Effects in production of word pre-activation during listening: are listener-generated predictions specified at a speech-sound level?

It has been demonstrated that listener-generated predictions of upcoming material can be specified to a phonological level, such that a specific word onset is anticipated (e.g., DeLong, Urbach, & Kutas, Nature Neuroscience, 8, 1117-1121, 2005). In the present study, we investigated whether such word-form-specific predictions impact picture-naming latencies in a manner similar to that observed when a distractor word is actually presented. Participants were auditorily presented with high-cloze sentence stems, in order to elicit word-form predictions. The pictures for naming were presented immediately following the sentence stem. We systematically manipulated the phonological relationship between the predicted word and the picture name. Across three experiments, naming was facilitated when the picture name fully matched the predicted word. However, naming was neither facilitated nor inhibited when the picture name overlapped phonologically with the predicted word. This finding is in contrast to the known effects of phonological overlap when a distractor word is heard or read. Our findings suggest that words that are internally listener-generated (predicted) during comprehension are not robustly specified at a speech-sound (phonological) level.