Deviant Cohesion and Unauthorized Atrocities: Evidence from the American War in Vietnam

Abstract

Why do soldiers engage in unauthorized atrocities? This article explores this question by analyzing the use of postmortem mutilation by American soldiers during the Vietnam War. I show that such acts were remarkably frequent, despite being explicitly prohibited by military policy, and argue that individual-level variation in participation in such violence is explained by social dynamics within military units. Soldiers used mutilation mostly as a means of avenging enemy atrocities or deaths among comrades. Revenge motives were stronger when soldiers shared particularly strong social bonds. Whether these motives resulted in unauthorized atrocity, however, depended on the extent to which discipline was maintained within military units. In units characterized by “deviant cohesion”—strong social ties and weak discipline—informal combatant norms diverged from organizational policies and promoted unauthorized atrocities as a unit-level practice. Evidence for this theory comes from a combination of archival sources and survey data gathered from a representative sample of Vietnam War veterans. A case study of a single Army unit illustrates the mechanism implied by the theory.

Many wartime atrocities are “crimes of obedience,” acts ordered or authorized by political or military leaders (Kelman and Hamilton 1989). Scholars of political violence have explained such acts in terms of a range of organization-level strategic logics (Valentino 2014). At the same time, a substantial body of research shows that not all wartime atrocities result from organizational policies (Manekin 2020; Mitchell 2004; Weinstein 2007); some occur as part of an unordered “practice of war,” a form of violence that is “driven from ‘below’ and tolerated from ‘above,’ rather than purposely adopted as policy” (Wood 2018, 514). Accounting for such unauthorized atrocities requires consideration of the motives of rank-and-file “agents” whose preferences may diverge from those of the organizational “principals” whom they ostensibly serve (Mitchell 2004).

In practice, distinguishing empirically between violence that enacts organizational strategies and that which is driven by the personal preferences of combatants is often difficult, particularly in the absence of microlevel data on individual motivations (Shaver and Bollfrass 2023, 442). To address this challenge, this article focuses on a particularly transgressive form of violence: the mutilation of corpses by American soldiers during the Vietnam War. Because such acts were explicitly and unambiguously prohibited by US military policy, analysis of postmortem mutilations provides a unique opportunity to isolate and analyze the role of personal motives in the perpetration of political violence. ¹

Postmortem mutilation involves the deliberate severing of body parts from the corpses of enemy combatants or civilians. These body parts are sometimes collected or displayed as trophies (Harrison 2012). As this article shows, such behavior was remarkably frequent among American soldiers in Vietnam: data from one of the largest-ever surveys of Vietnam War veterans (Kulka et al. 1988) suggests that at least one in 10 American soldiers who served mainly or completely in combat roles in Vietnam either engaged directly in postmortem mutilation or served in a unit in which such acts occurred. This level of participation in postmortem mutilation is puzzling. Not only does such violence clearly violate the laws of war, it also has negative consequences both for the military and for individual soldiers. As noted in the US Army’s Leader’s Manual for Combat Stress Control (Department of the Army 1994, 51), mutilation of corpses “tends to provoke reprisals, alienate world and home front opinion, and contribute to guilt and post-traumatic stress symptoms when the soldier returns home.” For individual soldiers, participation in postmortem mutilation has been found to increase the risk of post-traumatic stress and attempted suicide (Green et al. 1990, 35; Hiley-Young et al. 1995, 135, 137).

Given these costs, why would individual soldiers engage in such behavior? I argue that revenge is a key motive for unauthorized atrocities, and that both the strength of that motive and whether it gives rise to prohibited forms of violence are likely to vary across military units. In contexts of counterinsurgency, in which insurgents largely avoid open battle, hide among civilians, and sometimes perpetrate atrocities against counterinsurgents themselves, combat losses among state forces can give rise to powerful feelings of anger and desires for vengeance. I hypothesize that these emotional reactions will be more intense when soldiers share particularly strong social bonds. Whether these emotions result in unauthorized atrocities will depend on the extent to which discipline is maintained within military units. In units characterized by “deviant cohesion”—strong social ties and weak discipline—informal combatant norms can diverge from organizational policies and promote unauthorized atrocities as a unit-level practice.

Evidence for this theory comes from several types of sources. Archival documents from the Vietnam War Crimes Working Group, a Pentagon task force established in 1970 to investigate alleged American war crimes, provide valuable insights into both the motives of individual perpetrators and the degree to which official policies prohibiting mutilation were (or were not) enforced during the war. Ex-combatant survey data enables me both to estimate the prevalence of mutilation among American troops, and to test different explanations of this practice. Consistent with the theory described above, I find that soldiers who frequently lost close friends in combat, or who directly witnessed atrocities perpetrated against Americans, were more likely than others to participate in mutilation. I also find that such participation was particularly likely when strong social cohesion coincided with weak enforcement of organizational policies. I find less or no support for alternative explanations centered on individual predispositions to violence, traditions of white supremacy, or the influence of hunting practices. A case study of a single Army unit illustrates the mechanism implied by the theory.

Explaining unauthorized atrocities by American soldiers in Vietnam can provide important insights into the occurrence of such violence across a broader set of cases. Acts of postmortem mutilation by state forces have been reported in a number of conflicts, including wars in Guatemala (CEH 1999, 363), Kosovo (Physicians for Human Rights 1999, 69–70), Chechnya (Reynolds 2000), and Myanmar (Amnesty International 1988, 13). As in Vietnam, these atrocities occurred in contexts of counterinsurgency warfare in which often poorly trained recruits operated mostly in small units. Such conditions inhibit top-down supervision by senior officers, and thereby enhance the importance of unit-level social dynamics in shaping patterns of violence.

Unauthorized Wartime Atrocities

Researchers have increasingly explored the challenges that armed organizations face in trying to control the violence of their members in wartime. Armed organizations recruit combatants with varying preferences and levels of commitment; in wartime conditions, the behavior of these recruits often cannot be easily observed by organizational leaders (Abrahms and Potter 2015; Mitchell 2004; Weinstein 2007, 130–31). In the absence of top-down supervision, divergence between the preferences of organizational leaders and those of rank-and-file combatants can lead either to an “underproduction of sanctioned violence” (i.e., combatants evading orders to commit violence) or to an “overproduction of unsanctioned violence” (Manekin 2020, 5). Existing studies have shown how armed organizations try to mitigate these challenges through a variety of institutional strategies, including selective recruitment and vetting (Forney 2015), military and political training, and military discipline (Hoover Green 2017).

A growing body of scholarship has focused on the role of socialization in shaping the character and level of violence used by rank-and-file combatants (Bell 2016; Checkel 2017; Hoover Green 2018; ICRC 2020; Manekin 2020). Socialization can help armed organizations to overcome what Hoover Green (2018) calls the “commander’s dilemma”: the need to condition combatants to use violence when ordered to do so, while simultaneously restraining them from violence that might undermine organizational objectives. Through intensive socialization, armed organizations seek to instill in their members “a set of norms that construct organizationally useful violence as appropriate and valuable, and organizationally useless violence as illegitimate” (Manekin 2020, 5). A number of studies have shown that organizational efforts to socialize combatants have a significant influence on their subsequent conduct, notably by promoting restraint toward civilians (Bell 2016; Kahl 2007; Oppenheim and Weintraub 2017) and diminishing the use of sexual violence (Hoover Green 2018; Wood 2009).

