Direct punishment and indirect reputation-based tactics to intervene against offences

Abstract

Punishment and reputation-based mechanisms play a major role in supporting the evolution of human cooperation. Theoretical accounts and field observations suggest that humans use multiple tactics to intervene against offences—including confrontation, gossip and ostracism—which have unique benefits and costs. Here, we draw a distinction between direct punishment tactics (i.e. physical and verbal confrontation) and indirect reputation-based tactics (i.e. gossip and ostracism). Based on this distinction, we sketch the common and unique social functions that different tactics are tailored to serve and describe information-processing mechanisms that potentially underlie decisions concerning how to intervene against offences. We propose that decision rules guiding direct and indirect tactics should weigh information about the benefits of changing others' behaviour versus the costs of potential retaliation. Based on a synthesis of existing evidence, we highlight the role of situational, relational and emotional factors in motivating distinct punishment tactics. We suggest that delineating between direct and indirect tactics can inform debates about the prevalence and functions of punishment and the reputational consequences of third-party intervention against offences. We emphasize the need to study how people use reputation-based tactics for partner recalibration and partner choice, within interdependent relationships and social networks, and in daily life situations.

This article is part of the theme issue ‘The language of cooperation: reputation and honest signalling’.

1. Introduction

Across societies, individuals and communities face challenges in terms of maintaining cooperation, deterring free-riding on public goods and ensuring adherence to social norms [1–3]. Theoretical models and experiments have shown that punishment via the selective imposition of costs on non-cooperators and norm violators can support the evolution of human cooperation [4–8]. In experimental settings, individuals punish offenders even at a personal cost, though there is substantial cross-cultural variation in punishment norms [2,9,10]. That said, multiple complementary mechanisms have been proposed to explain the evolution of cooperation, including reputation-based indirect reciprocity [11,12] and partner choice [13–15]. Experiments have provided evidence for these mechanisms in action showing that gossip and ostracism can promote cooperation [16–20], perhaps more efficiently than punishment ([21,22]; see also [23,24]).

This review aims to contribute to understanding the unique antecedents and consequences of the various punishment and reputation-based tactics that humans use to intervene against non-cooperators and norm transgressors. Based on prior work [5,25,26], we define punishment as a response to an offence via inflicting some costs on the offender¹. While punishment might be aimed at changing an offender's (future) behaviour, we do not consider deterrence as a necessary component of its definition. For example, punishers can aim at reducing disadvantageous inequality or creating advantageous inequality without deterrence [30,31], and they can reap reputational benefits independent of any recalibration of offenders' behaviour. Moreover, we use an inclusive definition of punishment that considers a host of tactics used to inflict costs on offenders, some of which require punishers to pay significant short-term costs, while others are less costly.

Based on the costs as well as the benefits of different tactics, we distinguish between two broad categories of punishment: direct punishment, which involves physical and verbal confrontation², and indirect (reputation-based) punishment, which involves gossip and ostracism. Punishing via direct confrontation has high costs—in terms of energetic expenditure, an increased risk of retaliation, and negative reputational consequences—but may also produce substantial benefits. Directly confronting offenders is more effective at changing their (future) behaviour in ways that fit punishers' interests [32,33]. In the context of status competition, direct confrontation may also bring some reputational benefits when there is value in building and maintaining a reputation of being a tough bargainer ([34,35]; see Garfield et al. [36] in this issue for an overview of relevant reputation domains). By contrast, indirect punishment tactics have lower costs [37,38], because they allow punishers to remain anonymous (at least to offenders) and minimize the risk of retaliation. However, using indirect tactics of punishment is less effective at changing offenders' behaviour, partly because offenders are unable to identify which of their behaviours has evoked punishment. Nevertheless, gossip and ostracism can impose significant reputational and relational costs on offenders [37,39]. That is, offenders who are gossiped about tend to acquire a negative reputation, and thus are less likely to attract potential coalitional partners in future social interactions. Similarly, offenders who are ostracized suffer costs in terms of losing potentially valuable interaction opportunities.

