Recognizing Spoken Words : Neural Mechanisms of Inhibition

A fundamental issue in spoken word recognition is the mechanism by which one word representation among tens of thousands gets identified as the optimal match to a speech signal. For correct identification, erroneous hypotheses must be inhibited. Competition among lexical hypotheses is evidenced by delayed reaction times to words that are preceded by similar sounding words. However, these inhibitory effects are ambiguous as to the processing level inhibited. We used magnetoencephalography (MEG) to investigate the neural mechanisms of inhibition and found that the cognitive level of similarity-induced inhibition depends on the type of phonological similarity that holds between the correct and incorrect lexical hypotheses: onset-matching similarity leads to inhibited activation of the target representation while non-onset-matching similarity leads to inhibited recognition. Our results suggest that the speech recognition system has a mechanism of complete suppression of activation targeting mismatching representations that initially constitute perfect matches to an