Reduction of consonants and vowels in the course of discourse

Abstract

There is a clear link between the discourse status of a word and the degree of reduction. For instance, Gregory [Dissertation (2002)] provided evidence that hearer knowledge affected reduction in production for discourse-old items. Lexical information, such as word frequency, also plays a crucial role in the degree of reduction [Fosler-Lussier and Morgan, Speech Commun. (1999)]. These previous studies either looked at short discourses or words isolated from context. Therefore, the current study investigates longer discourses, using the VIC Corpus [Pitt etal., Corpus (2007)]. The primary question is to what degree do repeated uses cause further reductions, and if the reductions are syllabically and segmentally uniform across a word. Many studies on reduction rely on the intuition that reductions occur when information load is light, such as when the word was repeated recently or has a high probability of occurrence, so the prediction is that unstressed syllables would show more reduction than stressed syllables, due to their lighter load of information. Likewise, as vowels vary more than consonants across dialects, the informational load of vowel quality may be less than that of consonant quality, so the prediction is that vowels would show greater reductions than consonants.