca \ =M(0)- .(6(0)-m,(0)) l  Date: May 12, 2005.  129  2  MACIEJ MIZERSKI FOR IAN WILSON  Then at any other time t, b(t) is given by the equation 2: b(t)  = rni(t)  + ciVi(t)  + c v (t) 2  + c V3(t)  2  =  3  rni(f) +  M{t)  • c  where  (z) Thus the problem is reduced to inverting the matrix A'/(0) which is a 3 by 3 matrix, you will find formulas for this in linear algebra books. There is an alternate method for finding c's. Use the fact that is orthogonal to v\ and v (that is v\ • i>3 = vo • i<2 = 0). Taking the scalar product of I with v\, V2 and 03, get 2  (fc(0) - mi(0)) • «i(0) = d(i/,(0) • i>i(0))  +c (v (0) 2  2  •  »,(0))  (i(0) - m , ( 0 ) ) • trj(O) = c,( (0) • »j(0)') + c (t> (0) • «a(0)) Wl  2  2  (t(0) - mt(0)) • « (0) = <a(t*(0) • t*s(0)) 3  Thus / c! \ _ / t. (0)  uia(0) \  u  _  / «M(0)\  1  1 ~ «u(0)«aa(0)-«ia(0) ' f «22(0) -tli (0) \ / «M(0) \ \, - u ( 0 ) i-n(O) J \, «62(0) J s  2  I 2  and t'fc3(0)  =  t*3(0)  0 3  where define «u(0) : = « i ( 0 ) •«i(0) W12(0) : = t n ( 0 )  • ^2(0)  V22(0)  • wa(0)  :=  vo(0)  vsa(0) := «3(0) •vs(0) fw(0) := (fe(O) - m i ( 0 ) ) . « , ( 0 ) for t = 1,2 and 3. 2. F R O M PROBE COORDINATES TO LAB COORDINATES  It Ls the same idea for the coordinate transformation. Instead of the 3 markings on the head we use: p (resp. mjj and me) the vectors in lab coordinates of the probe center (resp. marking #5 and  130  LINGUISTIC  PROJECT  3  #6). Define t»i  :=  ms -  p  «2 := m a -  p  U3 : =  l'i  X t>2  So the position of the bone b is given by b — p = oit'i + aovo. + a3«>3  but 03 = 0 as we assume that the bone is on the cross section (i.e. the plane spaned by vi and 1*2): b — p = a ivi + a 2 « 2  Take the measuremenet of b — p in the probe coordinates. Express v\ and vo, in probe coordinates. Take the scalar products: (b-p)  - t i i = ai(vi • vy) +a (v 2  v,)  2  (b — p) • V2 = a i ( v i • 112) + 02(^2 • v ) 2  And we can isolate Qi and ao as follows:  (4)  ( \ = ( ~ "2 • V • C ^ ^' ^ a i  \ / \ 5  v i  "2 /  \  vi  V\ • V2  1,1  Vo • Vo  _P  1  J  \ (b — p) • Vo, J  _  /  [vi • i.'i)(v • v ) - ("2 • vi) Now b in lab coordinates is given by  2  2  1,1  2  •u  -vi • v\ \ , (  2  \ -vt • v  2  - p) • vi \ ( ~P)-V2 J b  in Vi J \ (b  b = p + a\L<i + a U2 2  where all vectors are expressed in lab coordinates. 