12.4.1 A Risk Perspective

Given that humans are not privy to the intentions, motivations, or beliefs of another, almost every social decision we take is marked by uncertainty (FeldmanHall & Shenhav, Reference FeldmanHall and Shenhav2019). For example, a trustworthy or cooperative individual can easily be exploited by another, less trustworthy individual. It is therefore necessary to constantly estimate the possible risks associated with making a prosocial action. Given this natural mapping between prosociality and uncertainty, there has been some work that explores individual differences in risk attitudes and their effect on altruism, trust, and cooperation (Eckel & Wilson, Reference Eckel and Wilson2004; Fehr, Reference Fehr2009). Early research revealed mixed evidence, with some work demonstrating a positive relationship between prosocial actions and risk attitudes and other work showing no effect whatsoever (Fairley et al., Reference Fairley, Sanfey, Vyrastekova and Weitzel2016). One possible reason for these mixed results was that individual risk attitudes were measured in gambling contexts in which people had access to the known probability associated with a given outcome (e.g., knowing that the probability of winning a lottery is 80 percent). In the social world, we do not have access to the known (or exact) probability of whether another individual will exploit us. The “risk” we perceive in the social domain is more akin to ambiguity – a type of uncertainty in which the probability associated with each outcome is unknown. Once ambiguity, rather than risk, is measured at the individual level, there appears to be a robust relationship between those who can tolerate ambiguity and the tendency to behave prosocially (Vives & FeldmanHall, Reference Vives and FeldmanHall2018). This relationship vanishes once the ambiguity associated with other individuals is resolved, for example, by gaining access to a history of their behavior or by the existence of strong social norms.

12.4.2 A Matter of Reputation

Given that evolutionary forces reinforce the selection of selfishness, reputation concerns have been leveraged in an attempt to reconcile this finding with the existence of prosocial behaviors (Boyd & Richerson, Reference Boyd and Richerson1989; Nowak & Sigmund, Reference Nowak and Sigmund2005). According to the reputation account, prosocial behaviors occur to bolster one’s reputation and maximize payoffs – assuming repetitive interactions among the same people. The price an individual pays for being prosocial is offset by gaining a positive reputation, which likely increases the probability of being treated well by others in the future. In line with this notion, people do behave more prosocially when reputational concerns are at stake (i.e., when they think others are watching; Bradley et al., Reference Bradley, Lawrence and Ferguson2018). By deciding quickly, people can also signal to others that prosocial actions are their default behavior (Jordan et al., Reference Jordan, Hoffman, Nowak and Rand2016). This type of prosocial signaling is useful for encouraging repeated social interactions. People use speedy prosocial responses as a signal to figure out who they should interact with during economic games, preferring those who are “mindlessly” prosocial (Jordan et al., Reference Jordan, Hoffman, Nowak and Rand2016). Critics have argued, however, that reputation as a motivational concern cannot explain a number of findings, including the high rates of prosociality observed when reputation is not at stake (i.e., when nobody is watching or decisions are anonymous). To combat this critique, it has been argued that people internalize the value of prosociality under standard reputation concerns, and, as a consequence, prosocial behavior is automatically endorsed even when nobody is watching (Jordan & Rand, Reference Jordan and Rand2020).

12.4.3 Other-Regarding Motives

Perhaps the most accepted explanation regarding motives for prosocial behavior is the existence of what are considered pure, other-regarding concerns (Cooper & Kagel, Reference Cooper, Kagel, Kagel and Roth2016). The more an individual is concerned about the well-being of others, the more likely they are to act prosocially (Eisenberg et al., Reference Eisenberg, Fabes, Miller, Fultz, Shell, Mathy and Reno1989). This dovetails with common wisdom and everyday observations: People are more prosocial with those they care about (colleagues, friends, and family). It has been theorized that our capacity for prosociality first originated to take care of kin, and, over time, this motive has generalized and extended to other people outside our kinship circles (Christakis, Reference Christakis2019). Research further reveals that the degree to which people value others (as measured by the Social Value Orientation scale; see Murphy & Ackermann, Reference Murphy and Ackermann2014) predicts their prosocial behavior in the laboratory (Balliet et al., Reference Balliet, Parks and Joireman2009). In contrast, stable traits that reflect egoistic motives, such as narcissism, psychopathy, and Machiavellism, are inversely related to prosocial behavioral patterns (Thielmann et al., Reference Thielmann, Spadaro and Balliet2020).

12.5.1 Affective States

Broad affective states, such as positive affect, or transient moods that are not explicitly tied to the causes and consequences of a decision itself, can readily modify how individuals respond when faced with a prosocial dilemma. One early example of this, documented by Isen and Levin (Isen & Levin, Reference Isen and Levin1972), revealed that a positive, happy mood can lead people to help others (Curry et al., Reference Curry, Rowland, Van Lissa, Zlotowitz, McAlaney and Whitehouse2018). To induce a positive mood, half of the subjects discovered a dime in the return slot of a public telephone. Subjects who found the dime then encountered a stranger who dropped a heap of papers in front of them. Subtly inducing a good mood (by way of a found dime – it was the 1970s) led subjects to help the stranger pick up papers, more so than those who were not put in a good mood. In essence, the “glow of goodwill” promoted prosocial behavior. Follow-up studies reported a similar finding, whereby inducing positive affect through emotional comedy videos led people to make more prosocial decisions (DeSteno et al., Reference DeSteno, Bartlett, Baumann, Williams and Dickens2010; Valdesolo & Desteno, Reference Valdesolo and Desteno2006), and reducing negative emotions also increased prosocial behavior (Kemeny et al., Reference Kemeny, Foltz, Cavanagh, Cullen, Giese-Davis, Jennings, Rosenberg, Gillath, Shaver, Wallace and Ekman2012). Additionally, recent research from our lab found that emotion prediction errors play a large role in governing whether an individual will punish a norm violator, revealing that violations of emotional expectations play an outsized role in governing our willingness to act prosocially (Heffner et al., Reference Heffner, Son and FeldmanHall2021). The idea that affective states govern our perceptions, including our social interactions, is gaining traction in the field, with more formal models dissecting how exactly affective states shape how we navigate and interact with our world (Eldar et al., Reference Eldar, Rutledge, Dolan and Niv2016; Quoidbach et al., Reference Quoidbach, Taquet, Desseilles, de Montjoye and Gross2019).

12.5.2 Distress

Other affective states, such as distress, can be tied to the decision to display prosocial behaviors. Watching another experience physical or emotional pain, such as social rejection, is highly aversive and is known to cause distress in the observer (Decety et al., Reference Decety, Michalska and Akitsuki2008; Eisenberger & Lieberman, Reference Eisenberger and Lieberman2004; Singer et al., Reference Singer, Seymour, O’Doherty, Kaube, Dolan and Frith2004), which can in turn amplify prosocial actions. For example, in one study, subjects’ skin conductance responses (SCRs) were measured while receiving painful shocks, or when observing pain being inflicted on another (Hein et al., Reference Hein, Lamm, Brodbeck and Singer2011). Subjects could prevent pain being administered to another by choosing to endure the painful shocks themselves. Critically, the magnitude of the SCR – the body’s physiological response to stressful or arousing events, which was taken here as a proxy for distress – predicted whether a subject would altruistically take the painful shocks in lieu of another.

While the link between distress and altruistic behavior has been replicated in other contexts as well (FeldmanHall et al., Reference FeldmanHall, Dalgleish, Evans and Mobbs2015), there is early theoretical and empirical work that suggests that feelings of distress are also associated with the opposite behavioral response: avoiding helping altogether (Shaw et al., Reference Shaw, Batson and Todd1994). In this case, terminating one’s own distress is more pressing than helping another in need (Batson et al., Reference Batson, Fultz and Schoenrade1987; Eisenberg, Reference Eisenberg2000; Eisenberg et al., Reference Eisenberg, Fabes, Murphy, Karbon, Maszk, Smith, O’Boyle and Suh1994). An example is when a person crosses to the other side of the street after seeing a homeless person on the sidewalk, since avoiding the moral situation attenuates the discomfort that comes with seeing another in distress. This led some researchers to posit that there are two ways to experience distress, either in sharing the pain of another (Cialdini et al., Reference Cialdini, Brown, Lewis, Luce and Neuberg1997; Singer et al., Reference Singer, Seymour, O’Doherty, Kaube, Dolan and Frith2004) or in understanding that another is in distress but without necessarily sharing in the pain itself (Batson, Reference Batson2011; Masten et al., Reference Masten, Morelli and Eisenberger2011). Evidence that individuals who show little arousal to the distress of others also exhibit little physiological arousal when distressed themselves (Shirtcliff et al., Reference Shirtcliff, Vitacco, Graf, Gostisha, Merz and Zahn-Waxler2009), hints at the possibility that the capacity to respond prosocially may be the result of failing to understand what the other is feeling, rather than actually sharing in their distress, per se. However, taking a bird’s-eye view of the data, especially with a nod to more recent neuroimaging work (Zaki & Ochsner, Reference Zaki and Ochsner2012), reveals mixed evidence for this claim. Context, for example, appears to have a strong mediating effect on whether prosociality is driven by understanding or sharing in the pain of others (see Zaki & Ochsner, Reference Zaki and Ochsner2012, for a more in-depth discussion).

Populations who exhibit a reduced ability to understand the distress of another help to clarify this relationship between distress and prosocial choice (K. Blair et al., Reference Blair, Marsh, Morton, Vythilingam, Jones, Mondillo, Pine, Drevets and Blair2006). Alexithymia, a personality construct characterized by the inability to identify and describe emotions in oneself or others (Sifneos, Reference Sifneos1973), showcases the intimate relationship between understanding another’s distress and altruistic action (Bernhardt & Singer, Reference Bernhardt and Singer2012; Taylor & Bagby, Reference Taylor, Bagby, Bar-On and Parker2000). Our lab tested the link between the ability to identify and understand another’s distress and helping behavior. In our Pain versus Gain paradigm explained above, we found that those high on the alexithymia spectrum reported experiencing less distress when watching others in pain, which translated into keeping more money at the expense of the Receiver’s physical well-being (FeldmanHall et al., Reference FeldmanHall, Dalgleish and Mobbs2013). This effect was mirrored at the neural level, with diminished activity in key brain regions that encode for the experience of distress.

12.5.3 Empathy

Empathy is one avenue in which people can understand the emotional experiences of another (Decety, Reference Decety2011; and see Chapter 11 in this volume). The term empathy is applied to a large spectrum of phenomena (Zaki & Ochsner, Reference Zaki and Ochsner2012) and includes a range of emotional dimensions, such as feelings of concern, affect sharing, and cognitive perspective taking. In short, empathy is a multicomponent process that includes both affective and cognitive dimensions (Zaki, Reference Zaki2014). Early foundational research suggested that feeling empathic concern for another in need is the lynchpin of motivated helping (Batson et al., Reference Batson, Fultz and Schoenrade1987). Batson and colleagues tested this empathy-altruism hypothesis (Batson, Reference Batson2011) across a number of studies. They found that increasing the perceived similarity between participants and the person in need of help increases altruistic behavior (Toi & Batson, Reference Toi and Batson1982). Moreover, inducing a person to feel greater empathy in general increases cooperation in the prisoner’s dilemma, regardless of whether the economic game is framed as a social exchange or a business transaction (Batson & Moran, Reference Batson and Moran1999). Empathy might even explain a famous “in the wild” experiment, in which Darley and Batson demonstrated that those who were walking across a university green with ample (rather than no) time on their hands before their next meeting were more likely to offer help to another in need along the way (Darley & Batson, Reference Darley and Batson1973) – perhaps because the helper had enough time to experience or conjure up empathic feelings.

More recent research has focused on investigating empathy in the context of witnessing another in physical pain. The prevailing narrative is that empathy is a form of mirroring another’s distress with our own distress. When we experience physical pain, there are certain brain regions, including the anterior insula, that are routinely activated (i.e., the brain’s prototypical pain circuit). This neural network is also activated when seeing a loved one – or even a stranger – in physical pain (Morelli et al., Reference Morelli, Rameson and Lieberman2014; Singer et al., Reference Singer, Seymour, O’Doherty, Kaube, Dolan and Frith2004). Evidence of shared neural activity suggests that perceiving an emotion in another can activate the same emotion in the observer. This perception–action matching mechanism cannot, however, differentiate between whether the observer is aversively aroused and therefore sharing the distressing experience, or whether the observer is simply recognizing and understanding the distress of the other.

Work from our own lab suggests that understanding the distress of another, rather than sharing in their distress, is the mechanism that fuels the empathy–altruism relationship. Using the same Pain vs Gain paradigm, we found that a Decider’s trait empathy, their ability to express empathic concern toward another, correlated with the amount of money they were willing to forgo to attenuate the number of shocks administered to the Receiver (FeldmanHall et al., Reference FeldmanHall, Dalgleish, Evans and Mobbs2015). This relationship between trait empathy and costly altruism was not reflected by activity in the classic pain circuit (i.e., anterior insula and anterior cingulate). Instead, we found the empathy–altruism relationship to be indexed by activity in the caudate, ventral tegmental area, and subgenual anterior cingulate – regions critical for processing reward and social attachment.

12.5.4 Guilt

As with empathy, guilt can also motivate people to behave in prosocial ways, especially when trying to repair relationships (Donohue & Tully, Reference Donohue and Tully2019; Drummond et al., Reference Drummond, Hammond, Satlof-Bedrick, Waugh and Brownell2017; Vaish et al., Reference Vaish, Carpenter and Tomasello2016). Guilt is a painful feeling aroused by causing (or even anticipating causing) an aversive event (Ferguson & Stegge, Reference Ferguson, Stegge and Bybee1998; Vaish, Reference Vaish2018). Guilt proneness consistently correlates with measures of perspective taking and empathic concern and is inversely related to antisocial behavior (Tangney et al., Reference Tangney, Stuewig and Mashek2007). Neuroimaging studies reveal that when describing moral transgressions, feelings of guilt are associated with neural activation in a network that is also engaged when thinking about another’s feelings (Basile et al., Reference Basile, Mancini, Macaluso, Caltagirone, Frackowiak and Bozzali2011; Takahashi et al., Reference Takahashi, Yahata, Koeda, Matsuda, Asai and Okubo2004). This is taken to indicate that a key function of guilt is to promote perspective taking and increased social cohesion. In fact, someone who expresses feelings of guilt or shame is more likely to be perceived positively than someone who does not express such feelings (Stearns & Parrott, Reference Stearns and Parrott2012). Guilt has also been shown to promote social cooperation. For example, norm infringement is associated with greater experiences of guilt, especially when confronted by an angry face perceived to be criticizing the norm violation (Giner-Sorolla & Espinosa, Reference Giner-Sorolla and Espinosa2011). Because guilt is associated with breaches of moral rule and social standard, the existence of guilt (or even the anticipation of guilt) can foster trust and social reciprocity (Chang et al., Reference Chang, Smith, Dufwenberg and Sanfey2011; Vaish, Reference Vaish2018). Together, this research illustrates how emotions can powerfully influence prosocial behaviors. Simply put, emotions afford a useful and effective control system to stymie antisocial behavior and promote prosocial behaviors.

13.2.1 Evidence from Kohlberg’s Test of Moral Development

Initial research in this area focused exclusively on assessing the psychopathic individual’s developmental level of moral reasoning, as per Kohlberg’s (Reference Kohlberg1963) theory of moral development. This theory outlines three overarching moral stages: 1) preconventional, 2) conventional, and 3) postconventional, with each stage subdivided further into two substages. In each overarching stage, Kohlberg proposed that people invoke various considerations when making moral decisions. Individuals at the preconventional stage decide on moral issues based on the immediate consequences for themselves, while those at the postconventional stage make decisions based on relatively abstract moral principles independent of existing social rules, laws, and authority. Those in the intermediate, conventional stage, make moral decisions based on the expectations of social groups and society. One well-validated and widely used questionnaire measure within Kohlberg’s moral development framework is the defining issues test (DIT; Rest et al., Reference Rest, Copper, Coder, Masanz and Anderson1974), which presents five moral dilemmas. One well-known example, the Heinz drug dilemma, asks participants whether a man should steal a drug to save his dying wife’s life when there are no other options, although doing so is against the law. The DIT yields an overall moral development score, with higher scores representing a higher level of postconventional reasoning.

Researchers have used Kohlbergian measures of moral reasoning to understand which elements of a psychopathic individual’s moral cognition may be compromised. Given that psychopathy is marked by self-centeredness, some authors (e.g., Campbell et al., Reference Campbell, Doucette and French2009) have hypothesized that psychopathic individuals tend to prioritize self-interest over abstract moral principles when making moral decisions and, therefore, do not progress beyond the lower levels of moral development (Campbell et al., Reference Campbell, Doucette and French2009; O’Kane et al., Reference O’Kane, Fawcett and Blackburn1996). Surprisingly, the research evaluating this argument is dated and has yielded inconsistent results. Some studies have supported the hypothesis of lower levels of moral reasoning in psychopathic individuals (Fodor, Reference Fodor1973; Jurkovic & Prentice, Reference Jurkovic and Prentice1977), some argue for higher levels of moral reasoning (Link et al., Reference Link, Scherer and Byrne1977), while others reveal no significant group differences on moral development (Lose, Reference Lose1977; O’Kane et al., Reference O’Kane, Fawcett and Blackburn1996). Supporting the hypothesis that psychopathic individuals reason differently from others about moral issues, Campbell and colleagues (Reference Campbell, Doucette and French2009) found that such individuals tend to both prioritize lower moral considerations (e.g., potential threats to personal interest) and de-emphasize higher moral reasons (e.g., attempts to protect human rights). Adding to the confusion, Link and colleagues (Reference Link, Scherer and Byrne1977) found that psychopathic individuals received significantly higher moral reasoning scores than did nonpsychopathic individuals on the Kohlberg (Reference Kohlberg1958) test. However, later researchers raised the criticism that Kohlberg’s test relies heavily on good analytical skills and completely neglects the evidence for the powerful emotional and nonverbal determinants of morality (Garrigan et al., Reference Garrigan, Adlam and Langdon2018). On balance, these findings provide only a cloudy and inconclusive picture of the moral-developmental level of the psychopath.

13.2.2 Evidence from the Moral-Conventional Distinction

Another form of the moral reasoning hypothesis relies on the findings of Blair et al. (Reference Blair, Jones, Clark and Smith1995, Reference Blair, Jones, Clark and Smith1997) based on the moral/conventional task (MCT). In these studies, investigators assessed the ability of adults and children who were low or high in psychopathic tendencies to correctly classify eight hypothetical actions based on their moral content. Using the MCT, Blair et al. (Reference Blair1995, Reference Blair1997) asked participants to judge whether 1) the hypothetical action was acceptable, and 2) whether it would still be acceptable even if an authority figure permitted it under free response conditions. Conventional transgressions presuppose that the perceived wrongfulness is reliant on what the authority says (e.g., wearing pajamas at work). In contrast, moral transgressions are defined as acts that society considers wrong even if there were no rules, customs, or laws against them (e.g., hurting others). Participants low in psychopathy correctly distinguished moral scenarios that were authority-independent from conventional scenarios. However, participants with high psychopathy made no such distinction. It is notable that psychopathic participants who failed to make moral-conventional distinctions in Blair’s (Reference Blair1995) studies did not rate both types of scenario as acceptable. Rather, they tended to rate both types as unacceptable, regardless of the conventions laid down by authority, suggesting that these participants believed all the scenarios to involve moral violations. Blair (Reference Blair1995) concluded that individuals with psychopathy exhibited this counterintuitive effect as the result of attempting to appear socially desirable or “faking good” (see Blair, Reference Blair1995, p. 23; Blair et al., Reference Blair, Jones, Clark and Smith1995, p. 749) and that in reality they may have difficulty detecting or rating merely conventional transgressions.

