3.1 Frank’s commitment model of guilt

Frank (Reference Frank1988) includes guilt among a suite of emotions that, he contends, evolved to enable individuals to make credible commitments with one another, yielding long-term payoffs. In Frank’s view, these payoffs are more likely to be realized if individuals maintain a firm commitment to cooperative arrangements, even when one or more party stands to benefit more by pursuing a strictly self-interested course of action at the expense of other group members. According to Frank, emotions underwrite cooperation in two ways. First, emotions such as love, envy, and guilt incentivize individuals to follow cooperative terms and provide a counterweight to impulses to cheat or defect. For instance, the anticipation of guilt can diminish the allure of cheating for a larger individual payoff. Second, individuals who are recognized as disposed to experience emotions like guilt and sympathy will be sought out by others for cooperative ventures.

Frank suggests two evolutionary pathways by which emotions might have emerged. First, along the “reputation pathway,” individuals who consistently resist the strong urge to cheat acquire a good reputation, the transmission of which leads to further opportunities to benefit through cooperative ventures. Second, along the “sincere-manner pathway,” the experience of emotions is associated with involuntary, hard-to-fake facial expressions, which others can use to draw inferences about whether an individual is a reliable cooperator. In both scenarios, the choice of reliable cooperators increases the selective pressure on particular emotional dispositions.

Because Frank counts guilt among the emotions favored by selection to serve as commitment devices, we should ask whether either of his proposed pathways is a plausible evolutionary scenario for guilt’s emergence. Consider first the sincere-manner pathway. Empirical studies on emotions and their associated facial expressions have shown that, in contrast to other social emotions, there is little evidence that guilt is associated with a stereotypical bodily signal by which it can be readily identified (Keltner and Buswell Reference Keltner and Buswell1996; Wallbott Reference Wallbott1998). Prinz (Reference Prinz2004), for one, takes the lack of evidence for a distinctive physiological signal for guilt as a reason to suppose that guilt did not emerge along the sincere-manner pathway. However, as we will argue in section 4, a modified version of this account yields considerable explanatory power. Much turns on whether the sincere signal needs to be an involuntary physiological change. But for now, let’s consider whether Frank’s reputation pathway might fare better.

Although reputation might have played some role in the evolution of guilt proneness, Frank’s account does not explain why guilt proneness was itself a particular target of selection. Outside parties would be making inferences only about an individual’s adherence to cooperative terms or general disposition to experience some set of social emotions. But whether the anticipation of guilt—and not, say, sympathy or fear—induces commitment in a given instance of cooperation would be opaque to observers. By merely positing that guilt was among a host of prosocial emotions that came under selection pressure at some point because they contributed to beneficial cooperative ventures, the reputation pathway fails to explain why selective pressures would have targeted guilt specifically, leaving guilt’s function undifferentiated from those of other prosocial emotions. On this view of guilt, we are warranted only in concluding that these pressures targeted whichever dispositions happened to be present, guilt or no guilt. This seems correct, as far as it goes, but the reputational pathway account does not provide much by way of explaining why guilt itself evolved. This is because it dilutes the role guilt plays in prosocial behavior, among other social emotions. Reputations would only need to track the resulting cooperation, not the particular disposition or motivation to cooperate. In our critique of Joyce’s and James’s evolutionary accounts that follows, we provide reasons why a predefection role for guilt does not fully explain why it might be an adaptation.

But consider a more serious problem for this account. Frank’s reputational pathway account considers reputation to be an important evolutionary driver for guilt. Now, the cognitive load involved in receiving, retaining, and transmitting information about the shifting reputations of multiple potential cooperators seems to require the sort of complex psychological machinery that would have evolved only in the context of complex human social interaction (Deem and Ramsey Reference Deem and Ramsey2016; Sterelny Reference Sterelny2012). Thus, Frank’s view of the reputation pathway appears to presuppose that which it wishes to explain, namely, communal stability and commitment within complex social environments.

