Time-course of lexical knowledge use in processing pronunciation variants

Abstract

Three experiments examined how listeners process words that are produced with a context conditioned sound change, fricative assimilation. In fricative assimilation, a word final /s/ can change to a /∫ / when a following word begins with an approximate segment (e.g., dress yacht --> dresh yacht). Fricative assimilation is blocked in other segmental environments such as plosives (e.g., dress boat will not alter the word final /s/). Previous research revealed that phonological variation is acceptable as long as it is licensed by the context in which it is embedded (Gaskell &Marslen-Wilson, 1996). However, there is indication that such licensing effects are largely consigned to particular types of assimilation (e.g., place assimilation) and that other word-final variations (e.g., flap variants), which happen more often in natural speech, have a separate phonological representation that is dependent on the frequency with which the variation occurs (Ranbom, Connine, &Yudman, 2009). The present research addresses how assimilated speech is processed in recognizing spoken words, and, specifically, how different sources of knowledge (lexical and phonological context) are utilized at different times during processing. The results point out that word knowledge is not essential for utilizing the phonological context during recognition, but it permits earlier use of the phonological context in recognizing the pronunciation variant.