Dynamic information signals voicing for syllable-final stops

Abstract

Several acoustic properties can influence listeners’ judgments of voicing for syllable-final stops, including vocalic length. Developmental and cross-linguistic experiments have suggested that listeners with little experience about how vocalic length is related to final-stop voicing weight this property less than more-experienced listeners. For five (English, CVC) minimal pairs, we examined the acoustic consequences of final-stop voicing, as well as listeners’ weighting of two properties that signal voicing: vocalic length and syllable-offset transitions. First, an acoustic analysis of words showed that differences in vocalic length associated with final-stop voicing are attenuated when words occur in continuous speech, raising questions about how much experience any listener really has hearing vocalic-length differences correlated with final-stop voicing. Next, adults and children (3, 5, and 7 years old) labeled natural stimuli made from words that originally had voiced or voiceless stops, in which we modified vocalic length. Partial correlations revealed that listeners of all ages weighted offset transitions more than vocalic length. In fact, listeners were reluctant to label any stimulus that did not have transitions clearly indicating closure as ‘‘voiced,’’ and so the traditional ‘‘vocalic-length effect’’ was apparent only for originally voiced stimuli. We conclude that dynamic information takes precedence in this perceptual decision.