3. FROM LAB COORDINATES TO PROBE COORDINATES  Have rng, m«, p andfcin lab coordinates. Define uj, «2 as above. Compute v 1 • vi, v\ • V2, (b — p) • vi. (b — p) •  t>2  and compute a\ and ao using 4. Then after expressing vi, v and p in probe coordinates (p = 0 as it is the origin of tlu's coordinate system), get 2  fc — an'i  +  £12^2  the position of the bone in probe coordinates. U N I V E R S I T Y OF BRITISH COLUMBIA, D E P A R T M E N T OP M A T H E M A T I C S E-mail address: mizerskiOmath.ubc. ca  131  Appendix XI Means and standard deviations for all 24 subjects In the following tables, individual means and within-subject standard deviations are given for each component of ISP for each subject. Note that for each monolingual subject, these means and standard deviations are based on a number of tokens approximately equal to the number shown in the rightmost column of Table 2.2 in Section 2.1.3.2. For each bilingual subject, these means and standard deviations are based on a number of tokens approximately equal to the number shown in the third and fourth columns of Table 3.5 in Section 3.1.3.2. If there were cases where the tongue line was not visible on the ultrasound image, the number of tokens may be slightly fewer than the numbers shown in those tables. The numbers given here for the bilingual subjects are for monolingual mode only.  132  Bilingual (NS of neither)  Bilingual (NS of Eng only)  Bilingual (NS of Fre only)  Bilingual (NS of both)  Monolingual  Monolingual  Tongue and j a w Subject  TTht  TBht  TDht  TRrn  1 - Eng 2 - Eng 3 - Eng 4 - Eng 5 - Eng 6 - Eng 7 - Eng 8 - Fre 9 - Fre 10 - Fre 11 - Fre 12 - Fre 13 - Fre 14 - Fre 15 - Fre 21 - Eng - Fre 17 - Eng - Fre 22 - Eng - Fre 19 - Eng - Fre 18 - Eng - Fre 23 - Eng - Fre 20 - Eng - Fre 24 - Eng - Fre  62.97 (1.72) 64.68 (2.33) 58.02 (1.68) 64.94 (1.17) 72.88 (2.46) 58.33 (1.41) 65.35 (1.06) 55.46 (1.23) 56.56 (2.09) 61.30 (2.42) 64.50 (1.54) 53.96 (1.93) 61.12 (1.82) 57.24 (1.82) 56.66 (1.46) 68.92 (1.73) 67.55 (2.24)  61.99 (2.87) 61.72 (3.02) 67.26 (2.24) 67.94 (1.41) 71.98 (1.91) 62.08 (2.24) 71.88 (1.68) 60.96 (1.20) 61.44 (1.77) 63.31 (2.43) 72.19 (1.54) 53.50 (1.84) 68.23 (2.05) 63.44 (1.65) 63.55 (1.63) 69.41 (1.27) 68.98 (1.33)  53.93 (3.63) 56.11 (2.92) 61.10 (2.59) 58.20 (2.29) 64.98 (2.11) 42.62 (3.43) 70.38 (1.66) 55.78 (2.61) 54.11 (2.59) 61.94 (2.30) 61.12 (2.89) 44.97 (1.73) 62.40 (2.69) 53.66 (3.42) 57.18 (2.59) 61.86 (1.68) 62.05 (2.16)  45.23 (3.90) 43.64 (4.37) 51.42 (2.25) 44.85 (1.80) 58.25 (2.00) 34.26 (3.22) 62.57 (2.14) 47.21 (3.03) 43.54 (2.80) 59.40 (3.55) 54.68 (2.24) 38.34 (1.77) 51.15 (3.80) 44.91 (3.10) 50.27 (2.45) 46.86 (3.79) 49.21 (1.89)  JAW1 8.26 (0.96) 14.35 (1.98) 2.29 (0.70) 3.65 (0.96) 5.30 (0.73) 