The purpose of the study was to use multivariate multilevel techniques to investigate whether it was possible to separate different dimensions in grades that relate to subject-matter achievement and to other factors. Data were derived from The Gothenburg Educational Longitudinal Database (GOLD), and the subjects were 99,070 ninth-grade students born in 1987. The analyses were based on subject grades and scores on national tests in Swedish, English, and mathematics. The results showed that, at both individual and school levels, the greatest part of the variance in grades was due to achievement in the different subject areas. At both levels, it was possible to identify a dimension that cut across the grades in all 3 subjects, which suggests that grading is influenced by factors other than achievement. One of the most interesting results concerns the relation between parental education and the common grade dimension at the school level. (Contains 1 figure and 5 tables.)