In 2 field experiments, we relied on the very features of real testing situations--where both math and verbal tests are administered--to examine whether order of test administration can, by itself, create vs. alleviate stereotype threat (ST) effects on girls' math performance. We predicted that taking the math test before the verbal test would be deleterious for girls' math performance (ST effect), whereas taking the verbal test before the math test would benefit their math performance. We also explored whether ST (if any) may spill over from the math test to the verbal test in a real-world testing situation. The studies were conducted among French middle-school students (Ns = 1,127 and 498) during a regular class hour. In both studies, whereas girls underperformed on the math test relative to boys in the math-verbal order condition (ST effect), they performed as well as boys in the verbal-math order condition. Moreover, girls' math performance was higher in the verbal-math order condition than in the math-verbal order condition. Test order affected neither girls' verbal performance (no ST spillover) nor boys' verbal or math performance. In Study 2, additional measures pertaining to students' self-evaluations in and perceptions of the math and verbal domains provided complementary evidence that only girls who took the math test first experienced ST. Implications of order of test administration for women's experience in math, for ST effect and ST spillover research, and for educational practices are discussed.