As several scholars have noted, socialization of combatants can occur both through formal processes initiated by commanders (training, political education) and through informal practices among peers (Bell 2016, 507; Wood and Toppelberg 2017, 626). The latter, “horizontal,” form of socialization usually occurs within relatively small groups of soldiers, and may therefore vary considerably across military units. In some units, practices of informal socialization can contravene organizational policies (Winslow 1999), giving rise to norms and behaviors that violate organizational prohibitions. For example, Wood and Toppelberg (2017) show how informal socialization that includes sexualized hazing contributes to the persistence of sexual violence within the ranks of the US military. In some groups, participation in transgressive violence may itself serve as a means of socialization: according to Cohen (2016), armed organizations that recruit members through abduction or press-ganging use collective participation in rape as a means of building cohesion among combatants with initially low levels of trust. In combat situations, combatants serving in groups or units with more radical norms about violence (e.g., those endorsing the targeting of civilians) have strong incentives to conform to such norms, whether for instrumental reasons (maintaining acceptance and status in the group) or as a result of norm internalization (Cantin 2021, 1575–78).

This article builds on existing literature by identifying the conditions under which military units are likely to develop informal norms that endorse the perpetration of unauthorized atrocities. One of these conditions has to do with unit-level discipline: as Manekin (2020, 166) notes, soldiers involved in counterinsurgency often operate in conditions in which supervision by more senior officers is limited, and decisions about discipline made by junior commanders can have a determining influence on the norms and conduct of rank-and-file combatants. Yet weak discipline alone cannot explain why some units adopt unauthorized violent practices while others do not. Here I emphasize the role of revenge as a key motive for committing atrocities. While existing studies certainly recognize revenge as a motive for wartime violence (Butler, Gluch, and Mitchell 2007, 671; Mitchell 2004, 160; Wood 2006, 335), variation in the strength of that motive, and in the degree to which it leads to unauthorized violence, has so far been neglected.

A second contribution of this article is empirical: in testing my theory, I evaluate a crucial but rarely studied implication of the view that unauthorized violence is driven in part by “small-unit dynamics” (Wood and Toppelberg 2017, 625), namely that individual participation in such violence should vary according to the characteristics of particular military units. Systematic analysis of such individual- and unit-level variation in violence is rare in the literature, largely because of a lack of sufficiently disaggregated data on combatant behavior (Humphreys and Weinstein 2006). This article uses a combination of survey and archival data to identify and account for such microlevel variation.

Mutilation during the Vietnam War

American atrocities during the Vietnam War are well documented. ² Such atrocities were widely publicized beginning in the late 1960s following the exposure of the My Lai massacre (Hersh 1970). Subsequent investigations by government (Peers 1979) and civil society organizations (Duffett 1970) produced a wealth of information concerning the victimization of civilians by US forces. Scholars have disagreed about whether My Lai should be seen as an isolated incident (Allison 2007, 93; Solis 2017, 121) or, on the contrary, as symptomatic of a general climate of brutality that characterized much of American conduct during the war (Greiner 2009; Turse 2013). ³ Less attention has been paid to the specific phenomenon of postmortem mutilation. To the extent that mutilation in Vietnam has been studied systematically, it has been by psychologists seeking to identify the impact of such behavior on postwar mental health problems (Beckham, Feldman, and Kirby 1998; Currier et al. 2014; Dennis et al. 2017; Green et al. 1990; Hiley-Young et. al. 1995). Few scholars have systematically analyzed why American soldiers used mutilation in the first place. In this section, I show that mutilation was explicitly prohibited by US military policy in Vietnam, yet was frequently perpetrated nonetheless, suggesting de facto toleration of the practice by many lower-level commanders.

Policy

Mutilation of the living or dead was explicitly and unambiguously prohibited by American military policy in Vietnam. Rules of conduct for US forces, enshrined in the Army’s 1956 Law of Land Warfare field manual, prohibited both the “physical mutilation” of prisoners and the “maltreatment of dead bodies” (Department of the Army 1956, 36, 180). Despite not recognizing the insurgency in South Vietnam as an international conflict, the US decided early on in the war that captured Viet Cong guerrillas would be entitled to the full protections of the Geneva Conventions (Prugh 1975, 66). US military authorities in Vietnam issued directives that required all military personnel to report suspected war crimes to their commanding officer for investigation. These regulations explicitly listed “the maltreatment of dead bodies” as a war crime. ⁴ Other regulations established rules for the disposal of enemy dead that required that they be “handled in a respectful and reverent manner.” ⁵

In addition to being formally prohibited by policy, practices of mutilation were repeatedly condemned by the highest levels of the US military hierarchy. In October 1967, in response to news reports about American soldiers cutting ears off enemy dead, US Army Chief of Staff Harold K. Johnson wrote to the commander of US forces in Vietnam, General William Westmoreland, describing the mutilation of corpses as “alien to all civilized practice and the traditional attitudes shown by American soldiers” and ordering that “no incidents of this nature [be] permitted or condoned by any US Army officer or soldier.” ⁶ Westmoreland duly issued a message to all US commanders in Vietnam in which he described the practice of cutting ears off corpses as “subhuman” and “contrary to all policy and below the minimum standards of human decency” (quoted in Lewy 1978, 329). Condemnations of mutilation were repeated in subsequent years: in late 1970, the commander of the Army’s XXIV Corps in Vietnam, Lieutenant General James W. Sutherland, issued a letter on the subject in which he deplored “a general lack of understanding of what constitutes ‘maltreatment,’” and listed acts, including “defiling or ridiculing the dead,” that were specifically prohibited by military policy. ⁷

In addition to promulgating rules that prohibited mutilation, attempts were made to communicate these to soldiers arriving in theater: all newly arrived US troops received “Geneva Conventions orientation” and were given a card outlining rules for the treatment of enemy soldiers, including the specific requirement that captives “must not be tortured, killed, mutilated, or degraded” (cited in Prugh 1975, 144, emphasis added). While scholars have criticized the limited nature of instruction received by US soldiers concerning war crimes (Allison 2007, 92; Parks 1976, 20), the fact that such instruction existed at all shows clearly that mutilation was not authorized by organizational policy.

Prevalence

The very fact that American military leaders had to repeatedly condemn mutilation is evidence that such behavior was recurrent among American forces. Yet scholars disagree about how common it was. Lewy (1978, 317) argues that accounts of American atrocities in Vietnam have been frequently exaggerated, but acknowledges that “incidents” of unauthorized violence occurred: “We know that hamlets were destroyed, prisoners tortured, and corpses mutilated.” Bourke (1999, 387), in contrast, suggests that mutilation was common, claiming that “[n]early every diary, series of letters, or autobiography from the Vietnam War contains examples” of such behavior. Milam (2009, 115) calls mutilation a “somewhat common practice of many troops.” Hastings (2018, 401) claims that “the practice of mutilating enemy dead was widespread.”

Studies of Vietnam veterans seeking clinical treatment for post-traumatic stress disorder (PTSD) have found that a majority reported witnessing mutilation by American soldiers in Vietnam, and about a third reported having directly participated in such acts (Beckham, Feldman, and Kirby 1998, 780; Dennis et al. 2017, 193; Hiley et al. 1995, 132). Clearly, such estimates suffer from selection bias given that soldiers who engaged in mutilation were more likely to suffer from PTSD (Green et al. 1990, 35; Hiley-Young et al. 1995, 135). Fortunately, less biased estimates of the prevalence of mutilation among American forces can be derived from data from the National Vietnam Veterans Readjustment Study (NVVRS). Carried out in the mid-1980s as part of a congressionally mandated study of PTSD among Vietnam-era veterans, the NVVRS included the largest-ever survey of Vietnam War veterans. Drawing on a sampling frame assembled from military personnel records, the NVVRS provided “the most representative sample of all Vietnam-era veterans studied to date,” and thus “the best available basis for inferences about the entire population of Vietnam veterans” (Kulka et al. 1988, 28).

