Guest Blogs

Conflict and Cooperation

By Antonio Silva October 6, 2014 24 Comments

The idea of war and conflict as a driver of cooperation has regularly been discussed in this blog, and it is often assumed that people help one another in times of adversity, especially when in conflict with other groups. Inter-group conflict has been widely suggested as a key mechanism in the evolution of uniquely large-scale cooperation in humans. Specifically, several models based on cultural and genetic group selection propose that groups in which people are altruistic towards their fellow group members and hostile to outsiders are more likely to be victorious in inter-group conflict.

From an empirical perspective, there are two main ways of testing this idea. One can look at the past for evidence that more altruistic groups out-competed less altruistic ones through inter-group conflict. There have been some indirect attempts to do this, notably Bowles (2009), which showed that the levels of inter-group warfare in the Late Pleistocene and early Holocene were likely to be high enough to enable selective pressures required for the evolution of altruistic behaviours. Alternatively, one can look at the present and test whether inter-group conflict is associated with increased levels of altruism towards the in-group, and this was what my recent paper with Ruth Mace attempted to do.

We measured cooperative behaviour between and within two groups with a long and on-going history of conflict: Catholic and Protestant communities in Northern Ireland. We quantified cooperative behaviour using naturalistic measures: donations to schools and charity, and lost letters experiments. In the donations experiment, we ran a door-to-door survey with questions about individual exposure to conflict, such as the number of times people had experienced sectarian attacks. We also offered people £5 for their time in answering the survey, and allowed them to keep or donate the money to a local primary school, which was either a Catholic or Protestant school, or to a neutral charity, Save the Children (people were only given one randomly selected choice). This experiment was, in essence, a surreptitious dictator game and allowed us to measure in-group, out-group and unbiased cooperation. In the lost letters experiment, we dropped stamped letters on the pavement addressed to fictional Catholic, Protestant, and neutral charities (CatholicAID, ProtestantAID and CancerAID) and measured the neighbourhood’s cooperative behaviour by the amount of letters returned, similar to the original Stanley Milgram 1960’s experiments.

Sign up for our newsletters

I wish to receive updates from:

We expected to find that exposure to conflict would result in increased levels of cooperation towards the in-group and reduced levels towards the out-group, in what is often referred to as parochial altruism. Although we found that increased exposure to conflict did reduce the likelihood of donating to out-group schools and returning out-group letters, it had no effect on helping the in-group through either donations or lost letters returns. However, we did find that individuals’ and neighbourhoods’ socio-economic characteristics were the best predictors of cooperative behaviour. Wealthy people were much more likely to give money to in-group schools and people in wealthy neighbourhoods were also more likely to return an in-group letter. Interestingly, wealth had no effect in the cooperative behaviour toward the out-group.

In the paper, we discuss possible reasons behind these results. One possibility is that high status individuals are more likely to invest in conflict because, in these situations, they have more to gain and lose (as recently modelled by Gavrilets & Fortunato 2014). Here, I would like to discuss another possibility: that individuals may not be necessarily behaving altruistically in situations of conflict. Our study was only based on two measures – and others have indeed found some association between conflict and cooperation – but perhaps there are more evolutionarily parsimonious explanations for the behaviour of individuals during wartime, such as reputation concerns and enforcement mechanisms. For example, a quick look at large-scale conflicts with high casualty numbers and where good recruitment data is available does indicate the importance of enforcement in conflict. In WW2, 61% of all US servicemen were forced to fight in the war as draftees; in the UK, almost every man between 18 and 41 years old was forced to enlist. In Vietnam, despite only 25% of soldiers being draftees, a disproportionate 35% of the casualties were of drafted soldiers. In this case, the volunteers (non-drafted soldiers) were not necessarily ready and willing to fight for their country, but by volunteering they were able to choose their assignment, avoiding the most dangerous sections such as the infantry, which was then mostly composed of draftees. A recent review of inter-group warfare in small scale societies also found that individual benefits – mostly related to reputation and status – better explained intensity of conflict than group-level benefits, again demonstrating how individual altruism may not be an important motivation for conflict. These examples are not conclusive evidence for whether individual altruism plays a key role in conflict or not, but they suggest that perhaps we are too quick to attribute altruistic tendencies to individuals in situations of war.

The idea of altruism originating from war is perversely attractive. Yet, the empirical question of whether wars are, in fact, won because group members are more willing to sacrifice themselves for the group has not been satisfactorily answered. Mathematical models neatly draw out the battle-lines for alternative accounts for the evolution of unique cooperative tendencies. But it is only through continued empirical work – ideally using real world groups and naturalistic measures of cooperation – that we can settle on a shared conclusion.


