A Review Of Vowel Phonology

Published Date: 02 Nov 2017

ABSTRACT

Previous research with respect to second language acquisition reveals that language transfer in both the language perception and production from the first language (L1) to the second language (L2). The present study investigates the role of language transfer in native Cantonese (L1-C) and native English (L1-E) speakers who are graduates of a Canadian French Immersion program. Specifically, the experimenter examined the speech production of the front vowels /i/, /y/ and /u/, in all three languages except English, which only has /i/ and /u/. Speech was elicited in a picture-naming task. The goal of this study is to (a) investigate if there is evidence of an L1 language transfer in the production of the French vowels for either or both groups of participants. Given that all of the L1-C individuals also had acquired English, another goal of this study was to investigate (b) if this population demonstrated an L2 English transfer to the L3, French. Acoustic analyses revealed a positive L1 transfer from English to French for the L1-E group. For the L1-C group, an L2 (English) transfer to the L3 (French), for the vowels /u/ and /i/, was found. Individual analyses revealed a possible L1 to L3 transfer for the L1-C group for the production of French-/y/. This work contributes to the field of research investigating the acoustic-phonetic differences between L2 and L3 learners of French in a Canadian French Immersion learning environment, contributing to the understanding of acoustic-phonetic abilities of early bilingual and trilingual speakers.

INTRODUCTION

This study examines the production of the high vowels /i y u/, in two groups of female French Immersion Graduates. One group consists of native Cantonese speakers (L1-C) who have acquired Canadian English as a second language (L2), and Canadian French as a third language (L3) in the context of a Canadian French Immersion program. The other group consists of native Canadian English speakers (L1-E) who also learned Canadian French as a second language (L2) in the context of a Canadian French Immersion Program.

Over the past five decades, many studies have investigated the acquisition of second L2 phonology; however, the majority of these studies subjects who were immigrants, learning the language spoken by their new community. Examples of these study include, but are not limited to, Flege, Bohn & Jang’s (1997)  study on German, Spanish, Mandarin and Korean learners of English in the US, Flege, MacKay & Meador’s (1999) research on Italian immigrants in Canada, and Tsukada et al.’s (2005) study on Korean immigrants in the US. In each of these studies, ‘L2 experience’ is typically defined in terms of length of residence in the country where the L2 is spoken, the age of arrival in the L2-speaking community, or a combination of both factors.

In contrast to the language learning environment of immigrants, second language immersion students is restricted to learning the L2 in a more formal, classroom environment. In the literature, immersion students are also defined as learning their L2 in a Foreign Language Acquisition (FLA) context (Best and Tyler, 2007), since the target language is a foreign language, rather than the language spoken by members of the community outside of the classroom. As stated by Best and Tyler (2007): "FLA listeners, just like L2 listeners, but unlike monolinguals, have exposure to the target language. Yet unlike L2 listeners or monolinguals, FLA listeners have L2 exposure primarily through formal instruction in a restricted setting, with little or unsystematic conversational experience with native speakers" (P.19). Thus, since the language learning environment cannot be measured in terms of length of residence or age-of-arrival, an immersion student’s experience is defined in terms of amount of formal instruction (Simon and D’Hulster, 2012).

The ability to both perceive and produce sounds in the L2 which do not occur in one’s L1 is known to be one of the biggest challenges for an a L2 learner, and often even highly advanced learners often retain a ‘foreign accent’; however, learners’ perception and production of an L2 may improve as a result of phonetic training and/or extended exposure (Simon and D’Hulster, 2012. Bongaerts, (1999) reported that some L2 learners may even be able to acquire an L2 accent which is indistinguishable from native speakers of that language. In their review, Piske, Flege and MacKay (2001) reported a number of variables which have an effect on the degree of foreign accent, including age of L2 learning, length of residence, gender, formal instruction, motivation, language learning aptitude and amount of native language use). The present study focuses on the effect of formal instruction and exposure to the target language in an FLA context.

