This experiment focused on domain-specific anxiety as a mediator of stereotype threat. A large sample of boys and girls in Spanish high schools (N = 313) completed tests of mathematical ability and mental rotation (i.e., the ability to correctly judge the orientation of an object that has been rotated through the perceiver's imagination) as well as a questionnaire designed to measure math anxiety focusing on "various psycho-physiological reactions, beliefs, and behaviors related to the daily and academic use of mathematics". Prior to completing these tasks, half of the students were placed under stereotype threat by being told that "various research studies have found out that boys have better math ability than girls. We are trying to analyze whether this difference exists (stereotype threat). The other half of students heard no mention of gender differences in math (control). The stereotype threat manipulation did not affect performance on either the math test or the mental rotation exercise. However, analyses focusing on the role of math anxiety significantly qualified the effects of stereotype threat; girls under stereotype threat with low levels of math anxiety performed better on the math test than did girls with high levels of anxiety. Anxiety did not affect math performance when girls were not under stereotype threat, nor did it affect the performance of boys. These results show that gender-based stereotype threat in math might be displayed only by girls with high levels of pre-existing anxiety. Given that stereotype threat produced opposing effects in girls depending on the level of math anxiety, the authors suggest that the construct validity of stereotype threat should be questioned.