Although field studies in small- and large-scale societies point to the key role that indirect, reputation-based tactics like gossip and ostracism play in promoting cooperation or enforcing norms [40–43], the experimental literature has overwhelmingly focused on direct punishment via economic sanctioning. This focus can limit the ecological validity of research findings because (i) it remains unclear which real-world behaviours are captured by standard operationalizations of punishment in laboratory experiments and (ii) many frequent, consequential, but low-cost forms of intervention against offenders are often neglected. Here, we propose a framework that integrates a larger breadth of punishment and reputation-based tactics used to intervene against offences. We suggest that the typology of intervention tactics we use here has the benefit of bridging strands of research on direct confrontation, gossip and ostracism—behavioural phenomena that have often been studied separately. Considering the multiplicity of tactics that humans have available when deciding how to punish, along with the functions they serve (§2) and the mechanisms that motivate them (§3), highlights directions for future research on intervention against offences, in the context of partner recalibration and partner choice, within interdependent relationships and social networks, and in daily life settings (§4).

2. Common and unique social functions of distinct punishment tactics

Theoretical accounts of direct reciprocity, indirect reciprocity and reputation-based partner choice suggest that multiple mechanisms can effectively promote and help sustain human cooperation [27,44,45]. Mapping onto these accounts, empirical work has shown that people use a variety of tactics—direct confrontation, gossip and ostracism—in response to non-cooperation and norm violations in real-world settings [40,43,46]. For example, a study of responses to norm violations in a laboratory setting [47] found that around one-quarter of witnesses directly intervened against a confederate who engaged in theft (see also [48,49]). A field experiment by Balafoutas et al. [50] found that a similar proportion of third-party observers directly punished littering in a public space, though this rate of direct punishment dropped substantially when observers could indirectly punish the transgressor by withholding help. Consistently, a recent longitudinal study in daily life [51] showed that people intervene against offences via various tactics, with gossip being the most frequent response, followed by direct confrontation, and social avoidance³. Together, these findings highlight the importance of studying the use of indirect reputation-based tactics alongside direct punishment tactics, to better understand how people intervene against offences in daily settings and identify which goals punishment achieves.

Tactics of direct and indirect punishment are posited to serve similar broad functions: (i) promoting cooperation, (ii) competing for resources and/or status, and (iii) reducing inequality (figure 1). Seminal experiments have shown that punishment can be used to promote cooperation [6–8], though often at the expense of efficiency (for a review, see [25]; but see [53]). More recent experiments have instead demonstrated that punishment is, in many cases, motivated by revenge ([30,31]; but see [54]), status concerns [55] or aversion to (disadvantageous) inequality [30,56]. Traditional views of gossip and ostracism have emphasized the dark side of these tactics, seeing them as means to indirectly aggress against peers [37,39], and to impose status costs via reputation manipulation ([57–59]; especially in the context of resource competition, see [60] in this issue). Recently, though, researchers have proposed broader conceptualizations of gossip that highlight its potential to strengthen social bonds and promote cooperative behaviours [38,61–63]. In a similar vein, and despite research traditionally focusing on the negative emotional and social consequences of ostracism [64], experiments show that opportunities to choose some partners and exclude others can effectively promote cooperation [17,23]. Although confrontation, gossip and ostracism can be used to achieve similar goals, each of these tactics has unique benefits and costs and may be additionally tailored to serve unique functions, which we articulate below.

Figure 1.

Information-processing mechanisms underlying decisions to intervene against offences weigh recalibration benefits against retaliation costs. Direct punishment and indirect reputation-based tactics to intervene against offences serve common and unique social functions.

