Evolutionary Psychology

Evolutionary Psychology is moving to SAGE. The new address is evp.sagepub.com. Submissions here.
Robert Kurzban

The Evolutionary Psychology Blog

By Robert Kurzban

Robert Kurzban is an Associate Professor at the University of Pennsylvania and author of Why Everyone (Else) Is A Hypocrite. Follow him on Twitter: @rkurzban

Do People Punish Altruistically? A Film by Casy Neistat

Published 2 April, 2012

A couple of weeks ago, the video section of the New York Times presented a film (below) by Casey Neistat called “Bike Thief.” In it, Neistat stages the theft of a bicycle in broad daylight on the streets of New York City in order to see if anyone – onlookers or police – would intervene, filming the result.

The brief answer is: not so much. In the first scene, he uses a hacksaw to break a chain and free a bike, taking six minutes, then riding away. He repeats the stunt in front of the 9th police precinct, stealing a bicycle in under two minutes. In the third scene, a Black friend of his steals a bike. Same result. In the final scene, he uses a power tool to steal a bike chained to the entrance of a subway. After 9 minutes — the text overlay on the video says that “countless” people walked by — the police intervene.

I previously wrote about a similarly staged theft, in that case filmed for the television show, What Would You Do?

These videos speak to an ongoing debate about the extent to which people are inclined to intervene when other people have violated a social norm or broken some sort of rule. This type of behavior is occasionally referred to as “altruistic punishment,” and I’ve assembled a set of my favorite quotations drawn largely from papers that use experiments to look at this issue:

These experiments show that…a sizeable fraction continues to punish in the complete absence of return benefits. These experiments unequivocally demonstrate the existence of altruistic punishment (O’Gorman et al. 2005)

Compelling evidence for the existence and importance of such altruistic punishment comes from controlled laboratory experiments, particularly the study of public goods, common pool resource, ultimatum, and other games, from ethnographic studies of simple societies, from historical accounts of collective action, as well as from everyday observation (Bowles & Gintis, 2004)

….costly punishment is present across a highly diverse range of human populations… (Henrich et al., 2006)

In general, people retaliate against injustice even if they are not directly victimized. Sanctioning of norm-violations is vital for prosocial behavior to be sustained. (Nelissen & Zeelenberg, 2009)

In laboratory experiments, people punish noncooperators at a cost to themselves even in one-shot interactions  and ethnographic data suggest that such altruistic punishment helps to sustain cooperation in human societies (Boyd et al., 2003)

[I]n the public goods game with punishment strikingly high cooperation rates can be enforced through punishment…. It turns out that the punishment by third parties is surprisingly strong. Camerer & Fehr, 2003

Ultimatum and dictator games with children of different ages show that older children are more generous and more willing to punish altruistically (Fehr & Fischbächer, 2003)

Thus, in contrast to the conventional assumptions made in game theoretic analyses of sanctioning behaviour, there is little evidence of self-interested sanctions whereas there is much evidence for non-selfishly motivated sanctions. (Fehr & Fischbächer, 2003)

Why the apparent discrepancy between these broad, even sweeping, claims and observations in videos such as “Bike Thief?”

One possibility might be that there is something different in the populations in the laboratory as opposed to the populations in the real world. However, many of the experiments on which the above claims rest are conducted with Western industrialized samples.

Another possibility could be that people prefer other people to punish altruistically. Because these scenes were filmed in areas with a certain degree of pedestrian traffic, perhaps passersby all prefer to free ride on other people’s altruistic punishment behavior. That is, people don’t have a “taste” or a “preference” for punishment in the same way that they have tastes or preference for, say, ice cream; perhaps people only like to punish if no one else is doing so (even if they like ice cream even if everyone else is also having some.) Along similar lines, in the real world, people might expect police to enforce the rules; in the lab, there are no such expectations.

Yet another possibility is that the experiments in which moralistic punishment is observed have important demand characteristics. Perhaps subjects infer that the experimenter wants them to punish, and do so in order to conform with experimenter desires as opposed to what the subject themselves desires. This predicts that as anonymity goes up, moralistic punishment should go down, which is consistent with some work we have done in my lab.

One more possibility is that estimates of altruistic punishment have been over-estimated in the lab. In Fehr and Gächter’s (2002) paper, for instance, people punished low contributors in a public goods game. This could have been due to a desire for revenge, given that low contributors didn’t just violate a norm, they also withheld benefits from the person who punished. If people have a desire for revenge, then estimates of “altruistic punishment” will be overestimated in experiments in which punishment is imposed on those who harmed – or withheld benefits from – subjects

I probably should add that, on the one hand, I don’t doubt that under some circumstances, people will intervene in other parties’ disputes, even at some cost to themselves. Indeed, there are some uplifting moments in What Would You Do that show that this happens even in the field. Still, while it seems to me that the lab and field evidence for revenge is strong, the evidence for third party intervention is more ambiguous (as I’ve written about).

So, as someone who rides his bicycle to work everyday, I’m going to continue to rely on my Kryptonite lock, and not so much on the kindness of strangers.


Bowles, S., & Gintis, H. (2004). “The evolution of strong reciprocity: cooperation in heterogeneous populations.” Theoretical Population Biology, 65, 17-28. I removed the in-text references from the quote.

Boyd, R., Gintis, H., Bowles, S., & Richerson, P.J. (2003). “The evolution of altruistic punishment.” PNAS, 100, 3531-3535.

Camerer, C.F., & Fehr, E. (2003). “Measuring social norms and preferences using experimental games: A guide for social scientists.” Available here.

Fehr, E., & Fischbächer, U. (2003). “The nature of human altruism.” Nature, 425, 785-791.

Fehr, E., & Fischbächer, U. (2004). “Social norms and human cooperation.” Trends in Cognitive Sciences, 8, 185-190.

Fehr, E., & Gächter, S. (2002). Altruistic punishment in humans. Nature, 415, 137– 140.

Heinrich, J. et al. (2006). “Costly punishment across human societies.” Science, 312, 1767-1770.

Nelissen, R. M. A., & Zeelenberg, M. (2009). “Moral emotions as determinants of third-party punishment: Anger, guilt, and the functions of altruistic sanctions.” Judgment and Decision Making, 4, 543–553.

O’Gorman, R., Wilson, D. S., & Miller, R. R. (2005). Altruistic punishing and helping differ in sensitivity to relatedness, friendship, and future interactions. Evolution and Human Behavior, 26, 375–387.

  • asehelene

    Well, I’m thinking Cialdini and Latane and all those people here. It is New York. Diffusion of responsibility. People don’t know one another, you don’t know the story behind it. Incidentally, some of the intro undergrad did a similar study here in Lund, which is, of course, very very very VERY much smaller than New York (and Swedish, to boot). One of the students carried away a locked bike. (The hi status/low status didn’t seem to matter). Really, nobody reacted regardless.

    And, thinking about it, it is an ambiguous situation. I’ve actually had to saw off locks on a bike (that was mine), because I forgot the combination.

    Lund is also known as a place where bikes just get stolen. Might as well have communal bikes it seems….

    So, I would conjecture that there are important circumstances that influences this. I have been at the blunt end of people thinking I behaved against norms here in Sweden (to my actual enragement – one involved not having my then 4 year old son in my lap on the bus, so as to let someone else have a seat – an older person. Yes, I’m both enraged and embarrassed). And, this by someone who would not benefit. A pregnant woman once got upset with me for coughing away, and not into my hand (ok, so that is not quite altruistic punishment, but, well….) etc. So it happens, but this has been in Sweden.

    (Oh, the confessions I do in the name of scientific conjecture)

  • Rasmus Forsberg

    Seems like a big factor in the decision to not intervene should be the risk of getting beaten or killed by a stranger with a powertool in his hands.