Among the risk factors for PTSD that the investigators studied was exposure to wartime atrocities, including mutilation. Veterans were asked to rate their exposure to the “mutilation of bodies of the enemy or civilians” by American soldiers on a six-point scale, ranging from nonexposure to individual participation. Column 1 of table 1 presents the distribution of responses to this question for all Vietnam War veterans surveyed (excluding a small number of respondents who refused to answer the question). Overall, the survey found that two-thirds of veterans interviewed reported no exposure at all to mutilation, while about a fifth knew or had heard about the practice, and about 8% had personally witnessed it; only a small proportion (under 2%) admitted that members of their unit had perpetrated mutilation, and an even smaller proportion indicated having personally participated in or been responsible for such actions.

Table 1 Reported Exposure to Mutilation

Source: Kulka et al. 1988.

At first glance, these responses seem to confirm the rarity of mutilation among American forces. Three things should be kept in mind, however. The first is a potential for underreporting due to social desirability bias. Given the highly transgressive nature of postmortem mutilation, some perpetrators may have been unwilling to admit their direct participation in such violence to interviewers, and may instead have reported only lower levels of exposure. Second, the figures presented in column 1 include a large proportion of soldiers who had only minimal involvement in actual combat, and would therefore have had little or no opportunity to engage in postmortem mutilation. Moskos (1970, 139) suggests that about 70% of American soldiers in Vietnam served exclusively as “rear echelon” support forces and “[could] not be considered combat soldiers except by the loosest of definitions.” Column 2 in table 1 therefore presents rates of exposure to mutilation for only those respondents (27% of the total) who reported having served “mainly” or “completely” in combat roles. These figures suggest that mutilation was not as marginal an activity as it first appears: over 5% of the combat soldiers surveyed admitted direct participation in mutilation, and another 5% reported unit-level participation.

Finally, even these reported rates of participation likely underestimate the prevalence of mutilation among American combat soldiers because of survivor bias. The NVVRS’s sample necessarily excluded soldiers killed in action and veterans who had died before the survey was implemented. Previous research has shown that participation in mutilation was strongly associated with combat exposure (Hiley-Young et al. 1995). Because those most exposed to combat were more likely to be killed, and may also have experienced higher post-service mortality (CDC 1987), respondents included in the survey were likely less exposed to combat, on average, and therefore less likely to have perpetrated atrocities, than those who died before the survey was conducted.

Overall, the figures depicted in table 1 suggest that a significant number of American soldiers in Vietnam participated in mutilation. Applying the NVVRS’s reported rate of individual participation for all respondents (1.84% [95% CI: 1.13%–2.53%]) to the total estimated population of Americans who served in Vietnam (about 2.6 million) yields an estimate of 47,680 (29,380–65,780) individual perpetrators. Alternatively, applying the reported rate of participation among combat soldiers (5.58% [95% CI: 3.25%–7.92%]) to an estimate of the number of soldiers who served mainly or completely in combat (about 702,000, or 27% of 2.6 million), yields an estimate of 39,172 (22,815–55,598) perpetrators. Even the lower bounds of these estimates suggest a significant level of participation in mutilation among American forces, even when ignoring the reporting biases noted above.

Punishment

How did mutilation become this widespread despite being officially prohibited by policy? Wood (2018, 521) argues that unauthorized violence can become frequent when commanders “either will not or cannot effectively prohibit” it. One indicator of effective prohibition is the frequency with which perpetrators of unauthorized violence are punished. In the US military, such punishment could take the form of a judicial process (court-martial) or of nonjudicial punishment imposed directly by unit commanders. Though we know that the latter was far more common than the former (Allison 2007, 71), there is unfortunately no comprehensive record of either form of punishment in Vietnam (Solis 2017, 122), and thus no way of precisely estimating how often mutilation was punished and in what manner. Two partial databases of judicial punishments in Vietnam, provided by Parks (1976) and Lewy (1978), suggest that mutilation was rarely punished through court-martial: Parks (1976, 18) reports only three convictions for mutilation (out of a total of 259 courts-martial) from 1965 to 1973, while Lewy (1978, 348, 456) reports five convictions for mutilation in the US Army and one in the Marine Corps.

The relative rarity of judicial punishments for mutilation is confirmed by the files of the Vietnam War Crimes Working Group. The working group was established by the Department of Defense in the wake of the My Lai massacre and provides the most comprehensive documentation of internal military investigations of alleged war crimes (see Nelson 2008). In total, the working group investigated 36 cases in which American soldiers had allegedly engaged in mutilation (each “case” could involve numerous individual allegations). ⁸ Of these, the group found that suspects were court-martialed in only eight cases, and convictions and punishments were decreed in only six. ⁹ In a majority of cases investigated by the working group, allegations of mutilation did not give rise to official military investigations or trials, and only came to the attention of the investigators because of allegations made by ex-servicemen, or because body parts sent by mail from Vietnam were intercepted by US Customs. In almost all of these cases, it was not possible for the working group to substantiate the alleged acts of mutilation (because veterans making the allegations refused to collaborate with investigators, for example, or because it was impossible to prove that intercepted body parts were acquired through the deliberate desecration of a corpse), and no charges were therefore brought.

If mutilation was rarely tried by court-martial, this suggests that such violations were usually dealt with by means of nonjudicial punishment. This, in and of itself, would indicate that such acts were not perceived as particularly serious offences, nonjudicial punishments being specifically intended to allow commanders to “resolve allegations of minor misconduct against a soldier without resorting to higher forms of discipline, such as a court-martial” (Zurick 2010, 301). There is no way of estimating how common nonjudicial punishment of mutilation was in Vietnam. Milam (2009, 135–36), in his study of junior officers in Vietnam, argues that “most units discouraged [mutilation] and punished the perpetrators”; yet he also remarks that such acts “were often considered nuisances by junior officers” who “had to decide if this kind of violation was serious enough to warrant discipline, because some men who participated in such atrocious behavior were very good infantry soldiers who performed well in combat situations” (emphasis added). Thus, while some unit commanders clearly took exception to the practice of mutilation and warned their soldiers against it (see, e.g., Caputo [1977] 2017, 124; Marlantes 2011, 112), others adopted a more tolerant attitude and effectively “‘turned a blind eye’ to such antics” (Bourke 1999, 41). Greiner (2009, 131) notes testimony from Vietnam veterans who “speak of superiors … who tolerate[d] mutilation of corpses in every conceivable way.” Some unit-level commanders may even have encouraged the practice, though this appears to have been rare. ¹⁰

In sum, it seems clear that mutilation was rarely punished as a serious offence in Vietnam, leaving enforcement largely in the hands of unit commanders, at least some of whom tolerated the practice.

Deviant Cohesion

Top-down toleration alone, however, cannot explain why some soldiers engaged in mutilation while others did not. Explaining the emergence of a wartime practice requires consideration of the preferences and motivations of rank-and-file combatants. Here I propose an explanation centered on a desire to avenge combat losses among fellow unit members.

Revenge

American forces suffered an estimated 47,434 “hostile deaths” in Vietnam, a rate of combat fatalities considerably higher than that suffered by Americans in more recent conflicts in Afghanistan or Iraq (CRS 2020, 9). These combat losses occurred primarily in the context of a counterinsurgency campaign fought against an elusive enemy who relied mostly on classic guerrilla warfare tactics, including systematic concealment, intermingling with the civilian population, and the use of ambushes and booby traps (Biddle 2021, 275–84). American forces were generally poorly prepared for this kind of war (Krepinevich 1988), and wartime and postwar interviews with veterans reveal how profoundly disorienting many found the experience (Gault 1971; Lifton 1973; Shay 1994). As Lifton (1971, 45) described:

The average Vietnam GI is thrust into a strange, faraway, and very alien place. … Finding himself in the middle of a guerrilla war in which the guerrillas have intimate contact with ordinary people … [h]e experiences a combination of profound inner confusion, helplessness, and terror. Then he sees his buddies killed and mutilated. He may experience the soldier-survivor’s impulse toward revenge, toward overcoming his own emotional conflicts and giving meaning to his buddies’ sacrifices by getting back at the enemy. And in an ordinary war there is a structure and ritual for doing just that—battle lines and established methods for contacting the enemy and carrying out individual and group battle tasks with aggressiveness and courage. But in Vietnam there is none of that—the enemy is everyone and no one, never still, rarely visible, and usually indistinguishable from the ordinary peasant. The GI is therefore denied the minimal psychological satisfaction of war, and, as a result, his fear, rage, and frustration mount.

