Punishment sustains large-scale cooperation in prestate warfare

Abstract

Understanding cooperation and punishment in small-scale societies is crucial for explaining the origins of human cooperation. We studied warfare among the Turkana, a politically uncentralized, egalitarian, nomadic pastoral society in East Africa. Based on a representative sample of 88 recent raids, we show that the Turkana sustain costly cooperation in combat at a remarkably large scale, at least in part, through punishment of free-riders. Raiding parties comprised several hundred warriors and participants are not kin or day-to-day interactants. Warriors incur substantial risk of death and produce collective benefits. Cowardice and desertions occur, and are punished by community-imposed sanctions, including collective corporal punishment and fines. Furthermore, Turkana norms governing warfare benefit the ethnolinguistic group, a population of a half-million people, at the expense of smaller social groupings. These results challenge current views that punishment is unimportant in small-scale societies and that human cooperation evolved in small groups of kin and familiar individuals. Instead, these results suggest that cooperation at the larger scale of ethnolinguistic units enforced by third-party sanctions could have a deep evolutionary history in the human species.

Over the last 50,000 y, humans have come to dominate the world's biota, in part because we cooperate on larger scales than other mammals. Recent models suggest that informal systems of punishment can maintain cooperation in large groups (1, 2). However, this work leaves two important questions unanswered. First, does punishment actually play an important role in sustaining human cooperation in the absence of formal coercive institutions? In laboratory experiments third parties—individuals who are not the primary party injured by a defection—bear costs to punish defectors (3, 4). However, scholars have questioned whether such punishment exists outside of laboratory settings (5) and, even if punishment does occur, whether it effectively promotes cooperation (6–8). Second, what is the scale of human cooperation? Punishment can sustain cooperation on scales ranging from small kin-based hunter-gatherer bands to large modern states (9). A common view is that the psychology that sustains human cooperation evolved in small foraging bands characterized by modest levels of genetic relatedness and repeated social interactions (10, 11), and this led to a psychology that supports band-level cooperation. According to this view, large-scale cooperation occurs only when this psychology “misfires” in novel modern social environments with coercive institutions. An alternate view is that humans have evolved to cooperate in ethnolinguistic groups (groups with shared cultural norms and language), comprising thousands of unrelated strangers, even without formal coercive institutions (12, 13). According to this view, the scale of human cooperation has been shaped by competition between culturally distinct groups, which led to sanctioning systems that enforce cooperation at the scale of cultural variation. Because much cultural variation is maintained along ethnolinguistic boundaries (14), these models predict cooperation to occur at the scale of ethnolinguistic units. Cooperation on the same scale can also be promoted through genetic evolutionary processes that favor ethnic parochialism (15).

Existing data do not answer these questions. Studies of contemporary hunter-gatherers show that they cooperate extensively in small face-to-face groups (16, 17), and it is evident that politically centralized societies can generate cooperation among thousands of strangers (18). However, numerous societies in human prehistory were larger than those of contemporary hunter-gatherers but still lacked centralized political authority, formal law enforcement, or other coercive institutions. Although ethnographic and oral historical accounts suggest that such societies could mobilize on large scales (19–23), there is little quantitative data on the nature of punishment and the scale of cooperation in these societies (5).

Here we report results from a quantitative study of warfare among nomadic Turkana pastoralists in East Africa that sheds light on these two questions. The Turkana are a large ethnolinguistic group with the social organization of a small-scale society. They are politically uncentralized, egalitarian, and economically undifferentiated. They lack formal or centralized institutions of leadership or coercive authority. They reside in nomadic settlements comprised of households that disperse and aggregate seasonally. The combination of large population size, small-scale social organization, and temporary encampments is not prevalent in contemporary hunter-gatherer populations (22), but did occur in some historically known hunter-gatherers (22, 24, 25). Warfare is a high-stakes form of cooperation. Individual warriors incur the costs of injury or death, but the gains from victory, such as defense, deterrence, or territorial expansion, benefit all. Even when the benefits of victory are not pure public goods (e.g., captured loot), there are still opportunities for shirking during combat that reduce individual risk but also lower the chance of victory. Although archaeological and ethno-historical data showed that war, raiding, and feuding were prevalent in societies without formal political structure (26, 18), few studies address how these societies solved the collective action problem in intergroup conflict.