    If the same situation took place, but either with a person known to not be violent or some way of guaranteeing that people don’t take a risk of unknown size, things might be a little different.

    • rkurzban

      In the “What With You Do” setting, they were able to monitor calls to 9-1-1. It’s hard to imagine a lower cost, safer way to intervene than that, I think.

  • A C Harper

    I don’t know about Lund but there are certainly university towns in the UK where bike theft is regarded as part of everday life – people just ‘borrow’ untended bikes. Part of the university sub-culture.

    I wonder if the ‘punishibility’ of the theft depend on, among other things, the closeness of the perceived relationships. A stranger stealing another stranger’s bike in a big city is easy to ignore because you know none of the players. Those people are just part of the social background noise.

    Someone you know in your village stealing your neighbour’s bike is a much more difficult event to explain away because you can more easily feel what it would be like to be the victim, and so your selfish altruism kicks in. Those people are part of the ‘150 people’ you are personally connected to.

  • Rick O’Gorman

    Hey Rob,

    Stumbling across your blog, I have a few comments. Firstly, you seem to have missed the most important message of the study you mention and the show ‘What would you do’. That is, you can’t rely on either people to intervene or your Kryponite lock. Most can be beaten in a couple of minutes, unless you get the super-expensive ones. Just get a cheap bike that is less attractive than others around you.

    More seriously, I don’t think the demos you discuss actually argue against costly punishing at all. Partly for reasons others here mention (diffusion of responsibility, ambiguity, lack of familiarity, etc.). Certainly it shows that people will not charge-in regardless, but I don’t think that the quotes above (chuffed to be first) nor the associated studies would claim that. I certainly wouldn’t. Moreover, the example situations above are not parallel to the studies that you cite–in those, players are engaged in a task together, have a shared interest, etc. Strangers on a street in New York don’t. We know this already–there are classic studies showing that people often don’t help others even in familiar areas if they are preoccupied, see the person in need as outgroup, and so on. So yeah, altruism is nuanced, and so is costly punishing likely to be nuanced.

    On a personal level, I’ve never seen someone try to steal a bike in front of me, but I have seen the odd person seem a bit shifty. I’ve lingered casually when parking my bike to watch them (always turned out that they just couldn’t find the bike–no surprise where we park here at my university, as there are masses of bikes together) but I’ve always felt uncomfortable because my assumption has been that they are not thieves and I feel bad even suggesting otherwise. I think you’d find the same for many passers-by–who expects a thief to actually try to steal a bike in broad daylight? Surely he (or she) is legitimately working on their bike.

    Finally (and this sorta follows on from my last paragraph), I am not sure I agree with you that the show ‘What would you do’ that looked at bike thefts had a particularly uplifting message. In one case, some men helped the bike thief (an attractive female! No surprise there for evpsychers…). And the strongest interventions against the thief came when the show had a black actor replace the white one to play the thief. There were much higher levels of intervention by passers-by. It was actually pretty depressing.

    Unless you are an attractive female white bike thief.

    • rkurzban

      Thanks for taking the time to comment, Rick.

      Regarding the claims in the quotes and the papers from which they are drawn, if they aren’t saying that people will endure costs to impose costs on those who violate norms under at least some circumstances, then I’m not sure what they are saying. I take the point that the models don’t say people always will; but they must be saying that people will under some set of circumstances. How would you characterize the boundary conditions of your claim? If not in broad daylight, when intervention is cheap and the violation is clear, then under what circumstances does the model say that people ought to intervene? (By the way, for readers who haven’t read the paper, O’gorman et al. write: “our result might be an artifact of using fictional scenarios that have no relevance to real-world interactions.” So I think that’s fair enough.)