(a) . The unique benefits and costs of direct confrontation

Among the repertoire of available responses to offences, direct confrontation seems better tailored to recalibrate offenders' current and future behaviour, in ways that benefit the punisher [32,33]. This is because physically or verbally confronting offenders is the most immediate and effective way to stop ongoing transgressions, and it allows the punished individuals to draw explicit links between instances of inappropriate behaviour and the elicited punishment. This is not the case when offences are met with gossip or ostracism, because these tactics do not convey information directly to the punished individuals about what they did wrong. By contrast, verbal confrontation not only imposes costs on the offenders, but can communicate valuable information to them [65,66]. For example, verbally confronting offenders can indicate which behaviours are perceived as offensive, how victimized parties are affected and how the offenders should change their behaviour to signal that they care about the punishers [67–69]⁴. Finally, compared with other intervention tactics, confrontation might be better suited to achieve retributive goals. Retribution involves a desire to balance or repay harm in a way that is proportionate to offence severity [72]. Arguably, when using direct confrontation against offenders, punishers have more control over the immediate outcomes of their behaviour, and they can adjust their responses more easily to fit the severity of the offence [73]. By contrast, the outcomes of gossiping about an offender are often delayed and more uncertain, and the spread of information shared via gossip (to other individuals and even to the target of gossip) may be harder to control.

In sum, direct punishment via physical and verbal confrontation seems particularly well tailored to achieve recalibration and retribution goals (see also [74]), as compared with indirect reputation-based tactics to intervene against offences. Notably, although directly confronting offenders can benefit punishers both in the short term, by putting an end to ongoing transgressions, and in the long term, by recalibrating offenders' future behaviour to fit the punishers' interests, it comes with substantial costs. Direct confrontation requires time, effort and energy; it bears the risk of counter-punishment and feuds [75,76], and it can result in incurring reputational costs (we return to this latter point in §4).

(b) . The unique benefits and costs of gossip and ostracism

Compared with direct confrontation, reputation-based tactics of gossip and ostracism seem well suited to minimize the risk of retaliation when intervening against offences, by obscuring the punishers' identity. However, as mentioned earlier, gossip and ostracism seem less effective at changing offenders' behaviour, compared with direct confrontation. If gossip and ostracism are not primarily aimed at recalibration, what can they accomplish?

First, gossip plays a key role in facilitating cheater detection and partner choice. Indeed, people share and use reputational information to selectively cooperate with partners who have positive reputations [77–81] and avoid partners who have negative reputations [17,81,82]. Further, although gossip itself may be less effective at changing targets' behaviours, the mere threat of being gossiped about by others motivates people to strive to build and maintain good reputations [16,18–20]. Second, gossip is ideal for communicating about norms of acceptable behaviour. It allows people to probe and safely test the limits of conventions, norms and prescriptions, and facilitates the formation of strategic coalitions around moral values that fit personal interests (see [83,84]) and/or group interests [85]. Compared with other intervention tactics, gossip thus appears better suited to achieve general deterrence goals and it can be used as a sanction against norm violations (for a discussion of distinctions between punishment and sanctions, see [54,74]). General deterrence involves a desire to prevent future offences, not only from the same perpetrator, but also from third parties (i.e. any potential perpetrators) [72,86]. Gossip can help achieve general deterrence by allowing people to coordinate with third parties and recruit punishment from them [82,87], lowering the otherwise high costs of uncoordinated direct punishment (see also [88–90]). Finally, gossip may represent one way for individuals to take revenge by imposing strong symbolic costs on an offender, while reinstating their own image in the eyes of their community [54,74].

Compared with gossip, ostracism can be more costly, especially if used against valuable relationship partners, because it can result in missed interaction opportunities and severed social ties [51,68]. Nevertheless, when used against partners with a lower relationship value, ostracism can achieve multiple goals. Given that there are cognitive and time constraints on the number and closeness of one's social relationships [91,92], the avoidance of offenders allows people to direct attention to more valuable social relationships. Indeed, reviews of the ethnographic record suggest that ostracism or avoidance is a common tactic to deal with norm violations [40,41,43], which limits the risk of conflict escalation. Finally, ostracism may be the most cost-effective way to incapacitate repeat offenders.

