Derived equivalence relations, it has been argued, provide a behavioral model of semantic or symbolic meaning in natural language, and thus equivalence relations should possess properties that are typically associated with semantic relations. The present study sought to test this basic postulate using semantic priming. Across three experiments, participants were trained and tested in two 4-member equivalence relations using word-like nonsense words. Participants also were exposed to a single- or two-word lexical decision task, and both direct (Experiment 1) and mediated (Experiments 2 and 3) priming effects for reaction times and event-related potentials were observed within but not across equivalence relations. The findings support the argument that derived equivalence relations provides a useful preliminary model of semantic relations.