Based on a representative sample of 88 recent Turkana raids, we find that informal punishment plays an important role in sustaining cooperation in Turkana warfare. Participation in raids has substantial private costs: warriors risk death during combat. The benefits include the jointly acquired loot and nonexcludable benefits, such as deterrence and territorial gain. Cowardice and desertion occur frequently and are punished by the community. Third parties play a critical role in adjudication and meting out punishment, which includes serious sanctions, such as collective corporal punishment and fines. We also find that cooperation occurs at scales much larger than that of foraging bands. Raiding parties are comprised of several hundred unrelated warriors. Participants in a raid are not day-to-day interactants. A vignette study suggests that Turkana norms governing warfare benefit the Turkana as a whole, not smaller social groupings, even though the Turkana number a half-million people, most of whom have never met. Our results indicate that human cooperation, sustained through sanctioning of free riders, could have evolved at the scale of ethnolinguistic groups comprising large numbers of unrelated and unfamiliar individuals and suggest that our cooperative and sanctioning psychology should be shaped for such an ancestral social environment.

The Turkana

The Turkana are nomadic subsistence pastoralists who live in the arid savanna of East Africa. They herd cattle, camel, sheep, goats, and donkeys, and subsist on milk, blood and meat from livestock, and maize flour obtained through trade (27, 28). They live in temporary camps, migrating 10 to 50 km in search of fresh pastures and water every 15 to 40 d. The Turkana were estimated to number 476,200 people in 2006 (29). Turkana speakers are divided into 20 geographic groupings, called “territorial sections”. Membership is acquired by birth and entitles herdsmen to graze and camp anywhere in the territorial section. Clans are based on patrilineal descent, but Turkana men do not reside with clan members for more than one or two generations. Much more important are age-groups that link men born within a 5- to 6-y period (30). Households that camp together comprise a settlement. In the wet season, households are dispersed. As the dry season progresses, settlements increase in size because households aggregate around scarce water. Political and military authority are not centralized. Senior age-groups and elders have decision-making privileges and arbitrate disputes. There is no hereditary leadership. Leadership roles are taken on by men who acquire prominence through their ability as diviners, their status as warriors, or capacity to make wise migration decisions. Although they play a role in coordinating people for action, such leaders are not vested with coercive authority. The Turkana believe in a supreme being, Akuj (31). Diviners and spirits are intermediaries between him and the people. Success and failure of raids can be attributed to the plans of Akuj, and warriors sometimes justify their conduct as having occurred according to the will of Akuj. The Turkana raid other ethnic groups (28, 32), a practice with a long history (33, 19), although fire-arms began replacing spears in the 1970s. Warriors do not train together, nor is there any formal military command structure. Although commercial cattle-raiding is practiced in some parts of East Africa, community-endorsed, noncommercial raiding is the predominant form of raiding where we conducted the study. Nation-state institutions have minimal influence in this region.

Results reported here are based on in-depth semistructured interviews with a representative sample of 118 men who were questioned about the most recent offensive raid in which they had participated. This questioning yielded detailed accounts of 88 different raids. We also conducted vignette studies to probe people's reactions to the acts of fictitious warriors, and reconstructed the family structure of 107 men and 121 women to estimate mortality from warfare. We describe methods, data quality, intersubject variability, and possible biases in SI Appendix.

Warfare Is an Important Source of Mortality

Warfare is a significant source of male mortality in Turkana society (Fig. 1). Between puberty and the start of their reproductive period, 14% of Turkana men die in warfare, accounting for 45% of mortality during that life stage. During their reproductive period, 9% of men die in warfare, accounting for 60% of mortality during that period. The reproductive period begins with marriage or the birth of a child and ends when a person no longer sires children. Twenty percent of all male deaths (including infants and children) are a result of warfare.

Fig. 1.

Percentage of the male population dying from raids and other causes. Estimates are based on survivorship of full siblings and offspring of a representative sample of 107 men and 121 women. The sample consists of 335 individuals in the middle childhood up to puberty category (∼6–16 y of age), 174 in the puberty up to reproductive category (∼16–32 y of age), and 213 in the reproductive category (∼32–65 y of age). Mortality from raids includes offensive and defensive warfare. The reproductive period starts with either marriage or having a child and ends when a person cannot sire children. Mortality in the “other” category is predominantly from disease. Error bars show one SE.

