Understanding class size effects on educational achievement remains a preoccupation of many economists. But empirical results are, to this far, still inconclusive. I use the two-stage least squares and the instrumental variable quantile regression methods on Lesotho’s grade 6 students maths and reading test scores to estimate, respectively, the mean and distributional effects of class size. I find strong evidence for putative class size effects on reading achievement, but not on maths achievement. Using the bounds methods of Conley et al (2012), I show that these findings are robust to weak violations of the instrumental variable exclusion restriction requirement. I also find that a small class size generally benefits the low-ability students the most.