According to Kornblum's (1992) dimensional overlap model, when an incongruent response to a stimulus is required, automatic activation of the congruent response must first be inhibited. Shiu and Kornblum (1996a) provided evidence for such inhibition in an incongruent symbolic negative priming task. Reaction time was longer when a trial's correct response was the name of the stimulus from the previous trial than when it was not. We report three experiments that test this inhibition hypothesis for spatial stimuli and responses. In Experiment 1, which used a spatial mapping analogous to the symbolic mapping used by Shiu and Kornblum (1996a), a similar negative priming effect was found. However, in Experiments 2 and 3, which used mappings that were conducive to simple transformational rules, a positive priming effect was obtained. The results suggest that inhibition in response selection may depend on the complexity of the relations between the stimuli and responses.