Perception of stressed vs unstressed vowels: Language-specific and general patterns

Abstract

Unstressed vowels are somewhat centralized (even full vowels such as the second in “city, taco"), reducing their acoustic distinctiveness. The current work compares listeners' perception of stressed and unstressed vowels in English and Dutch. The data come from two large projects on native listeners' perception of all possible diphones (CV, VC, CC, and VV sequences, all vowels stressed and unstressed) in English and Dutch. These datasets provide information about listeners' uptake of perceptual cues over time that is comparable across the two languages. Both groups perceived unstressed vowels less accurately than stressed, but this effect was far larger for English. English listeners showed a very large stress effect for lax vowels and a moderate effect for other vowels, while the Dutch listeners showed effects that were small and largely restricted to /a/. Dutch listeners may be able to identify unstressed vowels better than English listeners because the stressed-unstressed distinction has more informational value in Dutch than in English. However, both languages showed a larger stress effect just before a following consonant. This suggests that consonantal coarticulation obscures the quality of unstressed vowels in both languages. Thus, perception of stressed vs unstressed vowels demonstrates both language specificity and cross-language commonality.