Relating assimilation data to identification data in second language learners

Commonly cited models of bilingual phonology, e.g., the Speech Learning Model and the Perceptual Assimilation Model, project hypotheses about second language (L2) developmental patterns based on hypothetical assimilation patterns between categories in the learner’s first language (L1) and L2, with the addition of a learning component which differentiates an L2 and L1 system. This paper relates two experiments with two participant groups, one native Korean, and one native Taiwan Mandarin. Both groups identify English obstruents /p b t d f v θ ð/ in four prosodic positions (onset, coda, and intervocalic pre-stress and post-stress) using their native non-Roman graphemes in the first experiment, and using English labels in the second. The assimilation results generate predictions for discrimination performance in the second experiment, assuming only the L1 categories. The two groups show large differences in assimilation pattern due largely to the non-sibilant fricative /f/ in Mandarin, in opposition to the Korean system which has only non-sibilant /h/. Analyses reveal a strong correlation between L2 discrimination patterns and predictions based solely on the L2-L1 assimilation results, suggesting that the perceptual system of L2 learners can be surprisingly well-captured as a single perceptual system with a bilingual array of segmental categories.