Unfortunately, these results do not make the evolution of adaptations for collective action any less mysterious. Because punishing a free rider would generally have entailed some nontrivial cost, each potential punisher has an incentive to defect—that is, to avoid this cost by not punishing acts of free riding. Thus, the provision of punishment is itself a public good: Each individual has an incentive to free ride on the punishment activities of others. Hence, second-order free riders should be fitter (or better off) than punishers. Without a way of solving this second-order free rider problem, cooperation should unravel, with nonparticipation and nonpunishment the equilibrium outcome. Even worse, this problem reappears at each new level, revealing an infinite regress problem: Punishment needs to be visited on free riders on the original public good, and on those who do not punish free riders, and on those who do not punish those who do not punish free riders, and so on.
A different Cosmides-and-Tooby (and Michael E. Price) take: