Variation in Sexual Violence during War

ELISABETH JEAN WOOD

Sexual violence during war varies in extent and takes distinct forms. In some conflicts, sexual violence is widespread, yet in other conflictsincluding some cases of ethnic conflictit is quite limited. In some conflicts, sexual violence takes the form of sexual slavery; in others, torture in detention. I document this variation, particularly its absence in some conflicts and on the part of some groups. In the conclusion, I explore the relationship between strategic choices on the part of armed group leadership, the norms of combatants, dynamics within small units, and the effectiveness of military discipline. Keywords: sexual violence; rape; political violence; human rights; war

While sexual violence occurs in all wars, it occurs to varying extent and takes distinct forms. During the conflict in Bosnia-Herzegovina, the sexual abuse of Bosnian Muslim women by Bosnian Serb forces was so systematic and widespread that it comprised a crime against humanity under international law. In Rwanda, the widespread rape of Tutsi women comprised a form of genocide, according to the International Criminal Tribunal for Rwanda. Yet sexual violence in some conflicts is remarkably limited, despite widespread violence against civilians. Sexual violence is relatively limited even in some cases of ethnic conflict that include the forced movement of ethnic populations; the conflicts in Israel/Palestine and Sri Lanka are examples. Some
I am grateful for research support from the Yale Center for International and Area Studies and the Santa Fe Institute, and for research assistance from Margaret Alexander, Laia Balcells, Karisa Cloward, Kade Finnoff, Amelia Hoover, Michele Leiby, Amara Levy-Moore, Meghan Lynch, Abbey Steele, and Tim Taylor. I also thank the many people who commented on earlier versions, particularly Jeffrey Burds, Christian Davenport, Magali Sarfatti Larson, David Plotke, and Jeremy Weinstein.
POLITICS & SOCIETY, Vol. 34 No. 3, September 2006 307-341 DOI: 10.1177/0032329206290426 2006 Sage Publications

307

308

POLITICS & SOCIETY

armed groups engage in relatively little sexual violence; Sendero Luminoso was deemed responsible for more than half the deaths and disappearances reported to the Peruvian Truth and Reconciliation Commission but for only a tenth of the (few) reported cases of rape. In some conflicts, sexual violence takes the form of sexual slavery, whereby women are abducted to serve as servants and sexual partners of combatants for extended periods; in others, it takes the form of torture in detention. In some wars, women belonging to particular groups are targeted; in others, the violence is indiscriminate. In some wars, only women and girls are targeted; in others, men are as well. Some acts of wartime sexual violence are committed by individuals; many are committed by groups. Some acts occur in private settings; others are public, in front of family or community members. In some conflicts, the pattern of sexual violence is symmetric, with all parties to the war engaging in sexual violence to roughly the same extent; in other conflicts, it is very asymmetric. Some simple hypotheses do not explain the puzzling variation in the extent and form of sexual violence in war: sexual violence varies in prevalence and form across civil wars as well as inter-state wars, across ethnic wars as well as non-ethnic, and across secessionist conflicts. The variation has not been adequately explained in the literature, much of which focuses on single cases rather than comparison across cases.1 Focusing on sexual violence against civilians by combatants, I first show that sexual violence indeed varies in extent and form across several war settings. I focus in particular on the absence of sexual violence in some conflicts and on the part of some groups. I then discuss the methodological challenges to advancing our understanding of this variation and show that, despite these challenges, the subject merits further comparative analysis because sufficiently large variation occurs across well-documented cases. Distinguishing between distinct patterns of sexual violence, I then assess the extent to which the arguments advanced in the literature (often implicitly) explain the variation. In the conclusion, I focus on the relationship between strategic choices on the part of armed group leadership, the norms of combatants, dynamics within small units, and the effectiveness of military discipline, and suggest some promising explanatory hypotheses.
VARIATIONS IN WARTIME SEXUAL VIOLENCE: SELECTED CASES

Following the definition used by recent international war crimes tribunals,2 by rape I mean the coerced (under physical force or threat of physical force against the victim or a third person) penetration of the anus or vagina by the penis or another object, or of the mouth by the penis. Thus rape can occur against men as well as women. Sexual violence is a broader category that includes rape, coerced undressing, and non-penetrating sexual assault such as sexual mutilation. The

ELISABETH JEAN WOOD

309

sexual humiliation and abuse inflicted on prisoners by U.S. troops at Guantnamo, Abu Ghraib, and other prisons is thus a form of sexual violence.3 In this section, I describe the pattern of sexual violence in several wars, including inter-state as well as civil wars, ethnic as well as non-ethnic conflicts, and wars where sexual violence was very prevalent and where it was not. I begin by discussing the patterns of sexual violence during World War II. World War II As the Soviet army moved westward onto German territory in early 1945, large numbers of women were raped.4 While the earlier Soviet offensives in Romania and Hungary had seen widespread rape of civilian women (particularly after the siege of Budapest), the practice intensified as the army moved into East Prussia and Silesia. Although women of various ethnicities were raped in the course of looting villages and cities, German women were particularly targeted. In German villages in East Prussia, it was not untypical for Soviet troops to rape every female over the age of twelve or thirteen.5 As the Soviet army occupied Berlin in late April and early May 1945, thousands of women and girls were raped, often by several men in sequence, often in front of family or neighborhood, sometimes on more than one occasion. Soldiers sometimes detained a girl or woman for some days in her home or elsewhere and subjected her to repeated rape. Even after occupation became more institutionalized, Soviet soldiers continued to rape girls and women. Sexual violence gradually subsided as occupation authorities realized the harm being done to the Soviet postwar political project and gradually instituted stronger rules against fraternization in general and rape in particular. The pattern of sexual violence during the Soviet offensive varied in different settings. Naimark notes the contrast between the exemplary behavior of Soviet troops in Bulgaria and the generally better behavior toward Polish and other Slavs, with the looting and rape that occurred in Germany and Hungary, both non-Slavic groups.6 However, sexual violence in Berlin and Budapest suggests as well another pattern: in European history there appears to be a pattern of rape (and looting) following prolonged sieges as a form of punishment for holding out rather than surrendering.7 Moreover, throughout the offensive, frontline troops were less prone to rape than troops that came through later.8 During the occupation, women and girls were more vulnerable in border towns, naval centers, and transportation centers than elsewhere. Local variations also emerged as some commanders enforced the regulations and others did not.9 This is a relatively well-documented case: historians draw on a wide range of sources including Soviet military and secret police reports, military reports, wartime memoirs and diaries, and German hospital and police records (many women did report the incidents). Even in this case, however, the frequency of

310

POLITICS & SOCIETY

rapeeven in Berlin itselfis difficult to establish.10 The best estimate appears to come from the two main Berlin hospitals: staff members estimated the number of rape victims as between 95,000 and 130,000.11 Taking 100,000 as a rough estimate of the number of victims and 1,500,000 as the number of women in Berlin at the time implies a prevalence (victims/female population) of roughly 6 percent.12 As the Soviet army moved westward toward Germany, propaganda posted and distributed along the way as well as official military orders encouraged soldiers to take revenge on and punish Germans broadly speaking, not just soldiers. On the eve of the offensive into Poland, the orders to the First Belorussian Front included, Woe to the land of the murders. We will get our terrible revenge for everything. On the eve of crossing into East Prussia, the orders included,
[O]n German soil there is only one masterthe Soviet soldier, that he is both the judge and the punisher for the torments of his fathers and mothers, for the destroyed cities and villages . . . remember your friends are not there, there is the next of kin of the killers and oppressors.13