Silva, A.S., & Mace, R. (2014). Cooperation and conflict: field experiments in Northern Ireland. Proceedings of the Royal Society B: Biological Sciences, 281, 20141435.

Gavrilets, S., & Fortunato, L. (2014). A solution to the collective action problem in between-group conflict with within-group inequality. Nature Communications, 5, 1–11.

Bauer M, Cassar A, Chytilova J, Henrich J. (2014). War’s enduring effects on the development of egalitarian motivations and in-group biases. Psychol. Sci. 21, 47–57.

Bowles S. (2009). Did Warfare Among Ancestral Hunter-Gatherers Affect the Evolution of Human Social Behaviors? Science 324: 1293-1298.

Choi J-K, Bowles S. (2007). The coevolution of parochial altruism and war. Science 318, 636–640

Bowles S, Choi J-K, Hopfensitz A. (2003). The co-evolution of individual behaviors and social institutions. J. Theor. Biol. 223, 135–147

Milgram, S., Mann, L., & Harter, S. (1965). The lost-letter technique: a tool of social research. Public Opinion Quarterly, 29(3), 437.

Glowacki, L. & Wrangham, R. W. 2013 The role of rewards in motivating participation in simple warfare. Hum. Nat. 24, 444–60.

National WWII Museum. Profile of US Servicemen (1941-1945). URL:

The American War Library (1997). Vietnam War Casualties by Volunteer or Drafted, Enlisted. URL:

Wikipedia (2014). Conscription in the United States – Vietnam_War. URL:


Published On: October 6, 2014

Antonio Silva

Antonio Silva

I am interested in trying to understand why and when people cooperate with each other. I use naturalistic methods and field experiments to determine the individual and ecological contexts that promote cooperative behaviour. I’m working in Northern Ireland investigating how inter-group conflict between Catholics and Protestants affects within and between group cooperation. My other work includes running RCTs on how social norms affect university tuition fees payment rates and looking at the role of fairness and conformist norms on 3rd party punishment.

I also work at the Behavioural Insights Team on a broad range of topics, from reducing loneliness, improving money management and raising educational attainment.


  • Scott Atran says:

    We are working with foreign fighters joining ISIS. 75-85% join up through friends (much the same as was the case for those seeking out Al-Qaeda), 10-15% join through family. Very few high status people in this, and very willing to die for their buddies, and couldn’t give a hoot in hell for the locals (which wasn’t the case before), and they could care less about money or other material comforts.

    We’re also doing fieldwork among the various confessional groups in Lebanon. As the threats from the Syrian war accumulate, there is increasing willingness to make costly sacrifices fro the in-group.

    We find in-group love seems to increase under isolation and threat

    Here’s the abstract of paper that will soon appear in PNAS

    Motive attribution asymmetry for love vs. hate drives intractable conflict

    Adam Waytza, Liane L. Young, Jeremy Ginges

    Department of Psychology, The New School of Social Research, New York, NY 10011

    Edited by Susan T. Fiske, Princeton University, Princeton, NJ, and approved September 22, 2014 (received for review July 25, 2014)

    Five studies across cultures involving 661 American Democrats and Republicans, 995 Israelis, and 1,266 Palestinians provide novel evidence of a fundamental bias, what we term the “motive attribution asymmetry,” driving seemingly intractable human conflict. These studies show that in political and ethnoreligious intergroup conflict, adversaries tend to attribute their own group’s aggression to ingroup love more than outgroup hate and to attribute their outgroup’s aggression to outgroup hate more than ingroup love. Study 1 demonstrates that American Democrats and Republicans attribute their own party’s involvement in conflict to ingroup love more than outgroup hate but attribute the opposing party’s involvement to outgroup hate more than ingroup love. Studies 2 and 3 demonstrate this biased attributional pattern for Israelis and Palestinians evaluating their own group and the opposing group’s involvement in the current regional conflict. Study 4 demonstrates in an Israeli population that this bias increases beliefs and intentions associated with conflict intractability toward Palestinians. Finally, study 5 demonstrates, in the context of American political conflict, that offering Democrats and Republicans financial incentives for accuracy in evaluating the opposing party can mitigate this bias and its consequences. Although people find it diffi-
    cult to explain their adversaries’ actions in terms of love and affiliation, we suggest that recognizing this attributional bias and how to reduce it can contribute to reducing human conflict on a global scale.