Three theories, Flege’s (1995) Speech Learning Model and Best’s (1994; 1995) Perceptual Assimilation Model, have attempted to account for differences in production and perception of sounds for L2 learners. Both of these theories make predictions based on L1-L2 segment inventories; however, there are important differences between the two models which are often overlooked (Best & Tyler, 2007). The SLM was designed to explain speech patterns in experienced L2 learners, while the PAM was designed to explain non-native speech perception; therefore, two chief differences between these two models include (a) the population they are addressing, and (b) what they are predicting. Although Best and Tyler (2007) proposed a model to address patterns in speech perception in L2 learners (PAM-L2), their model still does not take into account the perception-production interface for L2 learners. Since the present study involves analyzing the production of speech by French-Immersion students, who are experienced L2 learners, only the predictions of the SLM model may be applicable to present study.

Flege’s (1995) SLM model predicts that a novel L2 sound which is phonetically distinct from sounds in the L1, will be easier to acquire than an L2 sound that is phonetically similar to an L1 sound because it would be perceived as more obviously "different" by the learner. Further, the SLM makes different predictions for the acquisition of perceived phonetically new and phonetically similar sounds by less and more experienced learner: inexperienced learners may at first assimilate new sounds in their L2 to pre-existing categories in their L1; however, with more experience through continued phonetic input, this theory predicts that learners will create a new phonetic category for the new L2 sound. In contrast, sounds which are perceived as phonetically similar are predicted to be more difficult to acquire as the result of equivalence classification (Flege, 1987): the L2 sound is classified as the closest L1 sound, and category formation is blocked. Therefore, both less and more experienced learners are predicted to fail to build new phonetic categories for these perceptually similar sounds.

One study (Simon & D’Hustler, 2012) tested the predictions made by Flege’s (1995) SLM in order to see if the predictions held for learners in an FLA context, where the foreign language was English. In their study, three groups of participants, who differed in the amount of formal instruction in English, were recruited: the first group consisted of monolingual native speakers of Dutch who were studying other languages and were naïve to English; the second group of speakers consisted of non-native speakers of English who had just begun the English-Immersion university program; the third group consisted of L2 English learners, who were in their fourth and final year of their English University Education (Simon and D’Hustler, 2012). The objective of the study was to see if see if the participants could produce the English vowel contrast between the vowels /ɛ/ and / æ /. Since the Dutch vowel system only contains /ɛ/, the SLM predicted that those learning English would be able to make the contrast if they perceived /æ/ as a novel category, and that the participants with the most experience would produce more accurate /æ/ productions. The results of this production task revealed, interestingly, that experience obtained through English education for the L2 learners did not have an important effect on learners’ performance; that is, when compared to the non-native speakers, the L2 learners did not produce the English vowels significantly more target-like (Simon and D’Hustler, 2012). The authors note that all learners’ productions especially deviated for the production of the novel phoneme, /æ/, which they seemed to treat as a ‘new’ phoneme (Flege, 1997), but which they failed to implement in a native-like way. They posited that both groups of participants who were taking English at university had clearly created a new category for /æ/ because they produced it significantly differently from /ε/; whereas the naïve participants produced more overlap between the two vowels than learners who had chosen to study English. Moreover, no participants seemed to create a new category for the ‘similar’ English phoneme /ε/. Instead, the researchers found that the participants directly transferred /ɛ / from Dutch into their production of English. Thus, with the exception that there was no effect on experience between those beginning and those completing the English University studies, researchers concluded that the SLM (Flege, 1987; Flege, 1995) made accurate predictions for L2 learners in an FLA context (Simon & D’Hustler, 2012).

In contrast to the English FLA context which is examined above, the following section (1.1) provides some background information pertaining to Canadian French-Immersion programs, which is the FLA context examined in the present study.

1.1 Background: French Immersion

The first French Immersion (FI) program was implemented in the Canadian school system in 1965, in St. Lambert, Quebec (Barik and Swain, 1978; Cummins, 1983; Genesee, 1984; Roy and Galiev, 2011; Hermanto, Moreno and Bialystok, 2012). Classically, the program was intended to teach monolingual Anglophone children high levels of French proficiency by offering the entire academic curriculum in the French language, despite the fact that these children would normally hear no French within the home environment (Hermanto, Moreno and Bialystok, 2012).In the past five decades, French Immersion programs have expanded to every province in Canada with over 342,000 in 2011 students enrolled in elementary and secondary immersion programs across the country (Friesen, 2013), encouraging bilingualism in both official languages across the nation. According to the Program of International Student Assessment (PISA) conducted in 2000, the Maritime Provinces hold the highest percentage of students enrolled in French immersion programs, with lower percentages enrolled in the Western Provinces. Additionally, more females than males enroll in the programs on average, and overall, these students tend to come from families of higher socio-economic backgrounds (PISA, 2000).