Soldiers were instructed not to forget the violence wrought by the German military against both family and country. Naimark documents the tolerance of sexual violence against civilians on the part of the Soviet command structure, from field officers to Stalin himself, who responded to complaints from East Prussia with We lecture our soldiers too much. Let them have some initiative, and to those from German socialists with In every family there is a black sheep. . . . I will not allow anyone to drag the reputation of the Red Army in the mud.14 Did the Soviet troops engage in such widespread sexual violence in retaliation for sexual violence by German troops? The extent of sexual violence by German troops occupying Eastern Europe is not well documented; it appears to have been widespread in some areas.15 According to Wendy Jo Gertjejanssen,16 German soldiers raped girls and women of various ethnicities, including Jews, despite regulations against sexual relations with non-German women.17 Much sexual violence appears to have taken the form of forced prostitution as many girls and women were forced to serve in military brothels in cities and field camps. While some volunteered to serve in the brothels as a way to survive in the dire circumstances of the occupation, others were forced to serve under threat of death or internment. Gertjejanssen estimates that at least 50,000 women and girls served in military brothels throughout the Reich.18 German military authorities also organized brothels in labor and concentration camps, which were visited by favored prisoners, guards, and occasionally officers. Some girls and women were forced to serve in these brothels; others, when offered the choice of internment or service in the brothels, chose the latter. The scale of sexual violence in the camps (aside from the sexual humiliation of

ELISABETH JEAN WOOD

311

forced undressing and the violence against homosexuals, which often took the form of medical experiments) appears to have been limited, as the number of women in the brothels appears to be a small fraction of the number interned in the camps.19 Massive sexual violence also occurred in the Pacific theater. The rape of Nanking, the widespread violence by Japanese soldiers in the environs of the Chinese city of Nanjing for eight weeks beginning December 13, 1937, included extensive sexual violence. According to Iris Chang, 20,000 to 80,000 women and girls were raped and then executed, that is, 8 to 32 percent of the approximately 250,000 female civilians present in the city at the time of the takeover.20 Among them were pre-pubescent girls, pregnant and elderly women, and Buddhist nuns; most were summarily executed afterward. Sexual violence in Nanjing also included various forms of sexual abuse of men, including rape, the forcing of men to have intercourse with family members or the dead, and the forcing of celibate men to have intercourse. One result of the negative international publicity in the wake of the violence in Nanjing was the widespread implementation of the so-called comfort women system of military-organized brothels that accompanied Japanese forces.21 According to a 1993 study by the Japanese government that included a review of wartime archives and interviews with both military personnel and former comfort women, more than 200,000 women from across East and Southeast Asia were recruited by force and deception to serve as on-call prostitutes subject to immediate violence if they resisted. In establishing the comfort stations, Japanese officials sought
to prevent anti-Japanese sentiments from fermenting [sic] as a result of rapes and other unlawful acts by Japanese military personnel against local residents in the areas occupied by the then Japanese military, the need to prevent loss of troop strength by venereal and other diseases, and the need to prevent espionage.22

Most of the comfort women were between fourteen and eighteen years old, and most were Korean. According to the Korean Council for the Women Drafted for Sexual Slavery by Japan,23 perhaps a third of them died in the course of the war. Bosnia-Herzegovina Sexual slavery was also a prominent form of sexual violence in the conflict in the former Yugoslavia in the early 1990s. According to a European Union investigation, approximately 20,000 girls and women suffered rape in 1992 in Bosnia-Herzegovina alone, many of them while held in detention facilities of various types.24 According to the UN Commission of Experts to investigate violence in the former Yugoslavia, the vast majority of the victims are Bosnian

312

POLITICS & SOCIETY

Muslims and the great majority of the alleged perpetrators are Bosnian Serbs.25 The history of violence in the district of Foa illustrates a common pattern in this conflict.26 Before the conflict began, Muslims comprised 58 percent of the residents.27 From March to September 1992, Muslim girls and women were subjected to rape in the forests, in their homes, in detention centers, and in private flats. Of the sixty-three cases of rape and sexual assault in Foa compiled by the commission, about 55 percent took place in detention centers, including the local high school, a gym, and the workers barracks of a hydroelectric plant under construction. In such centers, members of the various Bosnian Serb forces walked in, chose from among the girls and women there, and raped them either on the premises or in nearby flats. Many of the women and girls endured gang rapes, repeated over days or weeks.28 The most authoritative investigation of sexual violence in the former Yugoslavia was carried out by a UN commission.29 The commission drew on two sources of evidence. The first was their analysis of tens of thousands of allegations contained in documents from a wide variety of sources from which the commission distilled 1,100 reported cases of rape and sexual assault (eliminating duplicate and unspecific allegations), including 800 identifiable victims, 700 named alleged perpetrators with another 750 identifiable, and 162 detention sites.30 Representatives of the commission also carried out interviews with 223 people who were victims of or witnesses to sexual violence in BosniaHerzegovina.31 The commission identified several distinct patterns of sexual violence: (1) by individuals and small groups in conjunction with looting and intimidation of the targeted group; (2) in conjunction with fighting, often including the public rape of selected women in front of the assembled population after the takeover of a village; (3) against some women and girls held in detention or collection centers for refugees; (4) in sites for the purpose of rape and assault where all women were assaulted frequently, apparently for the purpose of forced impregnation (women were told that was the case, and pregnant women were sometimes held past the point when an abortion was possible); and (5) in detention sites for the purpose of providing sex. Sexual violence against men of various ethnicities (castration, being forced to perform fellatio or to have intercourse in front of guards), while much less frequent than that against women, also occurred in camps and detention centers (examples given include camps run by Serbs, Muslims, and Croats). Among the characteristics stressed by the commission were an emphasis on shame and humiliation (many assaults occurred in front of family or in public), the targeting of young girls and virgins along with educated and prominent female community members, and sexual assault with objects. Moreover,
In both custodial and noncustodial settings, many victims report that the alleged perpetrators stated that they were ordered to rape and sexually assault the victims, or that they

ELISABETH JEAN WOOD

313

were doing it so that the victims and their families would never want to return to the area. Also, every reported case occurred in conjunction with an effort to displace the civilian population of a targeted ethnic group from a given region.32

For example, the commission interviewed nineteen women from Kotor Varos, of whom six had been raped, most gang-raped by guards in a sawmill, which had served as a temporary collection center. One woman was told by a rapist that he wanted to try a Muslim woman and that she should be honored; a second woman was told that he would make Cetnik babies in Muslim and Croat women; a third woman was told by a rapist that he had been ordered to do so.33 The commission concluded that while some cases were the result of the actions of individuals or small groups acting without orders, many more cases seem to be part of an overall pattern. These patterns strongly suggest that a systematic rape and sexual assault policy exists, but this remains to be proved.34 In drawing this conclusion, the commission relied on the fact that a majority of the cases (600 of the 1,100) occurred against people in detention, that similar patterns of sexual violence occurred in non-contiguous areas, and that sexual violence was often simultaneous with military action or activity to displace certain civilian populations. While not explicitly stated in the report, the inference is clear that the commission believed it probable that a policy of systematic ethnic cleansing including rape existed on the part of the Bosnian Serb forces.35 Direct evidence that Bosnian Serb and possibly Serbian forces planned a campaign of sexual violence as part of the ethnic cleansing of Serbian areas of the former Yugoslavia is lacking, but may emerge as the various trials at the International Criminal Tribunal for the Former Yugoslavia continue. Sri Lanka Like Bosnia-Herzegovina, Sri Lanka is also a case of a secessionist ethnic conflict, but in Sri Lanka the level of sexual violence appears to be dramatically less. It has generally been wielded by government forces against women associated with the insurgency. Police, soldiers, or security forces occasionally subject displaced Tamil women and girls to various forms of sexual assault, including gang rape and rape with foreign objects, after their arrest or detention at checkpoints on the grounds that they or family members are suspected members of the Tamil insurgency.36 I could not find estimates of the prevalence of sexual violence in this case, but it does not appear to be either widespread or systematic.37 Sexual violence against Tamil women by government forces is one reason girls and women volunteer to fight with the insurgents.38 I am not aware of any allegations of sexual violence by insurgent combatants against civilians, despite their frequent targeting of civilians with other forms of violence, including their

