Cantonese has a more complex vowel inventory than Mandarin, resulting in a many-to-one mapping between the two systems. While such mappings are generally not considered problematic for L2 acquisition, a noticeable Cantonese accent persists in L2 Mandarin speech. This research aims to investigate the acoustic basis of the accent, based on a systematically recorded dataset that includes Mandarin monophthongs produced by Cantonese speakers. Cantonese vowels were also recorded as references. Each vowel token was manually labeled and measured at the acoustic midpoint using Praat. The first three formant frequencies were extracted to capture the spectral characteristics that define each vowel’s quality. By documenting both durational and spectral properties, the dataset provides a preliminary view of the interaction between Mandarin and Cantonese vowel systems. The results indicate that while most Mandarin vowels produced by Cantonese speakers aligned with L1 production reported in the literature, the high front rounded vowel /y/ demonstrated significantly greater variability in L2 production. Further, the mid-back unrounded vowel /ɤ/ in L2 was produced with a more fronted articulation (higher F2) than in L1. These patterns suggest systematic L1 transfer even in the cases of many-to-one phonological mapping.
Liu et al. (Wed,) studied this question.