I study the effect of coeducation on academic performance and science course enrollment using student-level data from a South Korean high school that, at the beginning of each academic year, quasi-randomly assigned students to either single-sex or coeducation classrooms. This paper presents two main findings. First, among girls with high achievements in Korean, coeducation increases the Korean exam scores by 0.11 standard deviations. Second, coeducation increases a girl's probability of enrolling in Physics, a math-intensive and boy-dominated science subject, by 71%, and increases a boy's probability of taking Biology by 34%.