The Influence of L1 Prosodic Background on the Learning of Mandarin Tones: Patterns of Tonal Confusion by Cantonese and Japanese Naïve Listeners

Human perception of non-native sound contrasts are strongly influenced by their native (L1) phonological system (Best 1995; Flege 1995). For example, it has been well documented that Japanese speakers have difficulties discriminating English /r/ and /l/ (Goto 1971, Lively, Pisoni, Yamada, Tohkura & Yamada 1994). Likewise, English speakers have difficulty in discerning the difference between the Hindi contrasts: voiceless retroflex stop [] and voiceless dental stop [] (Werker 1991). Jamieson & Morosan (1989) also found that English consonant contrast // and // posed a problem for French learners of English. Among existing working models of L2 speech perception, such as Speech Learning Model (SLM; Flege 1995) and Perceptual Assimilation Model (PAM; Best 1995), PAM is a phonologically-based model that focuses on the effect of L1 phonological system on perception of non-native sound contrasts. It posits that listeners’ native (L1) phonology greatly affects their perception of nonnative phones, as listeners perceptually assimilate the non-native phones to their native phonemes (Best 1995; Best, McRoberts & Goodell 2001). A non-native phone may be perceptually assimilated as a Categorized exemplar for some L1 phonemes, an Uncategorized segment that falls somewhere in between two native phonemes, or a Nonassimilable nonspeech sound that has no identifiable similarity to any native phoneme. According to PAM, listeners’ abilities to discern the non-native phones are predictable and dependent on how the nonnative phones assimilate to native segments. There are a few possible types of assimilations. When two non-native phones assimilate to two phonetically similar native phones separately, Two Category assimilation (TC) occurs and leads to successful perception of the non-native phones. In contrast, when two non-native phones assimilate to a single native phone, Single Category assimilation (SC) takes place and hinders listeners’ perception of the non-native phones. In the situation that one of the non-native phones assimilates better than the other, it is termed Category Goodness difference (CG). In addition, when one non-native sound is not categorized and the other is categorized, it forms an Uncategorized-Categorized pair (UC). If both are not categorized, then they will be an Uncategorized-Uncategorized pair (UU). However, if the phone is completely dissimilar to any native phonemes, it will be perceived as a NonAssimilable (NA) nonspeech sound, and native phonology does not have any influence on it. Although PAM originally focuses on speech segments, Hallé, Chang & Best (2004) suggested that PAM can be applicable to suprasegments as well. The present study will explore the possibility of their proposition.