Perceptual Overlap in Classification of L2 Vowels: Australian English Vowels Perceived by Experienced Mandarin Listeners

Perceptual Overlap in Classification of L2 Vowels: Australian English Vowels Perceived by Experienced Mandarin Listeners Yizhou Wang Rikke L. Bundgaard-Nielsen School of Languages and Linguistics MARCS Institute University of Melbourne, Australia University of Western Sydney, Australia wyz2014tw@gmail.com rikkelou@gmail.com Brett J. Baker Olga Maxwell School of Languages and Linguistics School of Languages and Linguistics University of Melbourne, Australia University of Melbourne, Australia bjbaker@unimelb.edu.au omaxwell@unimelb.edu.au Abstract This paper presents a study examining the role of classification overlap in second language (L2) vowel perception. Eighteen experienced L1 Mandarin Chinese (MC) listeners completed a classification task on twelve Australian English (AusE) monophthongal vowels. It was found that the AusE vowels were not equally difficult for the MC listeners, and the vowels also showed various overlapping patterns in the classification space. Correlation tests revealed that for individual L2 vowels, the classification accuracy was negatively correlated with the mean reaction time (RT), r = -0.817, p = 0.001; the level of classification overlap was negatively correlated with the mean accuracy, r = -0.860, p < 0.001, while positively correlated with the mean RT, r = 0.760, p = 0.004. The findings suggest that classification overlap can inform how experienced L2 listeners’ first language (L1) phonology influences vowel perception in a non-native language. 1 Introduction To successfully understand speech, listeners must process phonetic information and match this information to the phonemic categories of the language in question. For non-native or L2 listeners, this task can sometimes be difficult because their perceptual system is shaped by the L1 phonology. Mismatches between the L1 and L2 phonological systems can lead to perceptual difficulty of various linguistic units in the target language, e.g., segmental (Bundgaard-Nielsen et al., 2011a; Højen and Flege, 2006) and suprasegmental structures (Hallé et al., 2004; Tremblay, 2009). One prevalent theory of non- native speech perception, the Perceptual Assimilation Model (PAM, and its extension, PAM-L2) (Best, 1995; Best and Tyler, 2007) attributes the perceptual difficulty to the interference of the L1 phonological system: L2 phones tend to be assimilated into the listener’s L1 phonological categories, which can at times cause a loss of phonemic contrast in perception. PAM and PAM-L2 posit that the discriminability between a pair of non-native (L2) sounds depends on the level of perceptual overlap between the two sounds, e.g., a pair of perceptually overlapping L2 sounds are more difficult to discriminate than those that do not overlap in cross-language perception (Faris et al., 2018). For instance, L1 Japanese listeners assimilate the AusE vowels [iː] and [ɪə] as their native bimoraic category [iː], while mapping the AusE vowel [ɪ] onto the Japanese monomoraic category [i]; and this perceptual overlap pattern has led them to fail to discriminate between AusE [iː] and [ɪə], though not between AusE [iː] and [ɪ] (Bundgaard-Nielsen et al., 2011a, 2011b). Since PAM and PAM-L2 provide precise predictions in terms of pairwise discriminability, previous research adopting the framework often uses discrimination tasks (e.g., AXB tasks) to examine perceptual performance by L2 learners. Besides, an assimilation task is used in association to …

Content truncated to 500 words