      On your point of players engaged in a task together, there I agree. As I indicate in the paper that I link to in the post, this is why I think it’s important to distinguish third party moralistic punishment from revenge. There is a great deal of evidence that revenge is found both in the lab and in the real world. My reading is that third party moralistic punishment is rare in both the real world and the lab, though it can be found in the lab under just the right conditions. (I find small amounts of it when decisions are not anonymous, the subject is the only person who can punish, when the violation is clear, etc…)

      Finally, your argument that the observations don’t argue against costly punishment doesn’t seem right to me. The fact that we can explain why we don’t observe it doesn’t diminish the fact that we don’t observe it. That is, it could be that “diffusion of responsibility” explains why people don’t punish in the real world. If that’s true, then it puts a boundary condition on costly punishment. Note that in public goods games, there is also this potential, since others can be relied on to punish; it remains to be explained why people punish in PG games but not the field. In my view, the reason is that in PG games, we’re really seeing revenge, not moralistic punishment. That view, it seems to me, accommodates nearly all the data.

      • http://profile.yahoo.com/B2BUPKFKLPLCQ2N62NVGJNW5XI Rick

        But revenge implies being wronged. Spitefulness is the usual label I think, and I find that a bit dubious also. I suspect there are multiple motives in an artificial setting. (I also think there is a bit of evidence that people are reluctant to engage in revenge unless highly motivated. I think the anonymous games would be a weird place for people to lash out (which is the only way to characterize the punishing behaviour in PGGs given that people can’t know who they are hitting). In fact, as I touch on in my first comment, and you note to a degree, the PGGs may be particularly suitable to costly punishing because of demand characteristics. In contrast, the real world is highly ambiguous, with the potential for embarrassment. I’d bet more people fear the embarrassment of falsely accusing someone of being a thief than of the satisfaction of catching one. That doesn’t mean costly punishing is weaker than embarrassment, just that embarrassment is the more likely from personal (for everyone) experience. Most times, when we think someone is doing wrong, there is an innocent explanation.

        What’s most worrying about the stuff you comment on, as I noted, is that making the ‘thief’ black did resolve the ambiguity. Of course, I’d like to see the same thing done with a black individual doing something honest that could be misconstrued. I think we know how that would play out.

  • R J King

    I think people are more inclined to intervene if the role is clear and the situation non-ambiguous. The fact that the guy walks up with a bolt-cutter unambiguously shown might prompt people to think that he is legit–has just lost his key. I wonder what would happen if someone shouted across from a cafe that a bike was being stolen?

    • rkurzban

      In the WWYD? program, actors were told to reply to questions about whether it was their bike or not with, “it’s going to be,” or something similar. In other scenarios, there seems little ambiguity, as in the one with two able-bodied girls parked in the handicapped spot. From the looks of passersby in WWYD, and the interviews, most people seem to know that the bike is being stolen, but of course it’s impossible to say for sure. I’d be interested in other ideas about situations in which violations are clear. In the real world, this is also potentially the case. It seems to me that in unambiguous situations, people don’t intervene, as in this tragic case: http://bit.ly/H4mzc6. (shortened link to CNN site).

      • R J King

        Oh there are some very grizzly cases–no argument there. However its worth being very careful with some of the intervention studies. The classic Kitty Genovese non-intervention turned out to be not quite what the police chief and his friend–as it turned out–Latane led folk to believe. There were many calls and attmepted interventions–but the police were slow to act. I actually asked Latane about this and his reply to me was classic. “Oh no [the chief–I cant remember his name] told me over dinner that this was not true”. “Oh, that’s ok then!”
        However, the replies to the questions in the WWYD seem pretty unambiguous here. A thing that people say (and possibly believe) is not just that they are frightened of being hurt but they are worried that they will be treated more harshly than the criminal if they do intervene. I am not saying that this is true, BTW, but that it is commonly given in excuse.