3. Information-processing mechanisms underlying distinct punishment tactics

Considering the differential costs and benefits associated with direct confrontation, gossip and ostracism can be used to develop hypotheses about the putative information-processing mechanisms underlying direct punishment and indirect reputation-based tactics. Upon experiencing a norm violation, individuals first need to make decisions about whether to intervene or not and assess which tactics are available to them. If they decide to intervene, one possibility is that they then use whichever tactics they have available in an unconditional manner. An alternative possibility is that, upon deciding to intervene, people condition their choice of specific punishment tactics based on various situational factors. If so, what are the decision rules that they use to determine how to intervene against offences, when both direct and indirect intervention tactics are available? Following previous research on punishment in daily life [51], we propose that individuals should consider and integrate (at least) two types of information when deciding on which intervention tactic to use: (i) information about the benefits of recalibrating offenders' behaviour, and (ii) information about the costs of being targeted by retaliation. We expect that, when the benefits of changing offenders' behaviour are high, people will upregulate their use of direct punishment tactics; by contrast, when the costs of potential retaliation are high, people will upregulate their use of indirect, reputation-based tactics.

(a) . Factors that shift recalibration benefits

We first consider several factors that can shift the benefits of punishment in terms of changing offenders' (future) behaviour. One key factor that determines the benefits of intervening against an offence is the extent to which it has been personally harmful (i.e. the self-relevance of the offence). All else being equal, individuals have more to gain from deterring current and future offences that are harmful to themselves. Indeed, multiple vignette studies have experimentally manipulated the self-relevance of offences and found that people respond differently to violations victimizing themselves compared with those victimizing third parties. Specifically, offences that are personally harmful are met with more direct, confrontational punishment (or with equally strong direct and indirect responses), whereas offences that victimize third parties are preferentially met with indirect, less costly punishment tactics ([93,94]; but see [95,96]). Experience sampling studies on punishment in daily life settings [51,97] have found similar patterns, suggesting that self-relevant offences evoke stronger desires to punish offenders, with whichever means possible, whereas other-relevant offences are preferentially met with indirect punishment, via gossip or ostracism.

Importantly, most research to date has compared how people punish offences that harm themselves with how they punish offences that harm strangers. However, in real-world ecologies, people interact and experience offences within diverse social relationships with kin, friends, allies, ingroup members and outgroup members. Considering the relationship context in which offences take place is key for improving the ecological validity of research on punishment [51,52,98], and specifically for drawing accurate conclusions about the prevalence and use of punishment in the field. Indeed, recent work has taken promising steps in this direction, showing that people condition their punishment tactics on their relatedness and emotional closeness to victims. Consistent with the idea that people have more to gain from deterring offences that are harmful to interdependent others (e.g. their kin, friends and allies), offences against close relationship partners (i.e. family and friends) evoke similar responses to self-relevant offences, eliciting more costly confrontation than offences against strangers [93,99]. Moreover, offences harming kin are met with harsher punishment than offences harming friends, pointing to the role of relatedness and special obligations towards family in determining punishment [93,100,101].

In a similar vein, people might condition their punishment tactics on their relationship with offenders, especially when violations are self-relevant (i.e. when they are the victims and act as second-party punishers)⁵. Specifically, individuals may prefer directly confronting offenders whom they value highly rather than gossiping about them or ostracizing them [51]. This prediction is based on several reasons. First, there is more to gain from adjusting the behaviour of highly valued partners with whom one shares future interdependence. By contrast, if there is no expectation of future interactions with offenders, little can be gained by investing time and effort to recalibrate their behaviour. Additionally, there is less uncertainty about how close others will respond to punishment. Finally, gossiping about valued partners can backfire if they find out, while ostracizing them can damage otherwise important social ties [58,68]. Importantly, additional prescriptions to intervene against offences perpetrated by one's family or allies may apply in societies with strong kinship ties and norms of corporate responsibility [103].

(b) . Factors that shift retaliation costs

When making decisions about how to intervene against offences, people should not only consider the potential benefits in terms of recalibration, but also weigh the costs of receiving retaliation from offenders and their allies. Arguably, such costs of intervention differ depending on the severity of offences, with more severe offences being associated with a higher risk of retaliation. This is because offenders who have engaged in more morally wrong or harmful offences may be perceived as more willing and able to retaliate if punished. Nevertheless, previous work has found that costly punishment increases with the severity of offences, with people imposing harsher punishment against transgressions that are perceived as more severe [3,72], or transgressions that deviate more from group norms of cooperation [6]. However, studies documenting how people use a broader array of intervention tactics in the field reveal that severe offences are more often punished indirectly via gossip, ostracism or withdrawal of help [51,104].

