1.IntroductionThe phonetic realization of a word varies depending on sex, age, regional dialect, and a speaker’s language background. Phonetic convergence (PC) describes the process by which the acoustic-phonetic form of an individual assimilates that of a person or model after recent or prolonged exposure. Speakers have been known to converge on both vowel formants, and vowel duration. Few studies have explored the process of PC in non-native speakers. Data on phonetic convergence in simultaneous bilinguals is scarce. A study found that speakers with the same native language (L1) were more likely to converge than speakers that did not share the same L1. Furthermore, no previous study has compared the degree of PC in the productions of language learners, bilinguals, and monolinguals; much less in French. As a consequence, most studies on vowel convergence limit themselves to socio-dialectal differences and F1 and F2 values.2.Objectives and Hypotheses The goal of this study is to measure PC at the segmental level in English and in French. We compare the shadowed productions of American language learners of French and simultaneous bilinguals to those of monolingual speakers of each language. Language learners are expected to deviate in French (L2) but not in English (L1). We also predict that bilinguals could deviate for a number of reasons tied to either extrinsic or intrinsic factors. Extrinsic factors could stem from the socio-dialectal differences between the stimuli and the bilingual speakers. Intrinsic factors could be revealing of the organization of phonological space in simultaneous bilinguals. If we accept that simultaneous bilinguals have two separate phonological systems and that those two systems exert mutual influence on each other as stated in the Interactional Dual Systems Model we can expect bidirectional influence in bilingual productions causing them to deviate from the target values in both languages. 3.Methodology F1 to F3 formants values of American English (AE) and French vowels were analyzed for 5 simultaneous bilinguals (BL) living in France having been exposed to both languages from birth, 5 American second language learners of French (LL) having learned the target language after the age of 13 and having lived in France for more than 3 years on average and, 5 monolinguals (ML) from each language having only lived either in France or in the USA. A shadowing task based on a version of Goldfinger’s shadowing paradigm was used. With the exception of vowel /œ/ in French, all vowels were recorded in three different phonetic contexts: isolated, /bVb/, and /pVp/. In the case of the vowel /œ/, an /ʀ/ replaced the final consonant in both /CVC/ contexts due to phonotactic constraints. We then picked productions resembling canonical examples of AE and French vowels for our experiment. Measurements for these vowels constitute the target values (TV) for the experiment. Praat software was used to annotate and extract semi-automatically formants using midpoint and duration values. Outliers were checked manually. The study included 15 women and 5 men, so all formant values were normalized using Labov’s method. The advantage being that physiologically-caused differences in formant values were factored out while retaining sociolinguistic differences. Statistical analyses were performed with the Jasp software. A 97.5% confidence interval was used in order to avoid type I error. Error bars seen in the graphs correspond to the standard deviation; values represent the means. 4.Results 4.1FrenchThe ANOVA test performed on F1 values between the speaker profiles did not reveal any significant differences. None of the speaker profiles deviated significantly from the TV. Bilinguals deviated more than both language learners and monolinguals. Monolinguals deviated the least out of all the speaker profiles. Both bilinguals and language learners deviated from the target F1 values for the vowel contrast /u – y/. F1 for /y/ was higher for bilinguals and language learners than it was for both monolinguals and the TV. Additionally, F1 values for the low vowel [a] in bilinguals were lower than the TV. Language learner F2 values for the vowel contrast /u – y/ show differentiation problems. The formantic space between the two phonological categories was reduced due to higher values for [u] and lower values for [y]. All other speakers converged. Language learners appeared to merge vowel [y] with vowel [u] in terms of F3 as well. Bilinguals and language learners did not mark the contrast /ø – œ/ in terms of the F3.4.2English An ANOVA test of all F1 values detected significant divergences between speaker profiles. Bilinguals did not deviate from the TV. Monolinguals and language learners deviated significantly from the TV. F1 values were higher for mid [ɛ ɔ] and low [æ ɑ] vowels. An ANOVA test of all F2 values detected a significant divergence between speaker profiles. Language learners did not deviate from the TV. Bilinguals and monolinguals deviated significantly from the TV; F2 was higher in all speakers; bilinguals especially. An ANOVA test showed significant divergence for F3 values. All three speaker profiles deviated significantly from the TV. F3 was higher for all vowel productions. This was especially true for high [i ɪ u], mid [ɛ ɔ], and low [æ ɑ] vowels in bilinguals; mid [ɛ ɔ] and low [æ ɑ] in the other speakers. An ANOVA test detected significant differences in duration between the speaker profiles. Bilinguals and monolinguals deviated with shorter vowels as a whole. However, the intrinsic duration-related distinction between tense and lax vowels was well maintained by all speakers. 5.Discussion and conclusions This study found that subjects deviated more in English than in French. The degree of divergence varied depending on language background. In English, monolinguals deviated the least out of all the speaker profiles and bilinguals the most; leaving the language learners somewhere in the middle: dominance in French could be influenced by the fact that all bilingual speakers were living in France at the moment the recording took place. Language learners were found to deviate from the TV for F1 and F3. F1 values showed that monolinguals and language learners tend to deviate from the TV when it comes to mid [ɛ ɔ] and low [æ ɑ] vowels. This contradicts the results from the study showing that low vowels are more likely to be imitated than high vowels, and especially when it comes to F1. Deviation in language learners in English could be influenced by socio-dialectal differences amongst speakers as stated in the previous study. In French, monolingual productions were, once again, the least divergent; followed by the language learners, and lastly, the bilinguals. Fine acoustic analyses revealed that the French vowel contrast /e – ɛ/ and the English vowel contrast /ɑ - ɔ/ were particularly difficult to maintain for most participants. Bilinguals and language learners relied heavily on F1 to distinguish /ø/ from /œ/. Language learners relied on F2 for the /u – y/ distinction but both French contrasts proved problematic for this type of speaker. The difference in the acoustic cues reproduced by both bilinguals and language learners could suggest a difference in production strategies between simultaneous bilinguals and language learners. An articulatory study would prove useful in confirming this hypothesis. This could also suggest graded levels of interaction between phonological systems for both bilinguals and language learners. If deviation is seen as proof of interaction between two phonological systems in simultaneous bilinguals, the preliminary results from this study could support Paradis’ Interactional Dual Systems Model.