Men experience mortality from both offensive and defensive warfare. Offensive warfare occurs in the form of raids initiated by the Turkana, with the goal of acquiring cattle from neighboring ethnic groups. These raids take two forms: stealth raids, which involve a few men who aim to find poorly-guarded livestock and take it, and force raids, which involve larger numbers of men who plan to engage in combat. These large raiding parties surround settlements at night and attack at dawn, or target grazing areas, watering wells, and migrations during the day. Neighboring ethnic groups are also armed and launch similar attacks against the Turkana. Among adult males, 54% of mortality due to warfare occurred in offensive combat and 46% in defense. Women and children do not participate in offensive combat or actively engage in combat during defense, but can be killed during attacks at settlements, at watering holes, and during migrations.

Participation in Warfare Is Costly and Produces Collective Benefits

A Turkana warrior who participates in a raid incurs considerable private costs, primarily the risk of injury or death. Based on casualty rates in 47 force raids, a Turkana warrior has a 1.1% chance of being killed on each raid (Fig. 2), and a 1.3% chance of being killed if the raiding party encounters the opposition and engages in combat (which occurred in 89% of the force raids). In our sample of 34 stealth raids, 41% involved combat, but none produced any fatalities among the Turkana.

Fig. 2.

The risk of dying is the primary private cost borne by those who participate in raids. A warrior's expected risk of death when he departs on a force raid is 1.1%. If combat ensues, his expected risk of death is 1.3%. Estimates of the per capita fatality rate are based on a representative sample of 34 stealth raids, 47 force raids, and 41 force raids in which combat ensued. Error bars represent 68% confidence interval.

The primary goal of an offensive raid is to acquire livestock, which is divided after the raid. The average gain per participant is 11 cows in force raids and 3 cows in stealth raids (Fig. 3). Combatants may also acquire the firearms and personal belongings of enemies they kill in combat. Rarely, a warrior may kidnap a young boy for use as labor in herding.

Fig. 3.

The primary benefit of joining a raid is the livestock acquired. The average per capita gain in a stealth raid is 3 cows and in a force raid is 11 cows. Estimates are based on the share of livestock acquired by the focal warriors in 34 stealth raids and 53 force raids. Error bars represent 68% confidence interval.

Raids also create collective benefits that flow to everyone in the community. Retaliatory raids produce deterrence, a nonexcludable good. Such raids are not just a byproduct of a desire for loot; revenge is also an important motive. One participant said that he joined a raid because a neighboring ethnic group had just raided the Turkana. He did not lose any animals, but said that his brother-in-law lost his, and added, “Even if it's not your family, they are Turkanas, like me. So we have to go fight back [against] the enemy.” In one raid, the primary goal was revenge, not animals. The raid was initiated quickly after a settlement was attacked so that the enemy would know this attack was retaliation for the recent raid. The warriors set out knowing that they were not likely to acquire animals, and they considered the raid a success even though they returned empty handed.

Large-scale force raids can increase access to grazing areas and crucial dry-season watering holes. The precipitating event for one of the largest raids in the sample was a hunter's discovery that herders from another ethnic group had settled at a watering site that is typically used by the Turkana. The stated goal was to drive them away from Turkana territory. Furthermore, settlements move away from sites of raids, and this can lead to substantial shifts in the territorial ranges of ethnic groups over time.

Cooperation in Warfare Occurs at Large Scales

The Turkana mobilize a large number of warriors for force raids (Fig. 4). On average, 315 men participate in a raid, and the median is 248 men. Furthermore, participants in force raids are not all close associates: they are drawn from an average of five age-groups, four settlements, and three territorial sections (Fig. 5). Territorial sections number about 25,000 people, settlements and age-groups a few hundred. Although there are social ties that bridge these groupings, day-to-day interactions typically occur within them. When asked whether they recognized the men gathered for a raid, participants typically responded that there were some men they knew and some men they did not recognize. A warrior has on average only four of his close kin—father, son, full or half brother, brother-in-law, or full cousin—with him on a force-raid. Stealth raids are much smaller, averaging 12 warriors. A warrior has on average one close kinsmen with him on a stealth raid.

Fig. 4.

The number of warriors in a force raiding party. The average size of a force raiding party is 315 warriors and the median is 248.

