Previous work has indicated that ethanol is a potent stereotypy-inducing agent. At least this is the case for spontaneously emitted instrumental behavior. The present experiments were undertaken to determine if spatial variability could be generated by drugged rats when it was enforced by reward contingencies. With a reward nonreplacement rule in force, four arms of an 8-arm radial maze were baited on every trial. Rats injected with 0, 0.75, 1.5, or 2.0 g/kg ethanol were required to run to the same set of arms from trial-to-trial and session-to-session. Efficient performance depended upon their running to the correct set of arms as well as meeting a "win-shift" demand which proscribed returning to previously visited arms during a given trial. Although all groups were eventually able to run to the correct set of arms, alcohol, especially at higher doses, promoted repetition. The inability to refrain from reentering arms prevented many alcohol-injected animals from obtaining the four rewards in the allotted time. In Phase 2 of Experiment 1, the baited arms were rotated 45 degrees. Now the formerly empty arms contained pellets and rewards were withdrawn from the previously correct arms. Adjustment to this shift was rapid for 0 and 0.75 g/kg groups, but an increasingly severe perseveration was observed across the higher ethanol groups. Experiment 2 reproduced the results of Experiment 1 under different circumstances. While trained as before to run to a specific set of four arms in Phase 1, Phase 2 presented the rats with rewards in all eight arms of the maze. With higher doses of alcohol an increasing persistence in running to the original four arms was observed. Saline-injected animals, on the other hand, rapidly doubled the number of pellets taken. Taken together, and in view of earlier findings, the results suggest that alcohol interacts with previous training as well as recent choices with the result that spatial dispersion is restricted in spite of explicitly opposing reward contingencies.