Predicting L2 vowel identification accuracy from cross-language mappings between L2 English and L1 Korean

The L1 mapping model developed by Park and de Jong (2008) for English consonants was shown to predict Korean listeners' identifications of English stops with Korean analogs with great success while identifications of some English fricatives were much better than predicted, suggesting the creation of new L2 categories for these English fricatives. Given the different auditory-perceptual and functional nature of vowels versus consonants, the current study investigates to what extent Park and de Jong's L1 mapping model predicts native Korean listeners' L2 English vowel identification. This study also explores whether the extent to which pairs of Korean and English vowels are judged to be variants (or else an English vowel is judged to be “new”) predicts English vowel identification. Korean listeners participated in English vowel identification and cross-language mapping tasks. Results indicated that the L1 mapping model very closely predicted identification of English vowels analogous to Korean vowels (e.g., /ɛ/, /oʊ/, and /u/). This phenomenon was also observed when the listeners' goodness ratings were incorporated. However, it was found that the predictive power of the L1 mapping model for vowels, which was based on the new and similar distinction, was reduced relative to that for consonants due to the less categorical nature of vowels. Thus, the new-similar distinction in the Speech Learning Model (Flege, 1995) may not be so crucial to understanding the impact of vowel mapping in cross-language perception. Nonetheless, the results showed that Korean listeners may be developing new vowel categories for English vowels that are considered to be similar as well as new to Korean vowels.