    • Your work in ISIS and in Lebanon sounds interesting, what kind of measures of costly sacrifice are you using? It is indeed true that there are groups of people that are willing to sacrifice for the good of the group, suicide bombers being the usual example (but see Gallup & Weedon, 2013 for alternative explanations) but there are plenty of other cases.

      The issue though is whether these altruistic behaviours are strong and prevalent enough in populations to play a role in the evolution of cooperation through inter-group conflict, and at least in my sample I didn’t find that.

      Gallup, G. G., & Weedon, S. L. (2013). Suicide Bombers : Does an Evolutionary Perspective Make a Difference ? Evolutionary Psychology, 11(4), 791–794.

  • Robin says:

    As an ex-soldier, a native of Northern Ireland and a psychology post-grad I would add the following non-academic observations. – This is not really my area of knowlege.

    The Welfare State and the NHS could probably not have been created in five years without the collectivized combat experience and demobilized military middle management that were consequences of WW2.

    The NI “Troubles” can be described as a corporal’s war, with no virtually no military operations being bigger than company size on the State side and fewer than 500 members of the IRA who would actually pick up a gun or plant a bomb. Whilst it shaped attitudes it did not have much effect on most people’s behaviour.

    It is only above the Company size (150 men) that altruism truely comes into play. In order to work as a Battalion or Regiment (600-1000) then you have to stop working on the basis of mutual benefit of friendship and use unconditional loyalty/trust/obedience. I think that at this point a soldier stops using mentalizing or ToM and gains a perceptual and systemising edge which enables a bigger, faster thinking, unit. Soldiers call this change of mindstate “putting your regimental head on”

    It may only be this experience of living without using ToM for an extended period that is conducive of increased social trust and co-operation.

    • Your point about group size in situations of conflicts is important, as at a small scale reciprocity and reputation are sufficient mechanisms to maintain cooperation. In large scale warfare different mechanisms would probably be required. The point about unconditional trust and obedience makes sense in those cases, but while it may benefit the group, it may well benefit the survival of the individual as well; an analogy could perhaps be made with the behaviour of fishes in shoals under attack (

    • EALTurner says:

      //It is only above the Company size (150 men) that altruism truely comes into play. //

      If you don’t already know this –

      “Dunbar’s number is a suggested cognitive limit to the number of people with whom one can maintain stable social relationships … By using the average human brain size and extrapolating from the results of primates, he proposed that humans can only comfortably maintain 150 stable relationships.”'s_number

      There are alternative numbers. The number might also be a threshold that is higher for some individuals than others, and be influenced by available technology and ideology.

      Q1. Does “Dunbar’s number” suggest that stable social relationships can occur up to a certain level regardless of conflict?

      Q2. When people are required to handle more associations than they are cognitively able to – move beyond Dunbar’s number – is there a jump point when they need to acquire some cultural app (technology, ideology etc.) to make that possible, or else they become hostile? Thus an important distinction is not just family vs stranger it is below and beyond Dunbar’s threshold.

      • Robin says:

        One of Robin Dunbar’s suggestions for explaining group size is that alliance building is mediated in some way by ToM, and that above a certain group size we can no longer monitor the intentions of others. He sometimes uses the term intensionality. My guess is that to operate above the Dunbar’s number then you need to stop monitoring individual mental processes and use relationships that are unconditional rather than reciprocal.

        A selfish individual will always have an advantage over a selfless individual, but 1000 selfless people will always outperform 1000 selfish people. I don’t know where the crossover is but, at an anecdotal level, most people who have been in the Army know that the crossover exists, and that without selflessness you cannot have large scale collective action.

        As the Russian’s used to point out in evaluating the quality of a fighting force; quantity has a quality all of its own.

        • EALTurner says:

          My opinion – below e.g. 150 the community or army Company is essentially reciprocity taken for granted. in a situation where everyone knows everyone the free riders can be easily spotted and, if necessary, prodded into action. if you do something for someone, before long you have confidence the community will give something back. your memories of previous experience says this.

          members of this small community will also have an understanding of the community as a unit which does things for them (not necessarily represented in the same help-giver each time) and so there will be altruism for the community itself (rather than anyone specific in it). helping to dig trenches, or build a defensive wall around the village, or helping to dig a well or scout out enemy positions, could be an example of doing something for the group as a whole.