3.2 Joyce’s and James’s self-recrimination models of guilt

Like Frank, Joyce and James conceptualize guilt primarily as a kind of internal check on urges to defect or cheat on cooperative ventures. In contrast to Frank, Joyce and James seek to develop in more detail the specific role guilt plays in cooperation rather than leave its function undifferentiated from that of other social emotions. Joyce characterizes guilt as an “internal self-punishment system” (Reference Joyce2006, 70) that “guides action ‘from the inside’” (101). The action-guiding element of guilt stems from what Joyce sees as its close association with moral judgments about particular types of actions as deserving of punishment. On his view, this package of moral judgment and guilt was selected for because it “reinforced in a motivation-boosting way” (113) other social emotions, increasing the “likelihood that certain adaptive social behaviors [would] be performed” (114). James (Reference James2011) follows Joyce’s self-recrimination model, suggesting that guilt involves “feeling our wrongdoing deserves punishment” (56), functioning as a “check” (75) on temptations to transgress norms. On the self-recrimination model, individuals wish to avoid the painful experience of guilt, making them more compliant with group norms and, consequently, better cooperators.

Joyce’s and James’s accounts face two problems, however. First, they portray guilt as straightforwardly adaptive, underwriting an individual’s prosocial behavior. But the question of whether guilt really is adaptive for the individual is more difficult than either lets on. Guilt has the potential to incur significant burdens for the individual, even as it plays the self-recrimination role with which Joyce and James associate it. Clinical studies of guilt proneness in individuals show a significant correlation between guilt experience and individual psychopathology, including depression, self-loathing, and heightened anxiety (Harder Reference Harder and Fischer1995; Zahn-Waxler and Van Hulle Reference Zahn-Waxler, Van Hulle, Oakley, Knafo, Madhavan and Wilson2012). Any individual-level account of guilt, then, must explain how these ostensibly maladaptive effects on the individual were offset by the benefits accrued by being guilt prone.

Perhaps one might argue that it is enough that guilt disposes individuals to resist temptations to defect on cooperative arrangements, thereby further strengthening cooperative tendencies that are underwritten by other prosocial emotions. This line of argument faces a significant challenge: Why would selection favor a novel, complex emotion with a presumably high maintenance cost just to reinforce these tendencies? Consider that recent neurobiological and primatological research show guilt to be a cognitively complex emotion that might be unique to humans. For example, neurobiological research takes guilt experiences to be produced by complex subcortical and neocortical processes, which are associated with reduced asymmetry in right and left cortical activity and indicative of a unique simultaneous orientation toward withdrawal and approach behaviors (Amodio, Devine, and Harmon-Jones Reference Amodio, Devine and Harmon-Jones2007; Moll et al. Reference Moll, Ricardo de Oliveira-Souza, Grafman and Sinnott-Armstrong2008; Panksepp and Biven Reference Panksepp and Biven2012). This research supports both primatological research and philosophical analyses that posit guilt as a relatively late phylogenetic addition in humans, which emerged only after early human communities had developed sustained cooperative structures and cultural systems of transmission and enforcement of social norms (Boehm Reference Boehm2012; Fessler and Gervais Reference Fessler, Gervais, Kappeler and Joan2010). Moreover, as the contemporary psychological and legal literature we discussed in section 2 show, an important background condition for guilt experience is the individual’s capacity to recognize norms of behavior and evaluate and take responsibility for one’s actions. As several biologists and philosophers have argued, the capacities to accept social norms and make evaluative judgments about one’s actions and those of others likely emerged comparatively late in human evolutionary history (Deem Reference Deem2016; Laland and Brown Reference Laland and Brown2011; Silk and Boyd Reference Silk, Boyd, Kappeler and Joan2010; Sterelny Reference Sterelny2012). As human social arrangements achieved sufficient complexity, phylogenetically older tendencies instilled by, say, kin selection or reciprocal altruism likely became insufficient for generating cooperation on large scales.

If, as this broad set of research suggests, guilt is a neurobiologically complex and cognitively demanding emotion that emerged phylogenetically late within systems of human cooperation and cultural transmission, then it seems unlikely that selection would favor a novel and complex emotion simply to serve as a psychological reinforcement of cooperative tendencies that were underwritten by a suite of other, more phylogenetically ancient, emotions, such as sympathy, empathy, or fear. This is not to mention the potential fitness costs that guilt incurs before and after transgression. For this scenario to be plausible, we need to know the ways in which the fitness landscape changed, rendering these more phylogenetically ancient emotions and tendencies less reliable for underwriting cooperative arrangements and preventing individual defection. Importantly, Joyce and James ascribe to guilt the primary evolutionary function of reinforcing fixed cooperative tendencies, but they do not show how this outweighs the costly behaviors associated with guilt experiences that seem to have the most important social ramifications for the individual, namely, postdefection expressions of guilt. One current role of guilt may well be the regulation of norm transgression, but in order for this to be a plausible candidate for guilt’s evolutionary function, we would need to know how it yielded sufficient biological benefit to offset the potential fitness costs of postdefection guilt-induced behaviors.