Fig. 5.

The number of distinct social units from which participants of a force raiding party are drawn. Settlements, age groups, and territorial sections are social divisions in Turkana society within which day-to-day interactions are most likely to take place. Force raiding parties are comprised of men drawn from an average of four settlements, five age-groups, and three territorial sections.

Free Riding Occurs During Raids

A Turkana man is regularly faced with the option of joining a raid or staying back. Recruitment is informal: a raid is initiated by one or a few men who send word out, encouraging men from other settlements to join, and over the course of a few days warriors from various settlements trickle in to the settlement that has initiated the raid. A man can refuse to join the raiding party, but he needs a good reason, such as the lack of a suitable person to take over herding duties, the need to defend the settlement, being ill, not having a rifle, or having joined recent raids.

The raiding party travels on foot for one or more days to the place where scouts have located an enemy settlement. Along the way, men turn back. They may escape at night unnoticed, or tell their age-mates that they are ill, worried about their herds back at home, or have a premonition that they will die on this raid. Such desertions occurred in at least 43% of force raids (Fig. 6). This estimate is a lower-bound because it only includes desertions of which the interviewee was aware. Participants said that they would not know of discreet departures by men outside their age-group or settlement.

Fig. 6.

Proportion of force raids in which free riding in the form of cowardice and desertion was noticed and norms regarding loot-sharing were not followed. The estimate for desertion is based on 49 force raids that were initiated, cowardice is based on 40 force raids in which combat occurred, and loot division is based on 32 force raids in which the raiding party was successful. Error bars show one SE.

Warriors have many opportunities to reduce their risk during combat. The fighting begins when the raiding party surrounds the settlement, pasture, or watering point and opens fire. As soon as the offense gains the upper hand, fast young men begin to drive the cattle toward Turkana land, as seasoned warriors engage in a holding action to keep the pursuit at bay, continuing to fight as they retreat. During combat, men constantly urge each other to stand their ground, fire their weapon, and not to run away. In 45% of force raids in which combat occurred, the focal warrior knew of men who lagged behind others during combat, failed to fire their weapons, ran away when the fire-fight began, or retreated too quickly (Fig. 6). Again, this estimate is a lower-bound because warriors report that they mainly know about the behavior of men standing alongside them.

Finally, the jointly acquired loot can be unfairly appropriated. The men driving the cattle homeward are supposed to continue until they are safely within Turkana territory and then wait for the rest of the warriors to rejoin them. Once regrouped, they divide the spoils before they disperse to their respective settlements. Norms specify that members of senior age groups get a larger allocation than members of junior age groups, and that men within an age group get roughly equal shares. However, the loot-sharing system failed in 56% of the force raids, and some participants took whatever livestock they were able to drive off (Fig. 6). Some men position themselves in the rear, a safe distance away from combat, wait for the vanguard to start releasing the livestock, and appropriate the best of the spoils.

Free Riders Are Sanctioned

Informally enforced norms allow the Turkana to partially solve the collective action problem in warfare. In 47% of the force raids in which desertions were reported, at least one of the deserters was sanctioned, and in 67% of the force raids in which cowardice was reported, at least one of the cowards was sanctioned (Fig. 7). There are two levels of sanctions. When a warrior's behavior in a raid deviates from that of his comrades, he is subjected to informal verbal sanctions by his age-mates, women, and seniors. If there is consensus in the community that the act merits more serious sanctions, corporal punishment is initiated. Corporal punishment is severe: the coward or deserter is tied to a tree and beaten by his age-mates. One participant had scars on his torso from being whipped by his age group more than a decade earlier. During this process the violator is told not to repeat this mistake. Corporal punishment often culminates with the violator pleading for forgiveness and sacrificing an animal from his herd. Sanctioners do not confiscate the violator's share of the loot. Instead, the loss of the animal represents a “fine” that is then consumed by his age-mates and other older men. In force raids, there were 20 violations (combining desertion and cowardice) in which at least one of the violators was sanctioned (from Fig. 7). In nine of these cases corporal punishment or fines were imposed. In one case only 3 d had passed since the raid, and there was talk that further sanctions would soon be initiated. Of the nine cases in which serious sanctions were imposed, six involved corporal punishment followed by fines, two involved only corporal punishment, and one involved only a fine. In addition to these nine cases, in 2 of the 20 sanctioned cases the coward's share of the loot was reduced at the time of loot division. In stealth raids, only one of the five sanctioned violations involved a fine, and none involved corporal punishment. In a second case there was talk that further sanctions should be imposed but this had not yet been carried out at the time of the interview (20 d after their return). It is possible that the mechanisms maintaining cooperation in force and stealth raids differ, and collective corporal punishment may be more important in force raids.