314

POLITICS & SOCIETY

deployment of suicide bombers and forcing non-Tamil populations to leave areas of their control. Despite the frequent recruitment by force of girls as combatants, the group does not appear to engage in sexual abuse within its own ranks.39 Israel/Palestine In the Israeli-Palestinian conflict, also an ethnic conflict characterized by the increasing separation of ethnically defined populations, sexual violence appears to be extremely limited. While the forced movement of Palestinians out of some areas in 1948 was accompanied by a few documented cases of rape,40 at present neither Israelis nor Palestinians carry out sexual assaults despite the killing of Israeli civilians by Palestinian groups and of Palestinian civilians by Israeli security forces. In December 2003, I asked representatives of three human rights organizations (two Israeli and one Palestinian) whether they believed sexual assault was occurring but was not reported, or was not in fact taking place. They independently and unanimously stated that they received information for almost no cases of sexual assault and that they believed they would hear of it occurring as they did receive reports of lesser instances of sexual harassment (for example, during pat-down searches at checkpoints). It could be the case that the intensive international monitoring of the conflict deters the practice of sexual violence, but both sides do not appear much deterred in their other practices by their frequent condemnation by international actors. Sierra Leone Sexual violence during the war in Sierra Leone, in contrast to BosniaHerzegovina, did not involve explicit ethnic targeting.41 According to the Truth and Reconciliation Commission of Sierra Leone, sexual violence was carried out indiscriminately on women of all ages, of every ethnic group and from all social classes.42 Some women suffered rape by members of several armed groups. The commission noted two particular patterns: armed groups targeted young women and girls and also those girls and women associated with other armed groups. Young women and girls were targeted particularly because they were presumed to be virgins; female rebels occasionally checked the virginity of detained young women.43 Less frequently, older women also suffered sexual assault, including post-menopausal women for whom it broke a particular cultural taboo against sexual activity among this group. On occasion, rebels broke other taboos as well, forcing male family members to rape female family members or to watch them dance naked or be raped by others.44 Sexual violence was widespread among those internally displaced by the war, which comprised approximately a quarter of the population by 2001. According to a survey of 991 internally displaced women carried out by Physicians for

ELISABETH JEAN WOOD

315

Human Rights, 9 percent of the respondents had suffered sexual assault during the ten years of the war.45 Of the respondents who were sexually assaulted, 89 percent reported being raped and 33 percent reported being gang-raped.46 Of the human rights abuses suffered by household members, 40 percent were alleged to have been carried out by the rebel group Revolutionary United Front (RUF), 34 percent by unknown groups, 16 percent by unspecified rebels, and 4 percent by mixed groups.47 Sexual violence in Sierra Leone was also extremely brutal.48 Gang rapes often took the form of very young victims enduring gang rapes, with rebel combatants lining up to take turns. Many of those who suffered sexual assault did so on multiple occasions. The extreme violence with which girls and women were raped often resulted in severe initial bleeding; tears in the vagina, anus, and surrounding tissue; long-term bleeding and incontinence; and sometimes death.49 A particular form of sexual violence in Sierra Leone was the detention of girls and women, often for long periods of time, as slaves serving and sexually servicing a rebel camp or a particular rebel.50 In some cases, they underwent forced marriage with a particular person. Of the internally displaced women who suffered sexual assault, 33 percent of the respondents were abducted, 15 percent were forced to serve initial as sexual slaves, and 9 percent were forced to marry a captor.51 Escape was reportedly very difficult, and attempts were severely punished. At wars end, some wives were not willing or able to leave their spouses.52 The Commission found
that all of the armed factions, in particular the RUF and the Armed Forces Revolutionary Council, embarked on a systematic and deliberate strategy to rape women and girls, especially those between the ages of ten and 18 years of age, with the intention of sowing terror amongst the population, violating women and girls and breaking down every norm and custom of traditional society.53

However, the commission did not analyze patterns of sexual violence in detail and therefore makes a less compelling case for sexual violence as a systematic strategy than that advanced by the commission on the former Yugoslavia, which laid out specific patterns not easily accounted for except by such a strategy. U.S. Troops in Vietnam In contrast to the absence of ethnic targeting in Sierra Leone, US troops carried out an unknown number of ethnically targeted acts of sexual violence during the Vietnam War. The best documented is that which occurred in March 1968, when nine helicopters dropped soldiers of Charlie Company near the hamlet My Lai.54 By the end of the day, 128 to 500 civilians had been executed. Approximately twenty girls and women were raped, some by groups of soldiers,

316

POLITICS & SOCIETY

and then murdered. The orders for the operation, according to the Peers Report, an internal investigation of the subsequent cover-up, included the destruction of all infrastructure and food; soldiers claimed they understood the order to include the elimination of all residents, civilian or not. To my knowledge there has not been a scholarly study of sexual violence during the Vietnam War and many documents remain classified. The massacre at My Lai only became public knowledge because of the insistence of two helicopter pilots, one who witnessed the killings and another who flew over the area a few days later. The publication in Life magazine of photographs of victims also undermined the feasibility of the initial cover-up efforts. However, it is well documented that another platoon, the Tiger Force of the 101st Airborne, also carried out acts of sexual violence, mutilation, and execution of civilians over a seven-month period in the same province the year before the events at My Lai. According to an investigative report published in 2003,55 the first incidents were sexual abuse of both male and female prisoners held by the platoon. The violence dramatically increased after two members of the Tiger Force were killed and many wounded in an ambush. Much of the literature on My Lai attributes the violence against civilians to poor leadership and morale. Despite their many differences (Charlie Company was an ordinary grunt unit, while the Tiger Force was an elite one), both factors may have played a role in Tiger Force violence as well. Setting aside the difficulty of empirically establishing issues of leadership and morale without tautologically inferring their absence by the presence of violence, whether poor morale comprises an adequate explanation is difficult to judge in the absence of studies of other U.S. military units that also engaged in sexual violence and studies of those that did not. El Salvador Sexual violence during the civil war in El Salvador, a non-ethnic conflict pitting a leftist insurgency against an authoritarian government, was one-sided, and very low in comparison to Bosnia-Herzegovina and Sierra Leone. Government soldiers and security forces occasionally engaged in sexual violence, including gang and multiple rapes, against some suspected insurgent supporters (including some men) detained in both official and secret detention sites. There are also isolated reports of government forces carrying out sexual violence while on operations early in the war. For example, according to Mark Danner, some of the nearly 1,000 people killed by the Salvadoran military at El Mozote in 1981 were raped.56 And two of the four U.S. churchwomen detained and killed by National Guardsmen in 1980 were raped. The final report of the UN-sponsored Truth Commission mentions only one incident of rape, carried out by government forces in a village in eastern El Salvador in 1981.57 However, the unpublished

ELISABETH JEAN WOOD

317

annex to the commissions report discussed sexual violence in some detail.58 Sexual violence (alone or in conjunction with some other abuse) comprised 4 percent of human rights violations reported to the commission. The majority of incidents reported took place in the first few years of the war; all were reported to have been carried out by state forces or agents. Sexual violence appears generally to have varied over time with other forms of violence against civilians, steeply decreasing after 1983 in response to the U.S. conditioning its military aid to the government on improved human rights performance. No incidents of sexual violence in the annex were attributed to the insurgent force. In the human rights and ethnographic literature analyzing the conflict, there are very few reports of sexual violence by insurgent forces.59 Sexual violence in the Salvadoran conflict was thus asymmetric, distinctly low compared to other cases, and declined over the years of the war. Peru Sexual violence appears to have been relatively infrequent in the Peruvian conflict, comprising about 2 percent of the human rights violations reported to the Truth and Reconciliation Commission. Aggregate sexual violence against civilians in Perus civil war generally co-varied with other forms of violence against civilians: the frequency of different forms of violence varies similarly across time and space, according to the data compiled by the commission.60 As with other forms of violence, sexual violence was concentrated in the indigenous highlands of Peru. However, the pattern of responsibility for sexual violence diverged sharply from the pattern for other forms of violence, according to commission data. While Sendero Luminoso, the Maoist insurgent group, was responsible for more than half of the reported deaths and disappearances, they were responsible for only 11 percent of the (few) reported cases of reported rape.61 In contrast, state agents were responsible for about a third of the deaths and disappearances but about 85 percent of the reported rapes. Summary of Observed Patterns Sexual violence in these cases appears to vary substantially in prevalence; in form; in who is targeted (all women, girls and men as well as women, or particular persons, perhaps members of an ethnic out-group); in whether it is exercised by combatants from a single party or more generally; whether it is pursued as a strategy of war; where it occurs (in detention, at home, or in public); in duration; whether it is carried out by a single perpetrator or by a group; whether victims are killed afterward; and whether its incidence varies with other forms of violence against civilians or occurs in a distinct pattern. In some wars, armed groups mirror the use of sexual violence by committing their own; in other

