Talker and speaking rate variation affect lexical neighborhoods

Abstract

Words that are similar to a syllable can influence listeners perception of phonemes in the syllable. This influence of lexical neighborhood is part of how the words of a language influence speech perception. In previous studies (ASA, June 2000 and Dec 2000), this effect was shown to vary with talker voice, indicating that the similarity of words to nonsense syllables (lexical neighborhood) varies with talker. In a new series of studies, token variation in the nonsense syllables was introduced. Listeners identified the syllable initial stop in bowth-powth and powse-powse series. In these series, bowth and powse are similar to more real words. One group of listeners heard these series as spoken by four different talkers. The tokens of the different talkers were randomly intermixed. A second group heard tokens from one talker at three different speaking rates: fast, normal, and slow. Again, these were randomly intermixed. To the extent that token variation alters the influence of lexical neighborhood, factors other than simple acoustic–phonetic similarity must play a role in determining lexical neighborhood. The implications of these data for models of word recognition will be discussed. [Work supported by NIDCD Grant No. R01DC00219 to SUNY at Buffalo.]