This study tested the effects of exercise on eating behavior. The transfer hypothesis proposes that exercise leads to a generalization of healthy behavior and therefore an improved diet. The compensation hypothesis assumes that exercise leads to increased caloric intake in order to 'compensate' for the energy expenditure. We tested these hypotheses for actual as well as imagined exercise. Female university employees or students (N = 227) were randomly assigned to three experimental groups: actual exercise vs. imagined exercise vs. control. After baseline data had been obtained, the participants engaged in a 5-minute experimental task and were then left alone with unhealthy snacks. Participants who had imagined themselves exercising (M = 101 kcal, SD = 128 kcal) consumed significantly fewer calories than did controls (M = 129 kcal, SD = 142 kcal), consistent with a transfer effect. Participants who had engaged in actual exercise, but had been distracted from thinking about exercise, consumed quantities (M = 127 kcal, SD = 111 kcal) similar to those consumed by controls. This study suggests that transfer effects are underpinned by psychological processes, such as goal activation, which should be investigated in the future.