Distinctiveness of spoken word context predicts visual lexical decision time

We review various dimensions along which words differ and which, sometimes as part of a word recognition model, have been claimed to predict performance in the visual lexical decision task. Models of word recognition have typically involved inadequate, or non-existent, semantic representations and have dealt with words existing in isolation from any context. We propose an alternative perspective in which it is the relationships between words - reflecting usage and meaning - rather than the discrete entities themselves, that are fundamental to lexical processing. We present Contextual Distinctiveness (CD), a corpus-derived measure of the plurality of the different content-word contexts in which a word occurs in speech, and demonstrate that it is a significant predictor of response times in a simple visual lexical decision task. We argue that LDT effects previously attributed to Age of Acquisition and word frequency should be reinterpreted in terms of CD. As well as subsuming a number of other lexical variables, we detail further advantages of CD, in terms of computational tractability, objectivity, relation to real language, and relation to formal linguistics.