Research is presented on the attributional gender bias: the tendency to generate different attributions (explanations) for female versus male students' performance in math. Whereas boys' successes in math are attributed to ability, girls' successes are attributed to effort; conversely, boys' failures in math are attributed to a lack of effort and girls' failures to a lack of ability. This bias has been shown in previous research to be committed by teachers, parents, and students themselves. The present work sought to investigate whether this bias among secondary school math teachers might be reduced over time through adoption of an incremental theory of intelligence. Findings revealed at baseline, teachers committed the expected bias in reference to their high-achieving students' math performance. Following exposure to stimuli, teachers in both experimental and control conditions reduced this bias. Unexpectedly, teachers across conditions showed a type of compensation for the bias by reversing stereotypical attributions for girls' and boys' successes and failures in math. Further, participants relapsed to the original bias nearly a year later. Findings indicate the potential to modify attributional gender bias, but also the challenges for achieving long-term changes within school contexts and for emphasizing effort beyond ability in math performance.