Lexically-driven perceptual adjustments of vowel categories

Abstract

We investigated the plasticity of vowel categories in a perceptual learning paradigm in which listeners are encouraged to use lexical knowledge to adjust their interpretation of ambiguous speech sounds. We tested whether this kind of learning occurs for vowels, and whether it generalises to the perception of other vowels. In Experiments 1 and 2, Dutch listeners were exposed during a lexical decision task to ambiguous vowels, midway between [i] and [e], in lexical contexts biasing interpretation of those vowels either towards /i/ or towards /e/. The effect of this exposure was tested in a subsequent phonetic-categorisation task. Lexically-driven perceptual adjustments were observed: Listeners exposed to the ambiguous vowels in /i/-biased contexts identified more sounds on an [i]-[e] test continuum as /i/ than those who heard the ambiguous vowels in /e/-biased contexts. Generalisation to other contrasts was weak and occurred more strongly for a distant vowel contrast (/α / vs. / ͻ /,) than for a near contrast (I/ vs. / ε /). In Experiment 3, spectral filters based on the difference between the exposure [i] and [e] sounds were applied to test stimuli from all three of the contrasts. Identification data of these filtered stimuli suggest that generalisation of learning across vowels does not depend on overall spectral similarity between exposure and test vowel contrasts.