This is not to deny that guilt can serve as a motivational counterweight to temptations to defect on cooperative schemes. Indeed, as we noted in section 2, the anticipation of guilt often modifies our motivational profile, dampening the allure of violating norms. But it is to cast doubt on the notion that guilt was favored by selection primarily to serve this purpose or merely to increase the aggregate strength of the set of more phylogenetically ancient prosocial emotions. When we consider guilt’s late evolutionary emergence in humans along with its unique social and behavioral profile after transgression, we see that attributing the biological function of guilt solely to its role as a motivational counterweight leaves significant explanatory gaps in the evolutionary account.Footnote ⁴

This leads to a second and more significant problem for Joyce’s and James’s accounts: neither considers how the action tendencies of guilt in postdefection scenarios would have been adaptive for individuals. Recall from the discussion in section 2 that guilt frequently induces a number of potentially costly behaviors for the individual, including public confession to wrongdoing, submission to punishment, and self-penance. In contemporary social and legal settings, guilt-induced behaviors tend to be met by specific responses from the community that benefit or mitigate harm to the individual who exhibits them. Neither Joyce nor James considers how guilt-induced behaviors render the individual vulnerable to group response to expressions of guilt or why group response carries potential benefit to the individual. Why would members of early human groups respond positively to guilt-prone individuals? Guilt would hardly have been a boon to the individual if the expression of guilt were routinely discounted, ignored, or exploited by the community.Footnote ⁵

Alternatively, and we think more plausibly, guilt proneness might indeed reinforce these tendencies but was likely favored, at least initially, by selection for its role in restoring cooperative arrangements after transgression. If this is right, guilt would still largely fit the descriptive profile that Joyce and James sketch when an individual considers defecting on communal norms or moral commitments, but it would also enhance an individual’s derived benefits from cooperative enterprises, insofar as guilt-induced behaviors play important roles in the restoration of relationships that norm transgressions altered. Joyce’s and James’s accounts do, however, shed significant light on why individuals might have been disposed to reveal otherwise-concealed transgressions or perform actions such as apology and restitution, despite facing punishment from others. In these self-recrimination accounts, individuals often feel they ought to be punished. But even if this self-recrimination prompts such behaviors, we still need an explanation for why revealing one’s transgressions or placing oneself before the mercy and judgment of the group appears to be a stable adaptive strategy. In the following section, the account of the evolutionary origins of guilt we introduce uses an ingredient absent from the accounts of Frank, Joyce, James, and others: empathy. Empathy in humans, we argue, likely preceded the evolution of guilt, and this fact is a key to understanding of how guilt evolved. As we will show, the inclusion of empathy in an evolutionary account of guilt uniquely enables us to make sense of posttransgression responses to expressions of guilt, thereby laying the groundwork for an individual-level adaptive story for guilt’s origin and maintenance.

4.1 Explaining the group response to guilt

If guilt is potentially psychologically and socially maladaptive for the individual, as the empirical literature suggests, how might we explain the evolution of guilt without recourse to a group selection model? We are not assuming that group-level selection scenarios are outright untenable. Indeed, it may be that being composed of guilt-prone individuals provided groups with a competitive advantage. There are ongoing debates about the tenability of group selection models and whether group selection can, in some cases, swamp the effects of individual-level selection. Some group selection models have been strengthened by the addition of culture because culture can have the effect of increasing intragroup homogeneity and intergroup heterogeneity, increasing group-level selection pressures. Such models are increasingly used to account for the evolution of human cooperation (Henrich Reference Henrich2004; Richerson et al. Reference Richerson, Ryan Baldini, Bell, Karl, Hillis and Mathew2016). However, it is not yet clear that cultural group selection can provide adequate explanations of the evolution of guilt. As Nesse (Reference Nesse2016) argues, cultural group selection “has a hard time explaining the pervasiveness and intensity of guilt, motivations for reparations, extreme sensitivity to what others think, concern for others’ welfare, pity, commitment, empathy, philanthropy, and pride in generosity” (35). Furthermore, group-level explanations suffer from the fact that it is unclear whether group-level selection alone was strong enough for guilt to evolve (Deem and Ramsey Reference Deem and Ramsey2016). Such explanations appear better at accounting for the spread of guilt proneness through the species than the origins of guilt proneness in individuals.