Mounting feelings of anger could lead to what Shay (1994) calls “berserker rage,” a visceral need to “get back” at the enemy for the loss of close comrades. Mutilation of corpses often occurred in this context, as indicated by Dubberly (2011, 279): enemy attacks “killed or maimed many GIs and left their frightened and angry comrades with no means for revenge. … Retribution, or payback as it was known to GIs, took several forms. Mutilation was by far the most prevalent” (emphasis added).

Soldiers’ testimonies support the view that perpetrators of mutilation often viewed their acts as a form of vengeance. One veteran described how he “cut both ears off a dead man … because a close friend …, like a brother, had been KIA in Vietnam” (quoted in Milam 2009, 135). Another recalled the act of mutilation in the following terms: “[O]ut came the knife, and I started mutilating. It was … this overwhelming sense of, ‘I’ve got to. You’re gonna pay.’ … I guess my justification was revenge” (quoted in Appy 1993, 263). The files of Pentagon war crimes investigators confirm that revenge was often a motive for mutilation. Vengeance for fallen comrades was the most common motive cited by investigators in the cases in which they were able to substantiate allegations of mutilation. ¹¹ One soldier who admitted to investigators that he had engaged in mutilation explained that “he did it in revenge for the death of a close friend.” ¹² Another, when asked why he had mutilated an enemy corpse, explained that he had “been ‘out in the bush’ for three days and nights. … I kept thinking of all my buddies that had been killed and the hate just kept building up inside me.” ¹³ The intense feelings of anger that many soldiers felt after the loss of a close friend in combat can help to explain the extreme nature and apparent irrationality of the resulting violence (attacking the bodies of enemies who are already dead).

H1: Soldiers who experience more frequent combat losses among friends from their unit should be more likely to perpetrate mutilation.

In some cases, American soldiers were themselves victims of mutilations inflicted by the enemy. Appy (1993, 265) notes that veterans’ accounts of mutilation “often … begin by describing an operation on which they encountered bodies of American soldiers that had been mutilated by the Viet Cong or the NVA.” In these cases, mutilation of dead Vietnamese was seen as a direct response to enemy violations. Terry (1984, 293) cites one veteran explaining that he did not engage in mutilation “[u]ntil I had the misfortune to come upon those American soldiers who had been castrated,” after which point “it got to be a game between the Communists an ourselves to see how many fingers and ears we could capture from each other.” About half of NVVRS respondents reported that mutilations by enemy forces were “common” or “very common” during their tour in Vietnam, though only 11% reported having directly witnessed such acts. Soldiers’ testimonies suggest that directly witnessing mutilated bodies played an important role in spurring participation in mutilation. ¹⁴

H2: Soldiers who witness the mutilation of American soldiers should be more likely to perpetrate mutilation.

Cohesion and Discipline

If revenge is a powerful motive for mutilation, then direct participation in such violence should vary according to the strength of that motive. In general, soldiers with stronger bonds with other members of their units should be more likely to experience intense desires to exact revenge when their comrades are killed in battle. Thus, variation in the occurrence of mutilation should be determined in part by the strength of social cohesion within military units.

“Social cohesion” is a subtype of military cohesion defined by bonds of mutual liking and emotional closeness among soldiers (Kier 1998, 17; MacCoun and Hix 2010, 139). While early studies of military cohesion emphasized the importance of such bonds for military performance (Marshall 1947; Shils and Janowitz 1948), more recent research has questioned this view, emphasizing instead the role of “task cohesion” (shared commitment to group goals) as a determinant of combat effectiveness (MacCoun, Kier, and Belkin 2006). In some contexts, strong social cohesion has even been found to have negative effects on combat performance, as strong bonds among rank-and-file soldiers can sometimes subvert organizational policies and promote behavior seen as deviant from the perspective of the military as an institution (Finnegan 2021; Kier 1998, 15–16; King 2013, 31–32; McLauchlin 2020; Rielly 2001; Vennesson 2015). In the case of the Vietnam War, such behavior included drug use, “fragging” (assassination of commanders), mutiny, and desertion, all of which were generally perpetrated by groups of soldiers, rather than by isolated individuals (Faris 1977).

Whether social cohesion leads to deviant behaviors should depend in part on the extent to which the informal norms endorsed by members of a cohesive group are congruent with or diverge from official military policies and values. Informal norms are an important influence on the conduct of combatants (Anonymous 1946). As Henderson (1985, 5) argues, in a highly cohesive unit, the group “develops strong rules of behavior and expectations about individual conduct … and thereby becomes the immediate determinant of the soldier’s behavior.” Soldiers have strong incentives to conform to the norms of their unit, whether for instrumental reasons—to maintain the support of other group members and thereby enhance their chance of survival in a dangerous environment (Moskos 1970)—or because of deep emotional commitments to other group members. In situations of combat, strong unit norms can come to define the very standards of morality recognized by unit members. As one Vietnam veteran explained: “What matters [in combat] is how the people around you are going to see you. … This group of people was all that mattered. It was the whole world. What they thought was right was right” (quoted in Bilton and Sim 1993, 19).

As a result, when unit-level norms are congruent with the policies of the military as an institution, their influence on soldiers’ behavior should act to reinforce organizational prohibitions on unauthorized violence. On the other hand, when unit norms deviate from organizational rules, they can actively endorse unauthorized behaviors, including atrocities (Rielly 2001). Two key factors are likely to determine whether unit-level norms are congruent with or deviate from organizational policies. The first is official pre-deployment socialization aimed at inculcating organizational norms and values among rank-and-file recruits. Research suggests that political education aimed at socializing recruits in “norms of restraint” can shift combatant preferences for battlefield conduct in a manner that accords with organizational policies (Bell 2022; Hoover Green 2018). In the case of American forces in Vietnam, observers have noted that pre-deployment training in such “norms of restraint” (i.e., the laws of war) was often insufficient (Allison 2007, 92; Parks 1976, 20). With regard to the specific issue of the maltreatment of corpses, some soldiers claimed to have never received any relevant instructions on the matter (Milam 2009, 136). As noted above, high-level military officials as late as 1970 deplored “a general lack of understanding of what constitute[d] ‘maltreatment.’” ¹⁵

Second, and more importantly, the congruence of unit norms with organizational policies is likely to be determined to a large extent by decisions about discipline made by unit-level leaders. As Manekin (2020, 5) notes, once soldiers are actually deployed in combat, “the task of control [over rank-and-file violence] falls primarily to small-unit commanders, who are charged with reinforcing military norms regarding the use of force.” This is particularly true in counterinsurgency warfare, which is usually “based on operations in small teams, often far from the sight of supervising officers” (166). This was certainly the case among American forces in Vietnam, as Milam (2009, 139) describes: “Because the war was fought on a squad, platoon, and company level, the responsibility for the day-to-day behavior of the troops was most often [exercised by junior commanders] without senior officer review.” Faced with violations of organizational policy, junior commanders had to decide when these were “serious enough to warrant discipline” (136). These decisions likely had a significant impact on the character of informal norms that emerged in a given unit, and the resulting patterns of violence used by its members. Marlantes (2011, 112) provides evidence that commanders themselves recognized this influence: describing his time as a company commander in Vietnam, Marlantes recounts how he disciplined some of his men for the mutilation of corpses, not because he found the practice particularly objectionable, but because he understood the importance of discipline in shaping the “standards of behavior” within his unit.

