Use of L1 Categories in the Identification of English Posterior Obstruents

Abstract

The quantitative mapping model proposed by Park and de Jong (2008) predicted the identification of L2 English stops very well, especially when the L2 speakers’ goodness-of-fit was included. The current paper purports to provide a testing ground for the quantitative model by examining the model’s prediction for the identification of English posterior obstruents. To that end, 42 Korean speakers performed English-to-Korean mapping and English identification. Results showed that Korean speakers’ identification accuracy of English posterior obstruents was successfully predicted by the quantitative model except English voiced fricative /ʒ/. The prediction of accuracy was slightly improved when the speakers’ goodness ratings were included into the model. Also, error rates were well predicted except English /ʒ/ and /ʤ/. Further, Korean speakers were found to have much confusion between English /ʒ/ and /ʤ/, possibly due to the speakers’ having not developed firm sound categories for these sounds. Based on the results of the study, some implications for the mapping model and the Speech Learning Model were discussed.