Even if one thinks that one can produce tenable group selection models of the evolution of guilt, it is important to ask whether such models are necessary—whether, that is, there are viable evolutionary accounts of guilt that do not require the resources of a group-level selection framework. At any rate, our primary aim in this article is to identify the minimal components required for an individual-level selectionist account of guilt, remaining neutral on whether and to what degree guilt proneness as a trait gained an evolutionary foothold through genetic or cultural (e.g., individual-to-individual, across generations) transmission.

Perhaps the way in which an individual tends to alleviate guilt can provide some guidance here. There is evidence that the psychologically maladaptive effects of guilt are strongly mitigated by opportunities for the guilt-experiencing individual to make amends with those parties who were harmed by a particular transgression (Estrada-Hollenbeck and Heatherton Reference Estrada-Hollenbeck, Heatherton and Bybee1998). But the alleviation of guilt along this route presupposes that the group members are willing to restore the guilt-prone individual to some positive degree of social standing and reincorporate the individual in cooperative enterprises. This is precisely what the aforementioned individual-level accounts overlook: guilt-prone individuals would likely be at a significant disadvantage if guilt-induced behaviors were not met with positive responses from group members. In this scenario, it might indeed pay to be successfully deceptive or demure about one’s own transgressions while falsely signaling one’s acceptance of communal norms. Thus, merely pointing to the way guilt checks motivations to defect, or even noting the reparative behaviors it induces, is not sufficient to show why guilt was adaptive for the individual. Moreover, merely positing that guilt made individuals better facilitators of community benefits is too vague; it does little to differentiate the evolutionary and behavioral profiles of guilt from those of other social emotions and altogether ignores what is most puzzling about guilt’s evolution. The communal attitudes and responses to postdefection displays of guilt must be given an important place within any plausible individual-level explanation of guilt.

The foregoing discussion suggests that constructing an adequate individual-level account of guilt demands an explanation of both why guilt proneness is adaptive for the individual, despite its connection to maladaptive behaviors and psychopathology, and why others tend to forgive and reincorporate, rather than exploit or banish, individuals who perform costly reparative behaviors.

4.2 Behavioral regulation and posttransgression risks

Guilt’s role, then, as a regulator of norm transgression and indicator of such regulation, although important, is a poor candidate for the emotion’s main evolutionary function. If an account focuses only on how guilt affects the motivational and behavioral profiles of individuals, particularly as it mitigates the temptation to defect, then it neglects what is perhaps the more evolutionarily significant dimension of guilt—namely, how it elicits responses from conspecifics that benefit the guilt-prone individual. This is because experiencing and signaling guilt will not be favored by selection if they are met by negative responses from conspecifics that level high costs to the individual. Any plausible evolutionary perspective on guilt, then, must explain why individuals who experienced and displayed guilt altered the motivational and behavioral profiles of their conspecifics, effecting the individual’s posttransgression reintegration into communal life.

As we discussed in section 2, an individual’s expressions of guilt in contemporary social and legal contexts are often met with responses by others that produce some benefit for the individual (e.g., reincorporation into cooperative arrangements) or mitigate costs imposed on the individual due to others’ perception of the individual’s responsibility for a transgression (e.g., leniency in legal sentencing). An evolutionary account of guilt must consider these posttransgression benefits that expressions of guilt provide for the individual. But what accounts for the broadly positive attitudes of conspecifics toward an individual who, motivated by guilt, indicates that she or he is suffering the pangs of guilt or that she or he wishes to repair damage to relationships caused by norm transgression?

To answer to these questions, we must consider what dispositions were already present in individuals prior to the evolutionary emergence of guilt such that guilt-induced behaviors would have been regarded positively, leading to benefits for the guilt-prone individual. The earlier evolution of empathy in humans, we contend, provides a crucial piece of the explanation for how guilt might have been individually adaptive at its emergence.

