The target article's focus on mutualism is a welcome one. Recognizing that actors assort via partner choice, and that this can have a large impact on the evolution of cooperation, are good ideas. It is step away from the simple dyadic structure and Prisoner's Dilemma (PD)–centric worldview that has dominated theory for so long (Trivers Reference Trivers, Kappeler and van Schaik2006). This new understanding allows reasonably sized groups of potential partners.

The sorts of partnerships described ethnographically in the target article are not limited to hunter-gatherers. At bayside in the artisanal fishing village where I work in the Caribbean, on a daily basis men team up in cooperative partnerships in order to go fishing (Alvard, n.d.). These relationships are inherently mutualistic in the sense that the rewards obtained would otherwise be lost to those who go it alone. Baumard et al. say that people choose partners who are more cooperative and offer more in exchange. This is part of the story, but there is also more to it. In some cases, partner preference might be less about how generous or impartial a partner is and more about the extent to which the potential partner's understanding of how costs and benefits are allocated matches one's own understanding. Outcomes may be locally optimal but require a process of equilibrium (or group) selection to obtain the degree of morality discussed in the target article. For example, the standard distribution rule among fishers in Dominica first allocates the proceeds from the sale of fish to pay the cost of motor fuel. The balance is then divided with one share each for the owner of the boat, the owner of the motor, and each crew member. Fishers who do not follow these rules are no one's partners. The distribution norms appear to be designed to facilitate partitioning of resources in a way that reduces transaction costs (Allen Reference Allen1991; Ensminger Reference Ensminger1997; Young Reference Young and Hammerstein2003). Whether or not the rewards are proportional to the share owners' contribution to the hunt's success is an empirical question. I am not yet convinced that social selection via partner choice alone will always favor these sorts of proportional outcomes. Theoretical work shows that multiple, local optima outcomes often result in the context of frequency dependence (Boyd & Richerson Reference Boyd and Richerson1990). Social selection may result in local optima that might not be fair – just common. To the extent that partner choices are constrained by frequency-dependent processes, these processes may not be as immune from the folk theorem as the authors suggest, and a process of group or equilibrium selection may be required to produce the outcomes predicted by the models (Bergstrom Reference Bergstrom2002; Boyd & Richerson Reference Boyd and Richerson1990; Henrich Reference Henrich2004; Wilson & Sober Reference Wilson and Sober1994).

I have written elsewhere that cooperative systems are often usefully examined as coordination games (Alvard Reference Alvard, Stanford and Bunn2001; Alvard & Nolin Reference Alvard and Nolin2002). Baumard et al. refer to old views of mutualism, partner control, and PD, but surprisingly do not relate their views to games of coordination. Unlike the PD, where there is no cooperative Nash equilibrium, coordination games can have multiple equilibria (Boyd & Richerson Reference Boyd and Richerson1990). I suspect that the cooperative solutions produced by partner selection are similar. In a coordination game, being uncooperative brings lower returns, but is often less risky and this is a key difference with the PD (Skyrms Reference Skyrms2004). The stag hunt parable is the classic example where partner choice might facilitate cooperative outcomes. Stag hunting is a cooperative effort that requires a group of hunters because no one can take a stag alone. Hares, however, can be taken alone. The per capita returns from stag hunting are greater than those from hare hunting, and, of course, killing a hare is better than obtaining nothing – which is what one will get if one's partner goes for a hare. Hunters might be expected to select partners who will follow the one basic rule: Go for the stag, do not be tempted by the hare – unless of course, hare hunting is the norm. Solutions are frequency dependent. The best choice depends on the frequency of the strategies among potential partners, and it does not pay to hunt stag if it is difficult to find a stag hunting partner, just as it is not best to go for hare if most folks are hunting stag. In such cases, the outcome may be less about how fair the rule is than it is about finding a partner who shares the rule.

Such rules are not enforced by a state but are usefully viewed as institutions defined as “locally stable, widely shared rules that regulate social interaction” (McElreath Reference McElreath and Singer2008). Institutions can be large, complex, and imposed from the top down in the form of governmental regulations, or be locally generated and smaller scale. Among the Lamalera whale hunters, the rule describing the butchering and distribution of a whale is an institution. Rules are not negotiated each day on the beach, but rather are inherited culturally; clearly at some time in the past, however, agreements were made. Participants have expectations about how their partners will behave, and these expectations are so often met that an observer might assume they are implicit. Converging lines of theoretical research make the key prediction that social structure (i.e., nonrandom, assortative interactions) is fundamental to the evolution of cooperation (Boyd & Richerson Reference Boyd and Richerson2002; Fletcher & Doebeli Reference Fletcher and Doebeli2006; Reference Fletcher and Doebeli2009; Nowak et al. Reference Nowak, Tarnita and Antal2010; Pepper & Smuts Reference Pepper and Smuts2002; Rankin & Taborsky Reference Rankin and Taborsky2009; Sober & Wilson Reference Sober and Wilson1998). Assortment or partner choice brings together players who are more likely to share institutional norms like, for example, how to butcher a whale. Since there are many possible solutions, if one equilibrium has lower extinction rates or produces more migrants, the variants that characterize that equilibrium can spread to the population as a whole (Boyd & Richerson Reference Boyd and Richerson2010). I would encourage Baumard et al. to go even further and place their analysis within a larger context where groups are competing with other similar groups of partners (Wilson & Dugatkin Reference Wilson and Dugatkin1997).