Since French Immersion programs were implemented, a great deal of research has been performed on students enrolled in the program, monitoring their skills and progress throughout. With respect to linguistic research, there is abundant information pertaining to the syntactic speech production of French Immersion students (e.g. Mougeon & Beniak, 1995; Turnbull, Lapkin, & Hart, 1998; Nadasdi, 2001; Mougeon, Rehner & Nadasdi, 2004; Uritescu, Mougeon, Rehner & Nadasdi, 2004 ), but presently there is very limited research completed with respect to the production of French by Immersion students at the phonetic level.

Additionally, researchers have studied French-Immersion students who speak a language other than English or French at home (c.f. Bild & Swain, 1989; Dagenais, Day, & Toohey, 2006; Orpwood, 1980; Swain & Lapkin, 1991; Swain, Lapkin, Rowen, & Hart, 1990; Taylor, 1992). These students are in the minority, and in the past two decades concern has been expressed that such students may experience incomplete development of all languages if they are schooled in early French immersion programs (Hurd, 1993). In fact, according to the successful "immersion formula in the Canadian context (Taylor, 1992), children should be speakers of the majority-language, English. Genessee (2008) speculates that it is for this reason that Early French immersion has not been as popular as Middle French Immersion among heritage language children [1] in Toronto, likely because of a belief that these students should learn English first because it is the dominant language of the community.

That being said, one study examined the progression of a native Cantonese speaker who learned French in an early immersion context (Taylor, 1992). In this case-study, the researcher examined the academic, sociocultural, psychological and linguistic adaptation of French by a young boy named Victor, who learned English and French simultaneously in an early French Immersion program. Victor’s report cards though the years of the longitudinal study were used as an informal measure of the aspects of study, while the experimenter also had several meetings with the student, and administered formal measures, such as standardized tests. Although it was observed that Victor was initially hesitant to speak French, his teachers reported him to be flourishing, both linguistically and academically, after three years of the program. In contrast to Hurd’s (1993) remark noted above, Taylor (1992) reports that early French Immersion programs are suitable for children who do not speak the majority-language, given a supportive home and school environment. Moreover, Taylor (1992) acknowledges that Victor’s situation, as a minority-language speaker, is becoming more common place in multicultural cities such as Toronto, Canada.

In the present study, one group of participants studied is native Cantonese-speakers, and like Victor, they all reported completing an early French Immersion program. In the following sections (1.2a and 1.2b) a review of vowel phonology and previous studies involving the targeted languages in the present study, English, French and Cantonese, will be completed.

1.2a A Review of Vowel Phonology: Canadian English and Canadian French

According to Ladefoged and Maddieson (1996), three most basic ways to describe a vowel are in terms of (1) tongue height, (2) tongue backness and (3) lip-rounding. The acoustic correlates of tongue height, backness and lip-rounding are the first formant (F1), the second formant (F2) and the third formant (F3), respectively. As noted in Figure 1.0, The Canadian English vowel system contains four high vowels, /i/, /ɪ/, /ʊ/ and /u/. In comparison, the Canadian French vowel system additionally has two high front rounded vowels [y] and [ʏ]. In the present study, I will investigate the production of the vowels /i/, /y/ and /u/ in French and Cantonese, and the production of /i u/ in English.

Figure 1.2a: Canadian English Vowel Space (Left) and Canadian French Vowel Space (Right); vowels in parentheses are non-contrastive/allophonic vowels [Adapted from: Haimes-Kusumoto, 2010).

In this figure, one can identify that /i/ appears more front, while /u/ appears to be more back, in the Canadian French vowel space when compared to that of the Canadian English vowel space. Therefore, acoustically, one could remark that English /i/ would have a lower F2 than French /i/, while the value for F2 in English /u/ would be higher than that of the French /u/. These observations are consistent with the research completed by (MacLeod, Stoel-Gammon and Wassink, 2009) who found that the F2 measurements for native monolingual English speakers were lower for /i/ and higher for /u/, when compared to productions of native monolingual French speakers.

MacLeod et al. (2009) investigated the production of the high vowels present in both languages, Canadian English and Canadian French, [i, ɪ, u, ÊŠ], by a group of early bilinguals [2] compared to groups of monolinguals for both languages, English and French. By investigating these populations, these researchers were able to investigate two hypotheses put forth by the SLM: (1) the Equivalence Classification Hypothesis and (2) the Bilingual Category Hypothesis (Flege, 1995). The first hypothesis posits that bilinguals would produce merged categories for similar vowels as demonstrated by a lack of significant differences across the same vowel pairs in two spoken languages, French and English. The second hypothesis asserts that if an interaction exists between the productions in both languages, the bilingual would produce vowels which differ from monolingual productions in both respective languages. Interestingly, the results demonstrated that both hypotheses were rejected since early bilinguals’ vowels were produced in a nearly monolingual-like manner in both languages; thus, researchers concluded that the early bilinguals were capable of forming separate categories for acoustically similar vowels across their two languages.

However, when Levy and Law II (2010) studied the speech production by American-English learners of French in a French Immersion context, this was not what they found. In their study, which specifically looked at the production of the front-rounded vowels /y œ / in bilabial and alveolar contexts, they found this group of speakers produced the non-native vowels with no such accuracy, especially for the high vowel /y/. Although the group of American-English learners clearly did make a contrast between the productions of /u/ when compared to /y/, a contrast which is a well-documented to be difficult for English speakers both perceptually and productively, (e.g. Gottfried, 1984; Flege, 1987; Levy, 2009), they often did so in a non-native manner: producing /y/ as /ju/. Thus, these researchers concluded their study stating that accurate L2 vowel production may vary as a function of (a) the speakers’ language background (b) the vowel being produced and (c) the consonantal context in which the particular vowel is produced. That being said, it is important to note that the judgments of accuracy were based on the American-English Immersion students’ productions in comparison to Parisian French monolinguals, which is unwarranted especially since there is an unlikely chance that the dialect that this group of French-Immersion students learned was that of Parisian French. Therefore, the results of this study are questionable because the researchers did not take into account the importance of the dialect learned, which is crucial to the results of the study (O’Brien & Smith, 2010).

In the context of the present study, the dialects which are investigated are Canadian French and Canadian English (reviewed above), and Hong Kong Cantonese, which is reviewed below.

1.2b A Review of Vowel Phonology: Cantonese

The vowel space for Cantonese speakers is shown below, in Figure 1.2b (Zee, 1999). Since this figure was not created on the same scale as the figure 1.2a, depicting the vowel spaces of Canadian English and Canadian French, it would be ineffective to make observations between the two figures based on a visual inspection.

With respect to Cantonese, no research currently exists pertaining to the production of vowels between Cantonese and French speakers or for Cantonese learners of French. However, research involving the production of vowels and vowel contrasts by Cantonese learners of English and between Cantonese-English bilinguals does exist and will be reviewed in the following paragraphs to provide relevant background information for the participants in the L1-C group of the present study.

Figure 1.2B: Cantonese Vowel Space (Zee, 1999)

In their review, Meng, Zee and Lee (2007) derived salient vowel mispronunciations by native Cantonese speakers learning American English (AE). The aim of their research was mainly pedagogical, and their target learners were adults (university students), who were seeking to improve their pronunciation of AE. These researchers identified major disparities between the two languages, such as missing vowels and diphthongs in either of the languages, and then proceeded to identify production errors in the L2. From their analysis of vowel systems, they propose that vowels which are not within the Cantonese system are assimilated to vowel which is in their vowel system. A summary of their specific observations are found in Table 1.0:

Table 1.2: Salient vowel mispronunciations by (L1)Cantonese learners of (L2) English [Adapted from: Meng, Zee & Lee, 2007]

Additionally, these researchers identified vowel mispronunciations for vowel sounds which are in both languages. For example, both languages contain the lax vowels /ɪ/ and /ʊ/, but these researchers found that words containing these vowels, such as "sit" and "full", were often pronounced as "seat" and "fool". On the basis of this finding, researchers propose that the Cantonese AE learners make mispronunciation based on a tense-lax confusion (Meng, Zee and Lee, 2007); however, the researchers do not clarify if these observations could also be due to a confusion in (a) vowel height: /u/ is higher than /ʊ/, or (b) vowel length: in the case of broad transcription, /u/ is sometimes the longer counter-part of /ʊ/. Additionally they do not consider the possibility that the L2 learners are assimilating the AE vowel to more extreme areas of the vowel space, in order to make clearer distinctions in their perceptual and/or phonetic vowel space, which would be evidence of the Equivalence Hypothesis of Flege’s SLM.