7.45 (1.14) 3.20 (0.63) 4.85 (0.96) 4.68 (1.22) 9.96 (1.63) 6.22 (1.14) 5.30 (2.86) 11.59 (1.44) 4.20 (1.36) 5.48 (0.91) 5.45 (1.35) 3.97 (0.60)  n/a  n/a  n/a  n/a  n/a  60.83 (1.33) 55.50 (2.42) 55.86 (1.75) 61.00 (1.77) 60.60 (2.52) 63.19 (1.72) 64.65 (1.23) 64.16 (1.04) 62.63 (0.99) 63.00 (1.85) 66.43 (1.71) 65.86 (1.44) 65.23 (1.44)  62.81 (1.91) 57.94 (3.62) 57.37 (2.71) 70.14 (1.42) 68.56 (2.21) 68.48 (1.45) 65.86 (2.21) 75.14 (1.56) 71.95 (2.29) 66.56 (1.43) 67.53 (1.92) 64.24 (1.85) 62.18 (3.08)  56.64 (1.80) 56.53 (2.32) 55.50 (2.02) 67.03 (1.49) 65.49 (1.80) 63.89 (1.70) 61.92 (1.96) 66.19 (1.90) 64.88 (1.14) 55.88 (1.79) 54.66 (2.98) 55.65 (2.72) 56.26 (2.60)  55.98 (2.23) 44.90 (2.64) 44.87 (2.19) 48.19 (3.84) 49.34 (2.95) 56.61 (3.20) 54.28 (2.73) 57.09 (2.92) 56.13 (1.32) 45.92 (3.37) 44.57 (3.34) 43.25 (3.08) 45.62 (1.62)  5.79 (0.49) 2.28 (0.65) 2.30 (0.17) 3.40 (0.36) 3.92 (0.35) 2.05 (0.69) 3.61 (1.22) 14.63 (0.93) 13.82 (1.03) 11.07 (1.44) 10.27 (1.59) 8.82 (0.98) 7.80 (1.11)  16 - Eng - Fre  72.40 (0.92) 72.98 (2.01)  79.92 (2.67) 79.76 (1.61)  77.07 (4.00) 80.29 (2.92)  70.09 (2.45) 72.89 (1.50)  10.03 (0.56) 8.81 (0.63)  133  Bilingual (NS of neither)  Bilingual (NS of Eng only)  Bilingual (NS of Fre only)  Bilingual (NS of both)  Monolingual  Monolingual  L i p height and protrusion Subject  ULIo  LLIo  ULpr  LLpr  1 - Eng 2 - Eng 3 - Eng 4 - Eng 5 - Eng 6 - Eng 7 - Eng 8 - Fre 9 - Fre 10 - Fre 11 - Fre 12 - Fre 13 - Fre 14 - Fre 15 - Fre 21 - Eng - Fre 17 - Eng - Fre 22 - Eng - Fre 19 - Eng - Fre 18 - Eng - Fre 23 - Eng - Fre 20 - Eng - Fre 24 - Eng - Fre  78.99 (0.59) 70.83 (0.70) 75.67 (0.44) 76.70 (0.33) 74.80 (0.68) 74.64 (1.22) 76.14 (0.26) 74.70 (0.55) 65.34 (0.96) 80.97 (0.62) 70.98 (0.59) 66.47 (2.82) 71.92 (0.19) 76.13 (0.92) 72.52 (0.21) 78.17 (0.80) 77.96 (0.72)  99.45 (1.13) 98.63 (2.25) 91.11 (0.67) 98.19 (1.05) 96.06 (1.87) 95.02 (2.56) 101.52 (0.88) 95.83 (1.42) 83.47 (2.04) 105.16 (2.05) 102.53 (1.45) 90.22 (3.86) 100.00 (1.00) 96.63 (2.36) 95.44 (0.76) 99.56 (1.16) 98.61 (0.55)  44.71 (3.82) 24.72 (2.05) 23.84 (1.90) 32.96 (1.39) 32.75 (4.15) 29.79 (2.70) 33.82 (1.32) 18.37 (1.57) 17.38 (3.38) 27.21 (2.96) 27.73 (1.35) 21.32 (1.33) 24.90 (1.28) 30.39 (2.06) 21.49 (1.62) 27.04 (1.10) 26.05 (1.53)  48.93 (5.48) 30.55 (2.95) 28.10 (2.45) 39.06 (2.08) 37.13 (5.90) 32.09 (3.52) 36.14 (1.90) 20.11 (2.06) 19.50 (4.99) 30.34 (3.79) 30.62 (2.14) 27.76 (1.81) 27.28 (1.82) 34.47 (2.75) 25.96 (2.45) 28.53 (1.31) 27.31 (1.83)  n/a  n/a  n/a  n/a  66.59 (0.44) 71.11 (0.57) 72.17 (0.37) 68.84 (0.28) 68.77 (0.28) 75.74 (0.59) 76.36 (0.34) 71.21 (0.26) 70.86 (0.56) 73.73 (1.27) 73.37 (1.79) 71.29 (0.32) 71.45 (0.59)  96.63 (0.74) 96.38 (2.58) 94.80 (0.89) 94.33 (0.61) 95.00 (0.83) 94.56 (0.62) 95.41 (0.48) 90.34 (1.10) 89.40 (0.96) 102.25 (2.05) 101.26 (2.27) 98.88 (1.99) 98.23 (1.85)  26.01 (0.88) 27.90 (0.63) 25.64 (1.42) 26.02 (1.00) 25.22 (0.97) 24.09 (1.72) 28.41 (3.34) 31.42 (0.83) 28.05 (1.04) 23.91 (3.61) 21.72 (4.66) 21.46 (0.97) 21.93 (4.47)  30.93 (1.59) 28.38 (0.77) 24.45 (2.14) 29.51 (0.98) 27.98 (1.44) 27.12 (2.17) 32.53 (4.37) 39.44 (1.52) 34.86 (1.68) 26.57 (5.67) 23.31 (6.26) 26.95 (1.70) 28.01 (6.23)  16 - Eng - Fre  75.23 (0.62) 76.57 (0.44)  100.21 (0.56) 97.87 (0.58)  29.72 (1.73) 29.78 (2.03)  35.48 (2.26) 34.55 (2.44)  134  Bilingual (NS of neither)  Bilingual (NS of Eng only)  Bilingual (NS of Fre only)  Bilingual (NS of both)  Monolingual  Monolingual  Lip aperture and narrowing Subject  Lvap  Lhap  Lnar  1 - Eng 2 - Eng 3 - Eng 4 - Eng 5 - Eng 6 - Eng 7 - Eng 8 - Fre 9 - Fre 10 - Fre 11 - Fre 12 - Fre 13 - Fre 14 - Fre 15 - Fre 21 - Eng - Fre 17 - Eng - Fre 22 - Eng - Fre 19 - Eng - Fre 18 - Eng - Fre 23 - Eng - Fre 20 - Eng - Fre 24 - Eng - Fre  21.31 (1.35) 27.70 (2.45) 15.99 (0.80) 21.62 (1.19) 21.41 (2.28) 20.82 (3.39) 26.06 (1.00) 21.13 (1.83) 18.16 (2.67) 24.53 (1.77) 31.81 (1.94) 23.54 (2.67) 28.52 (1.01) 20.71 (3.03) 22.75 (0.91) 21.69 (1.22) 20.99 (0.67) n/a 30.31 (0.84) 25.81 (3.06) 23.32 (1.12) 26.00 (0.71) 26.81 (1.02) 18.88 (0.35) 19.11 (0.43) 19.12 (1.12) 18.53 (0.95) 28.75 (2.41) 28.14 (2.16) 27.59 (2.00) 26.80 (1.53)  58.63 (0.80) 69.16 (0.89) 59.71 (0.46) 59.64 (0.30) 60.93 (0.68) 58.94 (0.36) 66.73 (0.37) 55.99 (0.28) 59.53 (0.40) 72.43 (0.81) 58.10 (0.47) 55.84 (0.45) 65.21 (0.40) 57.76 (0.61) 61.62 (0.34) 65.34 (1.35) 65.96 (0.73) n/a 59.49 (0.38) 55.70 (0.20) 55.55 (0.30) 60.89 (0.26) 60.70 (0.18) 68.57 (0.29) 68.36 (0.63) 69.92 (0.59) 70.64 (1.11) 60.05 (1.58) 61.21 (2.34) 66.65 (0.34) 67.67 (0.66)  20.28 (0.80) 19.26 (0.89) 14.93 (0.46) 16.94 (0.30) 15.46 (0.68) 4.13 (0.36) 7.76 (0.37) 5.21 (0.28) 1.95 (0.40) 13.65 (0.81) 5.50 (0.47) 9.11 (0.45) 8.77 (0.40) 9.34 (0.61) 5.52 (0.34) 15.81 (1.35) 15.19 (0.73) n/a n/a n/a n/a 9.46 (0.26) 9.64 (0.18) 7.78 (0.29) 7.99 (0.62) 13.90 (0.59) 13.18 (1.11) 13.49 (1.58) 12.33 (2.34) 15.79 (0.34) 14.77 (0.66)  16 - Eng - Fre  25.19 (0.88) 21.47 (0.77)  62.89 (1.12) 61.73 (0.63)  13.47 (1.12) 14.63 (0.63)  135  Appendix XII Detailed statistics for Table 3.2 The following graphs and tables are extracted directly from JMP statistical analysis software. They show the results of t tests (assuming unequal variances) comparing the 7 individual English means to the 8 individual French means for each component of ISP. On the horizontal axis, "1" means English and "2" means French.  