Altruistic punishment
Also known as: Costly punishment
Paying out of your own pocket to penalize a cheater who never wronged you directly.
What it means
The willingness to incur a personal cost to punish norm violators even when the punisher gains no material benefit and was not personally harmed. In public-goods experiments, allowing players to pay to reduce free-riders' earnings dramatically raises and sustains cooperation, because the threat of punishment deters defection. The behavior is a puzzle for narrow self-interest — punishing is itself a public good, vulnerable to second-order free-riding — and is explained by strong reciprocity and negative emotions toward cheaters. It matters because such punishment appears to be a key enforcement mechanism behind human cooperation in groups beyond kin and repeated partners.
Why people pay to punish
The def's 'negative emotions' hides a two-part machine: anger starts the act, and satisfaction pays for it. Fehr and Gächter put hypothetical free-riding scenarios to subjects after the game: anger ratings climbed with the size of the free-rider's shortfall, and actual punishment climbed with the same shortfall. The two gradients match; the authors never measured whether a given person's anger predicted their own spending, and called the pattern only 'consistent with' emotions as the proximate cause. Punishing may also register as rewarding: de Quervain and colleagues (2004) used PET to scan people deciding how to sanction a defector, and found the dorsal striatum — a region tied to anticipated reward — more active when the punishment actually cut the defector's payoff than when it was merely symbolic; those with stronger activation paid more to punish. Anger as the trigger, satisfaction as the payoff — that pairing sustains the behavior without anyone deciding to enforce a norm. It also explains why punishment fires in one-shot anonymous encounters, where deterrence cannot repay the punisher. The person is settling a score; the group's benefit is a by-product.
What the evidence shows
The design is what makes the original result striking. Fehr and Gachter reshuffled strangers every period, so no reputational or repeated-game motive survived, yet 84.3 percent of subjects punished at least once, and 74.2 percent of punishment acts fell on below-average contributors and came from above-average ones. Balliet, Mulder and Van Lange's meta-analysis of 187 effect sizes put punishment's effect on cooperation at d = 0.70 — statistically indistinguishable from reward's d = 0.51, which is itself a finding the enforcement story tends to skip. The accounting is tighter than the headline, though. Money burned on punishing offsets the cooperation gains over short horizons; Gachter, Renner and Sefton (2008) found the net turns clearly positive only when groups play long enough — fifty periods rather than ten — for deterrence to amortize.
Where it breaks down
Two findings puncture the tidy story. Herrmann, Thoni and Gachter ran the same game in sixteen participant pools worldwide and found antisocial punishment — sanctioning generous contributors — was widespread, and in some pools strong enough to erase the cooperation benefit entirely. Weak civic norms and weak rule of law predicted it. Nikiforakis (2008) added a counter-punishment stage, closer to how retaliation works outside a lab: cooperators turned reluctant to punish, roughly a quarter of punishments were avenged, and groups earned less than in a no-punishment condition where free-riding ran unchecked. Peer sanctioning depends on institutions the experiment usually holds fixed and invisible.
Is it really altruistic?
The altruism label is contested. Dreber and colleagues found that punishment raised how often people cooperated but not what they earned, and the highest-scoring players punished least — hard to square with punishment as an adaptation for cooperation. Pedersen, Kurzban and McCullough report that victims of unfairness punished while mere witnesses largely did not, and that witnesses' reactions looked more like envy of the cheater's gains than moral anger; earlier third-party results, they argue, partly reflect people mispredicting their own feelings. Burton-Chellew and Guerin made a minority of players immune to punishment and watched them stop cooperating while still punishing others.
Using it in practice
For anyone designing a system — a marketplace, a team, a community platform — the lesson is not to add sanctions. It is that decentralized punishment works only where the norm is unambiguous, the target is visibly the violator, and retaliation is impossible. Where members can strike back, or argue about who was in the wrong, a sanctioning channel can cost more than the free-riding it deters. Where retaliation and ambiguity are live, an enforcer whose authority the group accepts tends to outperform peer sanctioning — which is the condition the lab quietly removes. The same machinery drives boycotts and reviewer pile-ons: cheap to trigger, hard to aim, and satisfying enough that accuracy is rarely what limits it.
Examples
In a public-goods game, a contributor spends some of their own money to dock a persistent free-rider, lifting everyone's future giving.
A shopper who has never been wronged by the brand stops buying it after a wage scandal, paying more elsewhere purely to make the company answer for it.
A colleague burns real political capital reporting a manager who took credit for someone else's work, knowing the only thing he earns is a reputation as a troublemaker.
A long-time forum regular with no moderator powers spends unpaid evenings compiling an evidence dossier on a spammer who never targeted her and pushing publicly for a ban she has no authority to impose, absorbing a stream of abuse in return, so that strangers can keep using the board.
A researcher who works in a different subfield spends a year assembling a misconduct case against a lab that fabricated its data, gaining no citations and several permanent enemies.
First described in Fehr & Gächter (2002).
Key references
- Burton-Chellew, M. N., & Guérin, C. (2021). Decoupling cooperation and punishment in humans shows that punishment is not an altruistic trait. Proceedings of the Royal Society B, 288(1962), 20211611. doi.org/10.1098/rspb.2021.1611
- Pedersen, E. J., Kurzban, R., & McCullough, M. E. (2013). Do humans really punish altruistically? A closer look. Proceedings of the Royal Society B, 280(1758), 20122723. doi.org/10.1098/rspb.2012.2723
- Balliet, D., Mulder, L. B., & Van Lange, P. A. M. (2011). Reward, punishment, and cooperation: A meta-analysis. Psychological Bulletin, 137(4), 594-615. doi.org/10.1037/a0023489
- Herrmann, B., Thöni, C., & Gächter, S. (2008). Antisocial punishment across societies. Science, 319(5868), 1362-1367. doi.org/10.1126/science.1153808
- Dreber, A., Rand, D. G., Fudenberg, D., & Nowak, M. A. (2008). Winners don't punish. Nature, 452(7185), 348-351. doi.org/10.1038/nature06723
- Fehr, E., & Gächter, S. (2002). Altruistic punishment in humans. Nature, 415(6868), 137-140. doi.org/10.1038/415137a
- de Quervain, D. J.-F., et al. (2004). The neural basis of altruistic punishment. Science, 305(5688), 1254-1258. doi.org/10.1126/science.1100735
- Gächter, S., Renner, E., & Sefton, M. (2008). The long-run benefits of punishment. Science, 322(5907), 1510. doi.org/10.1126/science.1164744
- Nikiforakis, N. (2008). Punishment and counter-punishment in public good games: Can we really govern ourselves? Journal of Public Economics, 92(1-2), 91-112. doi.org/10.1016/j.jpubeco.2007.04.008