In their study, Chen, Ng & Li (2012), challenged the SLM (Flege, 1995) postulate that sounds of a foreign language that are "familiar" can be produced with less accuracy than sounds that are "new" to L2 learners. This research involved the production of English vowels by Cantonese-English Bilinguals was conducted. In their study, new and familiar categories were determined first by a preliminary perceptual experiment, where a group of monolingual Cantonese speakers who were instructed to listen to tokens of the 11 English monopthongs: /i ɪ ʊ u ɔ ʌ ɛ æ e o ɑ/ produced by monolingual American English speakers in a hVd environment, and to and match the English vowels to Cantonese monophthong vowels in terms of perceptual "closeness" (Chen, Ng, Li 2002). As a result of the perceptual experiment, the English vowels /i ɛ æ ᴧ u ɔ/ were grouped as familiar vowels and / i e ʊ o ɑ / as new vowels to the native Cantonese (L2 English) speakers. Following the perception experiment, the production experiment involved a group of American-English and Cantonese-English bilinguals who produced all 11 vowels in the carrier sentence "Say hVd again". Acoustic analyses based on spectral measurements obtained from the English vowel sounds produced by Cantonese-English Bilinguals did not reveal favorable evidence which supported the SLM claim proposed by Flege (1995), since nearly all L2 phonetic inaccuracies occurred in the "new" vowel productions.

In the context of the present study, this postulate of the SLM will be challenged for both groups of speakers, the L1-C group and the L1-E group, with respect to their production of French-/y/, which can be categorized as a "new" sound for the L1-E group and can be categorized as a "familiar" sound for the L1-C group of participants.

1.3 Present Study

The goal of the present study is to answer the following question: Do the productions of vowels in French by French-Immersion graduates contain an L1 influence? In order to answer this question, I investigated the production of high front unrounded, high front rounded, and high back rounded vowels in Canadian French as produced by native Canadian English speakers and native Cantonese speakers. I also investigated the production of these same vowels in English, for all participants, and in Cantonese, for native Cantonese participants. The vowels targeted were the high front and high back vowels of Canadian English (i.e., /i u/), Canadian French (i.e., /i y u/), and Cantonese (i.e., /i y u/). These vowels are contrastive phonemes, forming minimal pairs in all three languages, as seen in Table 1.3.

Table 1.3: Minimal pairs in Cantonese, English and French words using the target vowels

/i y u/.

/i/

/y/

/u/

Cantonese

si4 [si:4]

[English Gloss: ]

syu6 [sy:6]

[English Gloss: tree]

[su:k1]

[English Gloss: grain]

English

seat [sit]

suit [sut]

French

si[si]

[English Gloss: if]

su [sy]

[English Gloss: knowledge]

sous [su]

[English Gloss: under]

1.4 Research questions

Two specific questions were targeted in this study involving a between-groups analysis : (1) Are the high vowels produced in the native languages for each group of speakers different with respect to F1 and F2?; (2) If these differences exist in the L1, are these same differences reflected in their productions of the same vowels in Canadian French? Two other questions were explored, involving a within-group analysis: (3) For the L1-C group, is there evidence of language transfer from the L2 (Canadian English) to their L3 (Canadian French)?; (4) For the L1-E group, do the Canadian French productions of /y/ support the hypotheses postulated by Flege’s (1995) Speech Learning Model (SLM)? In order to answer each of these questions, measurements at the mid-point of the vowel were used to investigate group differences in the normalized mean F1 and F2 values. I have limited the scope of this analysis to F1 and F2 measurements, as this approach has been found to be sufficient to capture differences in height and backness that contrast the high vowels targeted in this study (Ménard, Schwartz, Boë, Kandel, & Vallée, 2002). Due to the fact that there was a small sample size of five participants within each group, the first two questions will also be extended to address differences between individual participants. F1 and F2 measurements at the mid-point of each target vowel were used to answer each of the questions in the present study.

