Summary.In their recent paper, Gillet al.(1985) claim to have found evidence for “genetic factors” underlying variation in educational achievement. The conclusion, however, is based on a particular model of genotypes and phenotypes and the relation between them. When we come to consider learning and related characters, as evolved devices for coping with environmental change, other models of the genotype‐phenotype relation have to be considered. These generate expected mean squares in statistical models which can lead to conclusions different from those of Gill and colleagues. This paper examines some of these alternatives. Whilst we have a scientific responsibility to resolve the full range of possible models, before strong conclusions can be drawn formidable difficulties stand in the way. In the meantime, conclusions issuing from the application of one particular model must be treated with great ca