News & Views

Behavioural science: Homo reciprocans

Humans are often generous, but cooperation unravels when others take advantage of them. Many people punish such 'free riders', even if they do not benefit personally, and this 'altruistic punishment' sustains cooperation.

Garrett Hardin1 famously described a group of herders whose pursuit of self-interest leads to overgrazing of a pasture, driving it to ruin. His term for the process, the 'tragedy of the commons', underlined its inexorable nature. But herders, fishers and other users of common resources frequently avert the tragedy, devising ways to pursue common objectives and to curb free riding2. Indeed, it is a distinctive characteristic of humans that we cooperate in large groups of people to whom we are not related — giving to charity, participating in political movements, and conforming to social norms. On page 137 of this issue, Fehr and Gächter3 describe the outcome of an experimental game involving human players which suggests that a key ingredient in human cooperation may be the punishment of free riders by their more public-spirited peers.

Investigations using game theory and experiments show that contributions to a common project can be sustained in repeated interactions in small groups of two or more people by various self-interested means, such as the expectation that others will withdraw their contributions ('tit for tat')4 or the indirect benefits of building a reputation as a cooperative group member5,6. But humans often cooperate in 'one-shot' interactions in which large numbers of people contribute to a common project. In these situations there is little chance of direct or indirect reciprocation, so self-interest-based explanations of cooperation are unconvincing.

Cooperation in such situations can be sustained by the threat of punishment by peers. But if one incurs costs when punishing others, this merely displaces rather than resolves the puzzle. Punishing free riders in this type of situation is just as altruistic as actually contributing to the public good. The key difference is that the punishment is conditioned by the behaviour of others — hence our term Homo reciprocans. Cooperation can be sustained because altruistic punishers induce even selfish group members to contribute to the common project. But is such altruistic punishment common?

Fehr and Gächter3 show that it is. The results of their well-designed experimental 'public-goods' game (Box 1) reveal that most people incur costs to themselves in order to punish those who have shirked in contributing to a public good. This form of altruistic punishment is common even when it is unmistakably the case that there can be no subsequent (direct or indirect) material benefits to those doing the punishing, because the groups of players are shuffled after every round such that no participant encounters any other more than once. Those who are punished for shirking contribute more in subsequent rounds, with the result that high overall levels of cooperation are sustained.

Box 1: Box 1 'Public goods' experiments

In the 'public goods' game, each subject is given an endowment, w. Each subject may contribute any part of this endowment to a common account. Following this contribution stage, there is a pay-off stage in which the experimenter gives each of n subjects a fraction, q(1/n, 1), of the amount in the common account. Contributing is an altruistic act as it increases the average group pay-off (because q>1/n) at a cost to oneself (because q<1).

In the 'public goods with punishment' game, following the contribution stage there is a punishment stage in which the contributions of each subject, ai, are made public and each player is given the opportunity to reduce the pay-offs of subject j by mij at a cost to the punisher of cmij . After the punishment stage, the pay-off stage is as above. Thus the pay-offs to subject i are: where the summation is over all the group members. What the subject keeps is captured by the first two terms (wai); qaj is the amount received from the common account; ∑mji is the cost of being punished by others; and cmij is the cost of punishing others.

Interactions are anonymous and may last for a just a single round or a known number of rounds. In the 'partner treatment', membership of the groups remains unchanged during the game. In the 'stranger treatment', membership is shuffled after each round. The 'perfect stranger' treatment implemented by Fehr and Gächter3 ensures that no two subjects will interact more than once.

This result is striking in view of the fact that, in public-goods experiments with no provision for punishing free riders, contributions are substantial in early rounds of the game but dwindle to virtually nothing in subsequent rounds7. By showing that cooperation can be sustained by altruistic punishment where altruistic contribution may fail, the experiment will direct attention to why humans are so willing to punish those who violate norms, rather than focusing on why humans are also (sometimes) unconditionally generous to strangers.

