Probabilistic phonotactics and levels of processing and representation in spoken word recognition: When is a word not a word?

Abstract

Vitevitch and Luce (1998, 1999) hypothesize that two levels of processing—lexical and sub-lexical—are involved in spoken word recognition. Competition is observed at the lexical level: Words and nonwords with low phonotactic probability/sparse neighborhoods are responded to more quickly than stimuli with high-phonotactic probability/dense neighborhoods. Facilitation is observed at the sub-lexical level: Nonwords with high-phonotactic probability/dense neighborhoods are responded to more quickly than nonwords with low-phonotactic probability/sparse neighborhoods. Nonwords can be processed at either level, whereas words predominantly compete with each other at the lexical level. However, the present experiment demonstrates that words can be processed at a sub-lexical level. Target words varying in phonotactic probability/neighborhood density were presented in a same–different task with varying proportions of word and nonword pairs. When the stimulus set consisted mostly of word pairs, competition was observed: Words with low-phonotactic probability/sparse neighborhoods were responded to more quickly than words with high-probability/dense neighborhoods. When the stimulus set consisted mostly of nonword pairs, facilitation was observed: Words with high-phonotactic probability/dense neighborhoods were responded to more quickly than words with low-probability/sparse neighborhoods. The implications of these results for models of spoken word recognition are discussed. [Work supported by NIH-NIDCD Training Grant DC-00012.]