Recent studies on psychopathic individuals and the moral-conventional distinction have not reproduced Blair’s original results. Aharoni et al. (Reference Aharoni, Sinnott-Armstrong and Kiehl2012, Reference Aharoni, Sinnott-Armstrong and Kiehl2014) tested psychopathic individuals’ ability to distinguish between moral and conventional transgressions in a forced-choice setting, telling participants that exactly half the scenarios were considered morally wrong by typical members of society. Respondents were expected to make a forced choice between predefined moral and conventional transgressions, meaning overclassification would not be an effective strategy and allowing a purer assessment of moral reasoning. Under this methodology, if psychopathic participants truly lack moral understanding they should show less accuracy in making moral-conventional classifications than nonpsychopathic controls. Conversely, if psychopathic individuals are no less accurate in these forced-choice classifications, we could conclude that they do indeed understand moral wrongfulness. In both studies, no association was established between psychopathy scores and the percentage of acts correctly classified as moral or conventional. In sum, when factors relating to social desirability are removed, psychopathic participants in moral-conventional distinction studies perform as well as controls, providing additional evidence that individuals with psychopathy do indeed understand wrongfulness (Aharoni et al., Reference Aharoni, Sinnott-Armstrong and Kiehl2012, Reference Aharoni, Sinnott-Armstrong and Kiehl2014).

13.2.3 Evidence from Sacrificial Moral Dilemmas

Another commonly used method for measuring moral decision making draws upon the seminal work of Greene and colleagues (Reference Greene, Sommerville, Nystrom, Darley and Cohen2001). They used hypothetical moral dilemmas in which participants must decide whether to sacrifice the life of one person in order to save the lives of a greater number (for a broader discussion of moral dilemmas, see Chapter 5, this volume). In a version of the classic trolley problem (Thomson, Reference Thomson1976), the switch dilemma, participants must choose whether or not to pull a lever that would divert a train and save five others who are lying on the track ahead but also kill one person at the end of the diversion track. In another sacrificial dilemma, the footbridge dilemma, participants must choose whether or not to push a heavyset man off a bridge in order to stop the train, which would kill him but thereby save five others on the track ahead. Although the participants in both cases have to decide whether to cause the death of one person to save several, endorsement of the action in either situation indicates a willingness to engage in utilitarian decision making (i.e., saving five people instead of one), while avoiding action in either case indicates a deontological decision (i.e., considering it immoral to inflict harm). However, while avoiding action in the footbridge dilemma clearly aligns with deontological principles (i.e., considering it immoral to directly inflict harm), the deontological decision in the switch case is more complex. Pulling the lever can be seen as consistent with the maxim of universalizability and a duty to save lives, making the deontological judgment in this scenario less clear-cut.

Interestingly, most individuals deem the “pulling the lever” – the switch dilemma – to be more morally acceptable than pushing a person onto the tracks, despite both choices having identical outcomes in terms of numbers of lives lost (Cushman et al., Reference Cushman, Young and Hauser2006). One explanation for this discordance in moral decision making across these dilemmas postulates that most people think pulling the lever does not cause direct harm to someone else whereas pushing the man off the footbridge causes direct physical harm to another individual (Cushman et al., Reference Cushman, Gray, Gaffey and Mendes2012). This distinction is crucial because causing direct harm typically elicits stronger emotional responses, such as empathy and guilt. People are more likely to experience and be influenced by these emotions when contemplating direct harm. Psychopathic individuals, who have deep-rooted deficits in empathy and guilt (Berg et al., Reference Berg, Lilienfeld and Waldman2013), may not be as affected by the emotional weight of causing direct harm. As a result, they might adopt more utilitarian reasoning, focusing on the outcomes (i.e., saving more lives) rather than the means of achieving those outcomes. The reduced influence of emotions in individuals with psychopathy may enhance the more “rational” aspects of their moral decision making over those of nonpsychopathic individuals. The reduced influence of emotions in individuals with psychopathy may serve to enhance the more “rational” aspects of moral decision making over those of nonpsychopathic individuals (Pletti et al., Reference Pletti, Lotto, Buodo and Sarlo2017).

Supporting this prediction, Bartels and Pizarro (Reference Bartels and Pizarro2011) found that self-reported psychopathy was positively associated with utilitarian decision making about sacrificial dilemmas among undergraduates. Koenigs et al. (Reference Koenigs, Kruepke, Zeier and Newman2012) partially replicated this correlation in a sample of prisoners, albeit specifically in the case of the classical switch dilemma where only indirect harm is inflicted. Tassy et al. (Reference Tassy, Deruelle, Mancini, Leistedt and Wicker2013) replicated the study with French university students and showed that a high level of psychopathic traits predicted a greater likelihood of a utilitarian response to hypothetical choices (i.e., “Would you … in order to …?”) but not for the judgment question (i.e., “Is it acceptable to … in order to …?”). This finding confirmed the link between psychopathic traits and utilitarianism in choice of action but not in moral judgment.

13.2.4 Moral Foundations Theory

A final way in which researchers have examined the moral abnormality of psychopathy emerges from work on moral foundations. Six moral domains are argued to represent the various ethical considerations of decision making (Fernandes et al., Reference Fernandes, Aharoni, Harenski, Caldwell and Kiehl2020; Haidt, Reference Haidt2012): 1) preventing harm, which involves concern about the suffering of others; 2) preserving fairness, which involves following the norms of reciprocity, equality, and justice; 3) respecting authority, which involves moral obligations to hierarchies; 4) exhibiting in-group loyalty, which involves moral obligations to members of an identified group without betrayal and giving preferential treatment to in-group over out-group members; 5) practicing purity, which involves living in a noble manner with purity of body, mind, and soul; and 6) protecting liberty, which involves the feelings of reactance and resentment people feel toward those who dominate others and restrict their freedom.

In the context of moral judgments, moral foundations research is focused primarily on preferences in moral values as opposed to identifying deficits in moral judgment patterns. A study conducted by Blair (Reference Blair2007) showed that while individuals scoring high in psychopathy demonstrated a lack of concern about harming others, they had similar levels of concern within other domains, such as purity and loyalty in the MFQ (Graham et al., Reference Graham, Nosek, Haidt, Iyer, Koleva and Ditto2011). This finding was replicated in a sample of 222 incarcerated males, where it was found that psychopathic traits were negatively associated with concerns about harm and fairness (Aharoni et al., Reference Aharoni, Antonenko and Kiehl2011). Others, in contrast, have found that psychopathy is related to diminished moral concern in all five moral domains (Jonason et al., Reference Jonason, Strosser, Kroll, Duineveld and Baruffi2015). Using a large online sample, Glenn et al. (Reference Glenn, Iyer, Graham, Koleva and Haidt2009) found that individuals who scored higher on the psychopathic traits reported less concern about preventing harm and being fair than those who scored lower on psychopathy. The psychopathic individuals also showed slightly increased concerns about in-group loyalty, which may be due to their reduced concern about nongroup members and their desire for their in-group to socially dominate other groups. Although the body of work relating psychopathy to moral foundations theory (MFT) is relatively small, overall these studies suggested that psychopathy is associated with moral abnormality.

13.2.5 Meta-analysis on Moral Judgment and Psychopathy

Taking different perspectives and using various measures of moral judgment, the previous sections have tried to address the question of whether individuals with psychopathy can distinguish cognitively between right and wrong. Although the results of these findings are mixed, they generally point to the conclusion that psychopathic individuals have the capacity to make moral judgment (Borg & Sinnott-Armstrong, Reference Borg, Sinnott-Armstrong, Kiehl and Sinnott-Armstrong2013). Recently, Marshall et al. (Reference Marshall, Watts and Lilienfeld2018) conducted a meta-analysis of 27 studies, systematically analyzing the results of previous research to derive conclusions about the relationship between psychopathy and moral reasoning. Drawing from these studies (N = 4,376) of psychopathy using measures including self-reporting and common moral tasks (i.e., sacrificial moral dilemmas, Kohlbergian moral reasoning, and the moral foundations questionnaire [MFQ]), they detected a small but statistically significant relationship between psychopathy scores and the commonly used measures of moral decision making (rW = 0.16) and moral reasoning (rW = 0.10). However, very little evidence of “pronounced and overarching” moral deficits in conjunction with psychopathic traits is presented. It is possible that reports of marked deficits are overstated due to publication bias (Marshall et al., Reference Marshall, Watts and Lilienfeld2018). The authors suggested that their results raise the distinct possibility that the mental health profession and laypeople have underestimated the capacity of psychopathic individuals to appreciate moral responsibility.

Six studies within the meta-analysis used the MFQ created by Graham et al. (Reference Graham, Nosek, Haidt, Iyer, Koleva and Ditto2011). The results suggest that, like nonpsychopathic individuals, psychopathic individuals emphasize nearly all moral foundations they relate to equally. At the same time, while the slightly (but nonsignificantly) stronger scores on the harm subscale of the MFQ do not provide strong statistical support for the hypothesis (Glenn et al., Reference Glenn, Iyer, Graham, Koleva and Haidt2009) it hints at a trend that warrants further investigation. Overall, the meta-analysis from Marshall et al. (Reference Marshall, Watts and Lilienfeld2018) provides evidence against the idea that psychopathic individuals exhibit a pronounced or predominant deficit of moral reasoning (Furnham et al., Reference Furnham, Daoud and Swami2009), suggesting instead that psychopathic individuals exhibit subtle differences in moral judgment and emphasize slightly different moral foundations for making decisions than nonpsychopathic individuals.

In summary, as most studies fail to identify consistent or systematic differences in judgment capabilities between individuals with psychopathy and nonpsychopathy, there is good reason to believe that psychopathic individuals possess the capacity for making normative moral judgments. Though psychopathic individuals appear to have the ability to form a rational judgment as to whether an act is right or wrong, at a behavioral level they lack the capacity to follow through with what they reason to be right or wrong (or are prone to acting against moral norms out of self-interest). In other words, they are incapable of holding genuine moral concepts as a guide to behavior and thus fail to translate moral judgments into action. In Section 13.3, we move on from the role of reasoning and deliberation in moral judgment to the role of emotion processing and discuss how emotions contribute to moral decision making.

13.3.1 Shallow Emotional Experience

A fundamental reason why most individuals refrain from committing crimes relates to moral emotions. Feelings of anticipated shame, guilt, and remorse that are aversive and unpleasant serve to buffer individuals from offending (Rebellon et al., Reference Rebellon, Piquero, Piquero and Tibbetts2010; Tangney et al., Reference Tangney, Stuewig and Hafez2011). But what if an individual is unable to experience these emotions? Psychopathic personality is characterized by shallow emotional experiences and an inability to experience the full range and depth of normal emotions (Hare & Neumann, Reference Hare and Neumann2008; Ribeiro da Silva et al., Reference Ribeiro da Silva, Rijo and Salekin2012). Such emotional incapacity in psychopathic individuals may prevent them from relating empathically to others.

In an explication of early theories of psychopathy, Blackburn (Reference Blackburn and Patrick2006) put forward the idea that psychopathic individuals have a deficit in participating in role-taking within social groups and are insensitive to the needs of others. He argued that they are prone to ignore the rights of others and show a general inability to form lasting emotional bonds, all relevant to this key failure to assume the normal range of role-taking responsibilities that most people do quite easily. Without concern for others, individuals with psychopathy lack the capacity to make authentic moral judgments based on empathy and can only rationalize such judgments in the abstract. This basic ability to empathize with another can be viewed as foundational to all other emotions that one experiences in reaction to what people do to others. Essentially, the reason why we think it is wrong to harm others is because when we contemplate such harm, we feel part of the pain of the victim through an empathic affective reaction. This basic propensity therefore gives rise to other, more recognizable moral emotions, such as anger at injustice. A deficit in this ability to experience empathy among psychopathic individuals is thought to disrupt their emotional connectedness to others and interfere with the development of a moral conscience (DeLisi et al., Reference DeLisi, Vaughn, Gentile, Anderson and Shook2013; Poon & Ho, Reference Poon and Ho2015) and the internalization of moral standards of behavior (Blair, Reference Blair1995).

13.3.2 Facial Affect Recognition Deficits

Studies of emotion processing have shown that individuals with psychopathy not only display shallow emotional experience but also exhibit poor accuracy in their recognition of others’ emotions. Processing facial affect is crucial for socialization and normal social interaction (Corden et al., Reference Corden, Critchley, Skuse and Dolan2006; Fridlund, Reference Fridlund1991). Antisocial and aggressive behaviors may stem from an inability to correctly interpret the social and emotional cues of others (Blair, Reference Blair2003; Montagne et al., Reference Montagne, van Honk, Kessels, Frigerio, Burt, van Zandvoort, Perrett and de Haan2005; Poon, Reference Poon2016; Walker & Leister, Reference Walker and Leister1994). Some studies have suggested that this deficit appears to be specific to the emotions of fear and sadness, as demonstrated in a “violence inhibition” paradigm proposed by Blair and colleagues (Blair et al., Reference Blair, Mitchell, Peschardt, Colledge, Leonard, Shine, Murray and Perrett2004; Marsh & Blair, Reference Marsh and Blair2008).

In this model, the absence of fear and sadness when faced with others’ distress promotes violent behavior (Blair et al., Reference Blair, Budhani, Colledge and Scott2005). The core idea of Blair’s model (Blair, Reference Blair1995, Reference Blair2006) lies in classical conditioning theory. People typically find the fear and sadness cues of others to be innately aversive, so that when an antisocial behavior is followed by fear or sadness, the antisocial action itself becomes aversive and is inhibited. It follows that if an individual is impaired in recognizing fear and sadness, this allows them to behave in a more self-gratifying manner without the aversive effect of “feeling bad.” This explanation is supported by a meta-analysis conducted by Marsh and Blair (Reference Marsh and Blair2008), which suggests that psychopathy is associated with a specific deficit in recognizing fear and sadness, rather than a universal deficit in recognizing all emotions.

Some studies have found that psychopathic individuals showed impairments in the recognition of fear (Fairchild et al., Reference Fairchild, Van Goozen, Calder, Stollery and Goodyer2009) and sadness (e.g., Fairchild et al., Reference Fairchild, Stobbe, Van Goozen, Calder and Goodyer2010; Fairchild et al., Reference Fairchild, Van Goozen, Calder, Stollery and Goodyer2009; Hastings et al., Reference Hastings, Tangney and Stuewig2008) based on facial expressions. Others have also found deficits in recognizing fear by measuring callous-unemotional (CU) traits (Leist & Dadds, Reference Leist and Dadds2009; Muñoz, Reference Muñoz2009). There are, however, also numerous studies that did not find psychopathic traits to be associated with specific deficits in recognizing facial expressions of fear (e.g., Eisenbarth et al., Reference Eisenbarth, Alpers, Segrè, Calogero and Angrilli2008; Fairchild et al., Reference Fairchild, Stobbe, Van Goozen, Calder and Goodyer2010; Hastings et al., Reference Hastings, Tangney and Stuewig2008) or sadness (Del Gaizo & Falkenbach, Reference Del Gaizo and Falkenbach2008; Hansen et al., Reference Hansen, Johnsen, Hart, Waage and Thayer2008; Leist & Dadds, Reference Leist and Dadds2009; Muñoz, Reference Muñoz2009). Moreover, Del Gaizo and Falkenbach (Reference Del Gaizo and Falkenbach2008) found that psychopathy was associated with better recognition of fear, and a trend in this direction was observed by Woodworth and Waschbusch (Reference Woodworth and Waschbusch2008). Other studies have also found evidence that psychopathic traits are also associated with deficits in recognizing other emotions such as disgust (Hansen et al., Reference Hansen, Johnsen, Hart, Waage and Thayer2008), anger (in relation to CU traits; Muñoz, Reference Muñoz2009), happiness (Hastings et al., Reference Hastings, Tangney and Stuewig2008), and surprise (Fairchild et al., Reference Fairchild, Van Goozen, Calder, Stollery and Goodyer2009), calling into question whether psychopathic deficits really are specific to fear and sadness.

These mixed findings raise the possibility that deficits in facial expression recognition associated with psychopathic traits may be more pervasive than previously believed. A meta-analysis of facial expression recognition by Wilson et al. (Reference Wilson, Juodis and Porter2011) discovered that psychopathy was significantly associated with deficits in multiple emotional domains, although the effects were numerically small (N-weighted r = 0.06–0.12) and trended toward being largest for fear and sadness. These findings suggest that emotion recognition impairments in psychopathy are more pervasive than the more specific deficits outlined in Marsh and Blair’s (Reference Marsh and Blair2008) model.

13.3.3 Physiological Evidence of Emotion Deficits in Psychopathic Individuals

Psychophysiological research has convincingly supported the hypothesis that psychopathy involves a fundamental deficit in emotion processing. Studies in this area relate primarily to automatic affective responses consisting of a peripheral physiological response mediated by the autonomic nervous system such as the skin conductance response, eye blink reflex, and the approach-avoidance response (Levenson, Reference Levenson2014; Patrick, Reference Patrick2018). Early studies examined the change of participants’ skin conductance level to stimuli associated with the distress of others. In one paradigm, participants observed confederates who they thought were receiving electric shocks while their own skin conductance responses were recorded. When witnessing the distress of these confederates, offenders with a high level of psychopathy were found to show weaker autonomic arousal (i.e., lower skin conductance responses) than those without psychopathic traits (Aniskiewicz, Reference Aniskiewicz1979; House & Milligan, Reference House and Milligan1976).

Since then, numerous other studies have been conducted using the startle eye-blink response to test the hypothesis that psychopathic individuals are deficient in fear reactivity. The affective startle paradigm involves recording startle eye-blink responses to incidental noise bursts presented with emotional or neutral foreground stimuli. For the normal population, the startle blink response is an unconscious defensive reaction to aversive stimuli (Lang et al., Reference Lang, Bradley and Cuthbert1990), which is more potentiated during exposure to unpleasant images (e.g., violence, mutilation, murder) compared to neutral images (Vaidyanathan et al., Reference Vaidyanathan, Patrick and Bernat2009). A considerable body of research using this paradigm provides consistent and compelling evidence that, compared to normal individuals, psychopathic individuals are less sensitive (i.e., exhibit reduced or absent startle eye-blink responses) to aversive cues, indicating blunted emotions (e.g., Patrick, Reference Patrick2018) and heightened thresholds for reactivity to an aversive affective state (Levenston et al., Reference Levenston, Patrick, Bradley and Lang2000).

As one example, Patrick et al. (Reference Patrick, Bradley and Lang1993) reported that high-psychopathic male offenders have a heightened threshold for the startle blink during aversive picture viewing, indicating that it takes a stronger or more imminent threat to activate their defensive motivational system. Similar results were reported by Anderson et al. (Reference Anderson, Stanford, Wan and Young2011) in a sample of 76 undergraduate women assessed with high levels of psychopathic traits. The authors measured both affective startle blink responses and the P3 wave (i.e., an event-related potential component elicited in the process of decision making) using an affective picture-viewing task. Results showed that those scoring high on psychopathic traits lacked startle blink potentiation and demonstrated larger P3 amplitudes when faced with aversive stimuli. The P3 wave, is associated with attention and cognitive processing, with larger amplitudes indicating more cognitive resources devoted to the stimulus. In aversive contexts, larger P3 amplitudes suggest heightened arousal or emotional response, indicating greater engagement. For those with high psychopathic traits, larger P3 amplitudes in response to aversive stimuli suggest atypical processing, reflecting enhanced focus or heightened arousal that differs from typical responses. The study supports the generalizability of deficient startle potentiation of nonincarcerated females with psychopathic traits and adds to a growing body of research suggesting that psychopathic traits are associated with distinctive information-processing characteristics, as indexed by P3 amplitude.

In addition to the above paradigms, the approach-avoidance task (AAT) is another physiological reaction task used in psychopathy studies. In a study conducted by von Borries et al. (Reference von Borries, Volman, de Bruijn, Bulten, Verkes and Roelofs2012), the AAT was used to assess avoidance reactions to stimuli of potential threat among psychopathic individuals. During the task, stimulus photographs display facial expressions of actors with either angry, happy, or neutral expressions. Participants push a joystick either away or toward themselves as the faces increase or decrease in size. In this task, healthy individuals show a general tendency to move away from angry expressions and approach happy faces as in previous findings (e.g., Volman et al., Reference Volman, Toni, Verhagen and Roelofs2011). Psychopathic individuals, in contrast, show a total absence of avoidance tendencies toward angry faces (von Borries et al., Reference von Borries, Volman, de Bruijn, Bulten, Verkes and Roelofs2012), suggesting a lack of defensive responding consistent with findings from the startle blink paradigm. Moreover, their study found that avoidance rates were negatively correlated with aggression rates, suggesting that psychopathic individuals who show the lowest avoidance tendencies have the highest level of aggression. In summary, an absence of emotional reactivity seems to be a central deficit in psychopathy, leading to increased antisocial behavior due to diminished aversive arousal from punishment.