rjSOneway Analysis of TTht avg all contexts By Lang 1E;2F 75-  x  70H  c  o =  65-  Or  >  ra  Z 60H 55H  Lang 1E;2F •j Means and Std"lteVfaiionT Level  Number  Mean  Std Dev  Std Err Mean  1 7 63.8834 5.01672 1.8961 2 8 58.3507 3.56825 1.2616 •j t Test 1-2 Assuming unequal variances Difference 5.5327 t Ratio 2.429313 Std Err Dif 2.2775 DF 10.69191 Upper CLDif 10.5631 Prob > |t| 0.0340 Lower CL Dif 0.5023 Prob > t 0.0170 Confidence 0.95 Prob < t 0.9830  136  Lower  95%  59.244 55.368  Upper  95%  68.523 61.334  tfMOneway Analysis of TBht avg all contexts By Lang 1 E ; 2 F 70 H 1/1  QJ C  O  65-  av  OV  60-  *-*  oo l_  *  55H  Lang 1E;2F  • l Means and Std Deviations Level 1 2  Number 7 8  Mean 66.4098 63.3265  Std Dev 4.54969 5.45679  Std Err Mean 1.7196 1.9293  • i t Test 1-2 Assuming unequal variances Difference 3.0833 t Ratio Std Err Dif 2.5844 DF Upper CL Dif 8.6673 Prob > |t| Lower CL Dif •2.5008 Prob > t Confidence 0.95 Prob < t  1.193024 12.98153 0.2542 0.1271 0.8729  137  Lower 95% 62.202 58.765  Upper 95% 70.618 67.888  vHB Oneway Analysis of TDht avg all contexts By Lang 1 E ; 2 F 70-  s c o  6560-  "S  > 55ra  *-*  x: Q h-  504540-  Lang 1E;2F  Means and Std Deviations Level l 2  Number 7 8  Mean 58.1892 56.3939  Std Dev 8.83149 5.77336  Std Err Mean 3.3380 2.0412  Lower 95% 50.021 51.567  • i t Test 1-2 Assuming unequal variances Difference 1.795 t Ratio Std Err Dif 3.913 DF Upper CL Dif 10.500 Prob > Lower CL Dif - 6 . 9 0 9 Prob > Confidence 0.95 Prob <  0.458856 10.11401 0.6560 0,3280 0.6720  138  / -10  /  Upper 95% 66.357 61.221  ri S Oneway Analysis of TRrn avg all contexts By Lang 1 E ; 2 F 6560X Ol  c o  5550-  > 45C at  |—  403530-  Lang 1E;2F  • i Means and Std Deviations Level 1 2  Number 7 8  Mean 48.6031 48.6875  Std Dev 9.59189 6.64656  Std Err Mean 3.6254 2.3499  • i t Test 1-2 Assuming unequal variances Difference - 0 . 0 8 4 4 t Ratio Std Err Dif 4.3204 DF Upper CL Dif 9.4790 Prob > Ul Lower CL Dif - 9 . 