Conflicting predictions of reinforcement and neophobia-arousal theories were evaluated in a simple choice task. Four groups of C57BL/6J mice were administered daily two-trial tests in a uniform T-maze for 10 consecutive days. For three groups, the contingencies of footshock treatments were manipulated to reinforce alternation, perseveration, or both. A control group that was not administered footshock alternated, but all three groups that were stressed perseverated more and more across tests, despite the differences in reinforcement contingencies. These results are inconsistent with the predictions of reinforcement theory but consistent with the view that stressed or aroused animals are neophobic and use nonassociative learning (habituation) to distinguish between novel and familiar alternatives.