Thing or sing: Modeling the phonetic variation of the interdental fricative /θ/ in accented English

Abstract

Chinese learners of English tend to pronounce the English interdental fricative /θ/ as [s] because it is absent in Chinese (Rau et al., 2009). To investigate the phonetic variation of the interdental fricative /θ/, this study extracted the first four spectral moments of the target fricatives [θ] and [s] from fricative-initial words in an L2 English speech corpus, and then analyzed them using generalized additive mixed-effects models (GAMM) and random forest classification. The corpus we used contains 31.6 hours of recordings from 389 Chinese speakers from different regions in China (Chen et al., 2019). Classification accuracy rates suggest variation in the production of /θ/ among Chinese learners of English from different Chinese dialectal regions, and the computation of conditional feature importance sheds light on the relative impact of acoustic measurements in predicting the fricatives in accented English. The findings of this study can provide robust empirical evidence on the phonetic variation of the interdental fricative /θ/ in accented English by Chinese learners of English and shed light on a new way to probe into the phonetic variation in non-native speech.