Constraints upon semantic activation during sentence comprehension

Abstract

It is commonly assumed that the initial and automatic priming effect of a word reflects its “meaning” as represented in the mental lexicon, and as context-independently activated by that word. However, a number of cross-modal priming experiments are reported which demonstrate that while words in isolation or in random word lists immediately and automatically prime related targets, this effect disappears when the same words occur in sentence contexts, not only when these contexts are irrelevant to the prime-target relationships, but also when they are neutral and unbiasing. For relevant biasing sentences, target recognition was predominantly sensitive to the overall sentence meaning, although in the absence of strong priming at this level an effect of the prime word itself could be detected. These results are attributed to the nature of the prime-target relations used, since these reflected the functional relations between objects in the world (e.g. key-door). By assuming that such relations are supported by schematic world knowledge, rather than lexical meanings, it is possible to account for the results in terms of constraints upon, and attentional control over, the activation of background knowledge during sentence comprehension.