Overall, the above account suggests that variation in the probability of committing acts of mutilation should be shaped by a combination of social cohesion on the one hand, and unit-level enforcement of organizational policies on the other (table 2). In units in which disciplinary standards are actively enforced by unit leaders, informal unit norms are more likely to be congruent with official organizational policies, and therefore to act as a restraint on unauthorized violence, including mutilation. In contrast, in units in which disciplinary standards are not successfully maintained, unit norms can diverge from organizational policies. Whether such norms endorse the perpetration of atrocities is likely to depend in part on the strength of social ties among unit members. All else being equal, soldiers with weaker emotional bonds should be less likely to experience intense desires for vengeance in response to combat losses in their unit. The death of fellow unit members to whom one feels little emotional attachment may evoke fear, but should be less likely to motivate personally costly acts of vengeance than would the death of a close friend. As social cohesion increases, however, revenge motives should become more prevalent and powerful. Without consistent enforcement of discipline, highly cohesive units can develop norms that endorse atrocities as a legitimate form of “payback.” The existence of such unit-level norms can help to explain why postmortem mutilation in Vietnam was usually perpetrated by groups of soldiers rather than individuals (Hiley-Young et al. 1995, 132), and why its prevalence seemed to vary across units (Appy 1993, 265).

H3: Soldiers who experience stronger bonds with other members of their unit should be more likely to perpetrate mutilation, but only in units with weak discipline.

Table 2 Discipline, Cohesion, and Mutilation

Alternative Explanations

In addition to examining the influence of unit-level social dynamics on unauthorized atrocity, I also test several alternative explanations.

Bad Apples

One alternative explanation focuses on the role of soldiers with atypical predispositions to violence—the proverbial “bad apples.” In a large enough military force, some proportion of recruits will have atypical “preferences and talents for violence” (Mitchell 2004, 46), either because of abnormal psychology or due to prior habituation to violence. Particularly violent individuals may actively self-select into military service (Mueller 2004, 9) and go on to perpetrate a disproportionate number of atrocities, becoming what Mann (2000, 332) calls “disturbed killers”: men “whose hatreds, fed by disturbed personalities, [result] in sadistic behavior.” While existing scholarship mostly rejects the “bad apples” theory as a general explanation for wartime atrocities, arguing that the number of sadistic or otherwise abnormal perpetrators is generally small (Baumeister 1997, 252; Waller 2002, 75) and emphasizing instead the capacity of “ordinary men” to perpetrate extraordinary violence (Browning 1992; Waller 2002), such research has only rarely focused specifically on unauthorized forms of violence or acts like mutilation involving grotesque or “excessive” violence (but see Fujii 2013; Mitton 2015). It is possible that, while “ordinary men” can become reliable killers, it is a relatively small number of “abnormal” individuals who perpetrate the bulk of unauthorized extreme violence (cf. Valentino 2004, 54).

One implication of the “bad apples” theory is that perpetrators of mutilation should differ systematically from nonperpetrators with regard to some (set of) risk factor(s) associated with aggression and violence. Research in public health, psychology, and criminology has revealed a range of such potential factors, including childhood domestic abuse (Currie and Tekin 2006; Fitton, Yu, and Fazel 2020), parental mental illness (Dean et al. 2012; Tehrani et al. 1998), and parental criminality (Sivertsson, Carlsson, and Hoherz 2023; Tzoumakis et al. 2017). To the extent that these factors influence individual predispositions to violence, individual soldiers who experience one or more of them should be more likely to perpetrate violence not authorized by military policy, including mutilation.

H4: Soldiers exposed to risk factors associated with violence prior to deployment should be more likely to perpetrate mutilation.

Traditions of Racialized Violence

A second argument focuses not on individual pathology, but on long-standing traditions of racialized violence. As Harrison (2012, 4) notes, human trophy-taking “has been carried out, at least among European and North American military personnel, almost exclusively against enemies whom they have represented as belonging to ‘races’ other than their own.” In the American case, a tradition of military mutilation might be traced from the use of “scalping” during the wars of the colonial period through to the mutilation of Japanese war dead during World War II (Dower 1986, 66; Weingartner 1992). Mutilations in Vietnam are sometimes seen as a continuation of this tradition: Bourke (1999, 38) argues that the prevalence of “gruesome trophy-hunting” among American forces in Vietnam was due in part to a distinct “national narrative tradition” through which “Americans placed more emphasis on ‘scalping’ their enemies ‘like the Indians.’”

Postmortem mutilation has also played a prominent role in racialized violence within America itself, particularly in the context of late nineteenth- and early twentieth-century vigilante violence against African Americans (Garland 2005). Examining data on 3,767 lynchings in the American South between 1877 and 1950, Beck and Tolnay (2019, 327), find that the desecration of victims’ bodies (through mutilation, dismemberment, or burning) was a common feature of such events: in the 1930s, almost a quarter of lynchings featured such acts. Given the prevalence of such violence in the American South in earlier decades, it is possible that postmortem mutilation during the Vietnam War had some connection to these earlier atrocities. There is evidence that some American soldiers deployed in Vietnam identified openly with the tradition of white supremacist violence, flying Confederate flags, for example, or burning crosses in the style of the Ku Klux Klan (Phillips 2012, 223). If the practice of mutilation was a manifestation of white supremacy, such soldiers should also have been more likely to engage in such conduct.

H5: Soldiers who identify with traditions of white supremacist violence should be more likely to perpetrate mutilation.

Hunting

Finally, Harrison (2012, 10) proposes that wartime mutilation and trophy-taking are influenced by hunting practices, arguing that such violence is more likely when “the cognized boundaries between humans and animals, expressed in the activity of hunting, are shifted into the domain of human relations, and made to serve there as a model for violence between social groups.” The use of hunting metaphors to describe mutilation in the context of colonial violence against Native Americans (“scalp-hunting”), lynching in the American South (Chamayou 2012, 99–108), and atrocities in the Pacific (Weingartner 1992, 55) is consistent with Harrison’s theory. Harrison (2012, 10) speculates that the conceptual shift toward viewing enemies as quarry to be hunted and dismembered is “more likely to be made by men for whom hunting represents an important component of their social identity.”

H6: Soldiers with greater involvement in hunting prior to deployment should be more likely to perpetrate mutilation.

Empirical Analysis

Data

To test the above hypotheses, I examine patterns in individual participation in postmortem mutilation among American soldiers using data from the NVVRS survey (Brzezinski 2024). Conducted in the mid-1980s, this survey gathered data from a representative sample of 1,438 American Vietnam War veterans concerning a variety of prewar, wartime, and postwar experiences (Kulka et al. 1988). ¹⁶ Conflict researchers have long used ex-combatant survey data to analyze microlevel dynamics of armed conflict (Arjona and Kalyvas 2011; Humphreys and Weinstein 2006). Most have gathered their data during or soon after the end of conflict. The use of data gathered with a 10-to-20-year lag raises obvious concerns about the ability of respondents to accurately recall wartime conditions. On the other hand, such a delay has the advantage of allaying fears of possible punishment for respondents who engaged in criminal behavior, something which may be a serious concern for surveys carried out soon after conflict (Humphreys and Weinstein 2006, 434).