Fig. 7.

Rate of sanctioning of reported desertions and cowardice. Sanctions include verbal sanctions, reduction in a share of the loot, or corporal punishment followed by a fine of an animal from the violator's herd to be killed for his age-mates and seniors. The estimates for force raids are based on 19 raids in which desertions were reported and 18 raids in which cowardice was reported. For the stealth raids, there were three raids in which desertion occurred and six in which cowardice was reported. Error bars show one SE.

The patterns of sanctioning suggest that the system relies on third parties. First, community consensus determines whether someone deserves corporal sanctions. The violator's behavior in the course of the raid is discussed extensively, especially among his local age group. When people from his settlement see the violator, they ask him what happened and why he did what he did. They chastise him for endangering other men and remind him that his fellow warriors died that day in battle. Opinions about whether he should face further punishment often differ. Some may feel he should be excused this time as he has not acted like this before. Others will say his behavior was justified because it allowed him to save himself. Some may feel there is no use trying to change him because he is an inveterate coward. Still others will say that unless he is punished he will continue in these wrong ways. Second, once a consensus emerges, members of the violator's age group are responsible for administering punishment, even if they did not participate in the raid and did not experience the consequences of the violation. Third, the collective nature of the corporal sanctions is important. A sufficient number of age-mates must be present, and imposition of sanctions is often delayed because men are dispersed widely and busy with their herding duties. Fourth, both a failure to mete out justified punishment and second-party sanctioning without prior consensus are met with community disapproval. Finally, both deserters and cowards in force raids face similar corporal punishment. If only the victims of a violation initiate sanctions, we should expect that cowards would be punished by the men they endangered, and that deserters should escape sanctions. In raids numbering hundreds of people fighting without a well-defined unit structure, a desertion does not create a primary injured party with a sufficient motivation to impose sanctions.

The fact that direct punishment creates incentives to contribute to the combat efforts does not imply that it is the only factor sustaining the observed large-scale cooperation. We think that positive incentives—rewarding men who are brave in combat—also play a crucial role. However, rewards for bravery in this ethnographic context, are not associated with a single act. Instead, a warrior accumulates these benefits over a long period. Accurately measuring these diffuse benefits and the diffuse costs associated with indirect sanctions is difficult, and distinguishing the effect of behavior during warfare from the effect of other factors that affect a person's value as a social or mating partner is beyond the scope of the present study. However, we can conclude from the rate of direct punishment that, important as indirect sanctions and rewards may be, they cannot be the full story: if they created sufficient incentives, there would be no need for direct punishment.

Norms Governing Warfare Are Beneficial on Large Scales

A vignette study indicates that Turkana norms governing warfare benefit the ethnolinguistic group, not smaller social groupings in which people are more likely to know each other. Twenty-four subjects, all drawn from the Kwatela territorial section, heard two scenarios. In one scenario, two Kwatela warriors raid animals from the Toposa, another ethnic group, and bring them home to Kwatela land. In the other scenario, two Kwatela warriors raid animals from the Lukumong, another Turkana territorial section, and bring the loot home to Kwatela land. We alternated the order in which the vignettes were presented. Kwatela subjects had very positive attitudes to the warriors who went to Toposa land to bring animals. In stark contrast, they considered the raiding of the Lukumong to be wrong and were strongly motivated to sanction the warriors who did this (Fig. 8). Even though the subject's own territorial section would benefit from the act, most subjects thought the act was wrong, were displeased with the warrior, and thought he should be criticized and punished. Subjects also wanted to terminate social interactions with the warrior. Most subjects would not stand next to this warrior in a raid, entrust their herds with him, lend him a goat if he needed to borrow one, or let their daughter marry him.

Fig. 8.

Vignette study on the scale at which raiding norms are group beneficial. People from the territorial section that the subject belonged to either benefited when two warriors went to raid another Turkana territorial section, or when they went to raid another ethnic group. Bar lengths give the proportion of subjects agreeing with the proposition on the left. Each condition had 24 male participants in a within-subject design, with counter-balanced order. Error bars show 1 SE.

