In spite of the apparent success of affirmative action (AA) in the past, many oppose such policies. Opponents argue that the cost of attaining proportional representation by preferential policies is too high, reducing the quality of selected groups and stigmatizing members of the protected class. One way in which preferential policies might harm groups they are designed to benefit is by producing stereotype threat; that is, cueing a negative stereotype may lead individuals to conform to it. AA, by definition, singles out disadvantaged groups and therefore may unintentionally remind beneficiaries of relevant negative stereotypes and prime a stereotype threat effect. This paper investigates experimentally whether gender-based affirmative action, in the form of a quota, may have this unintended consequence. Using quantitative Graduate Record Examination (GRE) questions, we find that while affirmative action has a positive effect on the performance of non-high-ability women, it negatively affects high-ability women in our sample. Interestingly, there is no evidence that women (of any ability level) in the sample exert less effort under AA or in single-sex competition, or would benefit by doing so. Hence, these findings are consistent with affirmative action having an unintended negative consequence due to stereotype threat. Interestingly, men are not affected by the affirmative action policy regardless of their measured ability in this task. These results should be taken with caution, as this is the first study looking at the stereotype threat effect of AA.