Variables

My main outcome variable measures whether respondents reported individual participation in mutilation. As noted above, NVVRS respondents were asked to rate their exposure to mutilation by American forces on a six-point scale, with the last two levels denoting individual participation (“I participated” or “I was responsible”). ¹⁷ Overall, as indicated in table 1, 1.84% of respondents are coded as having participated directly in mutilation. About 1% of respondents did not answer this question, and are dropped from the analysis.

To assess the role of revenge as a motive for mutilation, I measure how frequently respondents saw close friends killed within their unit. This question (measured on a five-point scale from “never” to “very often”) has the advantage of focusing on the deaths of unit members with whom respondents had strong emotional bonds (“close friends”). I also code whether respondents reported having personally witnessed Americans mutilated by the enemy (as opposed to having merely heard of such behavior, or having neither seen nor heard of it).

To examine the effects of social cohesion and unit discipline, I use two variables. The first, unit social cohesion, is an ordinal variable measuring how “close or tight” respondents report having felt with other members of their unit. This variable reveals substantial variation in the strength of social ties among soldiers: while relatively few respondents reported very weak bonds with fellow unit members (2% reported feeling “not close at all”; 6% reported feeling “not very close”), significant proportions reported moderate levels of cohesion (32% felt “fairly close” with other unit members), strong cohesion (37% felt “very close”), or extremely strong cohesion (23% felt “extremely close”). My second key variable, decline in discipline, assesses the extent to which organizational rules and norms were enforced within each respondent’s military unit. Here I use a question that asked respondents about whether they experienced a “decreased emphasis in the field on military discipline and bearing.” The original purpose of this question was to measure how “satisfying” respondents found such a decline in discipline, but respondents were also given the option of indicating that they “did not experience” any such decline. I use these responses to code a binary variable measuring whether respondents experienced a decline in discipline. Because “discipline” bears directly on the following of organizational rules, while “military bearing” refers to “conducting oneself in a professional manner [and] uphold[ing] standards” (Grimmett 2018, 2), this variable provides a basic measure of the degree to which the policies and norms of the military as an institution were emphasized by a respondent’s immediate commanders.

To assess the “bad apples” argument, I measure several individual risk factors that have been found to be correlated with aggression or violence. Childhood domestic abuse measures the frequency with which respondents faced serious domestic physical violence between the ages of six and 16. Family mental illness measures whether respondents had family members who were hospitalized for mental illness or substance abuse when they were growing up. Parental imprisonment measures whether respondents had parents who were incarcerated.

Assessing the hypothesized relationship between wartime mutilation and American traditions of racialized violence is complicated by the lack of questions in the NVVRS survey directly measuring respondents’ attitudes toward race or racism. I therefore make use of two indirect measures of respondents’ identification with traditions of white supremacy. The first is a simple measure of self-reported racial identity: if mutilation in Vietnam was closely connected to traditions of white supremacy, then such violence should be particularly unlikely among African Americans—that is, among members of a group historically targeted by white supremacist violence. My second measure focuses on past histories of white supremacist violence, looking at the scale of anti-Black lynching in the regions in which white respondents were raised. Here I assume that, all else being equal, white respondents who were raised in regions with a tradition of large-scale lynching should be more likely to identify with that tradition, and thus more likely to engage in mutilation, than white respondents raised in regions with little or no history of lynching. To assess this possibility, I draw on Beck and Tolnay’s (1990) widely used inventory of lynching events in 10 southeastern states, supplemented with Seguin and Rigby’s (2019) dataset of lynching events in the rest of the contiguous United States. Because NVVRS data does not identify respondents below the level of US Census “divisions” (regions of between three and eight states), I aggregate the total number of Black male lynching victims reported by these two datasets in each such division between 1882 and 1930. I then use this aggregate number of division-level lynchings to code the scale of past white supremacist violence in each respondent’s region of upbringing. ¹⁸

To assess Harrison’s argument about the influence of hunting practices on wartime mutilation, I measure whether respondents were raised in a rural area. Though rural areas obviously include communities in which little hunting occurs, I assume that, all else being equal, hunting plays a more important role in the identity of rural populations than in the identities of urban ones.

Finally, I code a series of variables to control for likely confounders. Combat exposure measures self-reported exposure to combat (on a four-point scale from “almost none” to “heavy”). Combat exposure likely increases the probability of mutilation directly (more combat provides more opportunity to engage in mutilation of enemy dead), and is likely to be correlated with several of my other predictor variables. Notably, combat exposure likely contributes both to a decline in discipline (maintaining organizational norms becomes more difficult in intense combat) and to unit cohesion (cohesion will tend to increase as members of a unit face danger together). There is also evidence that Black soldiers were exposed to more intense combat on average than were white soldiers (Moskos 1970, 139). ¹⁹ I also control for deployment duration. Longer deployments not only provide more opportunity for the perpetration of mutilation, but may also reinforce bonds among soldiers (Rielly 2000, 62) while undermining discipline (Manekin 2013). I also control for whether respondents served in Vietnam before or after the Viet Cong’s 1968 Tet Offensive (post-Tet). Military discipline among American troops is thought to have significantly declined after this event (Moskos 1975, 25). Controlling for this factor helps to distinguish the effect of a unit-specific decline in discipline from that of a possible Army-wide change. Finally, I control for the age of respondents at the time of their deployment. In general, younger men are more prone to violence than are older men, and age may also influence other explanatory factors.

Results

Figure 1 depicts coefficient plots with 95% confidence intervals for a series of logistic regression models with individual participation in mutilation as the outcome variable. ²⁰ Each column in each plot includes results from a model estimated without control variables and another estimated with controls.

Figure 1 Individual Participation in Mutilation

Revenge and Deviant Cohesion

I begin by assessing the argument that explains mutilations in Vietnam as a form of revenge. Column 1 in figure 1a presents results for models estimating the effect of having had close friends killed in combat or having witnessed Americans mutilated on individual participation in mutilation. Consistent with H1 and H2, coefficients associated with both independent variables are positive and statistically significant at conventional levels, even when controlling for likely confounders. Importantly, these confounders include self-reported combat exposure, suggesting that neither independent variable is acting merely as a proxy for the general intensity of combat experienced by soldiers. Indeed, while combat exposure is a robust predictor of mutilation in most models examined in figure 1, the coefficient for this variable is not significant in this column, suggesting that much of the “brutalizing” effect of combat is explained by the experience of losing friends in battle or witnessing the mutilation of Americans by the enemy. Also, because close friends killed and witnessed Americans mutilated are closely correlated, inclusion of both variables in the models in this column ensures that neither is acting merely as a proxy for the other; the results indicate that both factors have an independent effect on individual participation in mutilation.

If the theory of deviant cohesion is correct, participation in mutilation should be systematically related to the strength of social cohesion within military units and the nature of unit-level discipline: desires for vengeance should be more intense for soldiers who have close ties with fellow unit members, and these desires should be more likely to give rise to mutilation when organizational norms are poorly enforced within their unit. These expectations are tested in columns 2–4 of figure 1a. Models in column 2 estimate the effect of unit social cohesion on mutilation, while those in column 3 focus on decline in discipline. Coefficients associated with each variable are not statistically significant at conventional levels once confounders are controlled for, suggesting that neither factor alone successfully predicts participation in mutilation. Instead, as shown in column 4, the estimated effect of each of these variables is moderated by the value of the other. Coefficients associated with the interaction term (unit cohesion × decline in discipline) are positive and statistically significant, suggesting that the effect of a decline in discipline on mutilation increases with unit social cohesion.

