Guala should be praised for having raised this debate about punishment experiments. I will focus on two main points. First, the target article claims that the empirical evidence on peer punishment is not enough to support theories based on strong reciprocity. As I argue below, behavior in peer punishment experiments cannot be entirely rationalized with self-regarding or weak reciprocity attitudes, and strong reciprocity is one model of other-regarding behavior among others currently under debate.

There is no lack of anecdotes about peer pressure and punishment in field settings, ranging from high school students to miners on strike (Francis Reference Francis1985) to fishermen communities (Bromley Reference Bromley1992) to workplaces (Kandel & Lazear Reference Kandel and Lazear1992). In the region studied in Casari (Reference Casari2007), costly punishment is still practised today. Recently, 1,800 young grapevines have been cut with pruning hooks and shears. Apparently two people acted overnight, causing damage in thousands of euros. In the last five years, there have been seven similar episodes in the same community. Generally the culprits remain unknown (Nardon Reference Nardon2011). The issue of peer punishment was raised after field research and was not born as a laboratory anomaly. Experiments helped to clarify the extent and drivers of peer punishment, because field evidence is often hard to interpret. There are nuisance factors and measurement limitations: The interaction may be repeated, the fine-to-fee ratio unknown, or institutions to promote cooperation may be present. Controlled experiments are useful because they remove many of these limitations. One robust finding is the willingness of many people to pay a personal cost to inflict a punishment on others, especially on free-riders. This result persists in one-shot situations when the punisher incurs a material loss. As in other experiments, the data point toward the existence of a mix of motivations in economic decision making. While most subjects exhibit exclusively self-regarding motivations, there are others who also exhibit an array of other-regarding motivations.

Weak reciprocity is simply not enough to rationalize the existing experimental results on peer punishment. For instance, subjects do not treat peer punishment as a second-order public good, that is, they do not employ punishment mainly to provide incentives for the free-rider to contribute, as a weak reciprocity argument would suggest (Casari & Luini Reference Casari and Luini2006; Reference Casari and Luini2009). One can also experiment settings with indefinite repetition, where weak reciprocators can support cooperative outcomes through a rational strategy different than costly peer punishment. When four subjects indefinitely played prisoner's dilemmas in random pairs, more than half of the time cooperators targeted defectors with peer punishment (Camera & Casari Reference Camera and Casari2009). Rational, self-regarding subjects had the alternative to support full cooperation through a simple grim trigger strategy. Instead, many still employed peer punishment. To sum up, experiments on peer punishment have shed light over important aspects of cooperative behavior that are likely to apply also in field situations. Yet, the existing evidence still leaves some deep questions open about the genetic versus cultural origin of other-regarding motivations; about the degree of external validity of experiments; and, about what model can fit the observed patterns of punishment with reasonable precision.

Guala's second main point is that cooperation in the field does not rely primarily on the forces uncovered in punishment experiments but is promoted by institutions that reduce the costs of decentralized punishment and facilitate the functioning of weak reciprocity mechanisms. I agree, although I will discuss two of Guala's related statements, which are based on unconvincing interpretations of the anthropological evidence: (1) Peer punishment does not occur in the field; (2) hence, it is irrelevant in a field setting. Guala argues that peer punishment is rarely employed and that some punishment acts are not costly, given that the cost to inflict punishment is claimed to be “low.” In the literature, what matters is the fine-to-fee ratio of a punishment act, not simply the absolute cost of a punishment request. Moreover, sanctions ought not to be always large but, rather, graduated (Ostrom Reference Ostrom1990). In the lab one observes a proportion between crime and punishment, that is, actions of full free-riding attract more punishment than actions of partial free-riding, and something similar may be expected in the field.

When extrapolating to field situations, one has to keep in mind that in laboratory experiments, people are forced to interact with others, have little control over the information flow, and have only few options available. In the field, people have multiple ways to inflict punishment and have strategies alternative to peer punishment. Instead of physically confronting a norm violator, a cooperator may decide to act to lower the cost to punish, to create institutions, or to move camp elsewhere. Hence, people can optimize over the many strategies available. A lower-than-expected frequency of peer punishment actions may simply reveal that there are better strategies in that situation, not that they are unavailable or irrelevant. For instance, speaking up against someone is costly because it exposes one to the risk of retaliation (Wiessner Reference Wiessner2005), as whistle-blowers know. To avoid counter-punishment, in the field people may increase the level of anonymity by spreading gossip instead of reproaching someone face-to-face. Another way to dilute the risk of retaliation is to punish in coalition with others, which is another form of decentralized punishment (Casari & Luini Reference Casari and Luini2009).

The success in overcoming a social dilemma situation may be due to multiple factors, and other-regarding attitudes may be one of them. Although in the field they can be sometimes hard to quantify, they may nevertheless play an important role. The analysis of the Carte di Regola followed a canonical model with identical, self-regarding agents out of parsimony in organizing the historical evidence (Casari Reference Casari2007). A related experiment examined in more depth a specific feature of the Carte system and uncovered the subtle role of subjects' heterogeneity in behavior for the success of the institution. Under a Carte system, the interaction of pro-social, self-regarding, and anti-social attitudes increased group welfare (Casari & Plott Reference Casari and Plott2003). Field and experimental evidence, therefore, complement each other.