A quantitative genetic model of two-policy games between relatives.

Equations are derived for the change per generation of the population mean of the probability that an individual adopts a policy 1 as opposed to a policy 2 in a behavioral interaction between two diploid individuals of the same generation in which two policies are possible. The probability is assumed to be a quantitative genetic trait determined by many additively acting genes of small effects and an independent environmental component. Equations are derived for the case that interactions occur at random between all members of the population and also for the case that interactions occur between relatives of the same average degree of relatedness. It is assumed that each group of relatives and the number of such groups is sufficiently large. For a quantitative genetic trait with the additional assumption of unlinked loci the latter equation can be heuristically derived from the first by substituting the corresponding inclusive fitness effects. When per locus selection coefficients are small and linkage equilibrium holds, the average degree of relatedness can be equated approximately with Wright's coefficient of relationship. Thus, the quantitative genetic model provides a genetic basis for the inclusive fitness approach toward games between relatives. By contrast, in a monogenic system with major gene effects we obtain substantially different results which contradict those obtained by the inclusive fitness approach in game theory. Applications are made to the hawk-dove game, and the simple and iterated forms of the prisoner's dilemma.