318

POLITICS & SOCIETY

wars, such tit-for-tat retaliation does not occur. In some conflicts, sexual violence increases over time; in others, it declines. The type of war (at the broadest level) does not explain the variation even among these few cases. Sexual violence varies in prevalence and form among civil wars as well as inter-state wars, among ethnic wars as well as non-ethnic, among genocides and ethnic-cleansing cases, and among secessionist conflicts.
CHALLENGES TO DOCUMENTING WARTIME SEXUAL VIOLENCE

Before continuing, however, a preemptive concern must be addressed. Perhaps the variation described above is merely an artifact of inadequate knowledge about the empirical patterns present in each case. The reported variation may reflect different intensities of domestic and international monitoring of conflicts rather than different prevalence rates: violence in some regions appears to garner more international attention than others. Even in peacetime and even in countries with well-developed infrastructure and liberal norms, the methodological challenges to gathering data concerning sexual violence are serious. For example, what counts legally as rape varies significantly among U.S. states depending on whether it is narrowly defined as forced penetration of the vagina by a penis or more broadly to include anal penetration and vaginal penetration by other objects, and whether rape requires forcible compulsion or merely lack of consent.62 The definitional ambiguity is still greater across societies; for example, societies differ in whether rape is considered possible between husband and wife. In some cultures, coerced vaginal penetration may be socially condoned in particular situations, with the result that an incident that would count as rape in other societies would not be considered as such.63 Whether persons who have suffered some form of sexual violence are willing to report it, whether to health workers, to police, to ethnographers, or in surveys, also varies substantially across societies. One reason that many do not do so, even in societies with liberal sexual norms, is that they feel shame and fear stigmatization. In most societies, male victims of sexual violence appear to be particularly reluctant to report it. And in societies where abortion is illegal, female victims of rape who abort may be particularly reluctant to report rape. These challenges are, of course, compounded during war when surveys are generally absent, police and health services are disrupted, and families and social groups are displaced and dispersed. The fear of reprisal for reporting sexual violence is likely greater in war settings, particularly if the perpetrator or his group is still present. Increased political polarization may intensify partisan bias in the reporting of human rights violationseven by non-partisan organizationsas violence and displacement may isolate some populations from services and intensify the counting of incidents in others. The destruction of rural infrastructure

ELISABETH JEAN WOOD

319

may reinforce urban bias. International organizations that document human rights violations tend to have limited resources and as a result focus their investigations on particular cases. And because many of the physical injuries sustained during sexual assault are to soft tissue, sexual violence does not always leave an observable trace in the long-run forensic record. As a result, the exhumation of massacre sites may not document sexual violence unless it took the form of mutilation or dismemberment, or other tissue damage likely to remain evident for many years. However, the disruption of war may also increase reporting. Sexual violence in the context of political conflict may be more likely to be reported as the stigma felt by its victims may be less, and displacement from home communities may loosen traditional norms and lessen the likelihood of reprisal. Health services may be more available, not less, to populations that fled to urban areas or in some refugee camps, compared to their place of origin.64 Human rights groups, womens organizations, and medical service groups may emerge or command more resources in wartime, enabling the compiling of reports and patterns and facilitating investigation by international commissions and human rights groups. Perhaps due to the strengthening of international norms against sexual violence during war, recent truth commissions tend to document sexual violence more carefully than earlier commissions. Another challenge to analyzing variation in wartime sexual violence is the fact that levels of sexual violence vary across countries in peacetime, making more difficult the interpretation of the wartime variation. Evidence for peacetime variation comes from studies that draw on two very different methodologies. The United Nations Interregional Crime and Justice Research Institute (UNICRI) uses crime victimization surveys in many countries to compile crossnational data on rates of sexual assault. In developing country capital cities, five-year prevalence rates in the mid-1990s for sexual assault varied between 0.83 percent, the average for the three cities at the low end (Manila, Gaborone, and La Paz), and 6.60 percent for the three cities at the high end (Rio de Janeiro, Tirana, and Buenos Aires), about eight times as high.65 In industrialized countries, estimated annual rates of sexual assault also vary, between 0.13 at the low end (the annual average for Japan, Ireland, and Scotland) and 1.03 at the high end (for Sweden, Finland, and England), with the high rate again about eight times as high as the low rates.66 Evidence for cross-cultural differences in the social regulation of sexual aggression also comes from analysis of ethnographic reports of practices in band and tribal societies before significant contact with modern societies.67 In her analysis of a subset of ninety-five societies of a standard sample of such societies, Peggy Sanday found that the rate of rape of women differed significantly: in nearly half of the sample rape was rare or absent, while in about a fifth of them, rape was moderately to highly frequent against women of that or other societies

320

POLITICS & SOCIETY

or it was an accepted way to punish women or occurred as part of ceremonies.68 (Sanday is not precise about what counts as moderately to highly frequent.) The significant correlates of rape among these societies were war, inter-personal violence (excluding rape), and ideologies of male dominance (women exercise little power or authority and do not participate in political decision making). Working with a different sample of thirty-five societies (randomly drawn), Patricia Rozee classified as rape patterns of sexual intercourse in which a woman would be punished or harmed if she refused to participate, and distinguished between normative rape (rape that occurs in circumstances condoned by that society) and non-normative rape (rape in situations not condoned).69 Based on her examination of 200 ethnographic sources, Rozee found that non-normative rape occurred in 63 percent of the societies.70 She also argued that the prevalence and form of normative rape varied among societies: marital rape occurred in 40 percent, exchange rape (in which women or girls are lent or given to guests or brothers, perhaps in the course of gambling or negotiations) in 71 percent, punitive rape (generally for transgressing gender norms) in 14 percent, ceremonial rape (such as ritual defloration) in 49 percent, and status rape (such as the right of chiefs to women) in 29 percent.71 Despite these empirical challenges, the variation in sexual violence is sufficiently well documented across enough wars and armed groups to suggest that it is real and not solely an artifact of bias in reporting and observation or a reflection of variation in peacetime levels. The variation in frequency among conflicts and among groups within a conflict appears to be large, with well-documented cases at both ends of the frequency spectrum. At the high end of the variation are some of the best documented cases, for example, Serbian forces in BosniaHerzegovina, for which it is also difficult to imagine a significantly lower rate given the numerous and mutually corroborating reports from dozens of investigations. And at the low end of the spectrum, it is difficult to imagine a high rate of sexual violence in the Israeli-Palestinian conflict going unreported, given the density of non-governmental human rights organizations and the intensity of international scrutiny of both parties behavior. Not only does the prevalence vary significantly, but the particular pattern of sexual violence does as well, which gives additional analytical traction. Finally, for some of the cases (World War II and Bosnia-Herzegovina, for example) it is evident that sexual violence was much more prevalent during the war than before. And in cases where it is unclear whether it was more prevalent during peacetime, the form of sexual violence changed during war, as in the case of sexual slavery in Sierra Leone.
EXPLAINING VARIATION IN WARTIME SEXUAL VIOLENCE

To address the central puzzlewhy the frequency and form of sexual violence vary across conflicts and across groups in a given conflictit is helpful

ELISABETH JEAN WOOD

321

to consider a prior question as well, namely, why sexual violence is often higher in wartime than in peacetime. The following framework will help organize potential explanations for variation in wartime sexual violence. First, individuals in peacetime differ in their interest in sexual violence, with some so interested in sexual gratification that they use violence to attain it, others associating sexual gratification with domination and perhaps violence, others seeking to engage in sexual violence as a form of control and power, and still others not interested at all in sexual violence. The expression of sexual aggression by individuals is regulated by a variety of social mechanisms that differ among countries, and often among within-country groups, with the result that the peacetime rate of sexual violence differs among countries and groups (as discussed above). (Societies vary, of course, in the extent to which their regulatory mechanisms in fact constrain illegitimate aggression.) Second, these regulatory mechanisms are often weaker during war, resulting in higher levels of sexual violence as the opportunity and/or incentive to engage in sexual violence increases. Third, the extent to which these regulatory mechanisms break down (and opportunity and incentive increase) varies across conflicts and groups. In some cases, regulation of sexual violence may be replaced by promotion of sexual violence as a strategy of war. In other cases, armed groups enforce effective sanctions against their combatants engaging in sexual violence, potentially even leading to reduced levels of sexual violence compared to peacetime levels. Opportunity War sometimes increases the opportunity for sexual violence. There are several possible reasons. Wars tend to be fought by armed young men in groups far from the normal social controls of their village or neighborhood. In these circumstances, sexual aggression is less regulated (the costs are lower) with the result that higher levels of sexual violence occur. Social controls may also be weaker among displaced civilians, particularly if communities are not displaced together. The dependence of some armed factions on the looting of civilian assets for combatant supplies may increase opportunity: entry into individuals homes to seize food (and often liquor) is an opportunity for sexual violence. Data to systematically test the hypothesis that sexual violence increases during war is limited. Neil Mitchell and Tali Gluch found that sexual violence was predicted by the presence of war in a statistical analysis that correlated a measure of sexual violence and the presence or absence of armed conflict.72 However, their finding was based on data for only one year and relied on a crude coding of limited human rights sources, principally U.S. State Department human rights reports. More adequate longitudinal data does not exist except for particular cases.73 Evidence that rates of sexual violence by U.S. troops in Europe

322

POLITICS & SOCIETY

increased during World War II comes from a study by Madeline Morris using data on founded investigations (meaning reported crimes that were not dismissed as unfounded by investigators) from the FBI and US military services.74 She found that the rates of rape by male U.S. military personnel during war were three to four times higher than the rate by male civilians of the same age (in contrast, military rates during peacetime were significantly lower than civilian rates). The increase in rape rates occurred as U.S. troops moved quickly through France in August and September 1944 and Germany in March and April 1945.75 Her interpretation is (in part) based on opportunity: such breakout periods are the relevant period of study, she argues, because it is then that soldiers have significant contact with civilians, as opposed to periods of intense fighting. It should be noted that opportunity arguments differ in the implied perpetrators of sexual violence. Some versions of the argument appear to assume that given the opportunity, all men will rape for sexual gratification. For example, evolutionary psychologists Randy Thornhill and Craig Palmer argue that men inherit a genetically transmitted propensity for rape, which was selected for because men with poor chances of reproductive success would have a better chance of reproducing if they raped vulnerable females than if they did not.76 However, there are several reasons to doubt this claim.77 For at least for the 100,000 or so years of (biologically modern) human history, it seems likely that the expected fitness gains to rape were offset by the cost, including lethal punishment by group members related to the victim. Adultery is a principal cause of homicide of males in some hunter-gatherer societies;78 the penalty for rape is presumably at least as severe. (However, an inherited propensity for rape of female outsiders would not be as vulnerable to the lethal punishment objection.) A universal male propensity for rape based on reproductive success does not easily account for the raping of girls under reproductive age (20 to 30 percent of rape victims in the United States)79 and elderly women, the excessive violence of many rapes, or the prevalence of gang rape.80 In any case, modern biology views genes as expressed only if environmental conditions are propitious, so the mapping from genotype to phenotype is much weaker than this theory presumes. Other versions of the opportunity argument assume merely that with an increase in opportunity men with a propensity to rape will do so more frequently or that more men (but not necessarily all) will rape.81 If sexual violence should vary with opportunity, as this reasoning suggests, there are two additional implications. The first is that if we assume a narrow notion of opportunity as access to civilians, such that opportunity for sexual violence against a civilian is also an opportunity to rob or kill that civilian, then sexual violence by that group should vary with other forms of violence. Groups that supply themselves by looting civilian homes have more opportunity for various forms of violence than groups that do not. The second is that sexual

ELISABETH JEAN WOOD

323

violence should not be targeted toward members of particular groups (unless opportunity varies systematically with groups). However, these predicted patterns based on opportunity as access to civilians are not sustained by the cases considered here. Co-variation in forms of violence sometimes occurs, but is often not the case: some armed factions such as the Sri Lankan and Peruvian insurgencies appear to strictly limit sexual violence but carry out other violence against civilians. Opportunity more broadly understood depends on armed group strategy as well as access to civilians. If an armed group strongly punishes combatants who engage in sexual violence, the opportunity to do so is less as the likely costs exceed the benefits. Opportunity broadly speaking also depends on the norms and practices of small units: if some members of a particular small group frequently carry out sexual violence, the social costs of sexual violence for other members may be lowered by conformity effects, with the result that individuals not particularly inclined to sexual violence may participate out of concern for his or her status within the group (a process similar to sexual violence on the part of some youth gangs). Moreover, in many conflicts, armed groups do not target all women with rape, but women (and sometimes men and children) of particular ethnicities or other social characteristics.82 While some sexual violence seems to be opportunistic (as in the rape of British and French as well as German women by U.S. troops in World War II), in other conflicts it is highly targeted, for example, on women of a particular ethnicity or ideology, as in Bosnia-Herzegovina where Serbs directed sexual violence exclusively at Bosnian Muslims and Croatians. It does not seem likely that differences in opportunity narrowly understood account for such ethnic targeting in societies where populations were very mixed before the war. Incentive A distinct approach is that war leads to an increase of sexual violence because wartime experience increases the incentive to engage in sexual violence, not merely the opportunity to do so. It is sometimes proposed that wartime increase in sexual violence is rooted in biology: wartime sexual violence is higher because of a putative link between the aggression necessary for combat and male sex drive (via testosterone). But the relationship between aggression, testosterone, and sexual drive is complex and, to the extent that it is understood, what we know gives little support for this proposal. Research in this area is difficult because testosterone levels in males vary over the course of the day, some findings in animals do not generalize to humans, social processes affect testosterone levels as well as vice versa, and the design of experimental studies is often inadequate.83 The salient findings to date appear to be the following.

324

POLITICS & SOCIETY

Testosterone appears to affect aggression in three ways. Before birth, high testosterone levels in fetuses with both an X and Y chromosome cause fetuses to become male and also organize the brain in a distinct way. Second, testosterone levels vary among males (and also over the lifespan of each) and there is evidence of a positive but weak relationship between aggression and testosterone levels among normal men. (One careful analysis of thirty-five studies of the relationship found an average correlation of 0.08 between testosterone levels and observed aggression.)84 Third, the relationship is apparently reciprocal: in male (but not female) individuals, testosterone levels increase in anticipation of physical competition such as sports and status contests such as chess games.85 Afterward, the winners level tends to remain high but the losers decreases. And the response to challenges to status appears to depend on culture: male experimental subjects from the U.S. South exhibited increased testosterone after being insulted, but males from the North did not.86 Thus an armed group might be particularly aggressive if its members had high testosterone levels or if combat had effects similar to physical competition and status contests, particularly among the winners. For aggression to take a sexual form, it would seem either that members of the armed group must also feel increased sexual desire or that they hold cultural beliefs rendering sexual violence a favored form of aggression even in the absence of desire (e.g., sexual attacks with weapons rather than the penis). Focusing on the former for now, the relationship between testosterone, sexual desire, and sexual activity appears to exhibit a threshold effect: if testosterone is below the threshold, an increase in testosterone has no effect on desire or activity.87 Above the threshold, in normal men an increase in testosterone increases sexual desire and to some extent sexual activity, which depends on social norms, status, and structure as well as on desire.88 Among convicted sex offenders, high testosterone levels were associated with more violent attacks and also with higher rates of recidivism (but only among those who did not complete an intensive treatment program).89 The fact that armed groups often engage in sexual violence during or after a successful engagement (for example, after the fall of a besieged city or in the course of forcing a civilian population to flee) might seem consistent with the proposed relationship between competition, increased testosterone, and engagement in sexual violence. And given that much sexual violence takes the form of rape of women by men using erect penises, it seems reasonable to suppose that sexual desire plays some role in sexual violence. But, not surprisingly, the proposed argument on its own cannot easily explain the observed variation in sexual violence, as it would presume (if it were true) significantly higher levels of testosterone in male combatants fighting for armed groups that engage extensively in sexual violence for no articulated reason. The burgeoning literature on the role of social processes in mediating the relationship between testosterone and aggression suggests that an explanation for the observed variation in sexual

ELISABETH JEAN WOOD

325

violence may lie in variation among armed groups in the social processes that regulate aggression. I return to this below. A very different explanation for increased incentive for sexual violence in wartime begins with an understanding of peacetime gender relations as patriarchal, in which womens inferior social status is maintained by the state and other institutions and by violence, including sexual violence.90 In wartime, the enforcement of gender relations by the state and other institutions tends to break down as their presence in war zones is weaker; in their absence, men resort more frequently to violence to enforce gender roles. The argument is similar to that often given to the rise in lynching after the Civil War in the U.S. South: with the outlawing of slavery, violence such as lynching of African Americans increased. Cynthia Enloe advanced a variation of this argument: sexual violence increases during war because gender roles become more polarized.91 Arguments based on patriarchal social relations imply that sexual violence should be more prevalent in wars in which traditional gender norms are more disrupted. In many civil wars, gender roles become less polarized because village hierarchies break down as the population disperses and women take on tasks normally carried out by men. It does not appear to be the case that sexual violence is higher when traditional norms are more disrupted. Contrary to the patriarchal thesis, in some conflicts patriarchal relations are so disrupted that there are significant numbers of female combatants in insurgent factions. Rather than the predicted high rates of sexual violence, rates appear to have been very low in two such cases, the insurgencies in El Salvador and Sri Lanka. And women sometimes participate in sexual violence as in Rwanda, where women sometimes incited men to rape, and in Iraq, Guantnamo, and Afghanistan, where U.S. servicewomen played key roles in the sexual humiliation of prisoners. Moreover, neither of these two explanations appears to account for the explicit targeting of certain people for sexual violence. A victory-driven increase in testosterone does not explain why enemy women and sometimes men, rather than women of ones own group, are often singled out for sexual violence. And based on the second argument, we would expect that women who broke traditional gender roles would be particularly targeted, but that does not appear to be the case. One reason often advanced for sexual violence based on increased wartime incentives that does account for targeting of enemy civilians is that of revenge. During war, combatants target enemy civilians with violence in revenge for the violence suffered by themselves, their family, or community members. However, why revenge takes the form of sexual violence rather than other violence should also be addressed. Sexual violence is sometimes said to occur in retaliation for sexual violence previously suffered (or rumored to suffer) by co-ethnics, but as our cases showed, sexual violence is often exercised by only one party to the war.

326

POLITICS & SOCIETY

Joshua Goldstein advances an argument for increased sexual violence during war that also accounts for the targeting of enemy women and men, and with specifically sexual violence.92 He argues that in order to persuade men to fight and endure all the terrors and hardships of war, societies need members willing to stand fast under fire. An extremely common way in which that is accomplished relies on the development of sharp distinctions between genders: to become men, boys must become warriors. Societies need for warriors therefore results in universal rituals of manhood that include tests of physical courage, endurance, strength, self-control, and obedience. The gendered formation of soldiers thus rests on particular ideas about manhood: leaders persuade soldiers that to be a real man is to assert a militaristic masculinity. One result of such practices is that soldiers then represent domination of the enemy in a gendered way, leading to the use of specifically sexual violence against enemy women and, occasionally, against enemy men who are dominated through male rape and castration.93 Goldstein also argues that it is loyalty to the small unit, not the army or the nation, that enables men to fight under the terrifying conditions of war; the bonding among members of the unit is therefore essential and usually takes gendered forms, reinforcing the militaristic masculinity advanced by military training. In her analysis of violence by the U.S. military, Morris similarly emphasizes the particular practices of the primary groups to which soldiers belong. A primary group is a small number of people who share a common ideology and among whom personal, affective bonding takes place; other bonds are undermined through initiation rituals.94 The sexual and gender norms imparted to recruits in their primary groups are inadvertently comprised largely of the sort associated with rape propensity, such as an understanding of masculinity as dominance, aggressiveness, and risk taking; adversarial sexual beliefs (both sexes manipulate and exploit the other); promiscuity; and general hostility toward women (including erroneous beliefs about rape, such as that women enjoy it).95 After documenting particular practices in the U.S. military in support of her argument, she reasons, like Goldstein, that this pattern is shared among military organizations generally. This argument emphasizing norms of militarized masculinity that rely on gendered representations of domination of the enemy accounts both for the targeting of enemy women and men and for the use of specifically sexual violence. Together with their emphasis on the importance of the bonding between men of the same unit, this might also account for gang rapes in wartime (as a form of male bonding among primary groups). However, if the militarized masculinity argument is to explain variation in wartime sexual violence, it would have to be the case that armies promote different notions of masculinity, with the armies that emphasize more militaristic notions of manhood responsible for higher levels of sexual violence. I am not

ELISABETH JEAN WOOD

327

aware of systematic comparisons of military training, norms, and practices across armed groups, but the variation in sexual violence among state militaries appears significantly greater than the variation in their training, which appears surprisingly similar; nor does training appear to vary much among insurgent groups. Moreover, there are obvious exceptions to the claimed relationship: the Salvadoran insurgency, one of the two most militarily effective guerrilla armies in Latin America, had little record of sexual violence despite their highly militarized notion of masculinity. Perhaps variation in sexual violence is better addressed by variation in military discipline than training and socialization. I return to this issue below. Sexual Violence as Instrumental for the Group In the previous explanations for sexual violence based on increased opportunity and incentive, sexual violence occurred for reasons of individual gratification or as a by-product of necessary training. In contrast, in some conflicts sexual violence is promoted or tolerated by (at least local) leaders of some armed groups as an effective means toward its goals. Such instrumental sexual violence may serve as a reward for participation. Or it may be tolerated as a form of small-unit solidarity promoting bonding of its members. Or it may be seen as a form of terror or punishment, as in Berlin and Nanjing, despite its undermining of military effectiveness. In Rwanda, pre-war propaganda denigrating and sexualizing Tutsi women created a climate in which mass sexual violence appeared to be an appropriate form of retribution for long-standing grievances.96 Or it may be pursued as a form of torture as in the Latin American instances discussed above and in detention sites in Iraq, Afghanistan, and Guantnamo, where US forces use sexual humiliation to soften up suspects for interrogation. In some cases, an armed group engages in sexual violence against civilian members of its own community, or its own combatants, as when such targets are suspected of collaborating with the enemy. However, the most prevalent form of selective violence against collaborators in civil wars is homicide, particularly in certain zones of war in which an army is in control but not dominant.97 Why some armies deploy sexual violence to control and punish collaborators while others do not remains unexplained. The most notorious instrumental use of widespread sexual violence against civilians occurs (sometimes) as part of ethnic cleansing, in which violence is used against entire populations to force their movement from particular regions claimed as the homeland, and as part of some genocides. The conditions for such instrumental promotion of sexual violence are not well specified in the literature. Several authors suggest that sexual violence is an effective form of wartime violence in particular cultural settings. For example, sexual violence may be an effective form of ethnic cleansing or genocidal violence,

328

POLITICS & SOCIETY

destroying the social fabric of a society, when used against groups that understand sexual violence against a woman as a violation of the familys honor as well as hers, and as humiliation of her male relatives.98 Cynthia Enloe suggests a more refined hypothesis along the same lines:
if military strategists . . . imagine that women provide the backbone of the enemys culture, if they define women chiefly as breeders, if they define women as mens property and as the symbols of mens honor, if they imagine that residential communities rely on womens workif any or all of these beliefs about societys proper gendered division of labor are held by war-waging policy makersthey will be tempted to devise an overall military operation that includes their male soldiers sexual assault of women.99