4.3 Empathic response as a key evolutionary driver

There is considerable variability within the empirical philosophical literature with respect to the precise nature of empathy, and we do not attempt to provide a complete descriptive account here.Footnote ⁶ However, there is some convergence within the contemporary literature on at least three key features of empathy, which we take to be important to explaining how guilt gained an evolutionary foothold. First, empathy has an affective aspect: an empathic state involves an experiencing of the positive or negative valence of another individual’s affective state (Coplan Reference Coplan2011; de Waal Reference de Waal2006; Hatfield, Rapson, and Le Reference Hatfield, Rapson, Le, Decety and Ickes2009). Whether this includes an additional epistemic state of being aware of how that individual feels or clear self–other differentiation is a matter of considerable philosophical debate, into which we do not enter here (Batson and Weeks Reference Daniel and Weeks1996; Coplan Reference Coplan2011; Smith Reference Smith2017). Second, empathy tends to be self-focused: the experience of empathy primarily involves a focus on one’s own experience of this negative or positive valence, as opposed to taking on another’s perspective or imagining oneself as if being in another’s position (Batson, Early, and Salvarani Reference Batson, Early and Salvarani1997; Miller Reference Miller2011; Snow Reference Snow2000). Third, empathy is associated with a behavioral response to one’s negative or positive affective experience. Some researchers claim that empathy typically motivates behaviors aimed at enhancing the welfare of another individual and producing a positive affect in one’s self (de Waal Reference de Waal2008; Eisenberg, Eggum, and Di Giunta Reference Eisenberg, Eggum and Di Giunta2010).

At the very least, it seems that the empathic response is motivated in large part to enhance one’s own positive affect or diminish one’s own negative affect, and this often involves actions directed toward another, insofar as the perception of the latter’s affective experience plays a determining role in one’s empathic experience (Batson, Ahmad, and Stocks Reference Batson, Ahmad, Stocks and Arthur2016). This minimal conception of empathy, which some social psychologists have called “empathic” or “personal” distress (Batson Reference Baston, Decety and Ickes2009), stands in contrast to more robust conceptions of empathic concern that include eliciting behaviors also aimed at relieving the distress of others.Footnote ⁷ For our purposes here, we assume what seems to be a baseline consensus that empathy involves at least the experiencing of the positive or negative valence of another’s emotional state and motivates behaviors that are associated with preserving or alleviating this euphoric or dysphoric experience. These two features of empathy, we contend, help to explain how guilt evolved in early human social contexts and rendered guilt adaptive for individuals.

For our claim to be plausible, we first need evidence that empathy preceded guilt on the evolutionary timescale and that empathy plays a significant role in the social restoration of individuals who transgress norms and subsequently express their guilt experiences. There is empirical evidence that the evolutionary emergence of empathy preceded that of guilt. Animal researchers claim that rudimentary forms of empathy are phylogenetically widespread, being found in a range of taxa (Langford et al. Reference Langford, Crager, Zarrar Shehzad, Smith, Sotocinal and Levenstadt2006; Povinelli, Nelson, and Boysen Reference Povinelli, Nelson and Boysen1992; de Waal Reference de Waal2008). De Waal, for example, claims that nonhuman primates exhibit susceptibility to other troop members’ negative affective states. This phenomenon, frequently dubbed emotional contagion, involves negative affective states in individuals inducing “a matching or closely related state” in others and motivating response behaviors aimed at relieving the distress caused by the shared states (de Waal Reference de Waal2006, 26). In contrast to empathy and its rudimentary forms across taxa, the cognitive machinery underlying guilt experiences is highly complex and perhaps unique to humans, suggesting that guilt might not be phylogenetically widespread (Amodio et al. Reference Amodio, Devine and Harmon-Jones2007; Boehm Reference Boehm2012; Deem and Ramsey Reference Deem and Ramsey2016). This provides evidence that guilt had a later evolutionary emergence than even the more robust form of empathy described by De Waal, and we can plausibly maintain that guilt in humans evolved within a social context in which, minimally, susceptibility to empathic distress was already established.

