Priming in sentence processing: intralexical spreading activation, schemas, and situation models.

A series of eye-tracking experiments investigated priming in natural language understanding. Intralexical spreading activation accounts of priming predict that the response to a target word will be speeded (i.e., primed) when strong associates appear prior to the target. Schema-based priming accounts predict that priming will occur when the target word is a component of an activated schema or script. Situation model accounts predict that priming will occur when a target word can be integrated easily into an evolving discourse representation. In separate experiments, we measured the effect of associated words, synonyms, and identity primes on processing times for subsequently encountered target words. Our designs crossed prime type (e.g., synonyms vs. unassociated words) with semantic plausibility (i.e., the target word was a plausible vs. an implausible continuation of the sentence). The results showed that identity primes, but not associates or synonyms, primed target words in early measures of processing like first fixation and gaze duration. Plausibility effects tended to emerge in later measures of processing (e.g., on total reading time), although some evidence was obtained for early effects of semantic plausibility. We propose that priming in naturalistic conditions is not caused by intralexical spreading activation or access to precompiled schemas.