This study investigated the benefits of all-girls' classroom instruction in math and/or science during Grades 9 and/or 10, within the context of a public co-educational high school. There were 118 participants in this longitudinal investigation: 26 girls in the all-girl classes, as well as 42 girls and 50 boys in the regular co-educational program. A range of pre-existing achievement, background, and psychological characteristics were included as covariates to ensure comparability of the groups. Individual growth curve modeling was used to examine how math achievement and perceived math competence changed over the course of high school and postsecondary education. Change was found to be nonlinear for both constructs (with math achievement starting high, dipping, and then recovering; and with perceived competence doing the opposite). The all-girl intervention appeared to act as a protective mechanism against the temporary decline in math achievement experienced by both co-ed boys and co-ed girls. (Contains 3 tables and 2 figures.)