Testing L2 speech acquisition models on Cantonese perception and production of English C-[r] clusters This study examines the acquisition of the English initial C[r] clusters by advanced Cantonese learners of English and puts the Perceptual Assimilation Model (PAM, Best 1995) and the Speech Learning Model (SLM, Flege 1987) to the test. While previous studies gave ample explanations of phonological deletion and substitution for the initial clusters containing /r/ as the difference between Cantonese and English (Chan 2006, Hung 2002), not much study has been done on the perception and production of the initial clusters by advanced learners of English. L2 sound categories in learners’ minds have long been viewed as discrete substitutions of L1 categories, as suggested by the time-honored Perceptual Assimilation Model. However, L2 realization of L1 sounds may also be gradient, as new categories, different from both L1 and L2, may emerge from learners’ productions (Flege, 2003; Lan & Oh, 2012). PAM predicts that learners of L1 would categorically map L2 sounds onto L1 ones if category goodness of the L2 sound ranges fair to good for L1. SLM predicts that experienced learners of L2 would acquire newly formed L2 categories by fine-tuning the stable perceptual distance between L1 and L2 sound categories. The present study contains two experiments examining whether perceptual assimilation is categorical or gradient, and whether perceptual distance between L1 and L2 would be stable for Cantonese advanced learners of English. In the first experiment, we examined Cantonese productions of C-[r] clusters to see if they phonologically delete /r/ or substitute it to [w] (e.g., cry as [kwa I]). Production tests show that the F3 values of /r/ in cluster (Kent & Reading, 2003) by Cantonese speakers is in between both native English and Cantonese speakers’ [r] and [w] realizations, which supports SLM’s idea of an emerging fine-tuned category. In the second experiment, whether the category is stable is tested, especially on the tr- cluster. Identification and discrimination tasks of tr- and ch- in five vowel contexts were carried out. Results showed that the L2 category of tr- may not be stable, and the clusters with the back vowels, particularly /u/, were much less accurately perceived than those with the front vowels. Production data by the same speakers also support the findings from the perception tests. The acoustic analysis of /r/ in the clusters showed that a native-like F3 cue was used with the front vowels (e.g., trees) but not with the back vowels (e.g., truth) where F2 was used as a cue. Findings from the perception and production tests suggest that L2 speech models should consider the gestural distance as well as the categorical one. I plan to conduct future researches on speech perception which involves manipulation of the [r] part in both F2 and F3 to give further proof to the hypothesis that the attention to acoustic cues for native English speakers and Cantonese learners are different. (455 words) References: Best, C.T. (1995). A Direct-realist view of cross-language speech perception. In Strange, W. (ed.). Speech perception and linguistic experience: issues in cross-language research, 171–204. Chan, Y. (2006). Strategies used by Cantonese speakers in pronouncing English initial consonant clusters: Insights into the interlanguage phonology of Cantonese ESL learners in Hong Kong, IRAL proceedings, 44, 331-355. Flege, J. E. (1987). The production of "new" and "similar" phones in a foreign language: evidence for the effect of equivalence classification, Journal of phonetics, 15, 47-65. Flege, J. E., Schirru, C., & MacKay, I. R. A. (2003). Interaction between the native and second language phonetic subsystems, Speech communication, 40, 467-491. Hung, T. N., Towards a phonology of Hong Kong English, in Bolton, K. (ed.), Asian Englishes today: Hong Kong English Autonomy and Creativity, 119-140, 2002. Kent, R. D. & Read, C. (2002). Acoustic analysis of speech, Thomson Learning. Lan, Y. & Oh. S. (2012). Distributions of [r]-deletion and [w]-substitution in English C[r] clusters by Cantonese speakers, Journal of the Acoustical Society of America, 131(4), 3272-3273.