The claim that empathy provided a pathway for guilt to evolve can be strengthened by considering current psychological research on the experience and effects of empathy. Empirical studies of empathy-related responses of children and adults show that among empathy’s primary functions are to render subjects sensitive to the emotional distress of others, to vicariously participate in this distress, and to prompt behaviors aimed directly at its alleviation in the agent or both parties, which frequently is achieved via the enhancement of the welfare of the party in whom emotional distress was initially detected (Eisenberg et al. Reference Eisenberg, Eggum and Di Giunta2010; Zahn-Waxler et al. Reference Zahn-Waxler, Cole, Darby Welsh and Fox1995). Successful alleviation of this distress is associated with experiences of positive affect, suggesting that an empathic response is associated with a benefit to the empathic party (Batson and Weeks Reference Daniel and Weeks1996). Experiences of empathy are also strongly associated with diminished anger and aggression toward others (Harmon-Jones et al. Reference Harmon-Jones, Kate Vaughn-Scott, Sigelman and Harmon-Jones2004; Jagers et al. Reference Jagers, Morgan-Lopez, Howard, Browne, Flay and Aya Coinvestigators2007; Strayer and Roberts Reference Strayer and Roberts2004), and there is evidence of a proportional relation between empathic capacity and guilt proneness (Treeby et al. Reference Treeby, Prado, Rice and Crowe2016).

From this psychological research on empathy, we can draw two plausible (but nondemonstrative) conclusions about the evolutionary interplay between empathy and guilt. First, as we observed in section 2, guilt experiences are negative affective states. From an evolutionary perspective, early experiences of guilt, if detected by others (more on this to follow), would likely affect others’ emotional states to some degree. Sensitivity to the emotional distress of guilt in others, then, would likely have influenced others’ motivational profiles via empathic distress at the very least, potentially prompting bystanders to behave in ways to alleviate or eliminate their own distress, perhaps along with the distress observed in the other. Second, this empathic experience in view of another’s distressing guilt experience would likely have dampened anger and aggression toward that individual. Although aggression and anger would have been responses to an individual’s perceived norm transgression—and again, we take no stand on whether what is represented is the transgression itself or a relation-dependent property of representing harm—the experience of empathy would potentially reduce urges to severely punish, return harm, or expel from cooperative arrangements.

An effect of empathy is that the perceived suffering of another individual causes one distress and compels one to relieve one’s own negative affect, often via attempts to relieve the other’s pain.Footnote ⁸ Now, this need not be motivated by an express concern for the other—indeed, we can imagine that relieving another’s distress could be taken merely instrumentally as a way to alleviate one’s own empathic distress. Our ancestors, then, would also have been compelled to reduce the suffering—physical and psychological—of others, either directly or indirectly. The emotional suffering of guilt, of course, would be something transgressors would try to avoid and ameliorate. But as long as others in the group were susceptible to either empathic distress or empathic concern, many would be inclined to aid in mitigating and eliminating this suffering. The group could do this in two ways: the anticipated pain of guilt would lead to group members encouraging others to avoid transgressions (e.g., “Think of how you’ll feel”; “How could you live with yourself?”), and detecting guilt experiences in individuals would reduce aggression toward transgressors, encourage group members to forgive transgressors, or relax the implementation of punishment norms on transgressors. Thus, according to our evolutionary perspective on guilt, the key affective and behavioral aspects of empathy noted in the empirical literature would have contributed to the individual-level benefit of the experience and expression of guilt at their evolutionary emergence. Furthermore, an individual’s expression of guilt in posttransgression scenarios could then be an effective adaptive strategy for the alleviation of affective distress and reincorporation in cooperative ventures.Footnote ⁹

4.4 Reliable guilt signaling as protection against exploitation

For guilt-induced behavior to be a fairly reliable indicator of whether individuals are experiencing emotional distress over transgressing norms and whether they are likely to recidivate, individuals’ empathic capacities and the group’s punitive system cannot be open to easy exploitation. After all, there may be large payoffs to individuals who can successfully feign experiencing guilt and gain forgiveness without the intent of changing their behavior—for example, by exhibiting a “hangdog” look without actually feeling any guilt. Psychopaths, to cite a contemporary example, have high recidivation rates, and yet they are the most successful at gaining conditional release when they go up for parole (Porter, ten Brinke, and Wilson Reference Porter, ten Brinke and Wilson2009). Intuitively, one might find it plausible that individuals who were particularly adept at concealing their transgressions or feigning guilt behaviors could reap the benefits of group forgiveness without suffering from feelings of self-recrimination or other maladaptive effects. Wouldn’t such individuals be better off than individuals who experience guilt and display guilt-induced behaviors? If all things were equal, this might indeed be persuasive. But all things are not equal. If guilt has a relatively late evolutionary emergence in hominins, as we have supposed, then it seems that a relatively stable suite of prosocial emotions—including empathy—would already have been established, driven in part by pressures favoring traits that secured and preserved cooperation. Presumably, this mitigated to some degree the threat of deception within cooperative arrangements for individual gain.

