Abstract

The purpose of this study was to critically evaluate the normalization models and to suggest the exemplar models as alternatives in explaining the processes of speaker variability in vowel perception. While the normalization models treat the speaker variability as a noise to be removed, the exemplar models use it as one of the important informations in vowel perception. A simulation study was conducted to quantitatively contrast the normalization models and the exemplar models(GCM and ALCOVE) by fitting them to the Peterson & Barney(1952)'s data. The quantitative fits of the exemplar models to the Peterson and Barney(1952)'s data of vowel and gender identification were compared with those of the normalization models. The results showed that predictions of the exemplar models are much better than those of the normalization models. It was suggested that the exemplar models can be more effective modules in processing of the speaker variability in vowel perception. Limitations of the present study and further research problems were discussed in the final section.