Many models assume that habitual human behavior is guided by spontaneous, automatic, or implicit processes rather than by deliberate, rule-based, or explicit processes. Thus, math-ability self-concepts and math performance could be related to implicit math-gender stereotypes in addition to explicit stereotypes. Two studies assessed at what age implicit math-gender stereotyping can be observed and what the relations between these stereotypes and math-related outcomes are in children and adolescents. Implicit math-gender stereotypes could already be detected with Implicit Association Tests (Greenwald, McGhee, & Schwartz, 1998) among 9-year-old girls. Adolescent girls showed stronger implicit gender stereotypes than adolescent boys, who, on average, did not reveal implicit gender-stereotypic associations. Girls also already showed an implicit affinity to language versus math at 9 years of age. In a regression analysis, implicit math-gender stereotypes predicted academic self-concepts, academic achievement, and enrollment preferences above and beyond explicit math-gender stereotypes for girls but (with the exception of achievement) not for boys. These findings suggest implicit gender stereotypes are an important factor in the dropout of female students from math-intensive fields. (Contains 4 tables, 3 figures and 8 footnotes.)