Previous research showed that compared to single-speaker stimuli and cue-manipulated synthesized stimuli input, multi-talker stimuli had an advantage in improving lexical tone categorization in both corrective-feedback and incidental-learning environment. Studies of how sampling distribution influences lexical tone categorization in incidental learning by bilinguals are limited. The current study replicated Liu et al. (2016) to examine how Portuguese-Cantonese bilinguals acquired Mandarin four lexical tones in a video game. The stimuli in the game consisted of tone tokens with both multi-talker stimuli (probabilistic-distribution) that were overlapped among tone categories and variance-manipulated stimuli (deterministic-distribution) that had no overlap among tone categories in terms of pitch height and pitch direction. Sixteen participants were exposed to the two sampling distributions respectively. Results showed that in contrast with the learning outcome of native English speakers in previous study, Portuguese-Cantonese bilinguals in the variance-manipulated condition generalized tone identification to new-talker stimuli better than those in multi-talker condition did. Moreover, the bilinguals in the variance-manipulated condition had a more nativelike cue-weighting in perception of Mandarin tones, and a higher production accuracy rate. The results are discussed in the dual-learning system framework (Maddox & Chandrasekaran, 2014; Roark & Holt, 2018) that the effect of sampling distribution on lexical tone categorization may be language-specific.
新闻报道