This interaction effect is illustrated in figure 2, which uses the results from the second model in column 4 (with controls) to plot the predicted probability of mutilation for soldiers who did and did not experience a decline in discipline within their unit across different levels of unit social cohesion. While the graph reveals no evident differences in the probability of mutilation in units with very low levels of cohesion (as noted, the number of respondents reporting very low cohesion is small), such differences are clearly visible when comparing respondents who reported moderate to high levels of cohesion. Among respondents who experienced a decline in unit discipline, those reporting “fairly” close bonds with other unit members had a predicted probability of participating in mutilation of 1.3%. This probability increases to 2% among those reporting “very” close bonds, and to 3% among those reporting “extremely” close bonds. While the overall probability of engaging in mutilation remains small even in the last group, the figure suggests a large (threefold) increase in the probability of participation in mutilation when comparing moderately cohesive units to highly cohesive units. Importantly, no comparable pattern can be seen among respondents who reported no decline in unit discipline.

Figure 2 Predicted Probability of Mutilation

In sum, the quantitative evidence examined here is consistent with the theory of deviant cohesion. Both the frequency with which soldiers lost friends in combat and the experience of seeing fellow Americans mutilated by the enemy are associated with a higher probability of individual participation in mutilation, and such participation is particularly likely when strong unit bonds coincide with weak enforcement of organizational rules.

Alternative Explanations

I now turn to alternative explanations. I begin by assessing whether mutilations in Vietnam can be explained as the actions of “bad apples.” Figure 1b depicts coefficient plots for models estimating the effect on mutilation of three individual-level risk factors associated with violence. I first estimate the effect of each factor individually, and then combine them, estimating the effect of each while controlling for the effects of the others. Of the three risk factors examined, only family mental illness has coefficients that are statistically significant at conventional levels across all specifications. These coefficients are positive, suggesting that soldiers with a family history of severe mental illness were more likely to perpetrate mutilation than soldiers without such a history. Insofar as such individuals may have had atypical predispositions to violence compared to other recruits, this finding is consistent with H4. It is worth noting, however, that such individuals account for only a minority (16%) of perpetrators of mutilation. (In contrast, soldiers serving in units with very or extremely high social cohesion and weak discipline account for 73% of perpetrators.) Overall, then, to the extent that these three risk factors provide an adequate measure of factors likely to predispose individuals to violence, these results suggest that the “bad apples” argument alone cannot account for mutilation in Vietnam.

Turning to arguments about racialized violence and hunting practices, figure 1c presents coefficient plots for a series of logistic regression models that include measures of racial identity, past regional histories of white supremacist violence, and rural background. The results presented in columns 1 and 2 provide no evidence for the hypothesized association between mutilation and traditions of white supremacist violence (H5). If mutilation were simply a manifestation of white supremacy, such violence should be particularly uncommon among Black soldiers. Yet the coefficient for Black soldiers in the first model in column 1 (without control variables) is actually positive and statistically significant. This is likely a spurious correlation driven by the greater involvement of African Americans in combat, and the coefficient associated with this variable is no longer statistically significant once confounders (including combat exposure) are controlled for. There is also no evidence for an association between past regional history of white supremacist violence and participation in mutilation among white soldiers. As shown in column 2, coefficients associated with an interaction term between division-level lynchings and white identity are statistically indistinguishable from zero. Finally, models presented in column 3 provide no evidence for H6: coefficients for rural background are positive but not statistically significant, providing no evidence for an association between hunting and individual participation in mutilations.

In sum, I find no evidence that variation in the probability that soldiers committed acts of mutilation in Vietnam can be explained by identification with traditions of white supremacist violence or by involvement in hunting. To be sure, the variables I use here are imperfect proxies for either hypothesized explanatory factor. Ideally, more accurate measures would be used to better assess these hypotheses. With regard to H5, moreover, it should be noted that, even if racial identity alone does not directly explain variation in the perpetration of mutilation, a tradition of racist violence may nonetheless have provided American soldiers with some of the specific techniques of violence used in Vietnam (e.g., scalping). It is striking, nonetheless, that such violence could be used by both Black and white soldiers, as testimonies by African American veterans make clear (see, e.g., Terry 1984, 293).

Robustness Checks and Reverse Causality

The online appendix provides a series of analyses checking the robustness of my findings (see tables A8–A15). First, I estimate models that include all of the variables associated with my alternative explanations as control variables, effectively treating these as potential confounders. Second, given the relatively small number of NVVRS respondents who reported direct participation in mutilation, I replicate the analyses in figure 1 using a penalized maximum likelihood estimator to correct for rare events (Woo, Berns, and Solanelles 2022). Third, given the probability of social desirability bias affecting respondents’ willingness to report personal participation in atrocities, I replicate the analyses in figure 1 using an alternative dependent variable that codes respondents who reported either individual- or unit-level participation in mutilation as possible perpetrators. Fourth, I replicate the analyses in columns 2 and 4 of figure 1a with an alternative operationalization of unit social cohesion, using a question that asked respondents how many members of their unit looked out for the welfare of other unit members. Statistical findings in all of these robustness checks are substantively similar to those reported above, though the precision of some estimates is reduced in the analysis performed with an alternative dependent variable (see table A12). Given the impossibility of precisely determining the size of the reporting bias affecting my main dependent variable (i.e., what proportion of direct perpetrators of mutilation reported only lower levels of exposure), this discrepancy does not warrant rejecting my key findings.

I also consider whether the direction of causality between social cohesion and atrocity might be the reverse of that assumed by my theory. Cohen (2016) argues that collective participation in rape can serve as a means of increasing group cohesion in armed groups that rely on the abduction or press-ganging of recruits. This argument poses a potential challenge to my explanation, suggesting that the experience of transgressive violence causes higher cohesion, rather than the reverse. While properly testing this argument using the NVVRS data is difficult, given that its measure of social cohesion does not specify whether strong social ties preceded or followed participation in atrocity, I note two reasons to doubt that Cohen’s argument about the causal priority of atrocity applies in the Vietnam case. First, as noted above, perpetrators of mutilation in Vietnam often understood their acts as a form of revenge for the loss of close friends in combat, suggesting that ties of friendship (i.e., high social cohesion) preceded the use of mutilation. (A similar temporal sequence—friendship, followed by loss, followed by mutilation—is suggested in the case study that follows.) Second, Cohen’s argument centers on the incentives that abducted soldiers have for engaging in transgressive violence as a means of signaling their reliability to other unit members and thereby building group cohesion. If this argument is extrapolated to the Vietnam case, we might expect greater participation in mutilation by recruits who most closely approximate abductees, namely those drafted into the military. With no prior connections to other members of their units, such soldiers might have had greater incentives to engage in costly signaling. In units that experienced a decline in discipline, such signaling might include participation in atrocities, including mutilation. As shown in table A16 in the online appendix, however, neither draftee status alone nor an interaction between draftee status and unit-level decline in discipline are statistically significant predictors of individual participation in mutilation.

Mutilation in the “Tiger Force”

The quantitative analysis above provides evidence consistent with the theory of deviant cohesion: strong ties among soldiers increase their probability of participating in mutilation, but only when organizational rules are poorly enforced within their unit. In this section I use case study evidence to illustrate the proposed mechanism. My focus is the so-called “Tiger Force,” a long-range reconnaissance unit in the 327th Infantry Regiment of the 101st Airborne Division that became notorious for atrocities, including mutilations, perpetrated by its members in northern South Vietnam in 1967. I draw on the work of journalists who have investigated this unit (Sallah and Weiss 2006), as well as archival documents. These sources provide insights that are consistent with my quantitative findings. First, while soldiers with atypical predispositions to violence were among the perpetrators of mutilation in the Tiger Force, this behavior was too widespread in the unit to be explained by the presence of “bad apples” alone, and in fact constituted a practice endorsed by unit-level norms. Second, the sources show that the emergence of this practice was influenced by combat losses and social dynamics within the unit, specifically the strong social ties that existed among some of its soldiers and the weak enforcement of discipline by unit commanders.