Moreover, as we and others have argued, for guilt to be reliably signaled, significant costs would attach to such signaling. It is reasonable to suppose that the motivation to take on such costs is more reliably produced by the feeling of self-recrimination (à la Joyce and James) than by coolly calculated deception. Further, as Sterelny has argued in response to concerns that deceivers would overrun cooperative systems that rely to some degree on group signaling, even if deception prevailed in at least some individual cases, its threat would have been relatively low within complex social systems in which information flow runs multidirectionally within groups and informational pooling precedes the use of information for planning and acting (Sterelny Reference Sterelny2012). A deceiver likely succeeds, in other words, when no one else checks on the lie.

But there are two ways by which fake signals of guilt may be rendered less effective within social contexts. First, the capacities for memory and the communication of fine-grained information about individuals’ reputation for cooperation render such exploitation much more difficult. Although someone may be able to get away with this deception within the relative anonymity of a large prison system, this would not be so easily accomplished in a smaller community of early humans with normalized social relations. If individuals were able to remember and communicate detailed information about the actions of others with whom they have interacted, they would be able to better distinguish earnest expressions of guilt from fraudulent ones, as well as move beyond firsthand experience in judging the relative degree of earnestness in attempts to seek forgiveness. However, there would be a significant cost in terms of time and biological maintenance of these memorial and communicative capacities, and it is reasonable to assume that competing adaptive trade-offs would result in a nonoptimal leveling of their power.

However, a second and perhaps more effective way to prevent both the exploitation by fakers of guilt proneness and the dissemination of inaccurate reputations about the guilt proneness of individuals would be through reliable signaling of guilt. Recall Frank’s sincere-manner pathway, along which hard-to-fake facial expressions serve as signals by which the group can accurately detect the presence of important social emotions in individuals. Guilt, we have seen, has no such telling facial or bodily expression, so Frank’s account does not straightforwardly explain the reliable signaling of guilt. But there is no reason to suppose that the primary source of evidence about guilt must come from bodily posture or facial expression. The risk and potentially enormous costs individuals incur by confessing their transgressions, submitting to potential punishment, and performing reparative actions can serve as credible signals of the guilt experience. Mimicry of guilt-prone individuals, then, would not only come at a high price to the individual who attempts to exploit the forgiveness of others and seek the dampening of their punishment after transgression but would also involve having to perform these costly actions presumably without the guilt experience that typically motivates and sustains them. Although this is no guarantee that guilt could never be successfully feigned, the performance of these potentially costly and maladaptive behaviors would have enabled members of groups to infer the high probability that an individual was actually experiencing guilt due to a transgression (Deem and Ramsey Reference Deem and Ramsey2016).

Other evolutionary accounts of guilt have also converged on the hypothesis that a costly signal might be required for expressions of guilt to yield individual-level benefits. Martinez-Vaquero et al. (Reference Martinez-Vaquero, Anh Han, Moniz Pereira and Lenaerts2015), O’Connor (Reference O’Connor2016), and Pereiro et al. (Reference Pereira, Lenaerts and Martinez-Vaquero2017a) provide a different avenue to this conclusion through the use of evolutionary game theory modeling. Their models of guilt show that apology after a broken commitment and the subsequent restoration of cooperative arrangements between transgressor and transgressed can yield fitness benefits to each under conditions of revenge, apology, and forgiveness. However, in order for the apology to function as a reliable signal of a willingness to recommit to cooperative arrangements, Martinez-Vaquero et al. (Reference Martinez-Vaquero, Anh Han, Moniz Pereira and Lenaerts2015) conclude that the apology must cross a “sincerity threshold … where the cost of apologising should exceed that of cooperation” (8), where the cost of cooperation is the risk of defection. In mixed-state games, where a population consists of both apology proposers and acceptors, Martinez-Vaquero et al. (Reference Martinez-Vaquero, Ahh Han, Moniz Pereira and Lenaerts2017) conclude that if the apology cost is too low, then apology defectors take over the population in repeated interactions. Similarly, O’Connor (Reference O’Connor2016) and Pereira et al. (Reference Pereira, Lenaerts and Martinez-Vaquero2017b) conclude that a significant cost to apologize is needed to lower the probability that fakers can exploit cooperative arrangements by feigning guilt and that dishonest apologizers will evolve. In addition to concluding that the costs of guilt expression to the individual must cross a high enough threshold in order to be reliable signals of cooperative intent, the Martinez-Varquero et al., O’Connor, and Pereira et al. models suggest that these costs must not be so high that they cannot be absorbed by the individual. In their discussion of their respective models, Martinez-Varquero et al. (Reference Martinez-Vaquero, Anh Han, Moniz Pereira and Lenaerts2015) and O’Connor (Reference O’Connor2016) conclude that the costs to the individual must be capped in some way. Otherwise, guilt expressions cannot be an evolutionarily stable strategy. Perreira et al. (Reference Pereira, Lenaerts and Martinez-Vaquero2017b) conclude that if this cost is too high, then revenge will dominate apology as a cooperative strategy.

