Factors shaping vowel perception biases in adults

Abstract

Vowel discrimination is often asymmetric such that discriminating the same vowel pair is easier in one direction compared to the opposite direction. The Natural Referent Vowel framework interprets these directional asymmetries as a universal bias favoring “focal” vowels (i.e., vowels with prominent spectral peaks formed by the convergence of adjacent formants). The Native Language Magnet model instead interprets asymmetries in terms of a language-specific bias due to distortion of perceptual space around native language vowel prototypes. To test these competing views, Masapollo et al. (2017) compared English- and French-speaking adults’ discrimination of synthetic /u/ variants; this was informative because English /u/ is naturally less focal than French /u/. Their findings revealed asymmetries to be predicted by focalization only; although stimulus limitations may explain the lack of prototype effects. Here, we synthesized a more refined series of vowel stimuli systematically varying in smaller psychophysical steps around the English /u/ and French /u/ prototypes to augment the measurement of focalization and prototype effects. Native English speakers completed a category goodness-rating task followed by an AX-discrimination task using these new variants. Results indicated effects of both focalization and prototype, which are moderated by the size of acoustic intervals along the stimulus series.