Is speech recognition automatic? Lexical competition, but not initial lexical access, requires cognitive resources

Current models of spoken word recognition suggest that multiple lexical candidates are activated in parallel upon hearing an utterance, with these lexical hypotheses competing with each other for recognition. The current project investigated the effect of cognitive load on initial lexical access and later lexical competition. In a set of priming studies, the lexicality of the primes (i.e., non-word vs. word) was manipulated to dissociate these two sub-processes. We tested performance on a semantic association task under conditions with no additional load, or with cognitive load that used cognitive resources that are either general or more specific to phonological processing. The results suggest that the initial access of lexical items is relatively automatic. In contrast, maintaining lexical candidates in competition requires cognitive resources, and these resources are specific to phonological processing. The overall result pattern provides insights into differences in the way that lexical activation and competition operate.