In this study, the effects of single-sex versus co-educational classes and schools on the progress in language and mathematics of boys and girls at the end of the second year of secondary education are investigated. Data from the Longitudinaal Onderzoek Secundair Onderwijs project are used. Multilevel analyses were carried out on a sample of approximately 4000 pupils, 330 classes (190 single-sex), 180 teachers and 50 schools (20 single-sex). The results indicate that for boys the gender composition of the classes has more impact than the gender composition of the schools, whereas for girls the gender composition of the schools is more important. Boys make more progress for language (and not for mathematics) in co-educational classes even after we have taken into account the selective nature of the classes. Girls, on the other hand, make more progress for mathematics (but not for language) in single-sex than in co-educational schools.