Van de Kragt, Orbell, and Dawes (1983) reported the results of a series of experiments in which n subjects each receive a monetary endowment and then may choose privately whether to contribute it to a monetary public good; the good is supplied if a prespecified number m < n of contributions or more is made. Decision policies maximizing expected value are derived for this experimental paradigm under three different assumptions about the expectations each individual has about the decisions of the other n – I members: A homogeneity assumption postulating that each other member contributes with fixed probability p, a heterogeneity assumption postulating that the n – I p's are independently selected from a subjective probability distribution, and a partial homogeneity assumption postulating that the n – I members are partitioned into distinct subsets with the same p for all members of each subset. The theoretical and social implications of these assumptions are briefly discussed.