Spectral patterns of the American English diphthong /ai/ as a function of coda voicing produced by native Korean speakers

Abstract

This study aimed to investigate how native speakers of Korean who learned English as a second language realize spectral differences in the American English diphthong /ai/ as a function of coda voicing. Ten Korean learners of English and eight native speakers of American English participated in a production experiment. The monosyllabic words “bite” (/bait/) and “bide” (/baid/) were read along with filler words in isolation and in a carrier sentence. The native group demonstrated significantly smaller F1 and larger F2 before /t/ than /d/ both in the nucleus /a/ and in the offglide /i/ (Moreton 2004). The learner group did not show statistically significant spectral changes in the nucleus and the offglide. Also, the native group significantly reduced the temporal distance between the nucleus and the offglide, and showed spectral peripheralization in the offglide before /t/ than /d/ (Pycha and Dahan 2016). However, the learner group did not show native-like reduction of the temporal distance between the nucleus and the offglide and spectral peripheralization in the offglide. Although the non-native speakers in this study exhibited some durational changes as a function of coda voicing, they did not learn the fine phonetic details regarding the gestural timing and spectral patterns in the diphthong. A considerable degree of individual variation in the learner group and speaking context effects were also found. It was interpreted that the Hyperarticulation hypothesis (Thomas 2000; Moreton 2004) and the Gestural Timing hypothesis (Pycha and Dahan 2016) could provide indices modelling the non-native phenomena found in this study as not attaining native-like phonetic values and gradual approach to the values concerning the spectral and durational aspects in /ai/ as a function of coda voicing. (Ewha Womans University)