Subcategorical mismatches and the time course of lexical access: Evidence for lexical competition

Participants ’ eye movements were monitored as they followed spoken instructions to click on a pictured object with a computer mouse (e.g., ‘‘click on the net’’). Participants were slower to xate the target picture when the onset of the target word came from a competitor word (e.g., ne(ck)t) than from a nonword (e.g., ne(p)t), as predicted by models of spoken-word recognition that incorporate lexical competition. This was found whether the picture of the competitor word (e.g., the picture of a neck) was present on the display or not. Simulations with the TRACE model captured the major trends of xations to the target and its competitor over time. We argue that eye movements provide a ne-grained measure of lexical activation over time, and thus reveal effects of lexical competition that are masked by response measures such as lexical decisions. It is now generally accepted that as listeners attend to a spoken word, they simultaneously entertain multiple lexical candidates, which compete for recognition (see Frauenfelder & Floccia, 1998, for a review). However, the mechanism by which competition among active candidates is realised and resolved remains controversial. In some localist connectionist models, such as TRACE (McClelland & Elman, 1986) and Shortlist (Norris, 1994), word candidates compete with each other via inhibitory lateral connections. Thus, the activation of a lexical candidate at a given point in time in the Requests for reprints should be addressed to Delphine Dahan, Max Planck Institute