The genetics of human family size is not readily susceptible to analysis. Some of the problems in its investigation are discussed, together with methods for overcoming them. In particular, the problem of obtaining a satisfactory model which will account for both the low heritability of family size and its over-dispersed distribution is examined in detail.