The question of“how a self‐fertile hermaphrodite will distribute the resources that it allocates to reproduction is studied by means of the ESS approach. Different models of the relations between allocation to male function, the male and female fertilities, and the selfing rate, yield different conclusions about how much resource should be allocated to male function. Values below a half are obtained with one model, while another can give values greater than a half. Even with no selfing, values other than a half are usually obtained; with both models studied, the values decrease with increasing selling. If the selfing rate is assumed to be independent of the fraction of resources allocated to male function, it can be shown that the ESS allocation to male function always decreases as selling increases. The types of relations that might be expected in species with different types of breeding biology, and some data on allocation to male function, are reviewed.The implications for the fitness of male‐ and female‐sterility mutations are discussed. It is argued that the concavity or convexity of the curve relating female fertility to male fertility is not a good guide to when hermaphroditism should exist when there is some selfing. Even with a concave relation, male‐sterility mutants can have a higher fitness than hermaphrodites, if there is some selling and inbreeding depression. Also, when the selfing rate depends on allocation to male I unction, an hermaphrodite ESS does not always exist when the function is concave (as it does when there is no selfing), and such an ESS may exist when the relation is convex. The fitness of male‐ or female‐sterility mutants may also depend on the existence of ‘fixed costs’. It is shown that these do not ailed the ESS alloca