Enloe makes a similar argument concerning sexual torture of suspected insurgents. It is especially likely, she reasons, when the regime is preoccupied with national security, a majority of civilians understand security as military security; security apparatuses are dominated by masculinist males; the definitions of honor, loyalty, and treason are derived from misogynous military and police cultures; men seen as threats are also seen as vulnerable through their roles as fathers, lovers, and husbands; and some local women are publicly visible as opposition leaders.100 These proposed conditions for the instrumental promotion of sexual violence are hypotheses generated by cases in which instrumental sexual violence occurred and have not been confirmed by careful empirical testing across cases. They appear to predict more sexual violence than is in fact observed: masculinist notions of honor are present in many societies that do not see massive sexual violence during conflict, as in El Salvador and Israel/Palestine. In addition, most instrumentalist accounts do not adequately address two issues important for the approach. The first is whether sexual violence as a strategy originated at the top of the command structure or at some lower level (and then may have diffused to other sites or groups). The second is whether the organization has a command and control structure sufficient to enforce a strategy of sexual violence if promoted by top leaders. Goals may diverge widely between leaders of an armed group and individual members,101 resulting in a potential gap between measures advocated at the top and priorities among small units on the ground. Sanctions against Sexual Violence The effectiveness of an armed groups command and control structure is also important for the opposite pattern, the effective prohibition of sexual violence by the groups leaders. Even if leaders were persuaded of the military effectiveness of sexual violence against particular groups, they might decide to prohibit it for normative, strategic, or practical considerations. For example, if an organization aspires to govern the civilian population, leaders will probably attempt

ELISABETH JEAN WOOD

329

to restrain combatants engagement in sexual violence against those civilians for fear of undermining support for the coming revolution. Similarly, if an armed group is dependent on civilians, leaders will probably attempt to restrain sexual violence. Jeremy Weinstein holds that an insurgent army whose military capacity depends on the voluntary and ongoing provision of intelligence and other services by civilians will only attract highly committed insurgents and is likely to limit its use of coercion and violence. In contrast, he asserts, armies that attract opportunistic not idealistic recruits, as when members have easy access to resources such as abundant, lootable natural resources, are less likely to constrain their use of violence against civilians.102 These considerations should extend to sexual violence: armies that do not depend on civilian populations will not limit their use of sexual violence. However, the Colombian leftist insurgent groups appear not to follow this pattern: despite their reliance on revenues from coca paste and kidnapping, they engage in relatively little sexual violence.103 Reasons for prohibiting sexual violence may reflect normative concerns as well as practical constraints. Revolutionary groups seeking to carry out a social revolution see themselves as the disciplined bearers of a new, more just, social order for all citizens; sexual violence may conflict with their self-image. For example, the norms and practices of liberation theology informed many of the practices and values of the Salvadoran insurgency;104 it is difficult to imagine the organization embracing liberation theology while violating one of Catholicisms central norms, the sanctity of womanhood. Similar normative considerations may account for the restraint of the Colombian insurgents. A norm against sexual violence may take a distinct form: in some conflicts, sexual violence across ethnic boundaries may be understood by leaders or combatants as polluting the instigator rather than humiliating the victim and the social group. New social norms against the use of particular forms of violence and in favor of others may also be actively cultivated by an armed group as a matter of strategy or principle. The Salvadoran insurgency attempted to shape individual longings for revenge toward a more general aspiration for justice because revenge seeking by individuals would undermine insurgent discipline and obedience.105 Despite systematic celebration of martyrdom in pursuit of victory, the insurgency did not endorse suicide missions and explicitly prohibited sexual violence. The Sri Lankan insurgents, in contrast, carry out suicide bombing and, arguably, shape desires for revenge toward that end, yet also do not engage in sexual violence toward civilians despite their practice of ethnic cleansing. Dependence on international allies may also constrain sexual violence if those allies have normative concerns about such violence. Even if neither the armed group nor its sponsor is itself normatively concerned, it may seek to avoid criticism by international human rights organizations. These examples suggest that armed groups may promote selected forms of violence and form able soldiers without tolerating sexual violence. An army for

330

POLITICS & SOCIETY

whom females comprise a high fraction of combatants may be particularly constrained in its use of sexual violence. This is suggested by the empirical pattern that female-intensive insurgencies in El Salvador, Sri Lanka, Peru, and Colombia appear to carry out less sexual violence. However, the mechanism is not clear, and these insurgencies share other characteristics as well, such as an unusual degree of internal discipline. Whether an armed group effectively enforces particular sanctions and norms decided on by leaders generally depends on the groups military discipline. In particular, prohibition of sexual violence based on practical constraints raises two issues. The first is whether the constraint operates directly on combatants as well as leaders. If soldiers do not themselves feel the direct causal pinch of the constraint, whether the constraint in fact constrains then depends on the degree of discipline within the organization. Many armies probably prohibit sexual violence yet do not in fact discipline soldiers who commit it. The second issue is methodological: independent evidence for the existence of constraints may be difficult to establish beyond the non-observation of the type of violence supposedly constrained. However, if combatants themselves have individual internalized norms against sexual violence or if small units share such a norm, small units may effectively enforce the norm without relying on the hierarchical discipline of the armed group.
CONCLUSION

The literature on sexual violence during war has yet to provide an adequate explanation for its variation across wars and armed groups. While many authors have distinguished between opportunistic and strategic sexual violence, the empirical pattern of variation is wider, including wars where sexual violence is remarkably low on the part of one or more parties to the conflict. In the light of comparative analysis, we do not adequately understand the conditions under which armed groups provide effective sanctions against their combatants engaging in sexual violence or those under which groups effectively promote its strategic use. In concluding, I identify the contributions of this article, suggest some possible patterns amid the variation, and outline a broader research agenda. Before proceeding, I should note explicitly that I have used a rhetorical device throughout the paper: I assumed a single, deterministic, causal path to a particular level or form of sexual violence such that a single instance that contradicted that path led to its rejection. It is likely that a range of causal mechanisms interact to create the variation in sexual violence and that a probabilistic rather than deterministic approach is necessary to account for the overall pattern of variation. The first contribution of the article is its emphasis on the neglected category of armed groups that do not engage in sexual violence, widening the variation that needs to be explained.

ELISABETH JEAN WOOD

331

A second contribution is the idea that addressing the puzzle of variation in sexual violence requires three levels of analysis, that of the armed group (an insurgent group or a national military), the small unit in which combatants have face-to-face relations, and the individual.106 An initial distinction is whether the armed group provides effective incentives that promote sexual violence or sanctions that prohibit it. If there are no effective sanctions either promoting or discouraging sexual violence (either because the armed group does not have an explicit policy or because there is no effective enforcement of that policy), the degree of sexual violence engaged in by combatants depends on whether the group has access to civilians (as when it loots kitchens and fields for food) or not, whether small units have norms prohibiting or endorsing sexual violence, and whether individuals have such norms. Strategic sexual violence appears to occur when an armed group believes it to be an effective form of terror against or punishment of a targeted group. While strategic sexual violence may not be explicitly ordered, it is (at least) tolerated; if any punishment occurs it is symbolic and limited, clearly for external consumption rather than deterrence. Such violence appears to take two broad forms. The first is sexual torture and/or humiliation of persons detained by an armed group (as in the treatment of persons detained by state agents in El Salvador and by U.S. forces in Iraq and Guantnamo). The second is widespread sexual violence against a targeted group, which frequently takes the form of gang (and often public) rape, tends to occur in a variety of settings, and usually occurs over extended periods of time. The latter form appears during some but not all ethnic conflicts; many but not all groups engaged in the forced movement of ethnically defined populations perpetuate it. The extent of opportunistic sexual violence depends on the absence of sanctions and norms (on the part of the armed group, the small unit, or the individual) that effectively prohibit it and on proximity to potential victims. Where individual and small-unit norms prohibit sexual violence, perhaps on the grounds that it is polluting to the perpetrator, sexual violence will not occur even if a unit has ready access to civilians and even if the armed group does not punish those who engage in it. Under what conditions armed groups, small units, and individuals develop sanctions and norms that effectively endorse or constrain combatants engagement in sexual violence is, of course, key to explaining the observed variation. While an adequate explanation is beyond the scope of this paper, some observations and hypotheses do arise from the analysis. Hypothesis 1. Where insurgent groups depend on the provision of support (supplies, intelligence) from civilians and aspire to govern those civilians, they do not engage in sexual violence against those civilians if they have a reasonably effective command structure. Suggestive evidence: Leftist insurgencies

