Immigrant children in the US tend to perform worse in reading, mathematics, and science compared to native children. This paper explores how much of such differences in achievement can be accounted for by a lack of English proficiency. To identify the causal effect of English proficiency on cognitive test scores, I use the fact that language proficiency is closely linked to age at arrival, and that migrant children arrive at different ages from different countries of origin. In particular, I instrument English proficiency by comparing children from English-speaking countries to children from non-English-speaking countries who migrated to the US at different ages. Using data from the New Immigrant Survey, I find that speaking English very badly or badly can explain 27-33% of the achievement gap between native and immigrant children in standardized language-related tests. However, I find no significant language effects for applied maths problems or calculations.