Doctor of Philosophy, The Ohio State University, 2002, Psychology

Locating word boundaries in continuous speech is a complex task that is completed effortlessly by listeners. Determining what sources of information are used by listeners to achieve successful segmentation is critical for developing models of word recognition. Six experiments were designed to look at the use of phonotactic information in speech segmentation. The first three experiments (Part I) replicated and extended the word-spotting findings of McQueen (1998), demonstrating that phonotactic information is used in the segmentation of words (i.e., CVCs) embedded within nonwords. The next three experiments (Part II) were designed to investigate whether phonotactic effects held up in the more naturalistic context of connected speech, where lexical information is also available to guide word segmentation. Phonotactic effects were present, though small, in comparison to lexical effects. Additionally, phonotactic effects generally were weaker in the context of words than in the context of nonwords, although this finding was not always statistically significant. The data suggest a secondary role of phonotactics when in conjunction with lexical information, and also suggest that maximal lexical activation may even diminish the effects of phonotactics in certain circumstances. Implications of these results for future research and for word recognition models are discussed.

Committee: Mark Pitt (Advisor) Subjects: Psychology, Cognitive