To conclude, intact emotional processes are theorized to be essential to moral behavior, providing immediate and salient feedback on behavior. Psychopathic individuals have difficulty responding empathetically, showing impaired processing of distress in others. Hence, we suggest that, while individuals with psychopathy understand in a rational sense what is right or wrong, they fail to experience this cognitive knowledge at an affective level because they lack the necessary emotional reactions to witnessing the harm caused to others. Nonetheless, how such deficits account for their deficient or absent moral sense still requires further investigation. We next take this physiological evidence further by addressing the neural foundation of the morality impairments found in antisocial and psychopathic individuals.

14.1.1 Moral Disengagement and Moral Exclusion

The first psychological writers on dehumanization began to publish on the subject in the 1970s. Their work made important contributions to the theory of dehumanization but relatively modest contributions to its empirical base. Perhaps as a result of this emphasis, the topic did not catch on in social psychology with the vigor that characterized the second wave of dehumanization scholarship, which arose around the millennium. Nevertheless, their work often foregrounded the moral dimensions of dehumanization in a way that later work did not. Among the key contributors of this period are Kelman (Reference Kelman, Kren and Rappoport1976) and Staub (Reference Staub1989), who wrote conceptually nuanced accounts of the role of dehumanization in war and in mass killing. However, the work of Bandura, Opotow, and Bar-Tal, and their colleagues, is of particular relevance to a chapter on moral psychology.

Bandura, one of the most influential psychologists of the twentieth century, is noteworthy for bringing dehumanization into the laboratory, for developing a theoretical formulation that emphasized how dehumanization overcomes inhibitions against aggressive behavior, and for examining individual differences in the tendency to dehumanize others. According to Bandura’s model, people are ordinarily restrained from behaving aggressively toward others by moral self-sanctions. When the potential targets of aggression are perceived as less than human, these self-sanctions are disengaged and those inhibitions are weakened. This “moral disengagement” account therefore presents an intrapsychic mechanism for enabling aggression, viewing it essentially as releasing the brakes of moral compunction and guilt. Bandura and colleagues carried out a series of studies that demonstrated greater punitiveness toward people described in a dehumanizing manner (e.g., Bandura et al., Reference Bandura, Underwood and Fromson1975; see Bandura, Reference Bandura1999, for an overview).

Whereas Bandura formulated dehumanization’s role in aggression in terms of loosened inhibitions on hostile behavior, Opotow’s somewhat related work on “moral exclusion” offered a more cognitive account that was not tied specifically to aggressive behavior. By her account (Opotow, Reference Opotow1990), individuals or groups can be shifted outside the moral sphere, where normal considerations of justice, fairness, and compassion apply. This active process of excluding persons from the moral domain effectively disqualifies them from moral standing. The result of that exclusion might be passive neglect or forms of active harming that extend beyond aggression in the narrow sense. Opotow did not privilege dehumanization as a phenomenon in immoral action or inaction but presented it as merely one among many forms of moral exclusion.

In a similar vein, Bar-Tal (Reference Bar-Tal, Bar-Tal, Grauman, Kruglanski and Stroebe1989) presented dehumanization as one of several forms of “delegitimization,” by which people withdraw legitimacy from groups they oppose, rendering aggressive behavior normatively acceptable. Legitimacy is commonly denied to opponents in violent conflicts, who are frequently denigrated as animals, demons, or monsters who are undeserving of concern and unfit for anything but the harshest punishments. Delegitimization and moral exclusion are therefore alternative ways of expressing the claim that people treat one another morally by default, but immoral treatment is enabled when their status as humans is denied or diminished.

14.1.2 Infrahumanization

Theories of moral disengagement, moral exclusion, and delegitimization understand dehumanization as one of a collection of related processes or mechanisms that serve to release moral restraints on aggression and other forms of ill-treatment. These theories typically understand dehumanization to be overt (e.g., revealed in language), its effects to be harmful, and its settings to be conflictual. The theories of moral exclusion and disengagement in particular also did not tie dehumanization primarily to intergroup contexts: One might exclude another from consideration on grounds other than their out-group status, and Bandura investigated whether delinquent schoolchildren’s aggressiveness might spring from greater tendencies to disengage morally. Research on “infrahumanization” (Leyens et al., Reference Leyens, Demoulin, Vaes, Gaunt and Paladino2007; Leyens et al., Reference Leyens, Rodriguez-Torres, Rodríguez-Pérez, Gaunt, Paladino, Vaes and Demoulin2001) differed from these earlier theories in all of these respects. The infrahumanization phenomenon was subtle and covert, was not necessarily associated with harmful behavior, occurred in the absence of major conflict, and was fundamentally an intergroup process.

Leyens and colleagues initiated the concept of infrahumanization not as part of a theory of dehumanization – they were at pains to differentiate the two concepts – but as a way to understand ethnocentrism, the tendency for ethnic groups to see their own cultural kind as the true humans. Pilot work suggested that humanness, conceptualized as contrastive with nonhuman animals, was understood by laypeople as constituted primarily by the capacities for language, intelligence, and uniquely human emotions (sentiments in French). Narrowing their focus to emotions, the researchers reasoned that if people see their in-group as more human than out-groups, they should reserve uniquely human emotions such as nostalgia and guilt for the in-group but not differentially attribute nonuniquely human emotions such as happiness and fear. This basic finding of differential attribution of uniquely human emotions, independent of emotion valence (positive versus negative), was replicated in many intergroup comparisons, including comparisons between groups with little or no contemporary conflict (e.g., Peninsular and Canary Island Spaniards). The ease with which infrahumanization could be studied in laboratory and field settings, using familiar survey and implicit social cognition tasks, ensured that it generated a substantial research literature.

Leyens and colleagues presented infrahumanization as a subtle but fundamental phenomenon that had more to do with how people view their in-group than how they view out-groups. Whereas dehumanization might normally be understood as a denial of humanity to others, generally accompanied by a negative view of them, infrahumanization was conceptualized as reserving humanity for the in-group, with the out-group thereby perceived as less human by this standard of reference. Infrahumanization does not entail dehumanization in a strong sense, but it might plausibly prepare the ground for it when conflict is stoked. If the out-group is already judged as less human than the in-group – further down the continuum from humans to animals than the in-group – then exacerbation of tensions might tip implicit deficiency in humanness into explicit denial. Leyens and colleagues showed that infrahumanization was not without consequences – it reduced out-group helping, for example, and deepened intergroup divides – but they gave less attention than earlier writers to the moral psychological dimensions of the phenomenon. Nevertheless, the fact that uniquely human emotions tend to be seen as more diagnostic of people’s “moral nature” (Demoulin et al., Reference Demoulin, Leyens, Paladino, Rodriguez‐Torres, Rodriguez‐Perez and Dovidio2004) than other emotions implies that infrahumanization is not a morally neutral bias in group perception.

14.1.3 The Dual Model

Infrahumanization theory offers a novel account of intergroup dehumanization that presents it as subtle, everyday, and falling on an implied continuum with more severe variants. Covertly perceiving out-groups as deficient in some of the emotional capacities that distinguish humans from nonhuman animals is a few incremental steps removed from overtly judging them to be animals. The theory also departs from prior accounts by understanding dehumanization processes in terms of human attributes rather than categories. Instead of understanding “human” as a sphere from which people could be categorically removed, as in moral exclusion, or “nonhuman” as a derogatory label such as “monster,” as in delegitimization research, Leyens and colleagues characterized infrahumanization as a lack of specified human characteristics relative to an in-group. Humanness was formulated as a dimension of psychological content along which groups can be located, such that dehumanization is a matter of people being judged less human or lesser humans, rather than as categorically nonhuman.

Haslam’s (Reference Haslam2006) dual model represents an extension of the approach pioneered by Leyens and colleagues. It retained their crucial idea that a satisfactory theory of dehumanization needs a theory of humanness, the quantity that is denied to people when they are dehumanized. It also retained their basic assumption that dehumanization can be subtle and should be conceptualized as a matter of degree. The dual model differed from infrahumanization theory in two key respects. First, it proposed that humanness could not be adequately defined only in contrast to nonhuman animals. Second, it expressly argued that dehumanization occurs on a continuum of severity or blatancy and that the contrast categories for defining humanness are implicated in that continuum.

On the first point, Haslam (Reference Haslam2006) argued that although humanness can be defined vis-à-vis animals as a set of uniquely human attributes, it can also be defined in a second salient way vis-à-vis inanimate objects. What makes us human is not only the uniquely human qualities of intellect, rationality, civility, self-control and the like that distinguish us from beasts but also the qualities of emotionality, warmth, and mental flexibility that people believe distinguish us from robots or automatons. Haslam referred to these latter qualities collectively as “human nature” and demonstrated empirically that the two senses of humanness were independent, were negatively associated as theorized with animals and objects, respectively, and had different correlates (e.g., human nature attributes were believed to emerge early in development and be widely prevalent in the population and were essentialized, whereas uniquely human attributes were believed to emerge later in development and to be uncommon, and were not essentialized). Haslam presented the two senses of humanness as encompassing dimensions that could be examined empirically by examining a range of psychological attributes, including emotions and personality traits.

On the second point, Haslam (Reference Haslam2006) argued that a theory of dehumanization must not only refer to these distinct senses of humanness but also remember that they are anchored by different kinds of nonhuman entity. Milder forms of dehumanization are likely to involve subtle denials of humanness to out-groups, but harsher forms will tend to involve explicit likening of those groups to the corresponding nonhuman. The standard infrahumanization effect is a subtle judgment that the out-group has a deficient capacity for experiencing refined, uniquely human emotions, but it lies at the mild end of a spectrum of animalistic dehumanization that extends to the use of vilifying animal metaphors at the other end.

Haslam’s dual model aimed to provide an integrative framework for dehumanization research in psychology by describing two such spectra. “Animalistic” dehumanization captures forms of dehumanization in which people are denied uniquely human attributes and/or likened to nonhuman animals and tends to occur when groups are seen as primitive, bestial, crude, uncivilized, barbaric, or stupid. “Mechanistic” dehumanization occurs when groups are denied human nature attributes and/or likened to inanimate objects. It tends to occur when people are judged to be cold, unfeeling, fungible, robotic, shallow, or literally objectified. An assortment of studies has found support for elements of this two-spectrum model, including the content and cross-cultural consistency of the humanness dimensions (Bain et al., Reference Bain, Vaes, Haslam, Kashima and Guan2012), and applied it to understanding dehumanizing perceptions of a range of groups (e.g., Loughnan & Haslam, Reference Loughnan and Haslam2007).

14.1.4 Dehumanization and Mind Perception

The dual model conceptualizes dehumanization as the denial of humanness to people along two dimensions. Although these dimensions would appear to have moral significance and should, in principle, capture the moral standing of humans relative to nonhumans, most research has assessed them using personality traits. Traits, and the emotions studied in infrahumanization research, are clearly germane to moral psychologists (e.g., Bastian et al., Reference Bastian, Laham, Wilson, Haslam and Koval2011). However, the moral psychological implications of dehumanization became clearer in another recent tradition of research on dehumanization that grounded it in mind perception. In this work (e.g., Gray et al., Reference Gray, Gray and Wegner2007; Waytz et al., Reference Waytz, Gray, Epley and Wegner2010), the account of dehumanization largely falls out of a primary focus on understanding the nature, causes, and consequences of perceiving mind in other entities, human and nonhuman. Dehumanization is understood as dementalization.

A key insight of this work is that perceptions of mind may not be unitary and different entities may be ascribed different kinds of mind. Work by Gray and colleagues (Reference Gray, Gray and Wegner2007), for example, found evidence of two dimensions of mind, which they labelled Agency and Experience. Agency represents the degree to which entities were judged to have the capacity for thinking, planning, and self-control, with adult humans ranking high and nonhuman animals ranking low. Experience, in contrast, represents the degree to which entities are judged capable of feeling, motivation, and subjectivity, with humans and other mammals ranked high and inanimate objects, robots, dead persons, and persons in a persistent vegetative state ranking low. The mapping of these dimensions onto the dual model’s human/animal and human/object contrasts is striking, but Gray and colleagues (Reference Gray, Gray and Wegner2007) drew out their moral implications more powerfully by linking their mind dimensions to the dimensions of moral agency and patiency. The degree to which people are seen as moral agents, capable of responsibility and blameworthiness, is driven by their perceived capacity for Agency, and the degree to which they are seen as moral patients, capable of having moral and immoral acts perpetrated on them, tracks their perceived capacity for Experience. On this understanding, dehumanization involves mind denial, and the form of that denial may have distinct implications for the moral credentials of the target.

Understanding dehumanization in terms of mind perception has benefits beyond providing clarity to moral psychologists. Research on the cognitive and motivational influences on mind perception (Waytz et al., Reference Waytz, Gray, Epley and Wegner2010) has direct relevance for understanding the causes and effects of dehumanization, connects it to important related literatures on empathy and theory of mind, and also helps to make dehumanization processes tractable for cognitive neuroscientists. Harris and Fiske (Reference Harris and Fiske2006), for example, evaluated dehumanization of an assortment of social groups by examining the absence of activation of social cognition centers in the brain, in the process supporting their proposal that groups stereotyped as low in warmth and competence, such as drug addicts and homeless people, are especially likely to be cognized as objects rather than social beings. This work has inspired several later studies of the neuroscience of dehumanization (e.g., Bruneau et al., Reference Bruneau, Jacoby, Kteily and Saxe2018; Jack et al., Reference Jack, Dawson and Norr2013). Although the specific validity of Gray et al.’s (Reference Gray, Gray and Wegner2007) dimensions of mind perception has been called into question (Malle, Reference Malle, Goel, Seifert and Freksa2019; Weisman et al., Reference Weisman, Dweck and Markman2017), and alternative structures might therefore imply different accounts of dehumanization, the idea of grounding an account of plural forms of dehumanization in an account of plural components of perceived mind has been a highly generative one. (See Chapter 9 in this volume for further discussion of the relationship between mind perception and dehumanization.)

14.1.5 David Livingstone Smith’s Account

We have examined several prominent psychological accounts of dehumanization, but it is also essential to review the most influential philosophical account, which has been set out by David Livingstone Smith in a powerful series of books (Smith, Reference Smith2011, Reference Smith2020, Reference Smith2021). Smith’s work places a high premium on applying his theoretical account of dehumanization to historical examples of racism, genocide, and enslavement, and it aims to explain the apparently paradoxical conceptual aspects of the phenomenon rather than to drive forward an empirical research program. In these respects, it tends to take a critical perspective toward psychological accounts of dehumanization, which in their efforts to bring the concept into the laboratory and make it tractable for social psychology researchers have often lacked historical depth, theoretical ambition, and philosophical sophistication.

At the core of Smith’s account is the recognition of an ambiguity or contradiction at the heart of well-known examples of dehumanization. He argues that perceiving others as humans is automatic (Smith, Reference Smith2021, p. 359). Dehumanization occurs when perceivers are led to believe by people with epistemic authority – whether they be scientists, ideologues, or others, and regardless of whether that authority is legitimate – that this human appearance is deceptive and the others are in reality subhuman. As a result, Smith (Reference Smith2021, p. 359) writes:

Smith uses the concept of “psychological essentialism” – laypeople’s tendencies to believe that certain categories are grounded in causal essences that define their members’ true nature – to explain this contradictory tendency for people to perceive others as simultaneously human and less than human. At the level of appearance, people are invariably perceived as human-like, but when they are dehumanized these “human-seeming beings” are ascribed a subhuman essence. Dehumanization therefore represents a belief that although someone might look human, they fundamentally are not: The target of dehumanization is both human and subhuman.

Smith takes issue with how some psychological writers employ the concept of psychological essentialism in their work. He argues, for example, that Leyens and colleagues’ work on infrahumanization, which formulates the phenomenon as a denial to out-groups of the “human essence” (represented by uniquely human emotions inter alia), is problematic because infrahumanization is conceptualized as a matter of degree whereas essences are either present or absent. To be infrahumanized should therefore constitute a categorical loss of humanness, not merely a partial or relative attenuation of humanness. Similarly, Smith criticizes Haslam’s account, which argues (and finds; Haslam et al., Reference Haslam, Bain, Douge, Lee and Bastian2005) that human nature attributes are essentialized and that mechanistic dehumanization therefore involves a denial of fundamentally human qualities to people. To Smith, phenotypical attributes such as traits are not the relevant locus of essentialist thinking in dehumanization and such attributes cannot qualify as causal essences regardless.

Ultimately, these critiques hinge on unresolved theoretical disagreements about the nature of psychological essentialism. At root, Smith’s account holds that essentialism operates at the level of the human category, that people construe essences in a categorical fashion, and that dehumanization is therefore a categorical perception that the target is nonhuman by virtue of having a subhuman essence. Leyens’ account also holds that essentialism operates at the level of the human category but conceptualizes the essence as a matter of degree: People are perceived to have the human essence to the extent they are ascribed uniquely human attributes. Haslam’s account places essentialism at the level of human attributes, not the category as a whole, and understands essentialist beliefs about human attributes to vary by degrees: The extent to which particular human qualities are seen as deep-seated, fixed, either/or, and defining falls on a continuum (Haslam et al., Reference Haslam, Bastian and Bissett2004).

The upshot of these disagreements is that Leyens and Haslam construe dehumanization as implicating essentialist thinking but occurring on a spectrum, a default assumption in quantitative social science if not in philosophy. By their accounts, dehumanization amounts to seeing others as less human, or as lesser humans, rather than as nonhuman. In contrast, Smith construes it as a categorical phenomenon: Either someone is dehumanized or they are not. This view raises challenging questions: If dehumanization is either/or, why does it appear to vary in severity or blatancy, and how can a person be dehumanized and yet still be perceived as human in some respects? Smith’s account resolves these questions by proposing that two simultaneous representations exist, one human (appearance) and one subhuman (essence). This account therefore has to posit dual representations of fluctuating relative salience to accommodate varying levels or severities of dehumanization, or it has to use “dehumanization” to refer only to the starkest examples and exclude subtler but otherwise very similar phenomena. In contrast, the psychological accounts can accommodate these variations more parsimoniously by invoking single representations that vary by degrees on a continuum from subtle to severe. Whether that parsimony is gained at the cost of a misunderstanding of psychological essentialism and an overly expansive definition of dehumanization remains open to debate.

14.2.1 Blatant versus Subtle

The most commonly discussed examples of dehumanization from history fall at one extreme on a dimension of blatancy or explicitness. When a human group is directly likened to a reviled animal and that likening carries an obvious implication for how the group should be treated, there can be no mistaking the severity of the dehumanization. As we have seen, however, some accounts of dehumanization refer to much subtler and milder phenomena. When a group is infrahumanized, there is no explicit likening to a nonhuman entity and both perceivers and targets may be unaware that targets are being perceived as less human humans. The phenomenon itself may simply be an unconscious and unmotivated tendency to not ascribe some emotions, both negative and positive, to the out-group. Other research on dehumanization deliberately examines unconscious or automatic processes by assessing implicit associations between groups and either humanness-related traits or nonhuman words, using the implicit association task or similar procedures (e.g., Loughnan & Haslam, Reference Loughnan and Haslam2007; Vaes et al., Reference Vaes, Paladino and Puvia2011). Research of this kind may reveal that people mentally associate out-groups with animals or do not associate them with uniquely human traits, but those associations are likely to be entirely unknown to the research participant and sincerely deniable by them. In cases such as these, dehumanization has occurred according to some existing accounts, even if they fall well short of prototypical cases from world history.