The scale at which norms are group-beneficial has a big impact on Turkana demography. Homicide or internal violence accounts for only 1% of adult male mortality within nomadic settlements, substantially lower than interethnic warfare that accounts for 50% of adult male mortality (from the same dataset as Fig. 1; “adult” refers to “puberty up to reproductive” and “reproductive” life stages). The suppression of within-group violence is noteworthy because most young Turkana men possess firearms, and have the opportunity to use force to resolve conflicts and acquire goods from other members of their society. We also think this pattern is not a recent feature of postfirearm warfare; ethnographers described norms that promoted raiding other tribes and deterred violence within tribes even during the period when combat was conducted with spears (19).

Conclusion

Although most scholars recognize that cooperation takes place in societies lacking a formal political structure, many think it occurs only at a modest scale and its evolution can be explained by kin-based altruism or reciprocity among familiar individuals. Contrary to this, Turkana sustain costly cooperation in warfare at a remarkably large scale without political centralization, formal coercive institutions, or even a political system based on descent (34). The punishment of free riders plays a role in sustaining this large-scale cooperation. Punishment is based on informal third-party sanctions. Although cultural norms specify how free riders should be treated, the enforcement of these norms requires the participation of the community at large, and no particular individuals are invested with the authority to implement sanctions.

The pattern of cooperation and sanctioning observed among the Turkana has important implications for understanding the origins of human cooperation. Sanctions can sustain cooperation on a wide range of scales. Thus, a full explanation of human cooperation must specify the processes that determine the scale at which norms sustain cooperation. Our results are not consistent with the view that human cooperative psychology—the mechanisms that motivate individuals to cooperate and to sanction free riders—was shaped by evolution at the scale of foraging bands. Turkana society lacks formal coercive institutions and the Turkana live in small nomadic settlements in which the cues of membership are much like those of foraging bands. Therefore, a psychology evolved to sustain cooperation in foraging bands should cause the Turkana to mobilize for raids at the scale of settlements. Instead, Turkana raiding parties are mobilized from multiple settlements, age groups, and territorial sections yielding large groups of warriors who are not kin or day-to-day interactants. Furthermore, norms governing warfare and their enforcement by third parties appear to create behavior that benefits the entire Turkana ethnolinguistic group at the expense of smaller social units. This finding is striking because the Turkana number around 500,000 individuals who commonly live hundreds of kilometers away from each other and never interact. Our results suggest that the human cooperative psychology was shaped by processes that have favored cooperation at the scale of ethnolinguistic groups comprised of large numbers of people who are not necessarily related to or familiar with each other. The evolution of cooperation at this scale requires a different kind of explanation. Cultural group selection is one possibility that predicts cooperative outcomes to occur at the scale of cultural variation, such as the scale of ethnolinguistic groups. An alternate solution relying on genetic transmission is presented in ref. 15.

These facts also shed light on the distribution of warfare across species. Warfare can create benefits in a wide range of ecologies. Because large war parties are likely to defeat smaller ones, on average, larger war parties will be more profitable if the collective action problem inherent in warfare can be solved. Thus, the scale at which the collective action problem can be solved constrains the scope of warfare. Patterns of warfare across animal societies are consistent with this claim. In most mammals, individuals participate in collective violence only in small groups, or with kin. For example, chimpanzee border patrols typically involve 5 to 15 males of the same community (35). In contrast, costly large-scale intercolony combat is common in a range of eusocial insects. The fact that humans are able to solve the collective action problem in large groups may explain why combat involving groups of hundreds of warriors on a side has been documented in hunter-gatherers (22–24, 36, 37) (see SI Appendix, Section 5 for more examples).

These results imply that large-scale cooperation may have been common for a long period of human evolutionary history. The Turkana cooperate in large-scale combat without political centralization and formal institutions. This fact suggests that early human societies could have done the same. Moreover, warfare is only one domain of collective action, and is particularly costly to individual participants. Although there could be something unique about warfare that facilitates cooperation, it is plausible that if societies can solve the collective action problem in large-scale warfare, they can also solve myriad lower-stakes collective action problems, and this ability likely played a critical role in the ecological success of our species.