This basic study was designed to explore the conceptualization of prejudice as a form of contextually controlled, derived, and arbitrarily applicable relational responding. Basic studies utilizing RFT methodologies have yielded examples of how stimulus functions of one set of stimuli, such as a stereotyped group, can transform the functions of another stimulus, such as an individual. Priming procedures, as contextual cues, have been used to affect prejudicial responding. Stimuli participating in relational frames have been shown to be sensitive to such priming procedures; however, the role of context in the priming of derived relational responses has not yet been established. In the present study, 11 participants were trained to respond to four 3-member equivalence classes, consisting of word-like stimuli, under the contextual control of two background colors. Participants then completed a single-word lexical decision task in which prime/target pairs, consisting of related and unrelated pairs, were presented with and without contextual cues. For participants who successfully completed the training phase, response latencies to identify related pairs were generally shorter than for pairs involving a neutral word. However, response latencies between related pairs and unrelated pairs, consisting only of previously trained stimuli, failed to meet statistically significant differentiation. Responses were also similar between contextually related and contextually unrelated word pairs. The results failed to indicate the presence of a contextually controlled semantic priming effect at a statistically significant level; however, these results do suggest the possible presence of an episodic priming effect.