Although researchers have tended to emphasize the more subtle forms of dehumanization over the past two decades, blatant forms have also received some attention. The important work of Kteily, Bruneau and colleagues is especially relevant (e.g., Kteily & Bruneau, Reference Kteily and Bruneau2017; Kteily et al., Reference Kteily, Bruneau, Waytz and Cotterill2015). They have shown repeatedly that many people are willing to adjudge some groups as less than fully human when asked to rate them on the well-known “Ascent of man” scale, which depicts an evolutionary sequence from ape (0) to modern human (100). This blatant form of dehumanization, which appears to be strongest toward groups with which study participants are experiencing conflict, such as Muslim terrorists in the US and Roma in Hungary, is only moderately correlated with prejudice toward those groups, demonstrating that dehumanization and humanness are not reducible to hatred and negativity. Equally importantly, blatant dehumanization measured by the scale predicts a range of aggressive, punitive, and callous responses to its targets independently of prejudice. Blatant forms of dehumanization have also been examined in studies of animal metaphors (e.g., Haslam et al., Reference Haslam, Loughnan and Sun2011), which make the important point that not all animal metaphors are judged to be dehumanizing when applied to people, and that animals whose metaphorical use is judged to be offensive tend either to be seen as disgusting or seen as hierarchically beneath humans (e.g., apes; Brandt & Reyna, Reference Brandt and Reyna2011) in a way that implies a dehumanizing comparison.

14.2.2 Absolute versus Relative

Just as contemporary research on dehumanization encompasses subtle as well as blatant expressions, it also encompasses cases in which an out-group is seen as lacking humanness in itself and also when that lack is simply relative to the in-group. The prototypical cases of historical dehumanization are absolute in this sense, as they viewed ethnic or racial groups not merely as less human than other groups but as categorically and intrinsically subhuman. However, some forms of dehumanization recognized by contemporary psychological researchers are entirely relative in this sense. The infrahumanization effect involves the lesser attribution of uniquely human emotions to the out-group relative to the in-group and therefore need not involve any absolute or wholesale denial of the out-group’s capacity to experience refined emotions. Misunderstanding this point has implications for some challenges to the role of dehumanization in violence, to which I will return later in this chapter.

14.2.3 Animal versus Object

Prototypical examples of dehumanization invariably involve animal metaphors. Although animals may constitute the most salient contrast to the human category, and hence a primary basis for dehumanizing comparisons, Haslam’s (Reference Haslam2006) dual model insists that dehumanization can occur on a second register. The human/object contrast may not provide as many ready-to-hand dehumanizing metaphors as the human/animal distinction. Nevertheless, people spontaneously define humanness relative to unfeeling machines, have cross-culturally consistent understandings of human nature that are orthogonal to their understandings of human uniqueness, and differentiate humans from objects on a dimension of perceived mind that is more powerful than the one that differentiates humans from animals, according to Gray et al. (Reference Gray, Gray and Wegner2007). Failing to recognize how humans can be denied human nature and likened to objects would overlook many proposed manifestations of dehumanization that do not fit the animalistic mold. These examples include the literal objectification of women (Loughnan et al., Reference Loughnan, Haslam, Murnane, Vaes, Reynolds and Suitner2010), the reduction of patients to unfeeling bodies in some medical practices (Haque & Waytz, Reference Haque and Waytz2012), and the treatment of employees as fungible instruments in organizations (Brison et al., Reference Brison, Stinglhamber and Caesens2022). The concept of dehumanization is necessarily broader than a single framing of humanness in opposition to animals allows.

14.2.4 Simple versus Complex

A final contrast is between simple models of dehumanization and more complex ones. The simple case that dominates standard conceptualizations of dehumanization is that dominant group A holds and expresses a dehumanizing view of subordinate group B. Recent psychological work on the topic complicates this picture in several ways: 1) A may dehumanize itself; 2) A and B may dehumanize one another reciprocally; and 3) A and B may both believe that the other dehumanizes them.

The first of these complexities may seem implausible, but researchers have documented several cases in which people hold transient dehumanizing self-perceptions, sometimes in response to externally inflicted events such as social exclusions (Bastian & Haslam, Reference Bastian and Haslam2010), sometimes as a side effect of engaging in violent behavior (Bastian et al., Reference Bastian, Jetten and Radke2012), and sometimes as an enabler of immoral action (Kouchaki et al., Reference Kouchaki, Dobson, Waytz and Kteily2018). A perceived need to restore humanity to the self or group may motivate moral conduct.

The second complexity, reciprocal dehumanization, has been studied primarily by Kteily and colleagues in recognition that dehumanizing perceptions and the actions they drive may engender a matching response. Kteily and Bruneau (Reference Kteily and Bruneau2017), for example, demonstrated that Latinos and Muslims in the US held dehumanizing perceptions (assessed on the Ascent scale) of Republicans and Republican candidates during the 2016 Republican Primaries, in response to accurately perceived dehumanization by them, and the former perceptions drove reduced willingness to assist with counterterrorism efforts and increased desires to engage in violent protest. Their work demonstrated that dehumanization can be cyclical, mutual, and self-reinforcing and that it may be symmetrical even in asymmetrical conflicts.

The third complexity is linked to the second: A major trigger of reciprocated dehumanization is the perception that one group has been dehumanized by the other. This “meta-dehumanization” – feeling or learning that one’s group is not seen as fully human by an out-group – has become a major focus of research, starting with the work of Kteily et al. (Reference Kteily, Hodson and Bruneau2016). Their studies found, for example, that feeling dehumanized was distinct from feeling disliked or hated (meta-prejudice) and that meta-dehumanization was associated with support for aggressive actions and policies toward out-groups independently of meta-prejudice. Recent work underscores the vital importance of meta-dehumanization, showing that it predicts hostility more strongly than meta-prejudice (Landry et al., Reference Landry, Ihm and Schooler2021) and that meta-humanization – feeling or learning that one’s group is seen as fully human – may effectively counteract prejudice (Pavetich & Stathi, Reference Pavetich and Stathi2021).

To summarize this section of the chapter, the recent psychology of dehumanization has extended its meanings well beyond the standard understanding embodied in well-known historical examples from the Holocaust and the Rwandan genocide. The prototype of dehumanization is blatant, categorical, and one-directional in applying an animal metaphor to a target group. However, dehumanization extends to cases that are subtle and unconscious, that are matters of degree defined relative to an in-group rather than absolute denials of humanity, that assimilate humans to inanimate objects rather than animals, and that occur in reciprocal cycles driven by meta-perceptions.

14.3.1 Before Harm

The role of dehumanization in conflict has often been thought of as preparatory. Dehumanizing a group is seen as creating the conditions under which violence flourishes. Several lines of theory and research help to comprehend how that preparation occurs. The most relevant approaches to dehumanization for this purpose are those that conceptualize it as an enduring set of beliefs or perceptions rather than as an episodic phenomenon or one that motivates specific actions.

The infrahumanization effect, for example, implies that ethnocentrism is a bedrock phenomenon in which groups perceive themselves to be more human than others. At the most basic level, Leyens and colleagues would claim, groups are predisposed to judge others as less good and less human than themselves. Intergroup relations start from a position in which a subtle degree of dehumanization is present even before conflict arises, and infrahumanization may make conflict more likely to develop than if that imbalance in perceived humanness did not exist.

Approaches to dehumanization grounded in trait-like dimensions of humanness or mind or in models of stereotype content are similar in this regard. In the dual model, out-groups can be located on dimensions of human uniqueness and human nature according to their members’ perceived attributes, and those locations may influence how more severe forms of dehumanization may arise under conditions of conflict. For example, out-groups perceived as lacking in uniquely human qualities (e.g., perceived as unintelligent and lacking in self-control) are at risk of being seen as bestial and barbaric during conditions of conflict, and the more they are seen as lacking those qualities the greater that risk may be. The same may be inferred from the mind perception account of dehumanization. Groups that are judged to lack one dimension of mind may be especially likely to be mistreated on account of their associated (perceived) deficiencies. A group believed to lack Experiential mental capacities should be especially likely to be treated badly when conflict arises, as it might be seen as lacking the capacity to suffer. This link between group stereotypes and vulnerability to mistreatment is clearest in Harris and Fiske’s (Reference Harris and Fiske2006) model of dehumanization, which identifies groups stereotyped as low in warmth and competence as at particular risk.

14.3.2 During Harm

Dehumanization in these varied theoretical models can help to explain how and why intergroup conflict may erupt in harmful behavior. Other models emphasize how dehumanization enables aggression in the present by releasing restraints on harming others. This element is most prominent in Bandura’s early work on moral disengagement (e.g., Bandura et al., Reference Bandura, Underwood and Fromson1975). In effect, Bandura proposes that perceiving the target of potential harm as less than fully human overcomes normal inhibitions that block it, such as anticipatory guilt and moral compunction. In the heat of conflict, dehumanizing the out-group may enable harmful behavior and dissolve objections to it if it is being carried out by members of one’s own group.

Delegitimization and inflammatory metaphors can also be understood to play a part in enabling harm or violence in the present. Bar-Tal’s (Reference Bar-Tal, Bar-Tal, Grauman, Kruglanski and Stroebe1989) work on intergroup conflict emphasized how withdrawing the legitimacy of an enemy group can take place in the emotional heat of conflict and is commonly expressed on vilifying labels. In this line of work, dehumanizing language expresses and reinforces a view of the other group as undeserving of concern in the here and now.

14.3.3 After Harm

Although dehumanization is usually examined as a process that prepares the ground for future harm and enables the commission of present harm, researchers have also examined how people dehumanize victims after harm has been perpetrated against them. In this work, people have several motives for dehumanizing harmed out-groups, including rationalizing the harm, minimizing its magnitude, and reducing guilt and collective responsibility for committing the harm. For example, Castano and Giner-Sorolla (Reference Castano and Giner-Sorolla2006) found that people whose nations had killed indigenous inhabitants of their colonies denied those inhabitants the capacity for uniquely human emotions more when reminded of past atrocities. Similarly, Cehajic et al. (Reference Cehajic, Brown and Gonzalez2009) showed that Serbians reminded of atrocities committed against Bosnian Muslims during the Bosnian War infrahumanized them. Other research has suggested that dehumanization may serve not only to whitewash the past but also to be an obstacle to future reconciliation. Tam and colleagues (Reference Tam, Hewstone, Cairns, Tausch, Maio and Kenworthy2007) showed that Northern Irish participants on the Protestant and Catholic sides who denied uniquely human emotions to their adversaries were less inclined to forgive, to abandon historical grievances, and to seek rapprochement.

14.3.4 The Case of Genocide

These discussions indicate that dehumanization may be thoroughly enmeshed in several aspects of intergroup conflict rather than merely setting the stage for it. The multiple involvements of dehumanization across the life history of conflict can be combined to offer an expanded framing of the role of dehumanization in genocide in particular (see Haslam, Reference Haslam and Newman2019, for a more fully developed analysis). According to one prominent model of the genocidal process (Stanton, Reference Stanton1998), dehumanization is a factor that is specific to the third of eight stages. Us–them classification of in-group and out-group (stage 1) is succeeded by the symbolizing of group identities by labeling or visible signs such as distinctive dress (stage 2). Dehumanizing expressions, taking the form of hateful animal metaphors, are then disseminated by perpetrator groups (stage 3) and promote the formation of organized militias (stage 4). Extremist voices polarize intergroup relations (stage 5) and extreme actions such as segregation of the victim group (stage 6) prepare the way for mass killings (stage 7), which are covered up after the fact by concealment of evidence and denials (stage 8).

This stage model may be accurate as far as it goes, but its delimited role for dehumanization rests on a narrow understanding of the phenomenon as blatant, absolute, animal metaphor-based, and one-directional. As I have shown, contemporary dehumanization theory and research provide a much more expansive account of dehumanization’s varied forms. Taking up that account allows for fuller understanding of the role that dehumanization might play in multiple stages of the genocide process.

For example, infrahumanization research indicates that even prior to conflict, in-groups tend to perceive out-groups as less human than they are, ensuring that even at stage 1 the us–them distinction incorporates a tacit assumption that the out-group is composed of lesser humans. Similarly, research indicates that the derogatory group labels and stereotypes of Stanton’s stage 2, which in theory occur prior to the overt adoption of dehumanizing animal metaphors, are often accompanied by covert and perhaps unconscious dehumanizing associations. The work of Goff et al. (Reference Goff, Eberhardt, Williams and Jackson2008), for example, showed that White Americans implicitly associate Black Americans with apes and that unconscious metaphorical association is implicated in hostile and punitive responses to Black targets. Implicit forms of dehumanization may therefore be consequential aspects of the symbolization of out-groups, before dehumanizing metaphors are spread by organized propaganda.

The processes of planning, polarization, segregation, and killing (stages 4–7) may also be propelled by the dynamics of dehumanization rather than merely set up by them, as Stanton’s stage model asserts. This view is shared by genocide scholars Haagensen and Croes (Reference Haagensen and Croes2012), who write: “Dehumanization … does not only play a role in preparing the ground for genocide; it also drives genocidal campaigns forward, sustaining killing operations over lengths of time” (p. 225). For example, Kteily et al.’s (Reference Kteily, Hodson and Bruneau2016) work demonstrates that the blatant dehumanization of out-groups is associated with endorsement of extreme or polarized policy positions, a finding also obtained by Maoz and McCauley (Reference Maoz and McCauley2008) in a study of the Israel-Palestine conflict. Moral exclusion surely plays a role in driving popular support for segregation of a targeted minority, and moral disengagement has been shown to release inhibitions on violent action, potentially including acts of mass killing. Stage 8 denial may also implicate dehumanization, which has been shown to help reduce collective guilt, rationalize past in-group violence, diminish support for providing reparations to victims, and minimize perceptions of their suffering in contexts including indigenous–settler relations, civil war, and religious strife.

In sum, it is possible that out-group dehumanization plays a variety of roles through the genocide process rather than being confined to a single stage. This speculation may even be too timid, as other forms of dehumanization may also be implicated. Self-dehumanization (Kouchaki et al., Reference Kouchaki, Dobson, Waytz and Kteily2018) may play a causal role in enabling the commission of violence or other immoral actions. Some descents into genocide may also be accelerated by reciprocal dehumanization between conflicting groups. When dehumanization is understood in an expansive way that is grounded in recent research and scholarship, the processes through which it might contribute to particular cases of violence are many and varied, extending well beyond the traditional emphasis on hateful propaganda.

15.2.1 Hunter-Gatherer Communities and Moral Criticism

Until about 10,000 years BCE, humans lived as hunter-gatherers in small bands of 25–50 (Boehm, Reference Boehm1999; Knauft, Reference Knauft, Sponsel and Gregor1994). We know this from some archeological finds (Bandy, Reference Bandy2004; Enloe, Reference Enloe2003), population genetic analyses (Atkinson et al., Reference Atkinson, Gray and Drummond2008; Henn et al., Reference Henn, Gignoux, Jobin, Granka, Macpherson, Kidd, Rodríguez-Botigué, Ramachandran, Hon, Brisbin, Lin, Underhill, Comas, Kidd, Norman, Parham, Bustamante, Mountain and Feldman2011), but predominantly from ethnographic research of hunter-gatherer societies over the past 100 years (Lee & Daly, Reference Lee and Daly1999; Wilson, Reference Wilson1988; Woodburn, Reference Woodburn1982). According to this accumulated evidence, hunter-gatherer communities were nomadic and thus highly mobile, changing camp every few weeks to months (Lee, Reference Lee and Spooner1972), mostly lacking possessions, and having little sense of territorial boundaries (Wilson, Reference Wilson1988). They were highly egalitarian and, without one supreme ruler, lawmaker, or judge, leadership was provided by different members for different tasks (Service, Reference Service1966).

The hunter-gatherer norm system was grounded in reciprocity, which regulated both community living and vital activities such as food acquisition and consumption. When a hunter shared his yield with another hunter one day, he could expect to receive part of the other’s yield another day. Indeed, sharing large hunts counted as a “virtually universal rule” among hunter-gatherer societies (Wilson, Reference Wilson1988, p. 37), and violating this sharing norm was met with sanctions.

In these highly interdependent bands, sanctioning was interpersonal. Most transgressions were easy to detect (Silberbauer, Reference Silberbauer, Leacock and Lee1982) because life was public and transparent. But sanctions of norm violators rarely consisted of punishment (Baumard, Reference Baumard2010; Guala, Reference Guala2012), since these egalitarian groups disdained assertations of power and coercion, and debilitating penalties would have hurt the community at least as much as the transgressor. Instead, communities favored communication, criticism, sometimes ridicule (Wiessner, Reference Wiessner2005), and gossip (Dunbar, Reference Dunbar1996). Repeat offenders may have been isolated or, as a last resort, expelled from the group (Woodburn, Reference Woodburn1982). But the typical way of responding to norm violations was a form of public criticism, a threat to the person’s social standing and reputation – which, in small, interdependent groups, largely ensured norm compliance.

These civil forms of sanctions, however, occurred within groups and did not prevent highly punitive actions between groups. How much warfare occurred in early human hunter-gatherer communities is debated. But few scholars doubt that intergroup violence gradually increased with population density and with the correlated expansion from extended families to clan-based networks (Flannery & Marcus, Reference Flannery and Marcus2012). These growth patterns set the stage for a transformation in the social organization of human societies.

15.2.2 The Emergence of Punishment Post Settlement

The second phase in the cultural evolution of sanctioning methods occurred after humans settled down, beginning around 12,000 years ago. Norm enforcement turned from an interpersonal process of criticism and reputation threat to an institutionalized process that confronted the perpetrator with the damaging actions of a ruling entity and its henchmen. The transition occurred in different places at different times, but the endpoint was, with few exceptions, a hierarchical society that centralized and legitimized harsh punishment. A few essential facts about sedentary life help explain this change (Lee, Reference Lee and Spooner1972; Peregrine et al., Reference Peregrine, Ember and Ember2007; Redman, Reference Redman1978).

Property. As hunter-gatherers, people possessed only tools, tents, and clothes; food was distributed and consumed right away; and land had no boundaries of ownership (Wilson, Reference Wilson1988). After settling down, more and more people had permanent housing, household effects, and eventually land, livestock, and crops. Quickly, some gained more property than others. Such inequalities were enlarged by inheritance rights, intermarriage, as well as reciprocal trade and protection among property owners (Peregrine et al., Reference Peregrine, Ember and Ember2007). The have-nots desired others’ property, and the crimes of robbery and burglary emerged. The haves feared losing their property, so the need for protection, law enforcement, and punishment increased.

Population Growth. Between 10,000 and 8,000 BCE, population growth was spurred by environmental changes (the end of the Last Glacial) and geographic opportunities (e.g., the fertile Near East; Aurenche et al., Reference Aurenche, Kozłowski and Kozłowski2013), turning camps into villages and towns (Atkinson et al., Reference Atkinson, Gray and Drummond2008; Gignoux et al., Reference Gignoux, Henn and Mountain2011). Settlements also increased the rates of offspring. Whereas nomad communities had to carry their newborns for thousands of kilometers a year, limiting child birth to once every 3–4 years (Lee, Reference Lee and Spooner1972), settled communities were able to dramatically increase the frequency of pregnancies (Buikstra et al., Reference Buikstra, Konigsberg and Bullington1986). Furthermore, plant and animal breeding experiments as well as improved food production and storage technologies fed more people. In a positive feedback loop, the resulting population growth put pressure on food production and demand for labor. Consequently, a second population growth turned towns into cities, several of which grew to 30,000 inhabitants or more between 2000 BCE and 1200 BCE. Another feedback loop then emerged between rapidly increasing built structures (Wilson, Reference Wilson1988) – from private homes to royal monuments and sacral edifices – and the massive opportunities for work to sustain this building boom (Boserup, Reference Boserup1965). That work was likely done by those who didn’t own property – the poor, foreigners, and conquered populations; thus, dependent labor and class stratification accelerated.

