Other things being equal, theory would suggest that students in smaller classes at school should do better in terms of attainment; convincing experimental evidence for this also exists in the US. However, a relationship between small classes and better outcomes has not generally been evident in individual-level studies, possibly because of endogeneity arising from low-attaining or otherwise 'difficult' students being put into smaller classes than their higher-achieving counterparts. The present paper uses data from the National Child Development Study to estimate the effects of class size. Ordinary least-squares estimates indicate that small classes are not related to attainment; however, instrumental variables estimates, with class size instrumented by the interaction between school size and school type, show a significant and sizeable association between smaller classes and higher attainment in reading in the early years of school. This effect is common to different groups of students, and for some groups (girls, and those from larger families), this association is also found to persist through to age 11.