Non-native Production of Thai: Acoustic Measurements and Accentedness Ratings

Abstract

This paper reports a study of the production of Thai vowels, consonants, and tones by native English speakers using two forms of evaluation: acoustic measurements and auditory evaluation by native Thai-speaking listeners. The investigation focused on (a) the acoustic parameters along which the two groups of speakers differ, and (b) which of these acoustic parameters influenced native listeners' judgments of perceived degree of accentedness. Three native Thai speakers and 6 native English speakers were tape recorded in an oral reading task. Speech analysis showed that the two groups of speakers differed more along the spectral dimension (namely formant frequencies and fundamental frequency) than the temporal dimension (namely voice-onset time and vowel duration). When the productions of both the native speakers and the non-native speakers were rated for accentedness, the rating data showed that non-native production can be readily distinguished from native production. Only some non-native tokens were judged as 'native-like'. Moreover, the rating scores for the non-native speakers were lower in level tones than contour tones, suggesting different degrees of difficulty for each tone. When the acoustic data were regressed on the rating data, significant predictors were spectral in nature and were found mostly for level tones. Moreover, no correlation between years of experience with Thai and the rating scores was found. Results are discussed in terms of a 'holistic' versus an 'analytic' approach in tone processing by adults, inherent acoustic characteristics of individual tones as well as quantity and quality of the native input. Furthermore, the fact that some non-native tokens were judged as 'native-like' seems to challenge the claim that segmental as well as suprasegmental errors arise from the loss of ability to learn non-native sounds.