Intergroup Conflict. By and large, hunter-gatherer societies were peaceful (Kelly, Reference Kelly2000; Silberbauer, Reference Silberbauer, Leacock and Lee1982). There was little reason to attack another band (what do they have that we don’t?). There was also little reason to fight over territory that constantly shifted (Woodburn, Reference Woodburn1982). But once there were settlements, some land was more fertile, some villages more productive, and fast-growing populations had a need to either intensify food production (Carneiro, Reference Carneiro1970) or expand their territory (Dumond, Reference Dumond and Spooner1972), especially because food resources may not have kept pace with population growth (Larsen, Reference Larsen1995). Thus emerged incentives for organized warfare (Milner et al., Reference Milner, Anderson and Smith1991; Redman, Reference Redman1978) to gain land, food, raw materials, tools, and to acquire further labor by subjugating other communities. As potential targets for attack and raids, settled communities bolstered their safety by developing weapons and defensive facilities, which further demanded an expanded labor force and eventually a professional army.

Hierarchical Social Structure. All these changes – expanded and unequally distributed property, population growth, dependent labor, and intergroup hostilities – demanded organization and administration, taken on by a few that turned into elites and authorities (Redman, Reference Redman1978). New crimes emerged when this authority was violated, such as treason, social revolt, or tax evasion. Leadership structures were also necessary to maintain within-group cohesion in light of new role differentiation and resulting conflict potential (White, Reference White2007). Settled humans were arranged from slaves to freemen to landowners, eventually aristocrats and priests, and almost always a supreme ruler. In the hundreds of thousands of years prior to settlement, primate hierarchies were transformed into egalitarian, cooperative human bands; within just a few thousand years, a new society was born, returning to hierarchical and often brutally competitive primate roots.

15.2.3 Escalation of Punishment

In small hunter-gatherer communities, a norm violator had equal-status relationships with other group members, and these relationships had to be maintained for continued group success. Sanction severity for norm violations was therefore restrained because harsh penalties might invite equally harsh responses in the future and harm community relationships. By contrast, in larger settlements, ties between norm violators and the ruling class became weak and asymmetric, promising scant reciprocal future interaction and setting few limits on sanction severity. Over thousands of years of growing social hierarchies, we therefore see both increasingly centralized and increasingly brutalized forms of punishment. Several causal processes contributed to such escalation of punishment.

Punishment became more severe when rulers defended themselves against threats to their position. For example, in eighteenth-century Tahiti, the chief left it to family heads to enforce rules about property or marriage but applied severe punishment when offenses were “sacred,” typically those against his interests (Claessen, Reference Claessen and Feldbrugge2003). In nineteenth-century despotic African Buganda, sacred crimes, especially those against the ruler, were met with punishments of maiming, cutting into pieces, and burning alive (Claessen, Reference Claessen and Feldbrugge2003). Political rulers in China’s hierarchical societies BCE (Bozan et al., Reference Bozan, Shao and Hu2001) also relied on stern rules and punishment, especially during times of civil unrest. The death penalty applied to many offenses against the emperor (e.g., treason, rebellion) and the state (e.g., malpractice, bribery, illicit sales). In the early Roman Republic, laws were in tight control of the ruling class and were even kept secret from the lower class. When the laws were finally publicized in the Twelve Tables (c. 450 BCE), they contained detailed civil statutes but also prescribed capital punishment for those who defamed others in song (VIII.1b) – to avert any chance of inciting the proletariat’s sentiments against the ruling elite. Over the centuries, wars and uprisings caused tyrannic rulers to take the reins of the Roman Republic, and punishments became excessively cruel, even public spectacles, to deter any threat to the elites (Moore, Reference Moore2001).

War is a driver of severe punishment. It brutalizes relations between communities, and it brutalizes relations within. When people are exposed to war, violence is part of life, and aggressive behavior becomes disinhibited (MacManus et al., Reference MacManus, Dean, Al Bakir, Iversen, Hull, Fahy, Wessely and Fear2012). In wartime, rulers and state governments also more easily justify violent punishment within the state, especially for crimes of treason or cowardice. For example, before and during World War II, Nazi Germany took capital punishment to incomparable extremes, with a huge increase in death warrants and executions (Messerschmidt, Reference Messerschmidt2005). The death penalty was applied to a wide range of criminal acts, various forms of military disobedience, and many forms of political resistance.

Despite the brutality of punishment in many early societies, it did not succeed in deterring crime, which then further incited harsher punishments. We see this cycle in modern times as well. For example, in the United States’ “war on crime” of the 1980s (Pratt et al., Reference Pratt, Brown, Brown, Wallsworth and Morrison2005), when punishment trends hardened in response to (perceived) rising crime rates but failed to loosen when crime rates declined. To this day, increased institutional punitiveness does not result in crime reduction (Cullen et al., Reference Cullen, Jonson and Nagin2011).

We also see detrimental escalation processes in today’s school environment (Darling-Hammond et al., Reference Darling-Hammond, Ruiz, Eberhardt and Okonofua2023). As teachers increasingly punish students for even minor misbehavior – often because they regard the students as “troublemakers” (Okonofua & Eberhardt, Reference Okonofua and Eberhardt2015) – students become defiant and misbehave even more (Pesta, Reference Pesta2022), which cyclically escalates punishment. This escalation occurs primarily for minoritized students (Amemiya et al., Reference Amemiya, Mortenson and Wang2020).

Thus, both past and present reveal that punishment is a questionable method of moral regulation; it often expresses power, social hierarchies, and their dominance-maintaining mechanisms (Sidanius & Pratto, Reference Sidanius and Pratto1999).

15.3.1 Distinguishing Punishment and Blame

Everyday moral sanctions range from raised eyebrows to muttered disapproval, from emotional complaints to thoughtful criticism; but there are also acts of physical aggression, withholding of resources, public shaming, and social exclusion. The first set is more representative of moral criticism (blame, for short) whereas the second set is more representative of genuine punishment.

Even though blame and punishment are not categorically distinct, a bundle of features allows us to distinguish them. Blame has a dual nature as a particular kind of moral judgment and as communicated moral criticism; punishment is always an observable act. When expressed, blame often requests or demands that the violator respond to the criticism and change their (future) behavior; punishment coercively imposes costs on the transgressor, often relying on power or status to legitimize the imposition of costs. Blame is typically reversible – one can admit that one unfairly blamed the other and take back the criticism – whereas the tangible costs of punishment can rarely be taken back. Finally, blame can be contested, leaving room for justification and negotiation; punishment leaves little room for negotiation and is often the last resort when moral criticism has failed. (For a different conception of punishment, incorporating the features of blame, see Chapter 16, this volume.)

In light of these distinguishing properties, we would expect that, in informal, everyday interactions, humans prefer to use moral criticism over punishment. Indeed, using daily surveys to study naturally occurring norm violations, Molho et al. (Reference Molho, Tybur, Van Lange and Balliet2020) found that people were generally unmotivated to engage in punishment as physical confrontation and somewhat more motivated to gossip or exclude the transgressor. Field studies show that people enforce norms with subtle verbal or nonverbal interventions but not active punishment (e.g., Przepiorka & Berger, Reference Przepiorka and Berger2016). In a cross-cultural study, rates of responding to violations with cost-imposing sanctions (e.g., yelling, insulting, hitting) were extremely low, between 1 percent and 6 percent (Pedersen et al., Reference Pedersen, McAuliffe, Shah, Tanaka, Ohtsubo and McCullough2020). Even though a number of behavioral economics experiments have suggested that people are willing to punish norm violators, this form of punishment is mild and indirect (taking a little money from a stranger). Moreover, when given a choice, people prefer not to punish but to criticize the violators to their face or to others (Feinberg et al., Reference Feinberg, Cheng and Willer2012; Kriss et al., Reference Kriss, Weber and Xiao2016; Xiao & Houser, Reference Xiao and Houser2005), to compensate the victim (Chavez & Bicchieri, Reference Chavez and Bicchieri2013; Hershcovis & Bhatnagar, Reference Hershcovis and Bhatnagar2017), or even to restore order by correcting the negative outcome (e.g., cleaning up other people’s litter; Berger & Hevenstone, Reference Berger and Hevenstone2016). As we move into more hierarchical relationships – parent and child, teacher and student, boss and employee – punishment becomes more likely, more consequential, and is often seen as more legitimate (Gershoff & Lee, Reference Gershoff and Lee2020; Mooijman & Graham, Reference Mooijman and Graham2018). Embedded in such hierarchies, “[p]unishment is not a behavior, but an institution” (Binder, Reference Binder2002, p. 321). Indeed, punishment is the predominant sanctioning instrument used by the state (Tonry, Reference Tonry2009).

15.3.2 Understanding Punishment

If interpersonal punishment is relatively infrequent in daily life, why is it so widespread in its institutionalized form? The brief cultural history of punishment has taught us that punishment has been used to maintain hierarchies, defend the powerful, and remove the unwanted. The severity of punishment has declined over time, but to this day, punishment is an instrument of power, oppression, and the maintenance of dominance hierarchies (Redford & Ratliff, Reference Redford and Ratliff2018; Sidanius et al., Reference Sidanius, Mitchell, Haley and Navarrete2006). And not only is punishment discriminatory, it is also ineffectual as an instrument of deterrence (Cullen et al., Reference Cullen, Jonson and Nagin2011). If punishment tends to fail in so many ways, what maintains the public support for institutional punishment?

15.3.3 Forces that Maintain Support for Punishment

Retribution. A first hypothesis is that people are fundamentally retributivists – desiring proportional and painful punishment of norm violators, even when it has no deterrent function (Carlsmith et al., Reference Carlsmith, Darley and Robinson2002; Goodwin & Gromet, Reference Goodwin and Gromet2014). But rather than explaining common support for institutional punishment, people’s retributive tendencies may be themselves explained by their exposure to institutions of punishment. Research on lay retributivism always probes whether people recommend or endorse retributive punishment for crimes, such as assault, rape, or murder. For those violations, people may endorse severe punishment because that, they have learned, is the proper response. However, people are not merely reflecting back the severity of punishment in legal reality; they are often manipulated by political rhetoric, more so from the right, that crime is rampant and society needs to be tougher on crime (Muhammad et al., Reference Muhammad, Gottschalk, Thompson, Travis, Western and Redburn2015; Tonry, Reference Tonry2009).

However acquired, what is it that human retributive tendencies really support? The current literature often defines retributive commitments as referring to “deserved punishment of a guilty offender” (Goodwin & Gromet, Reference Goodwin and Gromet2014, p. 562). People are retributivists, then, if they endorse a statement such as “by punishing offenders we give them what they deserve” (Nadelhoffer et al., Reference Nadelhoffer, Heshmati, Kaplan and Nichols2013, p. 244). But in what sense is deserved punishment “retributive” but a deserved award is not (Cottingham, Reference Cottingham1979, p. 239)? In both cases, the response is considered appropriate or justified, given what the person actually did. This notion of desert stands in contrast to the extreme consequentialist position that what the transgressor did (or thought while doing) is irrelevant; that the only thing that matters for punishment is its beneficial consequence (e.g., deterrence of crime, calming public outrage). Rejecting this position, people endorse retribution in the sense of paying back (Cottingham, Reference Cottingham1979), which is what the original twelfth-century English word referred to. Such retribution was a symmetric practice of moral accounting – both rewards and punishments were forms of paying back.

If retribution is the mirror image of the human disposition toward reciprocity for good deeds (Gouldner, Reference Gouldner1960), then being retributive need not imply cruel enjoyment of a wrongdoer’s suffering. Instead, a scale is being balanced between the transgressor’s behavior and the community’s norms; this scale, of course, is what the goddess Justitia holds in her hand to symbolize justice. Whether the repayment is cruel or civil – mutilations and executions, or monetary fines, apology, and atonement – depends on the historically and geographically specific currency in which moral debts are repaid. Thus, even if ordinary people’s support for retributive punishment is grounded in a general principle of reciprocity, it has undoubtedly been nourished by long-standing patterns of political and religious influence (Griffith, Reference Griffith2020).

Delegation. A second contributor to support for institutional punishment stems from the dangers of personal punishment. Punishment is typically an act of aggression and inflicts damage on the perpetrator, who may therefore retaliate. A perpetrator who has committed a serious transgression has already shown little commitment to the community’s norms; when punished, the person is likely to avenge the inflicted damage. In hunter-gatherer societies, personal interventions were more pressing and more likely in response to extreme violations (Kelly, Reference Kelly2000, p. 27). However, they came with significant risk. Lee (Reference Lee1979) reported that almost half of the !Kung community’s fatalities over several decades were bystanders or “peacemakers,” who died in the attempt to end a conflict. Retaliation against norm-enforcing punishment is even more likely to be committed by people outside one’s group because the perpetrators have no relationship with the norm enforcer or the community and thus no reason to show remorse, correct their behavior, and protect their reputation. Indeed, in human history, intergroup conflicts have often escalated from individual revenge to family- or group-based revenge, spiraling into blood feuds (Boehm, Reference Boehm2011).

In all these cases, the costs to directly punish a perpetrator are high, but the risks of retaliation can be reduced or even eliminated by handing the task of punishment to specific agents or institutions (Cushman, Reference Cushman2015). Such delegation of punishment protects ordinary people and allows them to safely condemn violators and support the punishment meted out by the authorities – even excessively harsh punishment. It is far easier to cheer on the torturer or vote for “tough on crime” policies than it is to break the prisoner’s arm oneself or stand guard at a solitary confinement cell. Delegated punishment, however, requires justification, which a ruler or state has in general and can award to particular officials, such as judges, prison guards, or executioners. Institutionalized punishment guarantees legitimacy without having to establish it anew each time (Binder, Reference Binder2002; Garland, Reference Garland1990). As long as the institution is seen as reasonably consistent and fair, the community will support it and grant it the permission to punish some of norm-violating members.

15.3.4 Does Punishment Foster Cooperation?

Leaving the institutionalized context, punishment in small groups can foster cooperation by decreasing the number of free riders and increasing the contributions individuals make to the group (Fehr & Gächter, Reference Fehr and Gächter2000; Yamagishi, Reference Yamagishi1986). Punishment works best when it articulates community norms (Andrighetto et al., Reference Andrighetto, Brandts, Conte, Sabater-Mir, Solaz and Villatoro2013; Xiao, Reference Xiao, Teitelbaum and Zeiler2018), teaches the violator (Cushman, Reference Cushman2015), brings about a response from the violator (Funk et al., Reference Funk, McGeer and Gollwitzer2014), and when it is gentle and deemphasizes power differences (Kochanska & Aksan, Reference Kochanska and Aksan2006).

But punishment can go awry. When cooperators are punished, costly punishment no longer promotes cooperation (Rand et al., Reference Rand, Armao, Nakamaru and Ohtsuki2010), and the option to punish can lead to retaliation (Hopfensitz & Reuben, Reference Hopfensitz and Reuben2009). When social communities view punishment as unwarranted, it is less effective at correcting social behavior (Herrmann et al., Reference Herrmann, Thöni and Gächter2008). Furthermore, impressions of third-party punishers are less than favorable (Dhaliwal et al., Reference Dhaliwal, Patil and Cushman2021), and even victims who punish are not seen very positively; instead, victims who forego punishment are seen as moral, trustworthy, and altruistic (Heffner & FeldmanHall, Reference Heffner and FeldmanHall2019).

Punishment may not even be necessary for cooperation (Baumard, Reference Baumard2010). Punishment’s apparent power to increase cooperation in mixed-motive games is matched (Feinberg et al., Reference Feinberg, Willer and Schultz2014) or surpassed (Wu et al., Reference Wu, Balliet and Van Lange2016) by that of gossip, and it is also matched by communicated disapproval (Masclet et al., Reference Masclet, Noussair, Tucker and Villeval2003). When we compare overall payoffs in these game contexts, both gossip and disapproval lead to better collective outcomes than does punishment (Dugar, Reference Dugar2010; Wu et al., Reference Wu, Balliet and Van Lange2016).

15.3.5 Punishment Judgments

Research has examined punishment not only as a behavior of imposing costs on another person but as ordinary people’s recommendations for the legal punishment a transgressor should suffer. Researchers have studied the kinds of information people are sensitive to and the various judgment biases they may fall prey to. On the information side, there is consistent evidence that people assign different degrees of punishment (e.g., length of prison term or monetary fines) as a function of violation severity, intentionality, foreseeability, and justification (Kneer & Machery, Reference Kneer and Machery2019; Robinson & Darley, Reference Robinson and Darley1995), largely consistent with what the law prescribes. In addition, however, studies document that punishment recommendations are susceptible to factors that should arguably not influence legal judgments – for example, the defendant’s ideology (Sood & Darley, Reference Sood and Darley2012), physical or social attractiveness (e.g., Nemeth & Sosis, Reference Nemeth and Sosis1973), stereotypically Black appearance (Eberhardt et al., Reference Eberhardt, Davies, Purdie-Vaughns and Johnson2006), or the juror’s prejudice (Gamblin et al., Reference Gamblin, Kehn, Vanderzanden, Ruthig, Jones and Long2021).

But while several studies do confirm biased recommendations for degrees of punishment, numerous other studies failed to find similar biases in people’s verdicts – that is, judgments of whether a person is guilty or not (Block, Reference Block1991; Stewart, Reference Stewart1985). Why might people be more calibrated in their judgments of guilt than in their punishment recommendations? First, punishment and sentencing recommendations are inherently more discretionary, comparative, and subjective than are guilty verdicts (Tata, Reference Tata1997), creating more room for errors and biases. Second, at least in the United States, people don’t normally make punishment recommendations – as jurors, their role is to come to a verdict, whereas judges decide on punishment. Thus, research that measures legal punishment recommendations may ask people for unusual, unfamiliar judgments, which may be more susceptible to biasing information. Third, previous studies often treated past transgressions, character, and motive as “biasing” factors when in fact they often are legally admissible for sentencing, if not for guilty verdicts (Guglielmo, Reference Guglielmo2015); so by and large, people’s judgments are consistent with the law.

In no way does this imply that verdicts by research participants or real jurors are generally unbiased. One of the most enduring biases in the United States criminal system is racial discrimination, which exerts its impact through police work, prosecution decisions, pretrial detention, guilty pleas, testimony, all the way to verdicts and sentencing (Kansal, Reference Kansal2005; Mitchell et al., Reference Mitchell, Haw, Pfeifer and Meissner2005; Sutton, Reference Sutton2013). Research is also growing to document similar biases toward other marginalized groups (e.g., Mirabito & Lecci, Reference Mirabito and Lecci2021). Such discrimination replays the hierarchical dominance patterns that have tilted punishment systems since early human settlement.

15.3.6 Understanding Blame

We have seen that people rarely mete out punishment in everyday life, and punishment recommendations occur more often in research studies than in the courtroom. By contrast, blame judgments occur frequently and naturally in everyday life. Interestingly, however, blame does not have the best of reputations in psychological research. Acts of blaming are often portrayed as distorted and dysfunctional. Some philosophical work, by contrast, has characterized blaming as more constructive moral criticism that calls out unacceptable behavior, affirms norms and values, and demands justice (Bell, Reference Bell, Coates and Tognazzini2012; Ciurria, Reference Ciurria2019). Research on “complaints” in sociology and pragmatics further shows the important function that criticism plays in regulating interpersonal behavior and the detailed work that goes into staging and warranting this kind of criticism (Morris, Reference Morris1988). Finally, recent research highlights the sophisticated information processing that seems to underlie blame judgments (Cushman, Reference Cushman2008; Monroe & Malle, Reference Monroe and Malle2019), at least when the community enforces standards of evidence (Malle et al., Reference Malle, Guglielmo, Voiklis and Monroe2022).

I now take a closer look at these opposing facets of blame, dividing the review into cognitive blame (judgments in the head) and social blame (acts of moral criticism).

15.3.6.1 Blame as Judgment

An Information Processing Model of Blame. A substantial literature shows that blame assigned to a person varies with a number of factors: the specific norm that was violated, the causal contributions the person makes, whether the person acted intentionally, what mental states or reasons the person had and how justified they were, and whether the person could have and should have prevented the norm violation (Cushman, Reference Cushman2008; Guglielmo & Malle, Reference Guglielmo and Malle2017; Monroe & Malle, Reference Monroe and Malle2019; Nadler & McDonnell, Reference Nadler and McDonnell2012).

