An experimental study of the effects of a didactic teaching approach and a constructivist teaching approach on 3rd- and 5th-grade boys' and girls' performance on arithmetic computation problems was conducted. Two groups of children, matched on the basis of initial computation performance as well as grade and gender, were taught how to solve arithmetic problems using one of these two instructional approaches. Analysis of subsequent computation test performance revealed that 5th graders scored higher than 3rd graders, and there was a significant interaction between gender and instruction group. After instruction, girls in the didactic group outperformed boys in both instruction groups and girls who had been taught using constructivist approaches. Gender differences in computation performance can appear relatively early, by 3rd grade, if didactic instruction strategies are used to teach computational rules. The pattern of gender differences suggests that didactic teaching does not handicap boys. Rather, it appears that didactic instruction enhances computational performance in girls.