Original ArticleHuman punishment is motivated by both a desire for revenge and a desire for equality
Introduction
Punishment typically involves paying a cost to harm individuals who harm or withhold benefits from the punisher (hereafter 'defectors', Clutton-Brock and Parker, 1995, Raihani et al., 2012; but see Irwin and Horne, 2013, Sylwester et al., 2013 for punishment aimed at helpful or cooperative individuals). Since punishment is costly to administer, both in terms of executing the punishment itself and in terms of the possibility of provoking retaliation from the target (Dreber et al., 2008, Herrmann et al., 2008, Janssen and Bushman, 2008, Nikiforakis, 2008), considerable effort has been expended in trying to understand the evolved function of punitive sentiments (McCullough et al., 2013, Price et al., 2002). Specifically, it has been argued that understanding the contexts that reliably motivate punishment can provide key insights into its likely evolved function (Price et al., 2002). Two broad functional explanations have been proposed. First, it has been suggested that punitive sentiment could confer a selective advantage if punishment deters targets (or bystanders) from harming the punisher in future interactions (e.g. dos Santos et al., 2011, Hilbe and Sigmund, 2010, McCullough et al., 2013). Under this hypothesis (hereafter the 'revenge' hypothesis), individuals should be motivated to reciprocally harm individuals that intentionally harm them, even if punishment cannot immediately equalize the payoffs between the defector and the punisher (Falk, Fehr, & Fischbacher, 2005). However, evidence that punitive sentiments are sensitive to the risk of suffering a fitness disadvantage relative to defectors (Dawes et al., 2007, Raihani and McAuliffe, 2012a) suggests an alternative explanation: that punishment primarily serves a fitness-leveling function, by reducing payoff differentials between defectors and punishers (Price et al., 2002; see also Carlsmith, Darley, & Robinson, 2002). Under this fitness-leveling hypothesis, punishers are expected to be motivated primarily by the desire to equalize payoffs, and any deterrent function of punishment would arise as a by-product. Here, we present an experiment to test whether punitive sentiment can best be explained in terms of desire for revenge or in terms of a desire to equalize payoffs in social interactions.
Interacting with a defector often reduces cooperators' payoffs and creates unequal outcomes. It can therefore be difficult to establish whether punishment of defectors is motivated by the disutility associated with receiving lower payoffs than a defector (‘disadvantageous inequality aversion’, (Fehr & Schmidt, 1999) or simply a desire for revenge (Raihani & McAuliffe, 2012b). A recent study attempted to disentangle these two possible motivations by asking whether, in the absence of disadvantageous inequality, experiencing losses was sufficient to motivate punishment (Raihani & McAuliffe, 2012b). Raihani and McAuliffe (2012b) found that defection, in the form of stealing money from the victim, did not motivate punishment when stealing resulted in equal outcomes or advantageous inequality for the victim. However, stealing did motivate punishment when it resulted in disadvantageous inequality for the victim (Raihani & McAuliffe, 2012b). These findings raise the possibility that individuals use punishment to restore equality in social interactions. However, the alternative possibility, that punishment is simply related to the disutility associated with experiencing disadvantageous inequality and is not tailored to achieve equal outcomes, could not be ruled out because players in this game were not allowed to tailor their investment in punishment.
Alternative studies have also suggested that investment in punishment is aimed at producing equal outcomes in social interactions. For example, in (Dawes et al., 2007) individuals were placed in groups of four and randomly allocated an endowment. Some players therefore started out richer than others in this game. Players were given the option to reduce (or increase) the income of others by purchasing negative (income-reducing) or positive (income-increasing) tokens and allocating these to other group members. In this setting, people allocated more negative tokens to the richest players and allocated more positive tokens to the poorest members of the group — suggesting that these behaviors were aimed at reducing outcome inequality. However, in this experiment, all four group members were able to purchase and allocate these tokens. Thus, it was impossible for players to predict how many tokens they would need to buy in order to achieve equal outcomes. Consequently, it is not possible to determine whether players adjusted investment in punitive behavior in order to achieve specific outcomes. Moreover, since initial payoff inequalities were exogenously determined rather than arising through some players defecting, the study could not test to what extent investment in income-reducing tokens was related to the target's behavior, as opposed to the outcome itself. In other words, since cooperation and defection were not possible in this game, any revenge-based motives of punishment could not be measured.
A more recent study by Houser and Xiao (2010) showed that players who were treated unfairly most commonly chose to punish as severely as possible and thus create inequality in their own favor. Although this seems to be more suggestive of punishment as a form of revenge rather than a fitness-leveler, it is important to take into account that in this study the severity of punishment chosen was not constrained by cost. In reality, imposing a larger cost on another individual is likely to also impose a larger cost on the punisher (Raihani & McAuliffe, 2012a). Since punishers have been shown to adjust their investment according to the costs associated with punishment (Anderson and Putterman, 2006, Bone et al., 2014, Carpenter, 2007, Nikiforakis and Normann, 2008, Ostrom et al., 1992), this creates a potentially important trade-off between maximizing income and achieving the desired punishment outcome.
