Language immersion education uses a non-English language for a significant portion of instruction. One goal is to develop bilingual students, which many argue provides cognitive benefits. Using data from school choice lotteries, I estimate the effect of random assignment to a language immersion school on achievement in math and reading. Winning the lottery has a statistically significant positive effect on math scores, and a positive but statistically insignificant effect on reading. Several peer and school characteristics are ruled out as non-language based alternative explanations for the higher math scores of lottery winners.