Fehr and Gächter's work goes beyond previous studies8,9 by reshuffling participants such that no subject interacted more than once with any other. This treatment precludes both reputation and repetition, the mechanisms underlying direct and indirect reciprocity and other self-interested explanations. As a result of this 'perfect stranger' treatment, subjects knew that punishing others could not raise their own pay-offs even if the shirkers they punished contributed more in later rounds. Rather, punishment per se provided the motivation, not some consequence anticipated by the player.

But why would the idea of punishing shirkers be a motivating factor? In the one-shot public-goods game, not punishing and not contributing maximize pay-offs irrespective of what the other group members do. So the willingness of subjects to reduce their own pay-offs by punishing shirkers but not to reduce their pay-offs by contributing to the public good in the absence of punishment needs to be explained in terms other than self-interest. Fehr and Gächter advance the view that punishment of shirkers is not evidence of a general-purpose predisposition to contribute to the public good, but rather reflects a negative emotional reaction to free riding. They stress a particular motivation — the desire to punish those who violate norms — rather than the more general motivation to contribute to the well-being of others. But the players are public-spirited nonetheless, as they do not themselves benefit from doing the punishing.

Shirkers, too, appear to have an emotional response to being punished. In other experiments10, even when punishment takes the form of verbal criticism rather than a pay-off reduction, shirkers contribute more in subsequent rounds, suggesting that punishment may evoke emotions of shame in the free rider.

Fehr and Gächter's experiment3 has implications for the design of constitutions and policies. It suggests that the objective should be to provide opportunities for the public-spirited to punish free riders, rather than to assume, as David Hume11 advised two-and-a-half centuries ago, that “every man ought to be supposed to be a knave and to have no other end, in all of his actions, than his private interest”.

Indeed, studies of contemporary hunter-gatherers and other evidence suggest that altruistic punishment may have been common in mobile foraging bands during the first 100,000 years or so of the existence of modern humans12. A plausible explanation of the evolutionary success of this strategy is that groups with a high fraction of altruistic punishers would have sustained cooperation more successfully than groups with fewer punishers, and so would have prevailed over them13. Fehr and Gächter's results suggest that explanations for cooperation based on indirect reciprocity, reciprocal altruism or other forms of self-interest are at best incomplete.


  1. 1.

    Science 162, 1243–1248 (1968).

  2. 2.

    Nature 340, 91–93 (1989).

  3. 3.

    & Nature 415, 137–140 (2002).

  4. 4.

    & Science 211, 1390–1396 (1981).

  5. 5.

    & Nature 393, 573–577 (1998).

  6. 6.

    , & J. Theor. Biol. 213, 103–119 (2002).

  7. 7.

    in The Handbook of Experimental Economics (eds Roth, A. E. & Kagel, J.) 111–194 (Princeton Univ. Press, NJ, 1995).

  8. 8.

    & J. Econ. Perspect. 14, 159–183 (2000).

  9. 9.

    , & Am. Polit. Sci. Rev. 86, 404–417 (1992).

  10. 10.

    & in Santa Fe Institute Working Paper (January 2002).

  11. 11.

    Essays: Moral, Political and Literary (Longmans, Green (original publication 1754), London, 1898).

  12. 12.

    Hierachy in the Forest (Harvard Univ. Press, Cambridge, MA, 2000).

  13. 13.

    Economic Behavior and Institutions: An Evolutionary Approach to Microeconomics (Princeton Univ. Press, NJ, 2002).

Download references

Author information


  1. Samuel Bowles is at the Santa Fe Institute, Santa Fe, New Mexico 87501 and the University of Massachusetts at Amherst, Massachusetts 01003, USA.e-mail:

    • Samuel Bowles
  2. Herbert Gintis is at the University of Massachusetts at Amherst, Massachusetts 01003, USA.e-mail:

    • Herbert Gintis


  1. Search for Samuel Bowles in:

  2. Search for Herbert Gintis in:


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.