1.5 Hypotheses

Based on the previous research (Chen, Ng & Li, 2012; MacLeod et al., 2009; Meng, Zee & Lee, 2007) I hypothesize that (1) there will be differences in the production of English vowels and Cantonese vowels produced by the native speaker groups. Moreover, I expect that (2) these differences will be reflected in the L2 for each respective group; that is, that there will be a positive language transfer demonstrated in French for the L1-E group and a positive language transfer in English for the L1-C group.

Seeing that (Leung, 2005) found that there is a linguistic advantage to learning English before French, (3) I suspect that the L1-C group will also show a positive transfer for the production of English vowels to French vowels. With respect to the L1-E participants’ production of French-/y/, (4) I predict that the speakers will make a category distinction between /u/ and /y/ as found in (Levy & Law II, 2009), supporting the predictions of the SLM in an FLA context as seen in Simon & D’Hustler (2012); since the amount of formal instruction experience in the L2 was over ten years for , I predict that the pronunciation of /y/ will not be /ju/ (Levy and Law II (2010), and will be closer to the predicted for monolingual French speakers, as found by MacLeod et al. (2009).

2. METHODS

Participants

Participants in the L1-E group were 5 females, between the ages of 21 and 30 years old, who reported being native speakers of (Canadian) English. The L1-C group consisted of 5 females between the ages of 18 and 21 years old, who reported being native speakers of Hong Kong Cantonese. All participants reported having completed a French Immersion program in Canada. At the time of the study, all participants were undergraduate or graduate students at the University of British Columbia, and reported having no speech, language or hearing disorders. Participants were compensated $10CAD for their time.

Materials

2.1 Target Words

The target words selected for this study were all real words from English, French and Cantonese that contained the vowels targeted in the study, /i, y and u,/, and two mid control vowels, /ɝ/, in English, and /œ/ in French and Cantonese. An additional set of fifteen control words were used using various low vowels was also implemented for the English word list, since this language does not contain the vowel sound /y/. Fifteen words were selected for each vowel in a variety of consonant onsets for a total of 60 words in each language (see Appendix A for a list of target words).

Although the phonological composition of the target words in the present study was not identical across the three languages, the words shared many phonetic and phonological features. Phonetic similarity was high for place of articulation of the initial consonant: in all three target languages there were 6 words which begun with a labial consonant; 4 were bilabial while 1 was a labio-dental consonant and 1 was a labio-velar consonant. Further, there were 17 words beginning with alveolar consonants in English, and 27 words in both French and Cantonese; 2 words began with a palatal consonant in English, with 1 in French and 6 words Cantonese ; and 4 words began with a posterior (velar/uvular or glottal) place of articulation in English, with 4 in French and 6 in Cantonese. There was also 1 word in English, 7 words in French and no words in Cantonese which began with the target vowel sound. By using target words which are essentially consistent for place of articulation for the onset of the vowel between all three languages, the possibility that the vowel formant pattern differences found in the data could be attributed to the phonological frame rather than the language being spoken is reduced (Hillenbrand, Clark, and Nearey (2001). In addition, the target words were similar in terms of grammatical class (noun, verb, or adjective), frequency of manner of articulation for both the onset and coda consonants surrounding the target vowel, and in terms of word final voicing.

2.2 Creation of Stimuli

Two sets of stimuli were created for each target language. The first set of stimuli created was for the first phase of the experiment, the learning phase. The stimuli created for the learning phase of the experiment were word-picture pairs presented on a slide which had a solid white background. The target word was centered at the top of the screen, with a picture centered below the orthographic representation of the word. Pictures for this experiment were found online using a Google image search, and were adapted from various websites. The words were generally nouns or associative descriptions of nouns, such as "soft" for a picture cotton balls. In all cases, the words presented were strongly associated with its respective picture pair, as judged by a native speaker of each of the target languages. The second set of stimuli were exactly like that of the first set, except the orthographic representation of the word was removed, leaving only the picture associated with the target word on the screen, as seen in Figure 2.2.

seat_w.jpg seat_p.jpg

Figure 2.2: An example of the stimuli: the example for the learning phase shown on the left, with the example from the test phase on the right.

These stimuli were originally created in the Microsoft® PowerPoint 2007 program. Each PowerPoint slide was then converted to a Bitmap (.bmp) picture file. The stimuli were then encoded into E-Prime® Experimental Software (Schneider et al. 2002), which was the program ultimately used in order to present the stimuli to participants.