6 4 7 8 Prob > t Confidence 0.95 Prob < t  -0.01954 10.51052 0.9848 0.5076 0.4924  139  Lower 95% 39.732 43.131  Upper 95% 57.474 54.244  r E Oneway Analysis of JAWI avg all contexts By Lang 1E;2F 15-  12.5107.5H  03 CSV >  I  2.5-  Lang 1E;2F  Vi Means and Std Deviations Level 1 2  Number  Mean  Std Dev  Std Err Mean  Lower 95%  Upper 95%  7 8  6.35707 6.53374  4.15665 2.71762  1.5711 0.9608  2.5128 4.2617  10.201 8.806  t Test 1-2 Assuming unequal variances Difference - 0 . 1 7 6 7 t Ratio Std Err Dif 1.8416 DF Upper CL Dif 3.9203 Prob > |t| Lower CL Dif - 4 . 2 7 3 7 Prob > t Confidence 0.95 Prob < t  -0.09593 10.11479 0.9254 0.5373 0.4627  140  r E Oneway Analysis of ULlo avg all contexts By Lang 1 E ; 2 F  LA  80 H  X  -  <v o u  75-  "ro  > ro o  _J  70-  65-  Lang 1E;2F  v Means and Std Deviations Level 1 2  Number 7 8  Mean 75.3947 72.3779  Std Dev 2.48669 5.07040  Std Err Mean 0.9399 1.7927  • i t Test 1-2 Assuming unequal variances Difference 3.0168 t Ratio Std Err Dif 2.0241 DF Upper CL Dif 7.5003 Prob > |t| Lower CL Dif - 1 . 4 6 6 7 Prob > t Confidence 0.95 Prob < t  1.49044 10.4556 0.1656 0.0828 0.9172  141  Lower 9 5 % 73.095 68.139  Upper 95% 77.695 76.617  • l Means and Std Deviations Level 1 2  Number 7 8  Mean 97.1414 96.1594  Std Dev 3.41264 6.90319  Std Err Mean 1.2899 2.4406  • i t Test 1-2 Assuming unequal variances Difference 0.9820 t Ratio Std Err Dif 2.7605 DF Upper CL Dif 7.0933 Prob > |t| Lower CL Dif -5.1294 Prob > t Confidence 0.95 Prob < t  0.355712 10.50062 0.7291 0.3645 0.6355  142  Lower 95% 93.985 90.388  Upper 95% 100.30 101.93  0 Oneway Analysis of ULpr avg all contexts By Lang l l j z F 454035-  n  30-  >  rt)  2520-  15  Lang 1E;2F  • l Means and Std Deviations Level  l 2  Number  Mean  Std Dev  Std Err Mean  7 8  31.7995 23.5994  6.95836 4.68467  2.6300 1.6563  •i't Test 1-2 Assuming unequal variances Difference 8.2000 t Ratio Std Err Dif 3.1081 DF Upper CL Dif 15.0969 Prob > It! Lower CL Dif 1.3031 Prob > Confidence 0.95 Prob <  2.638283 10.31259 0.0242 0.0121 0.9879  143  95%  Upper 95%  25.364 19.683  38.235 27.516  Lower  rj B Oneway Analysis of LLpr avg ali contexts By Lang 1E;2F 50451/1  OJ  c o  =  403 5  ~  ro  o> ro  a —j _i  30-  2520-  15-  Lang 1E;2F  Vj Means and Std Deviations Level 1 2  Number 7 8  Mean 36.0015 27.0052  Std Dev 6.89646 5.14720  Std Err Mean 2.6066 1.8198  TTrfest 1-2 Assuming unequal variances Difference 8.9963 t Ratio Std Err Dif 3.1790 DF Upper CL Dif 15.9911 Prob > Lower CL Dif 2.0016 Prob > Confidence 0.95 Prob <  2.829912 11.02861 0.0163 0.0082 0.9918  144  Lower 95% 29.623 22.702  Upper 95% | 42.380 31.308  H El Oneway Analysis of Lvap avg all contexts By Lang 1E;2F 30H  o  u  2 5  .  