My colleagues and I have aimed to integrate these findings in the Path Model of Blame (Malle et al., Reference Malle, Guglielmo and Monroe2014), depicted in Figure 15.1. The model a) specifies the various pieces of information that people normally process en route to blame; b) identifies a systematic order of processing – such that, for example, a question of intentionality comes up only when a causal contribution has been perceived (Guglielmo & Malle, Reference Guglielmo and Malle2017); and c) predicts patterns of blame updates when new information enters the processing stream (Monroe & Malle, Reference Monroe and Malle2019). A unique feature of the model is that intentionality does not merely influence blame judgments, like in other models (Alicke, Reference Alicke2000; Cushman, Reference Cushman2008), but it bifurcates the processing into distinct search paths (Monroe & Malle, Reference Monroe and Malle2017): either for the transgressor’s reasons that motivated and may justify an intentional violation; or for evidence that the transgressor could have prevented an unintentional violation.

Figure 15.1 The Path Model of Blame, adapted from Figure 2 of Malle et al. (Reference Malle, Guglielmo and Monroe2014). People form blame judgments by processing the depicted information components, typically in the four phases indicated. The fourth phase differs depending on whether the event is considered intentional (4i) or unintentional (4u).

The Path Model provides an orienting framework that allows us to examine a number of important questions about blame: to what extent it is motivationally biased, whether it is “intuitive,” and how it relates to emotions.

How Much Does Blame Suffer from Motivated Bias? All human information processing, moral judgment included, is imperfect: People have attentional lapses, forget information, or fail to integrate information under time pressure. Of greater concern in the moral psychology literature have been questions of motivated bias (Alicke, Reference Alicke2000; Ditto, Reference Ditto, Bayne and Fernández2009; Nadler & McDonnell, Reference Nadler and McDonnell2012).

Motivated moral judgment is usually described as a “desire to blame” (Alicke, Reference Alicke2000; Ames & Fiske, Reference Ames and Fiske2013), followed by information processing that strives to confirm the initial blame. However, when appropriate control conditions are implemented, people appear equally responsive to exacerbating information as to mitigating information (e.g., Monroe & Malle, Reference Monroe and Malle2019; Nadler & McDonnell, Reference Nadler and McDonnell2012, Experiment 2). Even without a general desire to blame, people might still show confirmation bias. Yet making an initial blame judgment and later revising it does not come with a confirmation bias (Monroe & Malle, Reference Monroe and Malle2019). Such a bias typically requires a more powerful source, such as the blamer’s relationship with the transgressor (Forbes & Stellar, Reference Forbes and Stellar2022), the blamer’s ideology (Niemi & Young, Reference Niemi and Young2016), or salient “extra-evidential” information (Alicke, Reference Alicke2000) – that is, information that one should not take into account when forming blame judgments.

The normative standards that determine what is extra-evidential, however, have been debated. Some scholars proposed that, even for everyday moral judgments, the standards should be set by “philosophers, legal theorists and psychologists” (Alicke, Reference Alicke2008, p. 179). Unfortunately, these experts do not necessarily agree on the pertinent normative standards, and such standards vary across cultures and historical times. So perhaps it is best to sidestep disputes over the normative standards themselves and ask, descriptively, which factors people take into account when forming blame judgments and how those factors influence blame. I will call these factors “exogenous” because they go beyond the canonical “endogenous” information components of blame, as assembled in the Path Model of Blame.

The current body of research appears consistent with the interpretation that, even when making blame judgments that some have called biased, people follow the information processing path depicted in Figure 15.1, though exogenous factors can alter steps in this path (Guglielmo, Reference Guglielmo2015). I briefly review research on two groups of exogenous factors.

As a first group, norms and values provide the standard against which people evaluate the initial violation (component 1 in Figure 15.1), which shapes the degree of downstream blame. In Sood and Darley (Reference Sood and Darley2012, Study 3), participants disapproved of a man who went to the supermarket in the nude, and they did even more so when he handed out flyers promoting a position on abortion opposite to their own. Participants’ blame (and recommended punishment) was greater presumably because they saw more or stronger norms violated. Likewise, people seem to apply different norms to intimate others than to strangers, finding strangers’ transgressions less acceptable than intimates’ (Weiss & Burgmer, Reference Weiss and Burgmer2021).

Norms also influence what people accept as justifications for a given intentional norm violation (component 4i in Figure 15.1). For example, the Southern and Northern US tend to differ in how justified they find violent acts in service of defending one’s honor after an insult (Cohen & Nisbett, Reference Cohen and Nisbett1994); hence these groups will differ considerably in how much they blame such violent acts.

Finally, people’s norms guide the obligations people impose on a person to prevent certain unintentional violations. For example, people with hierarchy-legitimizing ideologies believe that someone with implicit racial bias is not obligated to prevent their own discriminatory acts that follow from this bias; as a result of this normative belief about prevention, they blame a person less who unintentionally committed such discrimination (Redford & Ratliff, Reference Redford and Ratliff2016). Conversely, in liberal communities, people are obligated to avert microaggressions (Princing, Reference Princing2019) whereas others criticize this obligation as “ridiculous” (Dickey, Reference Dickey2019). In all these cases, blame varies considerably across groups of people, not because of motivated bias but because they interpret the same events in light of different norms.

Beliefs and attitudes toward the transgressor constitute the second group of exogenous factors. Several studies have shown that judgments of causality or preventability can be affected by negative character information (Alicke & Zell, Reference Alicke and Zell2009) or the perceiver’s attitude (Niemi & Young, Reference Niemi and Young2016). Moreover, character, past behavior, and stereotypes can affect people’s inferences of a transgressor’s intentionality or reasons for acting (Sood, Reference Sood2019). So while there is good evidence that exogenous beliefs and attitudes can insert themselves into the blame processing path, some authors have proposed that these exogenous factors directly influence blame (Alicke & Zell, Reference Alicke and Zell2009; Ciurria, Reference Ciurria2019, chapter 4), effectively bypassing the endogenous blame processing. The current evidence, however, does not provide convincing support for such direct effects (Guglielmo, Reference Guglielmo2015). Exogenous and endogenous factors would have to be properly compared in one and the same study, but they are rarely jointly measured (Dukes & Gaither, Reference Dukes and Gaither2017; Nadler & McDonnell, Reference Nadler and McDonnell2012); or, if they are measured, not all are included in model tests (e.g., Alicke & Zell, Reference Alicke and Zell2009; Nadler, Reference Nadler2012); or, if included, they do not yield the predicted direct effects (Mazzocco et al., Reference Mazzocco, Alicke and Davis2004).

There is no doubt that people sometimes blame unfairly, to protect their self-esteem, enact their prejudice, or protect their group. Extant evidence, however, suggests that even in these cases people rely on the same information components – causality, intentionality, reasons, and so on – on which they rely when blaming fairly.

Is Blame Intuitive? When scholars characterize moral judgments as intuitions, they rarely specify the type of moral judgment they are referring to. Some describe moral intuitions as feelings of “approval or disapproval” (Haidt, Reference Haidt2001, p. 818) or as some action being “bad” (Clark & Winegard, Reference Clark and Winegard2019, p. 14), which we may call evaluations (Malle, Reference Malle2021). To be moral evaluations, these intuitions rely on the relevant moral norms being activated in the observer, who then judges the relevant behavior as violating the norm. Evaluations incorporate explanations for simple causality (De Freitas & Alvarez, Reference De Freitas and Alvarez2018) and potentially about the intentionality at least of visibly performed behaviors (Decety & Cacioppo, Reference Decety and Cacioppo2012). By contrast, blame judgments also incorporate information about the intentionality of unobserved actions, the agent’s specific, potentially justifying reasons, and the complex assessment of preventability obligations and capacities (Monroe & Malle, Reference Monroe and Malle2019) – the kind of processing that typically does not fall under “intuitions.”

Suggestions that moral judgments are intuitions typically go hand in hand with the claim that people are “dumbfounded” when trying to justify their moral judgments (Haidt & Bjorklund, Reference Haidt and Bjorklund2008, p. 197). This claim faces a number of challenges. First, researchers routinely ask participants for simple wrongness judgments, which rely primarily on citing the relevant violated norm (Malle, Reference Malle2021). Proponents of the dumbfounding hypothesis treat a norm statement (e.g., “because it’s incest”) as a dumbfounding response, whereas others insist that a norm statement constitutes an actual justifying reason (Stanley et al., Reference Stanley, Yin and Sinnott-Armstrong2019). Second, proponents of the dumbfounding hypothesis instruct their experimenters to explicitly “undermine whatever reason the participant put forth in support of his/her judgment or action” (Haidt et al., Reference Haidt, Bjorklund and Murphy2000, p. 7), which arguably biases results in favor of the hypothesis (Gray et al., Reference Gray, Schein and Ward2014; Royzman et al., Reference Royzman, Kim and Leeman2015). Despite these favorable conditions, surprisingly few people give dumbfounding responses (e.g., 32 percent in McHugh et al., Reference McHugh, McGann, Igou and Kinsella2017; see Malle, Reference Malle2021, p. 309); and in less coerced situations, dumbfounding drops below 20 percent (McHugh et al., Reference McHugh, McGann, Igou and Kinsella2020). These weak dumbfounding rates can hardly support the claim that moral judgments are by nature “intuitive.”

Is there any dumbfounding evidence for blame judgments? Quite the contrary. Bucciarelli et al. (Reference Bucciarelli, Khemlani and Johnson-Laird2008, Study 3) showed that people had no trouble explicating their blame judgments in a think-aloud protocol, and Voiklis et al. (Reference Voiklis, Kim, Cusimano and Malle2016) documented that people provide rich and systematic explanations of their blame judgments. Importantly, these explanations referred to just the types of canonical information known to cause variations in blame judgments (e.g., the seriousness of the norm violation, intentionality, justified reasons).

How Does Blame Relate to Emotion? Determining the role of emotion in blame requires heeding distinctions between affect, evaluations, and emotions. Affect is often understood as a nonrepresentational valenced feeling state (Neumann et al., Reference Neumann, Seibt and Strack2001); evaluations are rapid valenced appraisals of some object (“this is bad”); and emotions are a large class of states (e.g., anger, resentment, or disgust) that are differentiated by appraisals – the cognitive processing of rich arrays of information (Scherer, Reference Scherer2013). Further, we must separate the different roles that these three phenomena can play in blame: They could fully constitute blame judgments, cause them, accompany (and possibly amplify) them, or be caused by them (Monin et al., Reference Monin, Pizarro and Beer2007; Strohminger, Reference Strohminger, Price and Walle2017).

Affect cannot constitute or cause blame because it lacks an object; but it can accompany blame. Evaluations of norm violating events lie at the beginning of any processing stream toward blame, so they make a causal contribution, but they do not process much of the information that blame judgments are normally based on and are therefore insufficient to fully cause, let alone constitute blame. Emotions such as anger rely on a number of appraisals (Ellsworth & Scherer, Reference Ellsworth, Scherer, Davidson, Scherer and Goldsmith2003) of just the kind of information that blame judgments respond to (e.g., causality, intentionality), so the two may co-emerge from this information processing. However, one can arrive at blame judgments by processing the relevant information without feeling angry, whereas it would be difficult to become angry before processing this kind of information (e.g., causality, intentionality). Emotions thus do not seem to constitute moral judgments but may accompany and amplify them.

Researchers have attempted to induce anger to test whether it indeed amplifies blame or related moral judgments. Although some studies showed such impact (Ask & Pina, Reference Ask and Pina2011; Lerner et al., Reference Lerner, Goldberg and Tetlock1998; Seidel & Prinz, Reference Seidel and Prinz2013), others did not (Gamez-Djokic & Molden, Reference Gamez-Djokic and Molden2016; Gawronski et al., Reference Gawronski, Conway, Armstrong, Friesdorf and Hütter2018). Conversely, judgments of blame and responsibility can alter emotions or mediate between norm violations and emotions (Quigley & Tedeschi, Reference Quigley and Tedeschi1996; Zajenkowska et al., Reference Zajenkowska, Prusik, Jasielska and Szulawski2021). Studies also found expressions of anger and moral outrage to arise after judgments of moral violations (Sasse et al., Reference Sasse, Halmburger and Baumert2020), and response times of expressing anger seem to be no faster or even slower than response times to blame judgments (Cusimano et al., Reference Cusimano, Thapa, Malle, Gunzelmann, Howes, Tenbrink and Davelaar2017).

We can conclude that when people encounter a norm violation, they are likely to morally evaluate it, they may or may not experience feelings along with it, and they routinely process a canonical set of information. This information processing guides blame judgments and may generate emotions, but which specific emotions arise (e.g., anger, resentment, indignation) will depend on the specific processed information. These emotions then help scale the intensity of a socially communicated blame judgment (Drew, Reference Drew1998). And this brings us to the phenomenon of blame as communicated moral criticism.

16.1.1 A Norm-Governed Form of Life

Human beings are distinctively normative creatures, investing in attitudes and behavior that are regulated, and so shaped, by a wide variety of socially developed and endorsed norms. By “norms,” we mean rules, principles, and standards (whether formal or informal) that govern what counts as appropriate or proper behavior in the various activities in which humans engage. This covers a lot of ground, much of it outside the moral domain. As many cognitive theorists are now beginning to stress, humans are a thoroughly enculturated species, developing many unique and sophisticated cognitive/behavioral capacities via the internalization of norm-governed social practices that are developed and passed down through successive generations. (For recent representative discussion, see Hutchins, Reference Hutchins2014; McGeer, Reference McGeer2020; Sterelny, Reference Sterelny2012; see also Chapter 4, this volume.)

To give just a taste of such normative enculturation: we learn to speak, read, and write a public language; we learn to enumerate and calculate an endless variety of things using a standardized number system; we learn to regulate our beliefs and desires in accord with communally endorsed evidential, practical, and inferential “rules of reason”; we learn to organize our days in accord with time- and calendar-keeping practices; we learn to recognize and manipulate a broad range of task-specific tools; we learn to greet each other, stand at an appropriate social distance from one another, manage turn-taking with one another in a huge variety of contexts, and so on in staggering detail. In short, we learn to converse together, work together, play together, eat together, and in general navigate complex built environments in relatively smooth, expectable, and cooperative ways by internalizing a remarkable variety of explicit and implicit norms and meta-norms (i.e., norms for resolving normative disputes, or norms for setting new norms, etc.).

Such norms may be shared across cultures or be deeply culturally inflected; they may be specific to particular tasks and/or types of agents occupying variously defined social roles; they may be formally codified or informally and implicitly adopted; they may be policed with rigor or with relative nonchalance, inviting greater plasticity and individual maneuverability; they may be central to our sense of social identity or relatively peripheral to that identity. But regardless of their variation along these different dimensions, norms are, for human beings, as ubiquitous and necessary as the air we breathe. We only become cognitively sophisticated, mutually recognizable and reliable agents, able to engage in complex and coordinated activities with one another, because of our unique capacity to generate and internalize the rich and intricate set of norm-governed practices that pervade our existence.

This richly elaborated norm-governed form of life implicates a distinctive kind of psychology. Human beings are not only sensitive to the presence of norms, acquiring them readily via environmental cues and feedback from others, they are motivated to comply with them and police transgressions against them in a variety of ways (e.g., by way of simply pointing out “that’s not how it’s done”).

A key indication of this sensitivity to norms can be found in the special qualities of human imitation. For instance, unlike other social animals (including especially other primates), children from the age of 24 months begin to engage in a phenomenon called “over-imitation”: faithfully copying others’ behavior even when that behavior is instrumentally unnecessary for achieving a goal. For instance, children continue a learned routine of tapping on the top of a transparent box before pulling a lever that releases a treat, even though it is visually manifest that pulling the lever is sufficient for getting the treat (Horner & Whiten, Reference Horner and Whiten2005).

While there is more flexibility in this early emerging behavior than originally supposed, research now shows that so-called over-imitation is significantly driven by imitators interpreting the modeled action sequence in explicitly normative terms: That is, imitators see the model as demonstrating the “right” or “proper” way to engage in the activity, regardless of the instrumental value of the actions undertaken (Hoehl et al., Reference Hoehl, Keupp, Schleihauf, McGuigan, Buttelmann and Whiten2019; Keupp et al., Reference Keupp, Behne and Rakoczy2013).Footnote ¹ Indeed, children seem biased toward a normative interpretation of others’ behavior (at least in many contexts, especially involving adult–child interactions), indicated not only by their own proclivity for over-imitation but also, more significantly, by the spontaneous protests and criticisms they make against those who engage in an activity without faithfully replicating what the model has done (“not like that!”) and by their own instructive modeling of the activity to others (Kenward, Reference Kenward2012).

Somewhat older children (4–5 years of age) will take account of moral costs in their imitative behavior. For instance, in contexts that would normally elicit normatively driven “over-imitation,” they are reluctant to reproduce a model’s actions, and object to others who do, if such actions entail the destruction of a valuable object belonging to the experimenter (Keupp et al., Reference Keupp, Bancken, Schillmöller, Rakoczy and Behne2016). These children show by their spontaneous protests and expression of moral concern (“she will be very sad if you do that”) that not all normative considerations are treated equally. This jibes with findings by Turiel and others that preschoolers are sensitive to a “moral/conventional” distinction, with this sensitivity emerging from about the age of 3 (Nucci, Reference Nucci2001; Smetana, Reference Smetana and Bennett1993; Turiel, Reference Turiel1983). Since children begin to engage in normatively driven “over-imitation” somewhat earlier (by the age of 2), this may indicate a more general sense of “normativity” that gets refined with experience and/or the development of linguistic competence. Findings suggest that language comprehension is related to an earlier emergence of this distinction (Smetana & Braeges, Reference Smetana and Braeges1990), possibly indicating that sensitivity to different kinds of norms is communicatively bootstrapped. Other lines of research highlight the importance of communication in supporting children’s development of myriad cognitive, social, and emotion-regulation capacities (Király et al, Reference Király, Csibra and Gergely2013).

16.1.2 Communicative Interaction of a Distinctively Human Kind Facilitates Normative Enculturation

Our readiness to detect norm-governed behavior in those around us, to engage in such behavior ourselves, and to police others when they fail to do likewise is a deep-seated feature of human psychology. But such normative proclivities are not an inflexible response to the world; rather, they constitute a predisposition or bias that can be triggered or inhibited by a variety of contextual cues, some of which may be discerned simply by observing the activities of others (Hoehl et al., Reference Hoehl, Keupp, Schleihauf, McGuigan, Buttelmann and Whiten2019). However, people generally do not rely on subtle observational cues to discern whether others’ behavior is norm-governed or not. Real-world interactions are typically rife with communicative cues, both linguistic and nonlinguistic, that explicitly mark how we are supposed to interpret what others are doing. And children are sensitive to such cues even from the age of 14 months, understanding the pedagogic intent behind a model’s behavior and revealing in their imitative responses a readiness to learn just what the model communicatively intends that they should (Gergely & Király, Reference Gergely, Király, Dukes and Clément2019; Király et al., Reference Király, Csibra and Gergely2013).

This turns our attention to the special qualities of human communication that make it particularly apt for facilitating normative enculturation. While the use of language itself is a uniquely human phenomenon, theorists have lately drawn attention to more fundamental features of human communication that distinguish it from the communicative practices of other animals. These are not obvious. For instance, in addition to vocalizing and displaying a range of relatively inflexible signaling behaviors in response to environmental cues, chimpanzees regularly gesture to one another to solicit nursing, food, or grooming in ways that are learned, flexible, and even idiosyncratic (Goodall, Reference Goodall1986). Such gestures are intentional acts of communication (Grice, Reference Grice1969; Moore, Reference Moore2016): They are directed toward particular others when, and only when, those others are attending and are appropriate targets of a particular solicitation; a given gesture may also be modified to suit the situation (Kaminski et al., Reference Kaminski, Call and Tomasello2004; Tomasello & Call, Reference Tomasello and Call2019; Yamamoto et al., Reference Yamamoto, Humle and Tanaka2012).

Still, while these are sophisticated communicative acts that depend on certain mentalizing skills (for instance, monitoring attention and likely even understanding what others can and cannot see), chimpanzees and other great apes will only use gestures in a proto-imperative way: to get others to do specific concrete things as a way of realizing their own goals.Footnote ² They do not use gestures – in particular, the pointing gesture – in a proto-declarative way: to show others something or simply to share attention for the pleasure of it (Tomasello & Call, Reference Tomasello and Call2019).

