In the separated arms conditioned cue preference (CCP) task rats are trained by confining them in one arm of an eight-arm radial maze with food and in another arm on the opposite side of the maze with no food on alternate days. After two such trials, rats prefer the food-paired arm when allowed to move freely between the two arms, neither of which contains food. However, if the rats are preexposed to the maze by exploring it without food before training, no preference is observed and at least 4 training trials are required to produce a CCP, suggesting that unreinforced preexposure to the maze latently inhibits acquisition. If this interpretation is correct, preexposure should reduce the size of the preference acquired with both 2 and 4 training trials. In Experiment 1, this prediction was replicated for 2 training trials; however, with 4 training trials, eliminating preexposure also eliminated the CCP. A previous finding that basolateral amygdala lesions impair the CCP with preexposure and 4 training trials was replicated in Experiment 2, but similar lesions had no effect on the CCP in non-preexposed rats given 2 training trials. In contrast, lesions of the central nucleus impaired the 2 training trial CCP but had no effect on the 4 training trial CCP. This double dissociation suggests that the BLA-mediated 4 training trial CCP may be due to learning about the reward features of the maze space, while the central-nucleus-mediated 2 training trial CCP may be due to a conditioned approach response.