I gathered 77 randomized experiments (with 111 treatment arms) that evaluated the effects of school-based interventions on learning in developing-country primary schools. On average, monetary grants and deworming treatments had mean effect sizes that were close to zero and not statistically significant. Nutritional treatments, treatments that disseminated information, and treatments that improved school management or supervision, had small mean effect sizes (0.04-0.06) that were not always robust to controls for study moderators. The largest mean effect sizes included treatments with computers or instructional technology (0.15); teacher training (0.12); smaller classes, smaller learning groups within classes, or ability grouping (0.12); contract or volunteer teachers (0.10); student and teacher performance incentives (0.09); and instructional materials (0.08). Metaregressions suggested that the effects of contract teachers and materials were partly accounted for by composite treatments that included training and/or class size reduction. There are insufficient data to judge the relative cost-effectiveness of categories of interventions.