This makes for a puzzling difference in the communicative behavior of great apes in contrast with humans, even at the nonlinguistic gestural level. In human interactions, the pointing gesture is naturally used and generally understood despite serving a variety of functions – proto-imperative (requesting something), proto-declarative (showing something), proto-informative (e.g., indicating the location of a hidden object), and proto-interrogative (e.g., soliciting more information about an indicated object). Such multifunctional use and understanding emerges in children as young as 12–14 months (Southgate et al., Reference Southgate, Van Maanen and Csibra2007; Tomasello et al., Reference Tomasello, Carpenter and Liszkowski2007) – and this despite the fact that the (prelinguistic) mentalizing skills of children at this stage of development are, in many respects, not hugely more sophisticated than those of chimpanzees and bonobos. So why this difference?

Drawing now on decades of comparative primate research, Tomasello and colleagues argue that a key determinant of this difference lies in one “small” cognitive and motivational factor that ramifies in powerful ways (Tomasello et al., Reference Tomasello, Carpenter, Call, Behne and Moll2005). Humans are distinctive insofar as their social activities are permeated by an irreducibly collective (shared, or “we”) intentionality (Gilbert, Reference Gilbert1994; Rakoczy & Tomasello, Reference Rakoczy, Tomasello and Tsohatzidis2007). When people interact with one another, they are not simply solo operators, aiming to engineer another’s behavior simply to forward their singular aims. Rather, humans often take themselves to be doing things together, naturally forming shared intentions to act in coordinated and mutually supportive ways in the context of joint activities: we are playing a game together; we are finding a hidden object together; we are looking at something together. And from here it seems a short step to structuring and understanding such activities in explicitly normative terms. Hence, the capacity for shared or collective intentionality, emerging early in development and potentially unique to the human species, undoubtedly plays a foundational role in explaining people’s distinctive proclivity to adopt norm-governed attitudes and behavior.

The capacity for shared intentionality thus heralds a fundamental change in the nature of human communication. The transformation is from an activity that is essentially behavior-focused and unidirectional (seen, for instance, in great apes) to an activity that is essentially attitude-focused and dialogical. Whether people communicate in language or in a variety of nonlinguistic ways (using ostensive eye contact, emotional expressions, prosody, bodily gestures and/or other symbolic acts), they invariably regard one another as potential candidates for shared intentionality, presupposing and reinforcing the joint activities and/or shared frames of reference that give meaning to their particular communicative acts (see, e.g., Tomasello et al., Reference Tomasello, Carpenter, Call, Behne and Moll2005). Hence, in communicating with others, people are not just aiming to shape or regulate others’ behavioral responses as a way of realizing their individual goals. People are aiming to make others understand the point of their communicative behavior – whether it be asking for something, sharing attention, sharing pertinent information, committing to some course of action, and so on in wide variety. Hence, in communicating with others, people invariably solicit a suitable communicative response from their interlocutors in turn – a reaction that makes sense in light of the shared understanding and collaborative support people aim to achieve. This is the sense in which human communication is essentially dialogical.

In the context of shared norm-governed activities, we see a special instance of this insofar as people’s communicative efforts are naturally bent toward signifying, in both word and deed, that there are proper ways of behaving that they expect others to understand and internalize. This message may be conveyed ex ante by instruction and modeling or ex post by protest and correction. But either way, people’s aim is not simply to elicit a suitable behavior response from the other. Qua communicators, people are dialogically focused on soliciting an appropriate attitudinal response – namely, others’ appreciation and acceptance that “this is the way we do things around here.”

16.1.3 Moralized Norms and Communication

Against this rich background of communicatively supported normative enculturation, we now pose the following two questions: What is distinctive about the norms people identify as “moral” – and/or attitudes and behavior governed by such norms? And is there anything distinctive about communication regarding such norms?

As a preliminary matter, we note there is a persisting tendency within the moral-psychological literature to treat norms as essentially dividing into two types, “moral” and “conventional” (Nucci & Nucci, Reference Nucci and Nucci1982; Smetana & Braeges, Reference Smetana and Braeges1990; Turiel, Reference Turiel1983). But, as the foregoing discussion makes clear, people are variously invested in a rich variety of norms as a matter of social and practical identity, and many of these norms defy easy binary categorization: Either they have elements of both the “moral” and the “conventional” (such as norms of “fair play”), or they belong to a category of their own (such as “norms of rationality”). This naturally complicates the search for criteria by which moralized norms can generally be distinguished from the rest. But, as our discussion in this section will show, it also places new emphasis on the nature of our communicative interactions in tracking this distinction.

In their quest to elucidate the distinctive nature of moralized norms, theorists have generally pursued two main lines of inquiry. The first considers the substantive content of the norms themselves. The second considers the psychological attitudes people take toward such norms, signaling a distinctive kind of normative commitment taken toward the content in question, whatever it may be. We discuss each of these approaches in turn, noting that a satisfying answer to the question may well appeal in some measure to both content and attitude.

With regard to the substantive content of moralized norms, theorists generally agree that this is something of a moving target. In any given society, ordinary people may make an intuitive distinction between moral norms and other kinds of rules, principles, conventions, or standards; but there are invariably borderline areas of disagreement. This problem is only compounded across different cultures; some communities operate with a relatively expanded conception of moral norms, others with a relatively contracted conception (e.g., as regards matters of food, clothing, or sexual preference). Yet even amid this variability, certain “themes” are discernible in the norms generally identified as moral. For instance, as Sripada and Stich (Reference Sripada, Stich, Carruthers, Laurence and Stich2006, p. 283) observe, most societies have “rules that prohibit killing, physical assault, and incest”; “rules promoting sharing, reciprocating, and helping”; “rules regulating sexual behavior among various members of society”; and “at least some rules that promote egalitarianism and social equality.”

Is there any underlying unity discernible in this motley collection of moralized norms? This is a matter of ongoing debate. At one extreme, reductive monists aim to show that all moralized norms are grounded in a dominant substantive concern. Accounts differ as to the nature of this concern: For instance, some argue that moralized norms are all concerned with prohibiting “harm” (Gray et al., Reference Gray, Waytz and Young2012); others, that they are concerned with prohibiting “unfairness/injustice” (Baumard et al., Reference Baumard, André and Sperber2013; Piazza et al., Reference Piazza, Sousa, Rottman and Syropoulos2019). But this has been a challenging position to defend, especially in a cross-cultural context (see for instance Berniūnas et al., Reference Berniūnas, Dranseika and Sousa2016).

A less extreme form of monism discerns some unity among moralized norms at a higher level of generality – for example, they are specifically concerned with placing a check on selfish goals or interests where such interests may threaten those of the community (Fehr & Fischbacher, Reference Fehr and Fischbacher2004). Undoubtedly this is true of many such norms; but, again, others seem to escape this characterization – at least in any direct sense. For instance, it is hard to see how moralized norms regarding treatment of the dead or eating certain kinds of food are specifically geared toward checking selfish goals or interests, though they may serve certain communal goals or interests (such as stabilizing a sense of group identity). Yet not all moralized norms are apt for serving communal goals or interests either, at least on a broad scale: For instance, many societies endorse “agent-relative” norms that permit (or even obligate) individuals to prioritize the needs of close relations over the community itself or other members within it, at least to some degree.

In light of these difficulties, other theorists take an avowedly pluralist approach to the content of moralized norms, arguing in addition that there is significant cross-cultural variation in the kind of “disinterested” concerns individuals are required to prioritize in their attitudes and behavior. Such moralized norms might prioritize a range of concerns for particular others (families, friends, strangers in need…), a range of concerns for maintaining the community as a whole (including its structure and operations), and even a range of concerns that relate to respecting or implementing some transcendentally determined cosmic order (such as “God’s will” or “Dhamma”) (McGeer, Reference McGeer and Sinnott-Armstrong2008; Shweder et al., Reference Shweder, Much, Mahapatra, Park, Brandt and Rozin1997). This pluralistic approach leads naturally to a different sort of question – namely, why human beings should have such a range of disinterested concerns, however much these require development or stabilization via communally endorsed norms. This is a challenging explanatory project in itself, perhaps explaining the continuing allure of monism for many theorists. It may also recommend taking a different approach to characterizing the distinctiveness of moral norms altogether – one that focuses less on their content and more on the psychological attitudes people take toward them (Nucci, Reference Nucci2001; Skitka et al., Reference Skitka, Hanson, Morgan and Wisneski2021).

In keeping with this second approach, there are a range of attitudes that theorists have identified as seemingly peculiar to moralized norms. Following Sripada and Stich (Reference Sripada, Stich, Carruthers, Laurence and Stich2006), we group these attitudes under three headings: motivation for compliance, the metaphysical status attributed to the norms themselves, and reaction to transgressions (cf. Skitka et al., Reference Skitka, Hanson, Morgan and Wisneski2021). To give Sripada and Stich’s summary characterization of the attitudes under each heading: 1) people find moralized norms “intrinsically motivating”; 2) they view them as having “independent normativity”; and 3) they react to their violation in a distinctively punitive way, characteristically involving a range of equally distinctive “reactive” emotions (condemnatory anger, resentment, indignation, guilt, shame, and remorse – depending, of course, on whether the transgressor is other or self) (Strawson, Reference Strawson1962). We discuss each of these features in turn.

To say that people find moralized norms (1) “intrinsically motivating” is to say they experience norm compliance as an end in itself. They do not comply with such norms simply for instrumental reasons – because of other benefits they may gain through norm compliance (e.g., reputational benefits, reciprocity benefits, realization of cooperation-dependent goals) – or, indeed, to avoid certain costs noncompliance may bring (e.g., retaliation, cooperation-dependent goal frustration). As Sripada and Stich (Reference Sripada, Stich, Carruthers, Laurence and Stich2006) emphasize, compliance is its own reward, other potential benefits notwithstanding.

The problem is that this psychological feature does not seem distinctively related to moralized norms. As our earlier discussion of normative enculturation suggests, humans are heavily dependent on developing normatively shaped attitudes and behavior of a more general kind, enabling a huge and sophisticated array of distinctively human capacities (including, in particular, linguistic capacities, but many others as well). Such deep enculturation would surely be impossible without some psychological machinery innately geared toward norm detection and compliance, where people regulate their behavior, not just in imitation of what others do, but in light of their explicit pedagogical and corrective feedback and where discovering the “right way” to do something is rewarding in itself. This, we think, is amply demonstrated in humans’ species-specific proclivity for over-imitation, emerging at least as early, if not prior to, a sensitivity to specifically moralized norms.

Next we consider (2) “independent normativity,” the idea that people view moralized norms as specifying what they ought to do “independently of any legal or social institution or authority.” Such norms are viewed as having the metaphysical status of “objective” truths, implying generality of scope across different times and places (Kelly & Stich, Reference Kelly, Stich, Carruthers, Laurence and Stich2007; Nucci & Nucci, Reference Nucci and Nucci1982; Skitka et al., Reference Skitka, Hanson, Morgan and Wisneski2021; Smetana, Reference Smetana and Bennett1993; Turiel, Reference Turiel1983).Footnote ³ It has been suggested that people are intrinsically motivated to comply with moralized norms because they view them as having independent normativity.Footnote ⁴ Be that as it may, we think these two phenomena are psychologically separable: People can be intrinsically motivated to comply with norms they do not regard as having independent normativity. Still, the question remains: Is independent normativity/objective validity really a distinctive feature of moralized norms? We think not. Rational norms – for instance, valid rules of inference – are generally held to have “independent normativity,” but it would be strange to classify them as moral norms (see Southwood, Reference Southwood2011 for a similar observation).

Finally, we come to the third psychological feature associated with moralized norms: the very strong disposition (3) to react to norm violations of this kind in a distinctively punitive way. Here there is some indication of a genuine difference in people’s attitudes to different sorts of norms. For while all norm violations attract notice, how individuals respond to such violations can vary widely depending on a range of factors, including how they are related to the transgressor, the nature of the norm (moral or nonmoral), the transgressor’s degree of normative understanding, the transgressor’s intention in violating the norm, the consequences of violating the norm, and so on. Such factors may influence whether any norm-enforcing action is called for at all (some norm violations simply call for understanding, or even appreciation, of the novelty so displayed). But even when the transgression calls for some norm-enforcing response, this too can take a variety of material and/or psychological forms, ranging from gently corrective to strongly sanctioning – what some might call “punitive.” However, this use of the term neglects an important dimension of norm-enforcement that is only incidentally related to the severity of sanction imposed – namely, what the sanction in question is meant to express.

We suggest that what makes the punitive reaction distinctive is a blaming attitude, or set of attitudes, that underlies it and that the punitive reaction is characteristically taken to express (resentment, indignation, personal offence, and condemnation) (Strawson, Reference Strawson1962). In the context of state-authorized sanctions, Feinberg (Reference Feinberg1965) argues that we make an intuitive distinction between “punishments” and “penalties” on this ground: Punishment is a legal sanction that expresses the state’s condemnation of both norm-violator and norm-violation, whereas “penalty” is a legal sanction that expresses no such thing; it is meant to operate on (potential) norm-violators as a purely instrumental deterrent.Footnote ⁵ Compare, for instance, the way we might respond to someone who violates a rational norm versus someone who violates (what we take to be) a moral norm. Even supposing the transgressor is at fault in both cases (i.e., we take them to be normatively competent in the relevant domain), the faulty reasoner may attract (at worst) derision or contempt, whereas the faulty moral agent attracts some degree of reprobative blame – that is, blame that characteristically expresses resentment or indignation. Why do these norm violations attract such different responses?

We think a satisfying answer to this question must finally appeal to the substantive nature of the norms themselves: in particular, to the idea that moralized norms generally require us to give due weight to a range of disinterested, mainly other-regarding concerns versus our own idiosyncratic interests or desires (whether “selfish” or not) (Voiklis & Malle, Reference Voiklis, Malle, Gray and Graham2017). To violate these norms is thus to prioritize our own interests and desires in a way that shows disregard or disrespect for a communally endorsed conception of the moral order. And though other kinds of norms are communally endorsed, violating them is not seen to be essentially self-prioritizing in the manner of a moral norm violation. In short, as moral philosophers are wont to emphasize, people view (normatively competent) violators of moral norms as expressing a distinctive attitude in their transgressive behavior – an attitude of disrespect, disregard, or even “ill will” toward others in the moral community, or to the community at large (Strawson, Reference Strawson1962). And it is this expressed attitude that calls for a distinctive response: one that communicates condemnation of the transgressor’s attitude, in addition to whatever (psychological or material) cost is imposed for the transgressive act itself.

But is the mere expression of condemnation sufficient to make punishment a genuinely communicative act? As we have argued, human communication is essentially dialogical in nature. Hence, if punitive responses to moral transgressions are genuinely communicative, they should be psychologically and/or normatively linked with an expectation of (perhaps even demand for) a suitable response from the target audience – for instance, from transgressors themselves, who as communicative targets of punishment are thereby called upon to regret and renounce their immoral acts and attitudes. A communicative view of punishment thus entails that punishment has a certain forward-looking drive or rationale, one that ultimately seeks the ratification of shared moral norms from its target audience. As we have noted, the target audience may – and perhaps should – include transgressors themselves; but as a communal or even state-authorized act, it is likely to include others as well, in whose name the punishment is implicitly or explicitly imposed (Feinberg, Reference Feinberg1965).

A communicative view of punishment has been explored and defended in the philosophical literature, most notably by Anthony Duff (Reference Duff2001) (for a sample of views, see also Bennett, Reference Bennett2008; Braithwaite & Pettit, Reference Braithwaite and Pettit1990; Feinberg, Reference Feinberg1965; Lacey, Reference Lacey1988; von Hirsch, Reference von Hirsch1993).Footnote ⁶ While this work focuses on seeking an adequate justification for the practice of punishment, our concern in the remainder of this chapter takes a more empirical direction. In what follows, we focus specifically on people’s actual motivations for punishment and how a communicative framework can forward psychological research on this topic.

16.2.1 A Brief History of Punishment Research

In the psychological literature, punishment has been somewhat broadly defined as “intentionally providing another person with negative or unwanted outcomes, as a motivated response to the perception that this person has violated widely shared norms, values, or rules” (van Prooijen, Reference van Prooijen2018, p. 10). (For a narrower treatment of punishment, see Chapter 15, this volume.) Decades of research into people’s punishment behavior indicate that it is motivated by several factors (for recent reviews, see Raihani & Bshary, Reference Raihani and Bshary2019; van Prooijen, Reference van Prooijen2018; Wenzel & Okimoto, Reference Wenzel, Okimoto, Sabbagh and Schmitt2016). Broadly, this research can be grouped into two different kinds. One broad kind comes from behavioral economics and evolutionary psychology and primarily focuses on the ultimate functions of punishment and its adaptive value in the context of evolution. Traditionally, this area of research has operationalized punishment as payoff reductions in the context of incentivized economic games. The most influential explanation for punishment as an adaptive mechanism over the past two decades is that it stimulates cooperation within groups by way of deterring wrongdoers, as well as communicating and reaffirming the group’s norms and values (Fehr & Fischbacher, Reference Fehr and Fischbacher2004). Other functional explanations highlight punishment’s competitive functions and beneficial effects for punishers, for instance, reestablishing their status and relative level of resources (Krasnow et al., Reference Krasnow, Cosmides, Pedersen and Tooby2012; Raihani & Bshary, Reference Raihani and Bshary2019), as well as signaling their trustworthiness to others (Jordan et al., Reference Jordan, Hoffman, Bloom and Rand2016).

The second broad kind of research on punishment comes from social psychology and focuses on the proximate mechanisms of punishment, examining why people punish in a given situation. In line with findings from economics and evolutionary psychology, social psychological research has identified the importance of rebalancing status and power and restoring value consensus (e.g., justice restoration theory, Okimoto & Wenzel, Reference Okimoto and Wenzel2008; Shnabel & Nadler, Reference Shnabel and Nadler2008). Still, over the last few decades this research tradition has been largely shaped by traditional philosophical theories regarding the appropriate justification for punishment. In this normative tradition, deontological justifications, historically associated with Kant, are purely retributive and backward-looking: The offender simply deserves to be punished as a reaction to their wrongdoing. By contrast, a consequentialist approach, historically associated with Bentham, justifies punishment in terms of specific forward-oriented consequences: For instance, it serves as a general deterrent by imposing a cost for criminal behavior and/or serves as a specific deterrent by incapacitating offenders; it may also prevent future crime by changing their behavior (rehabilitation). Numerous psychological studies have investigated to what extent retribution (as the deontological factor) versus deterrence (as the main consequentialist factor) drives people’s punishment-related reactions (for reviews, see van Prooijen, Reference van Prooijen2018; Wenzel & Okimoto, Reference Wenzel, Okimoto, Sabbagh and Schmitt2016), whereas other philosophical justifications (especially focusing on the expressive and communicative dimensions of punishment) have received comparatively little attention.

Results from these studies indicate more support for retribution than deterrence as punishment motive. For instance, in line with retributive theories, studies show that people punish proportionally to the seriousness of a transgression and that people’s punishment recommendations are less affected by factors central to deterrence theories, such as likelihood of detection or transgression frequency (e.g., Carlsmith et al., Reference Carlsmith, Darley and Robinson2002; Carlsmith et al., Reference Carlsmith, Wilson and Gilbert2008). In addition, if people can select which transgression-relevant information they would like to know before they recommend punishment, they select information on transgression seriousness first versus information related to deterrence, such as transgression frequency (Carlsmith, Reference Carlsmith2006; Keller et al., Reference Keller, Oswald, Stucki and Gollwitzer2010). A further set of studies has aimed to isolate retributive motives by studying people’s punitive responses toward animals that have attacked human beings (Goodwin & Benforado, Reference Goodwin and Benforado2015).Footnote ⁷ In line with the retributive view, the researchers found that people rate animals as “deserving” to be killed for their violent attacks; furthermore, people’s support for inflicting pain on animals during their execution increases with the perceived severity of the attacks. In sum, this body of social psychological research has been taken to support the view that people are first and foremost retributive in their punishment behavior, as they mainly care about giving transgressors what they “deserve.”

