Informational constraints on pre-lexical priming

A popular explanation of automatic semantic priming attributes those effects to the spreading of activation across a meaning-based network in lexical memory. A number of conflicting claims have been advanced regarding the kinds of information captured in that pre-lexical network. Lexical decision and naming experiments are reported which indicate (1) that automatic priming is supported equally well by a very broad range of relations, (2) that the quality of relation for individual prime-target pairs, as reflected in such measures as traditional associative strength, has no impact on the amount of priming produced, (3) that asymmetries between lexical decision and naming in the amount of priming they generate under stringent conditions of automaticity require the postulation of an extra-lexical source of automatic priming, and (4) the residual priming left for the putative lexical network to explain is at most a few milliseconds. Taken together, these findings suggest that explanations of automatic priming as a product of pre-lexical spreading activation are at best highly exaggerated and perhaps entirely incorrect. An alternative account appealing to retrospective but automatic semantic integration processes is discussed.