Is Automatic Priming Semantic?

Abstract
Experiments were conducted to examine the extent to which automatic priming reflects the retrieval of semantic information, as opposed to the associative/collocational relationships between words. When the target task was pronunciation of highly legible words, context effects were only obtained for normative associates that were also rated as close collocates. No effects were obtained for semantically similar pairs of low association strength. However, when target presentation was degraded, these latter pairs did support an effect. Further experiments found that semantically related but not associated pairs also produced context effects when the target task was lexical decision and the primes were masked. It is argued that, contrary to previous suggestions in the literature, automatic priming effects do have a pure semantic component, but it is meaning retrieval that is primed, not the lexical-level representation of the word. The manner in which meaning retrieval interacts with different target tasks, and the importance of response speed in determining the extent of such interactions, are discussed.