      • http://www.facebook.com/people/Robert-King/100000839600576 Robert King

        Looking at that CNN story–I am not sure any of those are appropriately described as bystanders. They seem more like perpetrators–or at least an appreciative audience to me. In this case they are not really bystanders, more like onlookers and rubber-neckers at a lynching. I will go away and watch the WWYD programme–thanks for sharing these.

  • Ganduixa

    I think, that punishment is not altruistic, anybody wants to be punished, I saw the video experiment and I think that the people thinks in owner situation first, with mirror neurons and think that the man(thief), lost the key of lock bike, second in full light day with many people around is not normal situation for steal, but if the same situation was at night was different and the thief was dressed with rag clothes was different too. if I recognize a bike of one friend and see the situation, I act questioning the thief, about his action. More, bike is an object, not a person. I think that people act quickly with punishment when is about persons, but not ever. Culturally the wrong punishment is an unpleasant act. In view of doubts better not act. Sorry for my English, I hope that understand it.

  • http://popsych.blogspot.com/ Jesse Marczyk

    This post, along with the comments, draws attention to some issues that are not often addressed by the literature on altruistic punishment:

    (1) The ambiguity of the violation: As highlighted in the comments, and perhaps by the recent George Zimmerman case, people have a habit of not being in agreement as to what counts as a violation. People also have a habit of not knowing about most violations firsthand; they have to take other people’s words for it.

    In order to get a clearer picture of the act in question, that requires a certain expenditure of time and energy devoted to active monitoring. Further, past certain points, no amount of monitoring will be able to conclusively determine whether an infraction occurred; it will simply be the case that people find it more or less probable that some action took place, was wrong, and should be punished..

    As monitoring the behavior of others has costs, we should not expect people to monitor the behavior of others indiscriminately. In experiments on punishment, there are no monitoring costs and no uncertainty regarding behavior. The information is often made freely available to all the participants (in a mandatory fashion, no less; people rarely have the choice to not find out in the lab, as they do in real life). One can only imagine how willing third parties would be to pay to find out if a violation even took place, relative to second parties, especially if there are tradeoffs between the two.

    (2) The risks of intervening: Stopping a man with a hacksaw, crowbar, or power tool who might be in the process of committing a crime entails a degree of risk, no doubt. The risks of acting are only one half of the equation, however. In real life, the costs of acting and what the effects of that intervention will be are both unknown variables. A low-cost action might result in a large benefit (you ask what’s going on, the thief runs), or what you thought was low-cost might end up with you getting killed and not stopping the crime anyway.

    In the research on punishment, the punishment rule is constant and known (you pay X to reduce the other’s payoff by 3X). Again, one can imagine how willing third parties would be to pay some cost for a chance at reducing some other player’s payoff, relative to a second party, as that equation changes (i.e. by paying X you have a Y% change of either reducing someone else’s payment by Z%, or reducing your own payment further by Z%).

    (3) What counts as an out-group: Some of the researcher on third party punishment lean heavily on the idea of group selection to explain their results. One of the (many) big issues there is that group membership can be quite fluid, depending on the context. In order to benefit your group, you need to at least know who your group consists of, and it should be stable. Further, you would need to ensure you aren’t benefiting an out-group at a cost to yourself. As subjects in the lab are not recruited on the basis of pre-existing group memberships, that makes such an explanation implausible from the gate

    Punishment in the lab is probably more like arousal to pornography; an artifact of evolved mechanisms for self-interested punishment being placed in a novel environment.

  • KMK

    Funny, thought-provoking post and video. Another factor that seems unmentioned so far is that there certainly seems to be at least two people standing around with cameras pointed at the would-be bike thief for, at least, the first effort. Doubtful that the presence of the camera explains too much, but… certainly could explain a portion of passersby’s disinterest.

  • Pingback: Punch-Ups In Bars | Pop Psychology

Copyright 2012 Robert Kurzban, all rights reserved.

Opinions expressed in this blog do not reflect the opinions of the editorial staff of the journal.