There is ample evidence today in the stereotype threat literature that women and girls are influenced by gender-stereotyped expectations on standardized math tests. Despite its high relevance to education, this phenomenon has not received much attention in school settings. The present studies offer the 1st evidence to date indicating that middle school girls exhibit a performance deficit in quasi-ordinary classroom circumstances when they are simply led to believe that the task at hand measures mathematical skills. This deficit occurred in girls working alone or in mixed-gender groups (i.e., presence of regular classmates) but not in same-gender groups (i.e., presence of only same-gender classmates). Compared with the mixed-gender groups, the same-gender groups were also associated for girls in the stereotype threat condition with greater accessibility of positive role models (i.e., female classmates who excel in math), at the expense of both stereotypic in-group and out-group members (i.e., low-math-achievement girls and high-math-achievement boys). Finally, the greater accessibility of positive role models mediated the impact of the activated stereotype on girls' performance, exactly as one would expect from C. M. Steele's (1997) stereotype threat theory. Taken together, these findings clearly show that reducing stereotype threat in the classroom is a crucial challenge for both scientists and teachers.