It is important to stress that most of these findings have been interpreted within a binary choice theoretical framework – retribution versus deterrence, thereby sidelining the potential role of communication as a key motivating factor. Nevertheless, researchers have recognized that communication may still play an important role in people’s punishment behavior. Subsequent work has thus aimed to control for this using carefully designed behavioral experiments. For instance, in so-called hidden punishment experiments, people are purportedly barred from using punishment in an immediately communicative way because the targets of their punishment will not be informed of the punishment they receive (Crockett et al., Reference Crockett, Özdemir and Fehr2014; Nadelhoffer et al., Reference Nadelhoffer, Heshmati, Kaplan and Nichols2013). Yet a small percentage of people still punish under these conditions. Similarly, studies have shown that people punish in repeated interactions, even if their targets will not be informed about whether or how much they have been punished until the very end of the experiment, apparently removing its immediate communicative rationale (Fudenberg & Pathak, Reference Fudenberg and Pathak2010). Many researchers conclude that this argues in favor of a purely, or at least predominantly, retributive punitive psychology. (For a critique of this conclusion, see Chapter 15, this volume, Section 15.3.3.)

We think the cited results do not make a strong case against the idea that punishment has a communicative dimension. In the first place, the putative isolation of retributive motives in controlled laboratory studies does not entail the stronger conclusion that people’s punishment motives are purely retributive. As an analogy, imagine a laboratory study that examines why people wear clothes. Assume that researchers want to study the effect of fashion preferences. If the results showed that people put on clothes of certain colors even if these clothes are hard to wear, it would be valid to conclude that fashion preferences affect people’s choice of clothing. Yet, it would not be a valid conclusion that comfort (or other factors such as warmth) play no role in people’s clothing decisions in everyday life. Indeed, one or more of these factors may remain an important, even dominating motivation in naturalistic conditions. Returning now to the punishment case, even if “hidden punishment” occurs, other factors may continue to play an important, perhaps dominating role, in more naturalistic conditions. Hence, it does not rule out communication as a proximate mechanism for punishment.Footnote ⁸

Second, if communication is an inherent feature of ordinary punishment behavior, studies on hidden punishment may simply demonstrate that people will engage in communicative behaviors even if they know their putative audience cannot hear them. Such behavior is a common occurrence in everyday life. People may show their communicative intent (verbally or otherwise) even when they know their target audience is not “listening” in any demonstrable way. They may do so for many reasons: to express who they are, to get things off their chest, to rehearse what to say if a suitable occasion arises, to solicit (imagined) advice, to (imaginatively) resolve disputes, or simply to seek comfort in the absent person’s (imagined) presence. Of course, “communicative behavior” could be defined in such a way as to require successful uptake in a target audience. But since our goal is to understand the psychological mechanisms motivating specific behaviors, we here take communicative intent to be independent of uptake. Building on our earlier discussion of human communication, the point we make here is that performing actions with communicative intent is likely to be a deeply rooted psychological proclivity in beings that are naturally oriented toward securing shared intentionality with others – a proclivity that is not easily “shut off” or subject to rational control. Thus, even in the context of anonymous single-shot punishment, it is certainly possible that punishers are likewise engaging in communicative behavior, even if potential targets of their communicative acts are not physically present.Footnote ⁹

These considerations point toward the same conclusion: Even if retributive aspects can be isolated experimentally, such findings do not establish that punishment is primarily retributive. Indeed, several research findings suggest that people’s punishment goals are context-dependent, as context changes the salience of punishment-relevant aspects of a given situation (e.g., salience of the crime itself, the offender, or the community; see Gromet & Darley, Reference Gromet and Darley2009; Twardawski et al., Reference Twardawski, Tang and Hilbig2020). For instance, people tend to put more emphasis on the need for deterrence and rehabilitation if the longer-term consequences of punishment are made salient. This already undermines the rationale for a binary choice between “the two” punishment motives (retribution versus deterrence). We think psychological studies would benefit from adopting a research framework that remains open to a potential range of several punishment motives (e.g., Fitness & Peterson, Reference Fitness, Peterson, Forgas and Fitness2008), while perhaps also discerning a unifying thread that may be present in most or all of them.

To this end, we think it helpful to regard punishment as a form of moral communication, capable of conveying a range of potential messages. As we have noted elsewhere (Funk et al., Reference Funk, McGeer and Gollwitzer2014; McGeer & Funk, Reference McGeer and Funk2017), a first advantage to this shift in framework is that it resolves a major ongoing debate between retributive and consequentialist theories of punishment: whether people’s punishment motives are, or should be, primarily backward-looking or primarily forward-looking. Under a communicative paradigm, punishment can be both retributive and instrumental (see also Duff, Reference Duff2001).

A second advantage of the communicative approach is that it frames punishment as a specifically moral response to wrongdoing – in that way, it satisfies retributivists’ normative ambitions; but it does so without relying on the strange moral alchemy that lies at the root of traditional retributivist views, that returning suffering for suffering somehow rectifies a moral wrong. The point of punishment, on the communicative view, is to convey a suitably targeted, expressively powerful, morally loaded message; it is not to inflict suffering as an end in itself (cf. Feinberg, Reference Feinberg1965).

Given these normative advantages, it is not surprising that the communicative view of punishment has received considerable attention and defense in the philosophical community. But, as a descriptive theory regarding people’s actual punishment motives, psychologists have given it surprisingly little attention in the design of their empirical studies over the past two decades. Luckily, times are changing.

16.2.2 Punishment as Communication: An Emerging Trend in Psychological Research

Recently, psychologists have started to study the communicative aspects of punishment experimentally and to differentiate them from retributive components. A first line of findings comes from research on the hedonic effects of punishment. From a retributive perspective, one could expect punishment in itself to be satisfying and restore people’s sense of justice. Yet this does not seem to be the case. In general, findings from social psychological studies on the hedonic consequences of punishment show that there are positive as well as negative effects of punishing (Eadeh et al., Reference Eadeh, Peak and Lambert2017). In particular, punishers are very sensitive to whether punishment sends a message and how that message is received. If punishers do not hear back from wrongdoers after punishment, they ruminate and are dissatisfied (Carlsmith et al., Reference Carlsmith, Wilson and Gilbert2008). Yet when punished wrongdoers indicate that they understand that they are being punished (Gollwitzer et al., Reference Gollwitzer, Meder and Schmitt2011) and especially if wrongdoers indicate a change in their behavior and attitude (Fischer et al., Reference Fischer, Twardawski, Strelan and Gollwitzer2022; Funk et al., Reference Funk, McGeer and Gollwitzer2014), punishers’ sense of justice is restored. Numerous studies have shown that people have less desire to punish offenders who are remorseful (e.g., Corwin et al., Reference Corwin, Cramer, Griffin and Brodsky2012). These findings fit well with the view that punishment is ideally a form of bidirectional (i.e., dialogical) moral communication: To wit, punishers intentionally express something via punishment that punished individuals are meant to understand – punishment carries symbolic meaning; and, likewise, punishers adjust their punishment behavior in reaction to intentional messages they receive back from the wrongdoers (for similar thoughts, see for instance Boon & Yoshimura, Reference Boon and Yoshimura2020; Nahmias & Aharoni, Reference Nahmias, Aharoni and Surprenant2018).

A second line of research explores the necessity of punishment’s materially punitive edge. In support of the communicative view, studies show that people expect harmless but symbolic punishment to be as effective as harmful punishment if its message is clear (Sarin et al., Reference Sarin, Ho, Martin and Cushman2021). Similarly, in real interactions, people accept trade-offs between punishment and other forms of moral communication. For instance, they reduce their punishment if they can also send along a note of disapproval (Xiao & Houser, Reference Xiao and Houser2005), and they opt for the less severe punishment if they can let the wrongdoer know why they are being punished (Molnar et al., Reference Molnar, Chaudhry and Loewenstein2023). Once again, these findings highlight that punishment behavior does not seem to be solely about punitive paybacks and proportionality as a purely retributive account of punishment would suggest.

16.2.3 Toward a Systematic Approach to Moral Communication: Content, Targets, and Means

We turn now to a consideration of how conceptualizing punishment as communication can provide researchers with a scaffold to systematically study and understand how people react to wrongdoing. If the underlying psychological mechanism of the urge to punish is communicative in nature, then three interrelated questions emerge as important to explore in depth: What is being communicated? Who is the target of communication? And how is this communicative content conveyed? Ultimately, a better understanding of these dimensions of punishment will allow theorists to question the efficacy of punishment in achieving its communicative goals and to invite comparison with other potentially more effective forms of moral communication.

The Content of Communication. What people aim to communicate via punishment is closely related to how people have construed the wrongdoing and what it meant to them psychologically. Did people construe it as a violation of their own personal status, for instance, or did they mainly see it as a violation of group norms? Punishment may communicate that someone has been wronged, or it may send a message reaffirming the victim’s standing, their values, or their moral identity (see, e.g., Okimoto & Wenzel, Reference Okimoto and Wenzel2008; Shnabel & Nadler, Reference Shnabel and Nadler2008). It may effectively “say” to others: “I won’t allow anybody to walk all over me” (Crombag et al., Reference Crombag, Rassin and Horselenberg2003), or “I’m a trustworthy person because I condemn such wrongdoing” (Jordan & Rand, Reference Jordan and Rand2020). It may communicate “I’m a member of a particular moral tribe”; “I care about a set of norms and their breaches, and I’m disposed to police the norms in question” (Shoemaker & Vargas, Reference Shoemaker and Vargas2019). It may communicate a desire regarding the future – for example, that the wrongdoer feels remorse and commits to change (Funk et al., Reference Funk, McGeer and Gollwitzer2014); it may even communicate a desire that wrongdoers suffer in light of their wrongdoing. And such desires might be essentially self-regarding (e.g., related to a victim’s need to restore their own standing) or they might be essentially other-regarding (e.g., related to a victim’s desire to prevent the victimization of others). There is a range of possibilities, with varying degrees of moral content. For researchers, it would be interesting to investigate factors that modulate expressed moral content; for instance, which message gets priority in what contexts – and why.

Importantly, once punishment is conceptualized as a nonlinguistic form of moral communication, it has one salient advantage over explicitly linguistic forms of communication: It may carry many symbolic messages simultaneously, either to the same target or to several targets, and potentially each with a different grade of urgency or importance to the punisher.

The Target of Communication. An interesting feature of punishment as communication is that the message may or may not be targeting the recipient of punishment. Obviously, punishment may communicate something to the wrongdoer, as previous research on the hedonic effects of punishment suggests (e.g., Funk et al., Reference Funk, McGeer and Gollwitzer2014; Gollwitzer et al., Reference Gollwitzer, Meder and Schmitt2011). But, as we noted earlier, the message may also be intended for others – for example, for particularly vulnerable members of the community (“You will be supported and protected”) or for the community as a whole (“Let’s all agree that this is a serious moral wrong that should be generally condemned”) (see also Feinberg, Reference Feinberg1965). Notably, the two consequentialist punishment motives, specific prevention and general prevention, can be accommodated, albeit reconfigured, within this communicative framework. While in the traditional picture, punishment is viewed as a relatively crude behavioral-conditioning device, it is here explicitly conceptualized as sending a message to the original wrongdoer or to the public at large that such wrongdoings will not be tolerated. A specific example on using punishment to communicate with various possible targets can be found within the legal system. While offenders are clearly the target of institutionalized punishment, the criminal justice system explicitly aims to communicate the validity of norms to several other targets, including the victims and their families, future potential wrongdoers, as well as potential voters and concerned citizens who are afraid of crime and who are supposed to get the message that wrong deeds are taken seriously (Feinberg, Reference Feinberg1965; Sunstein, Reference Sunstein1996).

The Means of Communication. There are many different forms of punishment and many ways in which punishment can be administered. Sometimes people react to wrongdoings with punishment that is proportional to the severity of the transgression (Carlsmith et al., Reference Carlsmith, Darley and Robinson2002) and may use retaliation as a quick default mode of communication. Within the communicative framework, this core retributive principle can be expressively reconceptualized. “Proportional” punishment may communicate to the community “this is how wrong it was,” or to the offender “this is how much you need to change” (Duff, Reference Duff2001). Still, the communicative framework can also explain why, at other times, a disproportional, more lenient reaction may be observed. In such cases, particular contextual factors may ensure that a more lenient punishment (or other symbolic gesture) is sufficient to convey a particular message effectively (see also Sarin et al., Reference Sarin, Ho, Martin and Cushman2021). For instance, community service is often sufficient in restorative justice settings to assure all the stakeholders that the crime has been taken seriously and suitable reparative measures have been undertaken (Johnstone, Reference Johnstone2013; Rossner & Bruce, Reference Rossner and Bruce2016; Shapland et al., Reference Shapland, Atkinson, Atkinson, Colledge, Dignan, Howes, Johnstone, Robinson and Sorsby2006). Or, to go to another example, a skeptically raised eyebrow may be enough to publicly lower a perpetrator’s status and reputation in the eyes of others. At the other extreme, punishers may decide not to interact again with the wrongdoer at all in order to communicate condemnation of the wrongdoer’s actions and attitudes (to the wrongdoer and/or to the group as a whole), to reaffirm solidarity with whatever group norms the wrongdoer has violated, and, finally, to endorse by way of enacting group norms for dealing with wrongdoers.

16.2.4 Putting the Communicative Framework to Work in Current and Future Research

Application to Current Findings. We here suggest that some otherwise puzzling results in punishment research may be profitably addressed by adopting the communicative framework we recommend – for instance, the mixed profile of findings related to harsher and/or more lenient punishment in intra- versus intergroup settings. On the one hand, studies show that deviant in-group transgressors are evaluated more negatively than deviant out-group transgressors (the so-called black-sheep effect; Marques et al., Reference Marques, Yzerbyt and Leyens1988). In addition, there is an independent effect of out-group leniency where punishers adjust their punishment downward if the transgressor belongs to a group that is of lower status than their own group (Braun & Gollwitzer, Reference Braun and Gollwitzer2012). On the other hand, there are also instances of in-group leniency (“in-group favoritism”), that is, in-group transgressors are punished less harshly than out-group transgressors (see, e.g., Sommers & Ellsworth, Reference Sommers and Ellsworth2000, or, for research findings from developmental psychology with children as participants, see Jordan et al., Reference Jordan, McAuliffe and Warneken2014). These apparently contradictory results may possibly be disentangled by way of identifying who the primary communicative target of punishment is meant to be (along with identifying the exact message it is meant to convey). For instance, is the primary communicative target the actual wrongdoer, in which case in-group leniency might be preferred if the message is regarded as effectively conveyed in a nonpunitive way (see, e.g., Sarin et al., Reference Sarin, Ho, Martin and Cushman2021)? Or are other in-group members the primary target, so that punishment would validate the group’s morals under attack, potentially resulting in harsher treatment of the black sheep? Or is punishment about confirming the victim’s membership status in the group? Or, finally, are out-group members the primary communicative target, which may result in either harshness or leniency, depending on what exactly the intended message is meant to be? Again, these examples highlight that punishment may be a means of communication, and the content depends on how the transgression has been construed.

In emphasizing that the punished entity is not necessarily the target of communication, research findings on punishing animals (Goodwin & Benforado, Reference Goodwin and Benforado2015) need not be interpreted in line with a purely retributive perspective. Instead, animal-punishers might be sending a message to those humans who observe the punishment, to the group as a whole, or even to the experimenter who will see participants’ response choices. Animals may also serve as a symbolic proxy for human beings, so that punishing them sends a message about what kinds of behavior in human beings would be considered wrong and subject to group condemnation. Similarly, identifying the target of communication helps in understanding the psychology of displaced punishment (i.e., when punishers do not punish the actual transgressor but someone else). Research on displaced revenge has illustrated that punishment can be satisfying if it does not punish the original wrongdoer but another member of the target group (Sjöström & Gollwitzer, Reference Sjöström and Gollwitzer2015) or even a person who is symbolically similar to the original transgressor (Washburn & Skitka, Reference Washburn and Skitka2015). These findings are hard to square with a purely retributive psychology of punishment but make sense on the communicative view.

Scaffolding Future Research. A communicative view of punishment invites researchers to further explore the content, target, and means of this robust human phenomenon, leading to a broader understanding and integration of the various findings uncovered thus far as well as inspiring work that still needs to be done. For instance, investigating punishment as communication generates interesting new research questions regarding the difference as well as the commonalities between second- and third-party punishment (i.e., punishment carried out by victims or observers). Is it the content, the target(s), and/or the means of communication that differ between them? Do second- versus third-party punishers, for instance, care differently about reaching various targets of communication (such as the transgressor and/or the community), potentially because they react more automatically versus deliberately in a given situation?

Such future research should also look into the ontogeny of punishment (and its alternatives) as communication. Adding to the many interesting studies about children as punishers (e.g., McAuliffe et al., Reference McAuliffe, Jordan and Warneken2015; Riedl et al., Reference Riedl, Jensen, Call and Tomasello2015), it would be worth studying at what age children start using punishment in a dialogical, communicative way. In addition, while gaining a better understanding about how socialization and cultural learning affect children’s displayed punishment behavior (see, e.g., Salali et al., Reference Salali, Juda and Henrich2015; Wu & Gao, Reference Wu and Gao2018) and how punishment behavior differs between children and adults (e.g., related to second- versus third-party punishment or in regards to being sensitive toward intent and outcome; see Bernhard et al., Reference Bernhard, Martin and Warneken2020), it would be worth studying how content, target, and/or means of punishment as communication may change over the course of children’s development.

Nonpunitive Forms of Moral Communication. Importantly, people can react to wrongdoing with communicative acts that do not involve retributive or punitive elements. Indeed, the prevalence of punishment behavior in economic games varies between subjects from different societies (e.g., Henrich et al., Reference Henrich, McElreath, Barr, Ensminger, Barrett, Botyanatz, Cardenas, Gurven, Gwako, Henrich, Lesorogol, Marlowe, Tracer and Ziker2006). In addition, if nonpunitive options are included in research paradigms, victims often prefer compensation over punishment (Heffner & FeldmanHall, Reference Heffner and FeldmanHall2019; for similar findings with children as participants see for instance Yang et al., Reference Yang, Wu and Dunham2021). Moreover, third-party observers often choose to help the victim instead of punishing the perpetrator (Chavez & Bicchieri, Reference Chavez and Bicchieri2013; for children as participants see Lee & Warneken, Reference Lee and Warneken2020), prioritizing communicative acts that acknowledge the victim’s value as a means of restoring justice. All of these reactions to injustice carry meaning and can be considered vehicles of communications.

Hence, a final advantage of the communicative approach we advocate here is that it has resources for explaining not just why people punish but why they sometimes refrain from punishing. Ideally, a theory of punishment will explain both. Notably, a purely retributive framework is not up to this task and cannot make any predictions. By its logic, so long as there are wrongdoers, they simply deserve to be punished. By contrast, punishment from a framework of moral communication may help predict when and why people might forego punishment. What was the wrong? What did the wrong communicate to begin with? How does it affect me as a victim or as an observer? What do I think about how others think it affects me? The construal of wrongs determines how people react to them. Future research would benefit from systematically using the framework of moral communication to identify the aspects that make punishment (versus compensation, helping, etc.) the most likely reaction to injustice, thereby more fully explaining why people do in fact punish.

Identifying the situations in which people’s underlying motives can also be addressed without punishment has important normative implications. If people’s need for punishment changes depending on the availability of other communicative options that are perceived to be suitable in a given setting, it would be desirable to educate people about these effects in order to de-escalate conflicts and to avoid a ratcheting-up effect of blame (McGeer, Reference McGeer, Coates and Tognazzini2013; McGeer & Funk, Reference McGeer and Funk2017). It would then directly benefit victims, transgressors, and observers of injustice to make such options available in the broader interests of “restoring justice,” as advocates of this explicitly named movement strongly recommend (Braithwaite, Reference Braithwaite2002; Braithwaite & Pettit, Reference Braithwaite and Pettit1990; Johnstone, Reference Johnstone2013; Strang, Reference Strang2002).