Another, more direct cue for assessing the risk of retaliation is the victim's relative power compared with that of the offender. Power can take many forms, including (i) one's privileged access to resources and the provision of benefits and costs, (ii) one's asymmetric control over one's own and others' outcomes, (iii) one's influence derived from prestige, and (iv) one's formidability based on one's strength or other physical attributes [105–108]. Individuals who experience high power relative to offenders—whatever the basis of this power—may be more willing to engage in direct, confrontational punishment [109–112], because they can afford the risk of retaliation. By contrast, individuals who find themselves in an unfavourable power position relative to the offenders are expected to be more cautious against potential retaliation. Consistent with these ideas, people who feel less powerful are more likely to respond to norm violations by gossiping or avoiding the offenders, rather than by directly confronting them [51]. Gossiping about transgressors also allows individuals who are less powerful to recruit punishment from third parties [40,82,87], potentially reducing individual costs of punishment and the risk of retaliation from powerful others.

In addition to the factors discussed earlier, the presence of bystanders may also influence the costs of third-party intervention against offences, and thus the likelihood of intervention. For example, contrary to the notion of diffusion of responsibility, a quasi-experiment conducted on a train suggests that the silence norm is more likely to be enforced when there are more passengers in a train car [49]. This could be because punishers expect others to take their side if the situation escalates, such that there are lower retaliation costs particularly when there are more bystanders present. Such a situation represents a volunteer's dilemma in which a single individual can maintain the (second-order) public good of silence in the train by punishing the norm violator (see also [90,113]). Future research needs to consider how the presence of others influences not only the probability that someone intervenes, but also the use of specific types of intervention tactics. It is plausible that in situations that resemble the volunteer's dilemma, individuals observing a norm violation first use gossip to coordinate and then rely on only one person to directly confront an offender.

(c) . Emotions as proximate motivators of punishment

So far, we have focused on the cognitive processes—whether conscious or unconscious—underlying decisions about how to punish offences. Importantly, though, punishment is often motivated by negative emotions, including anger, disgust and contempt [97,114–116]. Recent work has emphasized that different emotions may serve unique social functions [68,96,117], with anger and disgust motivating distinct responses to offences. While anger is associated with approach-oriented, aggressive behaviours [69,118,119], disgust has been seen as motivating social avoidance [96,117,120] and efforts to signal condemnation to third parties [94,121]. Consistent with these ideas, multiple vignette studies have shown that anger in response to offences is specifically associated with inclinations to punish offenders directly, via physical and verbal confrontation [93,94,122]. By contrast, moral disgust in response to the same offences is associated with inclinations to punish offenders indirectly, via gossip and ostracism. These findings are corroborated by studies on punishment in daily life settings, showing that anger predicts both direct and indirect punishment responses, whereas disgust is specifically associated with gossip and ostracism [51]. One potential explanation for why disgust motivates gossip against offenders is that sharing information about norm violations can effectively recruit subsequent ostracism from the receivers against the targets of gossip [82,87].

4. Addressing current debates and carving future directions

In the preceding sections, we have drawn distinctions between multiple direct and indirect tactics to intervene against offences—physical and verbal confrontation, gossip and ostracism. In what follows, we describe how these distinctions can help address ongoing debates regarding the prevalence and functions of punishment, as well as the reputational consequences of third-party intervention against offences.

First, various empirical studies have cast doubt on the generalizability and ecological validity of laboratory findings regarding the prevalence and use of punishment. Experimental research has shown that punishment may only promote cooperation under certain favourable conditions, such as when its cost-to-fine ratio is low [123] or when retaliation is not possible [75,76]. Further, reviews of the ethnographic record [41,43], as well as recent survey studies [51,52], suggest that punishment—especially as commonly operationalized in laboratory experiments (i.e. the imposition of monetary costs in response to offences perpetrated by strangers against oneself or other strangers)—is rarely observed in the field. Delineating between direct punishment and indirect reputation-based tactics can facilitate comparisons between the laboratory and the field and ensure that experimental findings can be generalized to equivalent real-world situations. Conversely, pinpointing which real-world intervention tactics are of empirical interest can inform decisions about how to operationalize punishment in the laboratory.