None of these models, however, specifies what would create this cap on the costs to guilt-prone individuals. Extensions of these evolutionary game-theoretic models in Rosenstock and O’Connor (Reference Rosenstock and O’Connor2018) and Pereira et al. (Reference Pereira, Lenaerts and Martinez-Vaquero2017a) provide some basis for explaining this posited cap on costs. In Rosenstock and O’Connor’s model, guilt-prone individuals are willing to pay a cost in order to apologize for defection and show a willingness to cooperate in future interactions. Using a model in which one player’s guilt expression occurs without the detection of guilt proneness in other players, Pereira et al. (Reference Pereira, Lenaerts and Martinez-Vaquero2017b) conclude that guilt-prone individuals will be exploited by non–guilt-prone conspecifics. Under such conditions, guilt proneness and its expression appear to carry a very high cost. In a second model stipulating that an individual will experience postdefection guilt when a co-player acts prosocially toward the individual or has also displayed guilt, Pereira et al. (Reference Pereira, Lenaerts and Martinez-Vaquero2017b) conclude that guilt proneness in a population will enhance cooperation and come to dominate.

Both the Rosenstock and O’Connor (Reference Rosenstock and O’Connor2018) and Pereira et al. (Reference Pereira, Lenaerts and Martinez-Vaquero2017b) models suggest that the cost of guilt expression is curbed by prosocial behavior toward apology on the part of conspecifics, but neither model provides any specification of what psychological trait might undergird that response. Empathic concern for the guilt-prone individual’s distress, we contend, is a good candidate for serving as this effective limit to the costs of a guilty apology because it explains both why players accept the apology and why guilt-prone individuals do not exact the kinds of high costs we outlined previously, such as exploitation, severe punishment, or social exile. Reliable signaling of guilt through costly apology induces empathic concern in the transgressed, motivating forgiveness and the recommencement of cooperation. Our account of the relation between guilt and empathy, then, has the virtue not only of being consistent with these game-theoretic models of the evolution of guilt but also of providing additional support to them by supplying a key condition for explaining how guilt-induced behavior could yield benefits to the individual.

Let’s take stock of the foregoing empirical and conceptual considerations. Our ancestors likely were empathic before they were guilt prone. Although the individual’s anticipation of guilt could decrease the motivation to violate normative standards, it was the empathic context in which guilt emerged that was the decisive factor in the evolution of guilt. Empathic concern for the emotional distress of guilt likely reduced aggression toward guilt-prone norm breakers and enabled others to participate vicariously in their emotional distress. Empathy, then, would have prompted behaviors aimed at mitigating guilt distress, including forgiving guilt-laden individuals and reincorporating them into cooperative arrangements. The high cost of expressing guilt, and the capacity to retain and communicate reputations of guilt proneness, would make guilt difficult to fake, and the benefits of genuine experiences of the emotion would amplify. The subsequent benefits of being forgiven and reincorporated into cooperative arrangements, then, would have made expressing guilt a stable adaptive strategy for the individual. Empathy, curiously absent from current evolutionary explanations of guilt, thus is likely a central component in the explanation of the evolution of guilt proneness.