          Beyond the Company or the small community unconditional trust must be required to “reach out” to another community but this is only the initial “breaking the ice”. for breaking the ice sharing the same religious practices or army uniform might provide a bridge. however, once established the bridge must be used. the communities don’t know much else about each other, other than their on-going relationship. so despite the initial altruism that provided the bridge for the relationship to be sustained there will need to be a lot of reminders that they are both on the same side – demonstrated reciprocity. this could be tribute in return for the larger community defending their borders. it could be joint training exercises or actual combat experience.

          in terms of the army, experience fighting usually makes for better soldiers who are more likely to obey orders and work as a team. this seems perverse logic as more time spent fighting, more likely they are to get hurt or killed. yet the more they fight the more they hold the line. even above Company (Dunbar’s number) the individual needs constant evidence and reminders that if they behave in a self-less manner, strangers will also do the same for them.

  • Where the people collecting the questionnaires while going door to door Northern Irish? If not could the lack of in group bias observed potentially be related to a general desire not to appear overly ‘ingroup biased’ after completing a questionnaire about the troubles? I realise that this would raise the question of why you still found intergroup bias but it could be that outgroup discrimination is harder to dissuade than ingroup bias. Just a speculation, curious to hear your thoughts.

  • The questionnaires were answered by the participants and filled in by the researchers while at their homes (by their door or sometimes indoors; interestingly quite a lot of variation of inviting us in the house between neighbourhoods but that’s another topic). I’m sure the presence of the researchers affected the results and we discuss this further in our paper, but we did find strong parochialism with Catholics and Protestants more likely to respectively donate to Catholic and Protestants schools. So people weren’t afraid of showing bias, it’s just that this bias was not mediated by their exposure to conflict.

  • Peter Turchin says:

    Antonio, thanks for this blog and a very interesting study. However, I’d like to level one criticism at you. The theory says that it is communities which have been exposed to intergroup conflict that become more cooperative, not individuals. Yet you apparently measured whether individuals who were more exposed to conflict had a greater tendency to be parochial altruists. This is a very important distinction, and it looks to me that you are not really testing the theory.

    • Peter, the theoretical models of parochial altruism that we tested in paper explicitly model individual behaviour (as individual agents in simulations). Also, I may be misunderstanding your point but communities are made out of individuals so my if on average individuals of a community don’t behave altruistically in conflict towards their conspecifics, the communities are not going be more cooperative either, right?

      • Peter Turchin says:

        Antonio, you are looking to capitalize on variation within groups, whereas my point is that the theory primarily speaks to variation between groups. It’s an issue of multilevel selection.

        • Yes, theoretically a trait could be selected at the group level if i) there’s greater between group variation than within and ii) the trait is advantageous to the survival of the group (plus issues around migration levels and few other things). We don’t formally test i) in our study as we only have two groups, but nevertheless the within variation in cooperative behaviour is larger than the between group variation. In relation to ii) and according to the multi-level parochial altruism models we should expect increased levels of ingroup altruism in a situation of conflict but we don’t find that. What do you think would be a good empirical test of the models?

          • Peter Turchin says:

            “within variation in cooperative behaviour is larger than the between group variation” – is this really true? When Catholics preferentially give to Catholic charities, and Protestants to Protestant charities?

          • Depends on how you define the trait under selection, but if you define the trait as parochialism/in-group bias then there’s is not much between group variation. I think this would be in line with the same approach that for example the Henrich et al papers use to find a lot of between-group variation in cooperative behaviour and in contrast Lamba & Mace, 2012, Holland et al, 2012, etc find comparable within-group variation.

            Although I do stress that my paper is not a formal test of the between/within group variation issue, I’m just pointing out that there seems to be more variation between individuals (based on SES) than there is between Catholics and Protestants.

            Silva, A. & Mace, R. 2014 Cooperation and conflict: field experiments in Northern Ireland. Proc. R. Soc. B Biol. Sci. 281, 20141435

            Holland, J., Silva, A. S. & Mace, R. 2012 Lost Letter Measure of Variation in Altruistic Behaviour in 20 Neighbourhoods. PLoS One 7, e43294.

          • Wrong citation above: Lamba, S. & Mace, R. 2011 Demography and ecology drive variation in cooperation across human populations. Proc. Natl. Acad. Sci. 108, 14426–14430.

  • nichola says:

    This paper (using real-world data) seems to partially support your idea that ‘altruistic’ contributions to war are at least in part driven by fear of punishment for defecting.

    • We refer to that Turkana paper in our article as an example of punishment driving cooperation. It still doesn’t solve the costly punishment issue but at least in the Turkana case the punishment cost is apparently diffused by everyone of the free rider’s age-group taking part on the punishment.

