We agree with Guala that regulation of cooperation by punishment is infrequent and often low-cost, at least in small-scale societies. Analytical models and experimental studies suggest that solutions to cooperative dilemmas do not depend on direct punishment if individuals can opt out of unproductive partnerships (Aktipis Reference Aktipis2004; Hauert et al. Reference Hauert, Traulsen, Brandt, Nowak and Sigmund2007) or assort with preferred cooperative partners whether kin (Hamilton Reference Hamilton1964) or non-kin (Barclay & Willer Reference Barclay and Willer2007; Noe & Hammerstein Reference Noe and Hammerstein1994). Guala cites Wiessner (Reference Wiessner2005), who observes that !Kung who shirk their responsibilities are ignored more often than they are verbally punished. Among traditional Tsimane horticulturalists of Bolivia, most conflicts are between close kin and regular cooperative partners (von Rueden et al. Reference von Rueden, Hooper, Gurven and Kaplan2009), who generally prefer reconciliation to revenge. Furthermore, defection among parties with few long-term shared interests is more often met with withdrawal and “voting with one's feet” than with punishment.

Guala does not distinguish second- from third-party punishment, but strong reciprocity theorists argue that both contribute to the maintenance of social norms (Fehr & Fischbacher Reference Fehr and Fischbacher2004). There is no consensus over whether third parties often punish or punish “enough.” In experimental games, third-party punishment is least common in small-scale societies (Marlowe et al. Reference Marlowe, Berbesque, Barr, Barrett, Bolyanatz, Cardenas, Ensminger, Gurven, Gwako, Henrich, Henrich, Lesorogol, McElreath and Tracer2008), and third parties may be especially wary of becoming involved in serious conflicts. A Tsimane man committed murder on two occasions, but punishment (public whipping) was administered only after the second murder. The community that sentenced and whipped him was not his resident community but a more acculturated community with more influential men. Nonpartisan members of the murderer's own community would not risk the threat of his retaliation.

Punishment occurs when there is minimal risk of (1) losing a valued exchange partner, (2) suffering reputational damage, or (3) provoking retaliation. For example, a low-status Tsimane man was in long-standing disputes with his neighbor over land and with his son-in-law over investment in his daughter. With few allies to support him, the man moved to another village with his family, with plans to return in a few months. The next day, the neighbor harvested the yucca from the man's field, and the son-in-law burned the man's house. The neighbor and son-in-law did not expect reputational damage or retaliation because they had strong kin support within the community, they could not be unambiguously identified as the punishers, and the punished man had few allies.

Guala identifies gossip as a low-cost alternative to direct punishment. Gossip can spread reputation-damaging information while obscuring the source of that information. Individuals may also gossip to gauge and build community support for punishment that is coordinated and more direct. As Guala argues, punishment that is coordinated carries less risk of retaliation and can be more effective at stabilizing collective action than distributed, individual acts of punishment (Boyd et al. Reference Boyd, Gintis and Bowles2010; Casari & Luini Reference Casari and Luini2009). Among the !Kung, Wiessner (Reference Wiessner2005) found that most harsh criticism was delivered by a coalition, and coalition-based punishment was twice as likely to provoke conformity in the accused. Among the Tsimane, most conflicts are confined to the parties directly involved, but on occasion a small, informal gathering of men will act as third-party adjudicators. The most serious conflicts among the Tsimane, such as those involving physical violence, are sometimes discussed in community-wide meetings in more acculturated villages, where influential individuals will try to generate consensus concerning the relative guilt of the parties in conflict. The community may decide to inflict punishment, usually verbal censure, community service, or public whippings on rare occasions. One village has a de facto rule that the whipper not yet be a father; he has no risk then of his children being targets of vengeance.

Coordinated sanctioning, however, may not be necessary to explain why individuals punish free-riders and non-punishers. Another explanation, which Guala does not discuss, relies on inter-individual differences in formidability, endowments, or in the expected gains from successful cooperation. Individuals with greater physical or social capital can punish with less risk of retaliation and with greater efficacy, and those who anticipate greater relative gains from cooperation are more willing to absorb costs of punishment to achieve those gains. In general, inter-individual differences can be powerful catalysts of cooperation, transforming prisoner's dilemmas into mutualisms and resolving second-order dilemmas of who punishes (Olson Reference Olson1965; Ruttan Reference Ruttan2008). Among the Tsimane, 66% of adjudicated conflicts were arbitrated by men in the top 10% of coalitional support within their community (von Rueden et al. Reference von Rueden, Hooper, Gurven and Kaplan2009). These individuals can steer conflict outcomes in their favor and their actions are less likely to be challenged.

Inter-individual differences in the costs to punishing contribute to the establishment of leaders and followers. Collective action, particularly in large groups, often depends on leaders bearing the costs of coordination and punishment in return for a greater share of the spoils (Hooper et al. Reference Hooper, Kaplan and Boone2010). Tsimane men do not gain more direct material benefits from organizing collective fishing events or acting as leaders in face-to-face collective action games (von Rueden et al. Reference von Rueden, Gurven and Kaplan2010), but long-term reputational benefits may be non-trivial. Positive reputations can serve as insurance against times of need (Boone & Kessler Reference Boone and Kessler1999; Gurven et al. Reference Gurven, Allen-Arave, Hill and Hurtado2000) or as signals to mates and allies of quality or cooperative intent (Bliege Bird & Smith Reference Bliege Bird and Smith2005). Where joint production is subject to greater economies of scale, such as in agricultural societies, coordination and punishment by leaders may pay even greater dividends.

We encourage more study of the role of inter-individual differences in the generation of punishment and cooperation. In the lab, players often feel equally entitled and motivated, while subject endowments are too often windfalls; these conditions rarely hold in natural settings. As Guala recognizes, context matters in shaping how social preferences impact behavior (Gurven & Winking Reference Gurven and Winking2008; List Reference List2006; Wiessner Reference Wiessner2009), so caution is required when making inferences from particular experimental games. Some experimental games, however, have introduced asymmetries into the effectiveness with which players punish (Nikiforakis et al. Reference Nikiforakis, Normann and Wallace2011), into decision-making authority over the distribution of public good shares (van der Heijden et al. Reference van der Heijden, Potters and Sefton2009), or into initial player endowments, as a function of individual inputs to joint production (Konigstein Reference Konigstein2000). With greater understanding of the pervasiveness of inter-individual differences and other cost-reducing conditions, punishment may not appear so altruistic after all.