CO  > £ >  20-  15-  Lang 1E;2F  • l Means and Std Deviations Level 1 2  Number 7 8  Mean 22.1306 23.8946  Std Dev 3.81637 4.41818  Std Err Mean 1.4425 1.5621  •I t Test 1-2 Assuming unequal variances Difference - 1 . 7 6 4 0 t Ratio Std Err Dif 2.1262 DF Upper CL Dif 2.8293 Prob > Itl Lower CL Dif - 6 . 3 5 7 4 Prob > t Confidence 0.95 Prob < t  -0.82967 12.99991 0.4217 0.7891 0.2109  145  Lower 9 5 % 18.601 20.201  Upper 9 5 % 25.660 27.588  r El Oneway Analysis of Lhap avg all contexts By Lang 1E;2F  7 0 H  x 4-"  c o _u  65-1  at > Q. ca  60H  5  55-  Lang 1E;2F  •rMeaHs and Std Deviations Level  1 2  Number  7 8  Mean  61.9613 60.8110  Std Dev  4.20848 5.62271  Std Err Mean  • l t Test  1-2 Assuming unequal variances Difference 1.1504 t Ratio Std Err Dif 2.5460 DF Upper CL Dif 6.6621 Prob > |t| Lower CL Dif -4.3613 Prob > t Confidence 0.95 Prob < t  0.451842 12.74012 0.6590 0.3295 0.6705  146  1.5907 1.9879  Lower  95%  58.069 56.110  Upper  95%  65.854 65.512  I B Oneway Analysis of Lnar avg all contexts By Lang 1 E ; 2 F 20H X  o  15-  o» 10>  ,-  re rd  S  5-  Lang 1E;2F  • l Means and Std Deviations Level 1 2  Number 7 8  Mean 14.1079 7.3814  Std Dev 5.98555 3.57144  Std Err Mean 2.2623 1.2627  • i t Test 1-2 Assuming unequal variances Difference 6.7265 t Ratio Std Err Dif 2.5909 DF Upper CL Dif 12.5383 Prob > |t| Lower CL Dif 0.9147 Prob > t Confidence 0.95 Prob < t  2.596244 9.527966 0.0277 0.0138 0.9862  147  Lower 95% 8.5722 4.3956  Upper 95% 19.644 10.367  Appendix XIII Questionnaire filled out by French subjects This page is part of a questionnaire created by Dr. Alain Desrochers (Cognitive Psychology Laboratory, University of Ottawa, 2003), and was used with his permission.  QUESTIONNAIRE SUR LA FLUIDITE LANGAGIERE DES PERSONNES BILINGUES FRANQAIS-ANGLAIS Section 1 Renseignements generaux Veuillez indiquer: Votre age Votre sexe Votre pays ou province d'origine Votre langue maternelle Vos langues secondes La langue maternelle de votre mere Les langues secondes de votre mere La langue maternelle de votre pere Les langues secondes de votre pere S'il y a lieu, a quel age avez-vous commence a : Le franca is  L'anglais  Autre : precisez  Parler Lire Fieri re Dans quelle(s) langue(s) avez-vous fait vos etudes primaires et secondaires? Veuillez indiquer la langue d'enseignement selon les niveaux specifies cidessous. Mat 1  2  3  4  5  6  Francais Anglais Autres  148  7 8 9 10 11 12 Se1 Se2 Se3 Se4 Se5 C1  13 C2  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0092890/manifest

Comment

Related Items