The fitness-leveling hypothesis predicts that individuals should only invest in punishment that is more costly to the target than to the punisher, and is therefore able to reduce any existing disadvantageous inequality. Nevertheless, empirical work has demonstrated that individuals are prepared to invest in punishment that is equally costly to the punisher and the target (Anderson and Putterman, 2006, Carpenter, 2007, Egas and Riedl, 2008, Falk et al., 2005, Nikiforakis and Normann, 2008) — or even more costly to the punisher (Anderson and Putterman, 2006, Carpenter, 2007, Egas and Riedl, 2008) — and so is unable to re-establish equality. These findings suggest that punishers are not solely motivated by a desire to remove fitness differentials and support the idea that punishers might instead be motivated by a desire for revenge against defecting partners. The predictions of the two hypotheses also differ with respect to whether the defection was performed intentionally or not. Specifically, the revenge hypothesis predicts that punishment should be focused on those who impose harm intentionally and can therefore learn to avoid repeating the harmful behavior in the future. Conversely, punishment aimed at removing fitness differentials should be less sensitive (or insensitive) to intentionality since the primary function is to reduce inequality rather than change the target's behavior. Evidence from empirical studies provides some support for both hypotheses. Whilst several studies have shown that individuals will punish in response to unequal outcomes created at random or unintentionally (Cushman et al., 2009, Dawes et al., 2007, Falk et al., 2008, Houser and Xiao, 2010, Kagel et al., 1996, Yu et al., 2014), individuals are significantly more likely to punish when unequal outcomes are created intentionally by the target (Falk et al., 2008, Houser and Xiao, 2010, Kagel et al., 1996).
Based on past research it is therefore unclear whether punishment is motivated by a desire for revenge or by a desire to equalize payoffs. We aimed to answer this question by investigating whether victims of cheats adjusted their investment in punishment in order to restore equality using a modified version of the game used by Raihani and McAuliffe (2012b). In the current study, one player could choose to steal $0.20 from their partner. Depending on the treatment, players interacting with a stealing partner experienced advantageous inequality, equal outcomes or varying levels of disadvantageous inequality. Players could punish stealing partners, but while some players had access to effective punishment (1:3 fee to fine) others could only use ineffective punishment (1:1 fee to fine). Players who had access to effective punishment could achieve equal outcomes by tailoring their investment in punishment: more extreme outcome inequality could be alleviated by investing more into punishment. However, under the ineffective punishment condition, increasing investment in punishment did not reduce inequality.
Although we suggest that revenge may serve a deterrent function, in the anonymous one-shot setting of our game, there is no scope for punishment to change the behavior of stealing partners (or bystanders). However, previous work has suggested that behavior may be constrained by psychological mechanisms that evolved in the context of non-anonymous repeated interactions and that responses that are attuned to these conditions may be invoked even in anonymous, one-shot settings (Ben-Ner and Putterman, 2000, Burnham and Johnson, 2005, Cosmides and Tooby, 1989, Delton et al., 2011, Hagen and Hammerstein, 2006, Hoffman et al., 1998, Johnson et al., 2003, Tooby et al., 2006). Thus, in our game a desire for revenge might reflect the desires of an evolved psychology that functions to deter cheats, even though this function is (due to the nature of the game) impossible to achieve. Nevertheless, we note that since deterrence is not the only possible function for this behavior we use the word ‘revenge’ in a purely descriptive sense.
The revenge hypothesis predicts that punishment will be used in both the ineffective and the effective punishment condition. Alternatively, if punishment is motivated by the desire to equalize outcomes, punishment should be used when it is effective but not when it is ineffective. Moreover, players should use the amount of punishment that is required to equalize payoffs (Table 1); not more or less.
Section snippets
Experimental protocol
This research was approved by the University College London ethics board project number 3720/001. Data were collected in October 2013 and July–August 2014. We recruited 4912 subjects (2856 males, 1967 females, 89 unreported) for our experiment using the online labor market, Amazon Mechanical Turk (AMT; www.mturk.com). Subjects were all based in the USA.
Of the 4912 subjects, 2456 were assigned the role of player one (P1). The remaining subjects were allocated the role of player two (P2). P1 and
Results
In both the effective and the ineffective punishment condition, P1 was significantly more likely to punish a stealing than a non-stealing P2 (Fisher's exact test, see Table 2 for p-values; Fig. 1). In the effective punishment condition, the tendency to punish a stealing P2 was increased significantly when stealing resulted in disadvantageous inequality (proportion punishing non-stealing P2 ± SE = 0.04 ± 0.01; stealing P2, no disadvantageous inequality = 0.19 ± 0.03; stealing P2, disadvantageous inequality
Discussion
In this study, P1 experienced the same losses when P2 stole ($0.20) in all five treatments. However, depending on the treatment this $0.20 loss resulted in P1 experiencing either advantageous inequality (treatment A), equal payoffs (treatment B) or disadvantageous inequality (treatments C–E) relative to P2. P2 stealing provoked P1 to punish even when stealing did not create disadvantageous inequality. Moreover, although relatively rare, P1 sometimes punished a stealing P2 even when punishment
Supplementary materials
The following is the supplementary data to this article.
