Minimizing the costs that others impose upon oneself and upon those in whom one has a fitness stake, such as kin and allies, is a key adaptive problem for many organisms. Our ancestors regularly faced such adaptive problems (including homicide, bodily harm, theft, mate poaching, cuckoldry, reputational damage, sexual aggression, and the infliction of these costs on one's offspring, mates, coalition partners, or friends). One solution to this problem is to impose retaliatory costs on an aggressor so that the aggressor and other observers will lower their estimates of the net benefits to be gained from exploiting the retaliator in the future. We posit that humans have an evolved cognitive system that implements this strategy – deterrence – which we conceptualize as a revenge system. The revenge system produces a second adaptive problem: losing downstream gains from the individual on whom retaliatory costs have been imposed. We posit, consequently, a subsidiary computational system designed to restore particular relationships after cost-imposing interactions by inhibiting revenge and motivating behaviors that signal benevolence for the harmdoer. The operation of these systems depends on estimating the risk of future exploitation by the harmdoer and the expected future value of the relationship with the harmdoer. We review empirical evidence regarding the operation of these systems, discuss the causes of cultural and individual differences in their outputs, and sketch their computational architecture.

Humans have the cognitive abilities to implement the revenge and forgiveness systems hypothesized by McCullough et al., but the evidence suggests that simpler processes may underlie most revenge cases in humans and other animals. The mediating role of emotions can be at the basis of the flexibility needed in the hypothesized systems and the associated assessment of social relationships.

We address how trait self-control and trait concern for others relate to the concepts of monitored and intrinsic Welfare Tradeoff Ratios (WTRs), respectively, and how recent work on personality, revenge, and forgiveness are informed by the adaptationist perspective proposed in the target article. We also discuss how the proposed adaptationist perspective provides clues to some previously puzzling findings on revenge.

The target article's important point is easily misunderstood to claim that all revenge is adaptive. Revenge and forgiveness can overstretch (or understretch) the bounds of utility due to misperceptions, minimization of costly errors, a breakdown within our evolved revenge systems, or natural genetic and developmental variation. Together, these factors can compound to produce highly abnormal instances of revenge and forgiveness.

There are interesting parallelisms between McCullough et al.'s article and studies of revenge presented by French legal anthropologist Raymond Verdier, particularly as regards the discussion of the increasing likelihood of revenge with increasing social distance. Additionally, the observation that many peoples speak of revenge in the language of debt and repayment, links it with exchanges of benefits as well as costs.

McCullough et al. argue that humans possess evolved computational systems for implementing retaliatory behavior that both deters aggression and promotes subsequent reconciliation. However, they do not apply this analysis to the sphere of intergroup relations. We believe their model can be usefully extended to this domain and discuss why this would be possible, pertinent, and productive.

General systems for reciprocity explain the same phenomena as the target article's proposed revenge system, and can explain other cooperative phenomena. We need more reason to hypothesise a specific revenge system. In addition, the proposed calculus of revenge is less sensitive to absolute magnitudes of revenge than it should be.

We focus on two aspects: First, we argue that it is necessary to include implicit forgiveness as an additional adaptive behavioral option to the perception of interpersonal transgressions. Second, we present one possible way to investigate the cognitive-affective underpinnings of revenge and forgiveness: a functional MRI (fMRI) approach aiming at integrating forgiveness and revenge mechanisms into a single paradigm.

McCullough et al.'s target article is a psychological version of the reputation models pioneered by biologist Robert Trivers (1971) and economist Robert Frank (1988). The authors, like Trivers and Frank, offer an implausible explanation of the fact that revenge is common even when there are no possible reputational effects. I sketch a more plausible model based on recent research.

We question whether the postulated revenge and forgiveness systems constitute true adaptations. Revenge and forgiveness are the products of multiple motivational systems and capacities, many of which did not exclusively evolve to support deterrence. Anger is more aptly construed as an adaptation that organizes independent mechanisms to deter transgressors than as the mediator of a distinct revenge adaptation.

McCullough et al. hypothesize that evolution has selected mechanisms for revenge to deter harms and for forgiveness to preserve valuable relationships. However, in highly dependent relationships, the more adaptive course of action may be to remain unaware of the initial harm rather than risk alienating a needed other. We present a testable model of possible victim responses to interrelational harm.

The target article's evolutionary approach provides an excellent framework for understanding when and why people retaliate or forgive. We argue that recent findings on the basic processes in forgiveness – particularly, the role of executive control – can further refine the authors' proposed model. Specifically, the lack of executive control may restrict the explanatory power of relationship value and exploitation risk.

This commentary discusses dozens of ecologically powerful social-psychological experiments from the1960s and 1970s, which are highly relevant especially for predicting the consequences of revenge. McCullough et al. omitted this work – perhaps because of its misclassification as “catharsis” research. The findings are readily accommodated by Konečni's anger-aggression bidirectional-causation (AABC) model and can be usefully incorporated in an adaptationist view of revenge.

We review two subjective (mis)perceptions that influence revenge and forgiveness systems. Individual differences predict more (e.g., narcissism) or less (e.g., empathy) revenge, with the opposite pattern for forgiveness. Moreover, differences in victim versus perpetrator perceptions can influence revenge and forgiveness systems, perpetuating never-ending cycles of revenge. These two examples point to the need for theories of revenge and forgiveness to address the role of cognitive and motivational biases in the functionality of such behavioral responses.

McCullough et al. suggest that revenge and forgiveness rest upon risk computation. Risk computation is implemented by emotions that evolved for additional functions, giving rise to phenomena such as betrayal aversion and taboo-tradeoffs, and specific patterns of forgiveness we have documented. A complete account of revenge and reconciliation should incorporate broader constructs from social psychology, including emotions and values hierarchies.

McCullough et al. propose adaptations that motivate forgiveness when the potential benefits of continuing the relationship outweigh the costs incurred by the transgression. The costs incurred are definite, whereas future benefits of forgiveness are only probabilistic. This situation exposes the forgiver to cheating in the form of repeat transgression. Adaptations motivating genuine forgiveness are therefore unlikely to evolve.

We propose that revenge responses are often influenced more by affective reactions than by deliberate decision making as McCullough et al. suggest. We review social psychological evidence suggesting that justice judgments and reactions may be determined more by emotions than by cognitions.

This commentary focuses on how third parties impact the course of acts of revenge based on their world views, such as belief in a just world. Assuming this belief to be true, the following questions could be asked: (a) What are the consequences of a third party's worldview in terms of secondary victimization? (b) Are bystanders actually aware of these consequences? (c) If so, then why do they let it happens?

Research indicates that individuals have incoherent intuitions about particular features of the criminal justice system. This could be seen as an argument against the existence of adapted computational systems for counter-exploitation. Here, I outline how the model developed by McCullough et al. readily predicts the production of conflicting intuitions in the context of modern criminal justice issues.

The psychology underlying revenge in an intergroup context is built around a small handful of recurrent interaction types. Analyzing the cost/benefit calculations of each agent's role within these interaction types provides a more precise way to characterize intergroup conflict and revenge. This in turn allows for more precise models of the psychology of intergroup conflict to be proposed and tested.