We examine the empirical determinants of student achievement in higher education, focusing our attention on its small-group teaching component (classes or seminars) and on the role of attendance, number of students per class, peers, and tutors. The empirical analysis is based on longitudinal administrative data from a major undergraduate program where students are allocated to class groups in a systematic way, but one which is plausibly uncorrelated with ability. Although, in simple specifications, we find positive returns to attendance and sizeable differences in the effectiveness of teaching assistants, most effects are not significant in specifications that include student fixed effects. We conclude that unobserved heterogeneity amongst students, even in an institution that imposes rigorous admission criteria and so has little observable heterogeneity, is apparently much more important than observable variation in inputs in explaining student outcomes.