Comprehending words in context : a reaction-time study of lexical ambiguity resolution

Abstract

Psycholinguistic research on semantic processing commonly distinguishes bottom-up lexical processes – involving the automatic retrieval of word senses stored in memory – from top-down contextual processes – involving the active imposition of situational constraints on the construal of lexical items (e.g. Andrews & Bond 2009). Extensive studies, since the 1970s, have sought to ascertain whether these two processes are largely autonomous mechanisms or interlinked sub-processes of a single mechanism. This has led to the proposal of three influential models of word processing – namely, the DIRECT ACCESS MODEL, the AUTONOMOUS ACCESS MODEL, and the REORDERED ACCESS MODEL – which critically differ in the extent to which they consider these two processes as largely interrelated or independent. Most recently, two studies (Peleg et al. 2001; 2008) claimed to have found compelling evidence disconfirming the DIRECT ACCESS MODEL and supporting the AUTONOMOUS ACCESS MODEL. The current study suggests theoretical and experimental reasons to doubt the above conclusion, and presents the results of counter-experiments which challenge the AUTONOMOUS ACCESS MODEL and support the REORDERED ACCESS MODEL instead.

Sixty native English speakers completed a series of timed lexical decisions across two experiments. In Experiment 1, each lexical decision probe was preceded by a priming passage which ended with a disambiguated homonym, to which it was either contextually-related or contextually-unrelated. In Experiment 2, each lexical decision probe was preceded by the same passages, except that the final homonym was omitted. A statistical analysis of participant reaction-times showed that lexical decision rates were significantly faster for contextually-relevant probes than contextually-irrelevant probes in Experiment 1 but not in Experiment 2, indicating that the homonym was crucial to the priming effect of these passages. Specifically, it demonstrates that prior context affected the semantic processing of these homonyms, which thereafter primed associated probes. Accordingly, this finding is problematic for the AUTONOMOUS ACCESS MODEL but fully consistent with the REORDERED ACCESS MODEL.