  • Jonas says:

    I’m sure you’ve seen this, but one of the best passages about the connection between war and societal cooperation is from Scott Fitzgerald. Talking about WWI:

    “That’s different. This western-front business couldn’t be done again, not for a long time. The young men think they could do it but they couldn’t. They could fight the first Marne again but not this. This took religion and years of plenty and tremendous sureties and the exact relation that existed between the classes. The Russians and Italians weren’t any good on this front. You had to have a whole-souled sentimental equipment going back further than you could remember. You had to remember Christmas, and postcards of the Crown Prince and his fiancée, and little cafés in Valence and beer gardens in Unter den Linden and weddings at the mairie, and going to the Derby, and your grandfather’s whiskers.”

    “General Grant invented this kind of battle at Petersburg in sixty- five.”

    “No, he didn’t — he just invented mass butchery. This kind of battle was invented by Lewis Carroll and Jules Verne and whoever wrote Undine, and country deacons bowling and marraines in Marseilles and girls seduced in the back lanes of Wurtemburg and Westphalia. Why, this was a love battle — there was a century of middle-class love spent here. This was the last love battle.”

    What self interest does the elite possess to treat the middle class with love? So they’d have an army to go in the meat grinder for them if need be.

    Of course, the Russians have never been known for treating their middle class or lower class with love, but seem to have plenty of bodies and resilient minds to go into a meat grinder. But in democratic societies, it’s one of the considerations.

  • O.Voron says:

    I read your paper published by the Royal Society.

    Your findings show high level of parochial altruism within communities and “reduced levels of out-group cooperation” as I would expect.

    “” The majority of people chose to donate (68.0%)…76.1% to an in-group school … We found clear evidence for the existence of parochialism…we found wealthy people and wealthy neighbourhoods associated with more help to the in-group…We also found that people with children were more likely to donate, but specifically to in-group not out-group schools (table 2). ”

    If I understand it correctly, your point is that people who had experienced sectarian violence and those who had not, demonstrated the same level of parochial altruism ( and it’s high ), the conclusion being that somebody’s exposure to a violent conflict does not increase his/hers cooperation with in-group.

    Can it be that _all _ people there are exposed to the conflict one way or the other? It is not necessary to be personally attacked to experience it. There are many other ways – through watching the news, through listnening to family stories, through feeling the rising tensions in the wake of the yearly Orange marches and so on. After all, although “The intensity of the conflict has eased since the Good Friday Agreement in 1998, but in 2011 alone over 130 sectarian bombings and shootings were recorded”, to cite your paper.

    I happened to visit Belfast once. It’s a relatively small place so it would be almost impossible for anyone living there to escape at least some exposure to the conflict. This might explain why there is no much difference in cooperation levels.

    If you repeat your experiments in some quiet and peaceful communities in England and Ireland and compare them to your Belfast findings, that will be interesting to see.

    • It’s possible, but there was quite a lot of individual variation of exposure to conflict, and crucially this predicted how people behaved towards the out-group which indicates that our measure does capture something about the conflict. What it didn’t affect was how people behaved towards their own group.

      • O.Voron says:

        ‘there was quite a lot of individual variation of exposure to conflict, and crucially this predicted how people behaved towards the out-group’

        It is kind of obvious that the more people are exposed to conflict with an out-group, the less desire they have to cooperate with the said group.

        ‘What it didn’t affect was how people behaved towards their own group.’

        My point was that their in-group cooperation is uniformly high according to your findings. Probably, as high as you can get assuming that 100% cooperation is not achievable. (Or is it?)
        If you research similar communities, Protestant in England and Catholic in Ireland, which are not under conflict, and they show the same degree of in-group cooperation that Belfast communities show, then one can say that exposure to conflict has no influence on people’s behaviour towards their own group.

  • Ross David H says:

    A hypothesis for consideration: communities under conflict change primarily not in the willingness to cooperate, but rather in the willingness to punish non-cooperators. This is harder to test, but I wonder if you could have researchers try to “cut in line” at restaurants or wherever people wait in line, and measure the willingness of locals to speak up or otherwise criticize.

    It would take some thick-skinned researchers, but it might give interesting results.

Leave a Reply to Chris Kavanagh Cancel Reply

<textarea name="ak_hp_textarea" cols="45" rows="8" maxlength="100" style="display: none !important;">

This site uses Akismet to reduce spam. Learn how your comment data is processed.