Electrophysiological evidence for the efficiency of spoken word processing

The Cohort model (Marslen-Wilson, Spoken Word Recognition, MIT Press, Cambridge, MA, 1987, pp. 71-103) proposes that spoken words are frequently recognized at the first point in the acoustic signal where a unique lexical representation is specified. This proposal was tested in two experiments. In experiment 1 participants made speeded lexical decisions to spoken words and pseudowords. In experiment 2 participants passively listened without making overt responses. In both experiments the recognition points for words (the point past which no other lexical item was consistent with the acoustic signal) and deviation points for pseudowords (the point past which no real word is compatible with the acoustic signal) were manipulated. An ERP negativity in the region of the N400 component and RT occurred sooner for items with early than late recognition/deviation points when measures were time-locked to stimulus onset. In experiment 1, when time-locking was to recognition/deviation points, early and late words produced N400s and RTs with indistinguishable latencies, while late pseudowords produced faster RTs and earlier N400s than early pseudowords. Experiment 2 replicated the N 400 effects for words, but only produced a trend in the same direction as experiment 1 for pseudowords.