While experimental studies point towards large benefits of highquality preschool programs, evidence on the returns to universal preschooling is rather ambiguous. Although the difference in returns is often attributed to the difference in the quality of care, aspects of preschool quality are still understudied. The study at hand aims at shedding light into the role of preschool quality by analyzing a low-cost and easy-to-implement language training program in preschool. Receiving the language training is associated with an increase in native language skill formation in four-year-olds by about 14 per cent of a standard deviation. Identification of the point estimates relies on a differences-in-differences matching strategy where preschools are compared based on a large set of covariates and the intra-skill difference between language and math skills is exploited to account for individual heterogeneity. The short-term effects are robust against partial identification and placebo checks and subsample evidence indicates that the returns are at least persistent until primary school.