Assortative mating is clearly important in determining the genetic structure of a population but studies of assortative mating in man have been restricted mainly to the examination of correlations between spouses for single variables. It has been assumed that assortation is based directly on the measured phenotype or on components of the phenotype whose determination is similar in both sexes. Recently, such assumptions have been questioned (see, for example, refs 1-3) because males and females may regard different aspects of the phenotype as important for mate selection. The term 'asymmetric assortative mating' has been used to describe such situations. It has also been suggested that the study of the spouses and offspring of twins1,2 may permit resolution of such components of the mating system. Ad hoc models have been proposed (for example, ref. 1) to specify the contribution of asymmetric assortment to the correlations between relatives, but these have not been internally consistent or sufficiently parsimonious. We show here how a simple model based on biological considerations offers a foundation for a consistent and economical theory of asymmetric mate selection which may be used for the analysis of patterns of familial similarity.