Beating the bounds: Localized timing cues to word segmentation

Abstract

Prosody facilitates perceptual segmentation of the speech stream into a sequence of words and phrases. With regard to speech timing, vowel lengthening is well established as a cue to an upcoming boundary, but listeners' exploitation of consonant lengthening for segmentation has not been systematically tested in the absence of other boundary cues. In a series of artificial language learning experiments, the impact of durational variation in consonants and vowels on listeners' extraction of novel trisyllables was examined. Language streams with systematic lengthening of word-initial consonants were better recalled than both control streams without localized lengthening and streams where word-initial syllable lengthening was confined to the vocalic rhyme. Furthermore, where vowel-consonant sequences were lengthened word-medially, listeners failed to learn the languages effectively. Thus the structural interpretation of lengthening effects depends upon their localization, in this case, a distinction between lengthening of the onset consonant and the vocalic syllable rhyme. This functional division is considered in terms of speech-rate-sensitive predictive mechanisms and listeners' expectations regarding the occurrence of syllable perceptual centres.