Second, delineating between direct and indirect tactics can contribute to our understanding of when and how punishment serves cooperative versus competitive goals (for a detailed discussion, see [25]). Confrontational punishment is largely evoked by offences that harm oneself or close others [51,93,94,97]; it involves aggressive inclinations, such as anger and revenge motives [30,31,116,124], and it can lead to feuds [75,76]. Thus, while confrontational tactics may be favoured in the context of status competition, using them among peers is often discouraged and, in some cases, even proscribed to ensure harmony within communities [125]. In real-world settings, individuals often seem to prefer indirect, reputation-based tactics to deal with free-riding and other norm violations [50,51]. Gossip, in particular, may be preferred over confrontational punishment because it allows individuals to first communicate about norms of acceptable behaviour, and then coordinate their behaviour with others, thus lowering the costs of intervention and the risk of conflicts.

Third, distinguishing between direct and indirect tactics can help address debates about the reputational consequences of third-party intervention against offences. That is because the reputational consequences of intervention seem to vary depending on the tactics that are used to impose costs on offenders. Experimental studies show that direct punishment, especially when imposed by third-party observers, effectively signals trustworthiness [126,127]. However, when other means of intervention are available (e.g. helping the victims of transgressions), direct punishment loses some of its reliability as a signal of cooperativeness [128–130]. Further, other findings cast doubt on the idea that second- and third-party punishment signals trustworthiness and show that generous but not punitive individuals tend to be trusted more [131]. Especially when intervention takes the form of physical or verbal confrontation, it may even backfire, because confrontational individuals appear aggressive and are seen as motivated by selfish concerns [121,132]. More work is needed to understand how observers perceive and judge intervention via reputation-based tactics of gossip and ostracism⁶. This is an especially interesting avenue for future work because different societies may deem different ways of intervening against offences as more or less appropriate. A recent cross-cultural study of meta-norms (i.e. social norms about how people should treat norm violations) has provided some initial evidence that the appropriateness of confrontation, gossip and ostracism differs across societies [125]. Undoubtedly, understanding the ecological and cultural origins of variation in such meta-norms is a fascinating puzzle to be addressed by future research on cooperation.

Before closing, we turn to three additional recommendations for future research based on the work that we have reviewed here. First, our analysis suggests that the same intervention tactic can be used to achieve multiple purposes. To illustrate, gossip can be used for general deterrence (i.e. to recalibrate third parties' behaviour) and can also be used to facilitate partner choice. Similarly, ostracism can represent an effort to recalibrate someone's behaviour (e.g. in the case of the ‘silent treatment’), but it is often used merely to navigate away from offenders and toward more valuable relationship partners. Future research on gossip and ostracism would benefit from studying when and how these reputation-based tactics are used for partner recalibration versus partner choice. Second, as noted earlier, third-party intervention in natural settings occurs within a rich relational context where the offender, the victim and the third-party observer have varying degrees of interdependence with each other. Different structures of interdependence may affect when and how third parties choose to intervene against offences [98,108,133]. For example, people may be more prone to intervene against offences that harm someone with whom they are mutually dependent [98], whereas they may be less prone to intervene against offences that harm someone they have conflicting interests with. Considering interdependence relations and the properties of the networks that people are embedded in (e.g. network centrality and relational mobility; [134,135]) is key to understanding third-party intervention. Likewise, our understanding of when people intervene against offences in ecologically valid situations can be ameliorated via the use of a variety of field methods. We believe that by revisiting the rich ethnographic record and by using novel, experience sampling techniques, future work can document a variety of intervention tactics in real-world settings and provide valuable insights into the factors that shape the use of confrontation, gossip and ostracism across social and cultural contexts.

Data accessibility

This article has no additional data.

Authors' contributions

C.M. and J.W. wrote and revised the manuscript.

Competing interests

We declare we have no competing interests.

Funding

C.M. acknowledges IAST funding from the French National Research Agency (ANR) under grant no. ANR-17-EURE-0010 (Investissements d'Avenir programme). J.W. acknowledges funding from the National Natural Science Foundation of China (grant no. 71901028).