Theories of arousal suggest that arousal should decrease performance on difficult tasks and increase performance on easy tasks. An experiment tested the hypothesis that the effects of stereotype threat on performance are due to heightened arousal. The authors hypothesized that telling participants that a math test they are about to take is known to have gender differences would cause stereotype threat in women but not in men. In the experiment, each participant took two tests--a difficult math test and an easy math test. Compared to women in a "no differences" condition, women in the "gender differences" condition scored better on the easy math test and worse on the difficult math test. Men's performance was unaffected by the manipulation. These data are consistent with an arousal-based explanation of stereotype threat effects. Data were inconsistent with expectancy, evaluation apprehension, and persistence explanations of the stereotype threat phenomenon.