Bad Apples and Unauthorized Practices

Individual soldiers with marked predispositions to violence likely played an important role in the perpetration of mutilations in the Tiger Force. Pentagon investigations into the unit were sparked initially by complaints about the behavior of a 22-year-old private by the name of Sam Ybarra, who was allegedly observed beheading a Vietnamese infant during a raid on a village in late 1967. Though investigators were unable to confirm this specific allegation, they did establish that Ybarra had “on numerous occasions cut ears from dead bodies; possessed a set of human ears and a jar containing two ears; [and] possessed a string with human ears which he wore on several occasions around his neck.” ²¹ Investigators also concluded that Ybarra’s close friend, Private Kenneth Green, also engaged in postmortem mutilations, among other atrocities. ²² Both men clearly fit the profile of “bad apples.” According to Sallah and Weiss (2006, 11), Ybarra was a profoundly troubled young man, known for his violent behavior prior to enlisting in the Army. Soldiers interviewed by Pentagon investigators described him as “crazy” ²³ and “a very cold blooded person.” ²⁴ Green was equally violence-prone, known for “getting into fights almost weekly” prior to enlisting (Sallah and Weiss 2006, 10).

Yet Ybarra and Green were not the only perpetrators of mutilation in the Tiger Force. Indeed, soldiers interviewed by Pentagon investigators reported that mutilation was a “common and accepted” practice in the Tiger Force, and was “condoned by the unit.” ²⁵ Asked by investigators why Ybarra was permitted to engage in mutilation, one soldier explained that “cutting the ears off the dead was an accepted practice within the Tiger Force … accepted by the men in the field that were actually doing the fighting.” ²⁶ Another stated that “it was a common practice to cut at least one ear off or at least one joint of a finger after a kill.” ²⁷ Among the perpetrators were soldiers who hardly fit the profile of “bad apples”: one, Private Terrence Kerrigan, was described as an avid surfer from southern California, known to his friends as “a beach kid … who never talked about fighting” (Sallah and Weiss 2006, 31). Within the Tiger Force, however, Kerrigan had befriended Ybarra and Green and, after Green’s death in combat in September 1967, quickly joined others in the unit’s practice of mutilating bodies (200).

Eventually, according to Sallah and Weiss (2006, 211), the practice of mutilation became so common in the Tiger Force that the unit’s medic had to hide his surgical blades to prevent their theft, and “just about everyone [in the unit] was carrying shriveled lumps of flesh in ration bags, openly and proudly.” Clearly, mutilation in the Tiger Force was more than a case of a few “bad apples”; the fact that unit members were engaged in such behavior “openly and proudly” indicates that mutilation became an accepted practice endorsed by unit norms.

Discipline, Cohesion, and Revenge

Perpetrators of mutilation in the Tiger Force appear not to have been seriously sanctioned for such behavior. ²⁸ According to Sallah and Weiss (2006, 86), Ybarra’s use of mutilation was tolerated by successive Tiger Force platoon commanders who chose not to punish him because of their admiration for his courage in combat. Soldiers interviewed by Pentagon investigators claimed that mutilation was “known to officials and ignored.” ²⁹ This toleration was part of a broader “hands-off” approach to discipline in the unit. The Tiger Force was designed to operate as a commando unit deep in enemy territory, maintaining only a weak link to battalion headquarters and taking orders almost exclusively from its team leaders or platoon commander (Sallah and Weiss 2006, 52). As Sallah and Weiss describe, these leaders had differing views on the limits of permissible behavior: while some remained committed to respecting the Army’s official rules of engagement, others advocated less restrained forms of violence. Over time, the latter group came to dominate the unit, as more restrained leaders were either removed by injury or transferred out of the unit.

Despite its internal divisions, the Tiger Force was described as having a high level of social cohesion, “a real bond” in the words of one soldier cited by Sallah and Weiss (2006, 170). As predicted by the theory of deviant cohesion, combat losses within the unit therefore gave rise to powerful desires for vengeance. Sallah and Weiss are explicit in linking the unit’s descent into extreme violence to the “deep and visceral” bonds that existed among its members, and the consequent “anger and a sense of revenge” that arose when the unit suffered losses (195). Several men in the unit had particularly strong ties: Ybarra and Green had been friends since high school, and had enlisted together as part of the Army’s “buddy system.” Green’s death in an ambush, coming after a series of other losses in the unit, sent Ybarra into a fit of rage, and also had a profound effect on other members of the unit (190). Sallah and Weiss describe how many Tiger Force members joined with Ybarra in a collective vow to avenge the deaths of their comrades (198). Within days, several unit members “were openly wearing necklaces of ears, and others were carrying severed ears in pouches.” Their primary motive, according to Sallah and Weiss, was not the collection of souvenirs, however: “[T]hey were mutilating bodies to deal with the rage” (203).

Conclusion

The experience of the Tiger Force illustrates how extreme forms of violence can emerge as an unordered practice in the context of counterinsurgency. Highly cohesive units in which unit commanders fail to systematically enforce organizational policies can develop norms that endorse unauthorized atrocities as a legitimate form of revenge. The survey data explored in this article show that the Tiger Force was hardly exceptional in this regard. A significant number of American soldiers in Vietnam engaged in mutilation, and individual participation in such violence was particularly likely when strong group bonds coincided with weak discipline.

To what extent can these findings be generalized to contexts beyond America’s war in Vietnam? There is at least anecdotal evidence for a connection between group cohesion, revenge, and unauthorized atrocity in other contexts. Russian soldiers involved in counterinsurgency in Chechnya in the 1990s reportedly perpetrated a variety of atrocities, including mutilations of the living and the dead (Human Rights Watch 2001). Soldiers interviewed about such acts explained them as an unauthorized “revenge ritual,” even claiming that severed ears were sometimes placed on the graves of fallen comrades as a “way of telling [the] deceased mate: Rest in peace. You have been avenged” (Reynolds 2000). Yet not all counterinsurgencies have featured such violence: with a small number of exceptions (see, e.g., Boal 2011), postmortem mutilation has been rare in recent Western counterinsurgency campaigns in Afghanistan and Iraq. Future research should systematically document variation in such violence across and within a broader array of conflicts, and explore to what extent other cases can be explained by similar motives and unit-level social dynamics.

This article has left the question of why social dynamics vary across military units largely unexplained. In the case of the Tiger Force, weak enforcement of discipline and the deviance of unit norms appear to have been linked to the unit’s operation in particularly difficult terrain and its consequent isolation from superior commanders. Extrapolating from this case, it may be hypothesized that counterinsurgent forces operating in especially “rough terrain,” in which top-down supervision is particularly difficult, will be more likely to adopt informal norms that deviate from organizational policies. Future research should test this implication. More broadly, my findings show that key characteristics shaping combatant behavior—such as group cohesion and discipline—can vary significantly across military units, even within a single military force, suggesting that future research should pay closer attention to the military unit as a distinct level of analysis.

Finally, the findings of this article also have implications for policy. Because military organizations continue to foster close social bonds among soldiers (Spencer 2022), military leaders need to be aware of the potential “dark side” of military cohesion (Rielly 2001). Though some have argued that social cohesion is less important in contemporary professional militaries than in conscript forces (King 2013), research suggests that “informal norms [among combatants] … can be as strong as formal norms, even within highly professional State armed forces” (ICRC 2020, 25). As suggested above, efforts to shape such norms may be particularly important among counterinsurgent forces operating in conditions that impede top-down supervision of discipline. Appropriately socializing junior leaders may be particularly important in such contexts, as such leaders will play a determining role in ensuring that cohesive groups of soldiers remain bound to the rules and norms of broader military organizations.