It has long been known that a word (e.g., BUTTER) presented shortly after a related word (e.g., BREAD) can be processed more rapidly than when presented shortly after an unrelated word (e.g., TABLE). This phenomenon has come to be referred to as "semantic" priming. To this date, however, only I. Fischler (1977, Memory & Cognition, 5, 335-339) has provided any evidence that this phenomenon is semantically and not associatively based. In the present paper six studies were undertaken in an attempt to generalize Fischler's findings to tasks other than the simultaneous lexical decision task he used. In Experiments 1, 2, and 3 it was determined that semantic category relationships, in which the two words named members of the same semantic category (e.g., DOG-PIG) did little to facilitate naming of the second stimulus. In Experiments 4 and 6, it was determined that a semantic category relationship did nothing to augment the priming from associative relationships in naming and lexical decision tasks, respectively. However, in Experiment 5, in a replication of Fischler's results, semantic relatedness alone did produce priming in a lexical decision task. These results appear to indicate that the role of semantics in the priming process is somewhat limited. Further, these results also indicate that the amount of priming observed is somewhat task dependent. Implications for models of "semantic" priming are discussed.