Vowel Discrimination by English, French and Turkish Speakers: Evidence for an Exemplar-Based Approach to Speech Perception

Abstract

Recent studies of speech perception have shown that speakers retain significant amounts of the phonetic detail of heard speech providing strong evidence for an exemplar-based model of the representation of speech sounds. This does not preclude the existence of a feature-based model as well; indeed many theories of speech perception advocate a feature- or contrast-based model for the discrimination of speech sounds. In this study, we provide evidence that an exemplar-based model of speech perception makes more accurate predictions for the performance of English, French and Turkish speakers in a vowel discrimination task. Participants were asked to discriminate the four high front vowels of German, which differ by both the rounding and tense/lax contrasts. Crucially, English has only the tense/lax contrast, while Turkish and French have only rounding. The results show that having one of the vowels in one’s language facilitates discrimination more than having the featural contrast supporting an exemplar-based model. Furthermore, different discrimination tasks were used showing that the effects of native language on vowel discrimination are mediated by task and by psychoacoustic similarity.