Teacher Mindsets Help Explain Where a Growth-Mindset Intervention Does and Doesn't Work

9,167 student records matched to 223 mathematics teachers, a subset of the National Study of Learning Mindsets sample

US ninth graders and their mathematics teachers.

Secondary moderator analysis of the NSLM randomisation. Not separately preregistered - the authors state they reused the 2019 preregistered model specification plus a conservative Bayesian analysis in place of a new preregistration. Teacher mindset was measured pre-intervention by teacher self-report, and the outcome was switched from overall GPA to ninth-grade mathematics grades.

Where the mathematics teacher held more of a fixed mindset the intervention did nothing (CATE = -0.02, 95% CI -0.074 to 0.038, t = -0.61, p = .542). Where the teacher held more of a growth mindset it worked (CATE = 0.11, 95% CI 0.046 to 0.167, t = 3.37, p < .001). The continuous interaction was b = 0.09 (95% CI 0.026 to 0.150, p = .006), robust to school-level moderators and to restricting to single-teacher students.

Low - the treatment contrast is randomised. The moderator is not randomised, so the teacher-mindset story is correlational within an experiment.

Unreplicated as a standalone moderator claim, and directly contradicted by Porter et al. 2022, which found the LARGEST effects where teachers held FIXED mindsets. Two developer-side papers, opposite moderators, same year.