Procedure

2.3 Data Collection

Participants were asked to give written consent for their participation by signing the participant consent form, and were also asked to complete a language background questionnaire prior to commencing the study. The participants were tested individually in a sound-booth within a quiet room at the Speech in Context Laboratory at The University of British Columbia, Vancouver campus. The L1-E participants were tested in a single 40 minute session, and the L1-C participants were tested in a single hour-long session.

All sessions were conducted by the author, who provided oral instructions to participants in English, prior to the start of the study. Participants wore a head-mounted microphone (MicroMic The Original, C520 Vocal Condenser Microphone), which was connected to a pre-amplifier (USB PRE 2), recording directly to a Lenovo Think Centre Desktop computer. In all sessions, the same head-mounted microphone (MicroMic The Original, C520 Vocal Condenser Microphone) was used. Recordings were sampled at a 705kbps bit rate and 44.1 kHz.

After the experimenter provided oral instructions for the experiment, the experimenter left the room, and participants followed a set of written instructions presented on the screen in front of them. Participants were first asked to read a short passage in the respective target language. These passages consisted of an introduction to a children’s fairy tale which were "Goldilocks and the Three Bears" and "Le Petit Chaperon Rouge" [English Gloss: Little Red Riding Hood], for the cases of English and French, respectively. For Cantonese, the passage which participants read aloud was a short poem titled "夜思" [English Gloss: Night Thoughts], which is a poem commonly recited by school-aged children (Sheik, 2009). The purpose of participants reading these stories was to ensure that participants’ first utterances in the target language were not during the experiment. Additionally, it allowed participants to gain comfort switching between speaking each language. This phase of the experiment was implemented to prepare participants to speak in the respective target language and was not used for the purpose of testing or analyses. A copy of the French and English story introductions and the Cantonese poem can be found in Appendix B.

After participants completed the short reading-tasks, they proceeded to complete the two phases of the experiment, the learning phase and subsequently, the test phase. Each phase consisted of viewing 60 slides, which were randomized to control for order effects. In the learning phase, participants viewed slides consisting of word-picture pairs, with one pair on each slide. Each slide was presented for 4000 milliseconds. The learning phase was implemented to participants as a preventative measure to reduce the number of non-target words produced by participants in the test phase. In the test phase, participants viewed a picture-only slide show. This slide show consisted of the same pictorial stimuli as seen in the learning phase; however, no words were shown. Participants were asked to say the word which was paired with the picture in the previous slide-show.

2.4. Data analysis

The target vowels selected for analysis were /i u/ in English and /i y u/ in Cantonese and French. Each participant produced 120 words in all three languages, English and French and Cantonese. Due to participant errors, which includes instances where the participant forgot the target word and said nothing or if the participant produced an incorrect word for a target, some productions were excluded; the total numbers of words included in the analysis for each group were: 538 English target words, 310 Cantonese target words and a total of 784 French target words were acoustically analyzed, yielding a grand total of 1632 tokens.

The audio files in "WAV" format from each participant's production of the carrier phrase task were analyzed using PRAAT 5.3.42 (Boersma and Weenink, 2013). The onset and offset of each vowel was identified and marked by adding boundaries in the "text" tier in PRAAT. The window size used to view the tokens and mark boundaries varied between 1.42ms and 1.33ms in size. The onset of the vowel was identified by finding the first regular voiced pulsation accompanied by the presence of formants in the spectrogram. The offset of the vowel was located by finding the last regular voiced pulsation accompanied sometimes by the presence of formants in the spectrogram or the absence of formants if the vowel devoiced word-finally. Once vowel boundaries were located visually using this procedure, they were re-checked by listening to the segmented vowel: if a non-vowel sound (e.g., a fricative noise) was heard, the vowel boundaries were re-evaluated to exclude this non-vowel sound; however, there were instances where the vowel itself sounded fricated or was spoken in a creaky voice, in which case they were retained. This procedure was sufficient for identifying the onset and offset of the vowels in words with an obstruent or fricatives in the onset and coda, and when the word had no coda. Additional guidelines for measurement were needed when the vowel was adjacent to a liquid or a nasal. In the case of a liquid, the measurement procedure was modified such that the vowel onset