Predicting asymmetries in vowel perception: Formant convergence succeeds where peripherality fails

Abstract

Vowel discrimination is often asymmetric such that discriminating the same vowel contrast is easier in one direction compared to the opposite direction. According to the Natural Referent Vowel (NRV) framework, these asymmetries reveal a perceptual bias favoring acoustic vowel signals produced with more extreme vocalic gestures, which act as natural referent vowels. The NR vowel within a contrast typically falls in a more peripheral location within articulatory/acoustic vowel space (defined by F1 and F2) and with F1 and F2 in closer proximity. However, these properties do not always align, as in the case of the /e/-/Ø/ contrast. We here compared findings across three studies where asymmetries were observed during discrimination of this contrast. Peripherality predicts an asymmetry such that perception would be better in the /Ø/□/e/ direction. However, all three studies showed better performance in the opposite direction, which also aligns with the prediction based on formant convergence (derived from reported acoustic measures for each stimulus set). These findings suggest that this perceptual bias is shaped by formant convergence, rather than peripherality per se, which is presumably tied to the extent of vocal-tract constriction. A follow-up study will obtain articulographic recordings of these vocalic gestures to quantify this articulatory-acoustic relation.