Acknowledgements
We would like to thank the editor, Robert Kurzban, and the three anonymous referees for their useful comments. This study was funded by a Royal Society University Research Fellowship to N.R.
References (66)
- et al.
The pleasure of being nasty
Economics Letters
(2009) - et al.
Do non-strategic sanctions obey the law of demand? The demand for punishment in the voluntary contribution mechanism
Games and Economic Behavior
(2006) - et al.
On some implications of evolutionary psychology for the study of preferences and institutions
Journal of Economic Behavior & Organization
(2000) The demand for punishment
Journal of Economic Behavior & Organization
(2007)- et al.
Evolutionary psychology and the generation of culture, part II: Case study: A computational theory of social exchange
Ethology and Sociobiology
(1989) - et al.
Testing theories of fairness–Intentions matter
Games and Economic Behavior
(2008) - et al.
Let me sleep on it: Delay reduces rejection rates in ultimatum games
Economics Letters
(2011) - et al.
Game theory and human evolution: A critique of some recent interpretations of experimental games
Theoretical Population Biology
(2006) - et al.
Inequality-seeking punishment
Economics Letters
(2010) - et al.
A normative explanation of antisocial punishment
Social Science Research
(2013)
Evolution of cooperation and altruistic punishment when retaliation is possible
Journal of Theoretical Biology
Fairness in ultimatum games with asymmetric information and asymmetric payoffs
Games and Economic Behavior
Resource scarcity and antisocial behavior
Journal of Public Economics
Punitive sentiment as an anti-free rider psychological device
Evolution and Human Behavior
Punishment and cooperation in nature
Trends in Ecology & Evolution
The promise of Mechanical Turk: How online labor markets can help theorists run behavioral experiments
Journal of Theoretical Biology
Bargaining under time pressure in an experimental ultimatum game
Economics Letters
Profit-seeking punishment corrupts norm obedience
Games and Economic Behavior
Controlling the false discovery rate: A practical and powerful approach to multiple testing
Journal of the Royal Statistical Society: Series B Methodological
Defectors, not norm violators, are punished by third-parties
Biology Letters
The effect of power asymmetries on cooperation and punishment in a prisoner’s dilemma game
PloS One
The biological and evolutionary logic of human cooperation
Why do we punish? Deterrence and just deserts as motives for punishment
Journal of Personality and Social Psychology
Punishment in animal societies
Nature
Accidental outcomes guide punishment in a “trembling hand” game
PloS One
Egalitarian motives in humans
Nature
The neural basis of altruistic punishment
Science (New York, N.Y.)
Is numerical comparison digital? Analogical and symbolic effects in two-digit number comparison
Journal of Experimental Psychology. Human Perception and Performance
Evolution of direct reciprocity under uncertainty can explain human generosity in one-shot encounters
Proceedings of the National Academy of Sciences of the United States of America
The evolution of punishment through reputation
Proceedings. Biological Sciences/The Royal Society
Winners don’t punish
Nature
The economics of altruistic punishment and the maintenance of cooperation
Proceedings. Biological Sciences/The Royal Society
Civic capital in two cultures: The nature of cooperation in Romania and USA
Cited by (37)
Punishment is strongly motivated by revenge and weakly motivated by inequity aversion
2021, Evolution and Human BehaviorCitation Excerpt :The deterrence hypothesis predicts that individuals should be motivated to punish those who intentionally harm them, regardless of whether punishment equalizes the payoff differentials between defectors and punishers. A second functional explanation is that punishment evolved to serve a competitive function, by increasing relative fitness differences between punishers and cheaters (Bone & Raihani, 2015; Price, Cosmides, & Tooby, 2002). The competition hypothesis predicts that individuals should punish in order to increase their relative payoffs and that any deterrence function would occur as a by-product (Raihani & Bshary, 2019).
Distinct affective responses to second- and third-party norm violations
2020, Acta PsychologicaWhen do we punish people who don't?
2019, CognitionCitation Excerpt :As humans, we often punish those who are antisocial: People who do not cooperate (Balliet & Van Lange, 2013; de Quervain et al., 2004; Fehr & Gächter, 2002; Gächter, Renner, & Sefton, 2008; Mathew & Boyd, 2011), who cause harm (Bone & Raihani, 2015; Buckholtz et al., 2008; Martin & Cushman, 2016b; McCullough, Kurzban, & Tabak, 2013; Morris, MacGlashan, Littman, & Cushman, 2017; Treadway et al., 2014) who behave unfairly (Cushman, Dreber, Wang, & Costa, 2009; FeldmanHall, Sokol-Hessner, Van Bavel, & Phelps, 2014; Henrich et al., 2010, 2006; Martin & Cushman, 2015; Mendoza, Lane, & Amodio, 2014) or who violate norms (Balafoutas & Nikiforakis, 2012; Carpenter & Matthews, 2009; Fehr & Fischbacher, 2004b).
Is increasing inequality harmful? Experimental evidence
2018, Games and Economic Behavior