A DISTRIBUTED-MEMORY MODEL OF SEMANTIC PRIMING

An alternative to semantic network models of lexical knowledge representation and access is described, in which knowledge about a word is represented as a pattern of activation across a collection of processing units. In this distributed memory model, semantic priming effects arise naturally from the similarity of the patterns of activation that represent a related prime and target. Priming effects can be reduced by an intervening stimulus that modifies the pattern of activation before the target appears. This process is demonstrated empirically with a word naming task. An implemented version of the distributed memory model is used to simulate these results, and results from previous research in which participants overtly responded to the item that intervened between a prime and target are also simulated. Comparisons with semantic network and compound cue models of priming are discussed.