Dividing students’ school days into subject-specific classes is a critical resource allocation decision in educational production. While the optimal allocation of instructional time depends on the nature of complementarity (substitutability) between subjects, intermediate analysis of potential allocations should measure gains and losses across subjects. Using regression discontinuity methods, I study a sample of middle-school students who were quasi-randomly assigned to take a second math class for one year instead of taking a class in foreign languages, PE, or arts. Students scored 0.12-0.15σ higher in math at the end of the treatment year, but the gain faded over the two subsequent years. At the end of high school, treated students were less likely to have completed two years of a foreign language, a common requirement for college entrance; but were not observably different from control students in exposure to PE and arts.