We conduct a randomized controlled trial of an Indian school library program. Overall, the program had no impact on students' scores on a language skills test administered after 16 months. The estimates are sufficiently precise to rule out effects larger than 0.053 and 0.037 standard deviations, based on the 95 and 90 percent confidence intervals. This finding is robust across individual competencies and subsets of the sample. The method of treatment, however, does seem to matter – physical libraries have no effect, while visiting librarians actually reduce test scores. We find no impact on test scores in other subjects or attendance rates.