332

POLITICS & SOCIETY

in Latin America, which typically have intensive socialization processes and effective command structures, do not engage in sexual violence. Hypothesis 2. Where the norms (either condemning or approving sexual violence) held by individual combatants and small units are the same and are also endorsed by the armed groups leadership, sexual violence by that group will be either very low or very high, respectively. Such alignment of norms may have a complementary rather than a merely additive effect: each norms effect is strengthened by the presence of the same norm at the other levels. Specifically, where armed groups reinforce cultural taboos against sexual contact with the potential target populations, sexual violence against that population will be low. And in the absence of such cultural taboos, where armed groups promote a policy of sexual violence against a population, violence will be high. Suggestive evidence: There was relatively little sexual violence (apart from sexual humiliation) in the labor and concentration camps of Nazi Germany. And the high marriage rate among ethnic groups before the conflicts in Bosnia-Herzegovina and Rwanda, according to this hypothesis, facilitated the widespread sexual violence during the conflicts. Hypothesis 3. States confronting an armed threat will tolerate (and possibly promote) a degree of sexual violence against suspected insurgent supporters during detention, constrained by the degree of accountability of state agents to civilian authorities and those authorities beliefs about the acceptability of sexual violence. Suggestive evidence: State forces engaged in sexual torture against detainees suspected of support for insurgent groups in El Salvador, Peru, and Argentina. And Jennifer Green found that state forces engaged in collective rape in twenty-three of the thirty cases of widespread sexual violence that she identified.107 Hypothesis 4. To the extent that military forces in democracies are more accountable for their practices to civilian authorities, sexual violence will depend on the norms and tactics of civilian leaders, who may endorse some types of sexual violence while effectively sanctioning others. Suggestive evidence: Democracies rarely engage in widespread sexual violence and generally punish rape for personal gratification, but limited sexual violence in the form of sexual humiliation against persons in detention by U.S. forces is an ongoing practice reflecting its effective endorsement by civilian leaders. Hypothesis 5. If an armed group prohibits sexual violence against a particular population, the less effective the military discipline of the group, the more likely it is to engage in sexual violence (unless combatants hold particularly strong norms against it). Thus ill-disciplined militias, ill-trained armies of conscripts, poorly trained military police, and little supervised service troops are more likely to engage in sexual violence than elite frontline troops (in the absence of a policy promoting sexual violence). Hypotheses 6. Armed groups with a high proportion of female combatants engage less in sexual violence. I am uncertain why this seems to be the case in Sri Lanka, El Salvador, Peru, and Colombia, among others. (The exception is U.S. forces, where women in the military police and military intelligence

ELISABETH JEAN WOOD

333

participated in sexual violence against detainees.) The presence of female combatants may disrupt the dynamics of bonding in small units, may displace patriarchal role models that support sexual violence, may constrain sexual violence for fear that the enemy could target fellow soldiers in kind, or may put in motion some other mechanism. However, the documented variation in patterns of sexual violence raises questions that far exceed these hypotheses. To what extent is sexual violence accounted for by a breakdown in command-and-control structure and morale versus a change in norms on the part of combatants? What accounts for the emergence of an organizational structure strong enough to enforce strategic decision by the leadership? How and why do small-unit norms evolve that enable sexual violence by its members? In what conditions do military victory, on the one hand, and military stalemate, on the other, contribute to sexual violence? To what extent do international norms and law constrain the practice of sexual violence? Why are men targeted in some settings but not in others? In particular, what accounts for the distinct forms of sexual violence? To what extent does the form of sexual violence reflect the type of war in which it occurs, with public gang rape tending to occur during ethnic cleansing, sexual torture in states confronting threats, and so on? More broadly, the following avenues of research may contribute to addressing the overall puzzle of variation in sexual violence. First, more research is needed to better document variation in the patterns of sexual violence across conflicts, including those analyzed here, and to assess whether the hypotheses above make sense in light of more cases. In particular, because the cases were chosen for their variation in sexual violence, additional research is needed to estimate the relative frequency of occurrence of different patterns in the actual universe of cases.108 Second, efforts to gather data on the extent and form of sexual violence should include cases where it appears to have been relatively infrequent. Such research on negative cases of groups or conflicts where sexual violence does not occur (or occurs at low levels) should illuminate cases where it does. Of particular interest are those conflicts where one party does not mirror the use of sexual violence by another party to the war and conflicts where sexual violence seems anomalously low in the light of high rates in similar conflicts. This article suggests a key distinction among such negative cases: whether sexual violence does not occur due to effective sanctions against it, individual norms against it, or small-unit norms against it, or because the group has little access to civilians. In particular, a comparison of the working of ideological and religious or other cultural mores against sexual violence might shed light on the character of many insurgencies. However, establishing the operative force of such mores poses a particular methodological challenge, namely, how to establish

334

POLITICS & SOCIETY

the causal force of a stated norm or sanction independently of the observed presence or absence of sexual violence. Such research requires access to detailed local sources, which is not always possible during or in the aftermath of war. However, wars differ in the availability of such records and the possibility of extended local field research.109 Fortunately, it is precisely such negative cases for which local research may be possible. Colombia may be a case of particular interest, for it appears that sexual violence was fairly frequent during the period of civil violence termed la Violencia (1948-1958), but has not been a significant element in the repertoire of violence of the insurgent groups today.110 Third, to explore the force of potential causal processes, within-case contrasts should be explored as the simplest way to control for many otherwise confounding variables. This approach is already proving very rich for the study of violence and participation in civil war, including in Greece,111 Rwanda,112 Peru,113 and El Salvador.114 Ideally, one could compare patterns of sexual (and other) violence not just between factions and over time, but across subunits of the armed factions, thereby clarifying the causal force of factors at different levels.115 The extent to which sexual violence varies with other forms of violence should also be analyzed as a way to identify particular strategies and norms of violence. A particularly interesting case would be that of U.S. forces in Vietnam, assuming that documents exist that report unit behavior and that they are or could be declassified. Comparing patterns of sexual violence in different colonies of the same empire would also be an illuminating variation on this research design. Fourth, the small-group dynamics that lead to unit norms promoting or constraining the occurrence of sexual violence appear a promising avenue of research. Relevant factors include recruitment of individuals who endorse the group norm and conformity to the norm once the individual is a member of the group. For example, there may be systematic differences between armed groups that rely on mercenaries, career professionals, and conscripts. In particular, the extent to which military training practices differ among armies in the degree of brutalization of recruits and in the activities to build bonds between members of the small units could be a fruitful avenue for further exploration. Comparison to small-group dynamics in other settings where group sexual violence sometimes occurs, such as fraternities, urban gangs, and sports teams, may prove fruitful. A fifth, related, issue that would be very illuminating is the study of perpetrators of wartime sexual violence. Although such research would be difficult to carry out, for human subjects concerns as well as practical reasons, it may not be impossible. Scott Straus was able to interview a particular subset of perpetrators of the Rwandan genocide, those who had been convicted, had confessed, and had been sentenced.116 A sixth avenue of research would focus more explicitly on dynamic interactive mechanisms. For example, patterns of sexual violence might be fruitfully

ELISABETH JEAN WOOD

335

analyzed with a model based on positive feedback mechanisms that amplify small initial differences between groups, units, or sites and result in large differences in the prevalence and form of sexual violence. Such models may illuminate the diffusion of decentralized norms that condone sexual violence, for example. One such mechanism is escalating revenge: if a member of one party commits sexual violence against a member of another group, a member of the other may retaliate in ways leading to a spiral of sexual violence. Or epidemiological models might be productive, in which if some members of a small group commit sexual violence, other members of that small group may do so as well; once that small group does, neighboring units may join in, leading to widespread sexual violence by that party to the war. In both cases the dynamic processes explaining the escalation or dampening of violence will be characterized by tipping points such that seemingly small differences in the causes of violence would account for large differences in the consequences. The ongoing brutality in Darfur reminds us that sexual violence remains a horrifying aspect of war. Understanding the determinants of its variation may help to define policies better able to curtail this savage form of violence that targets the most vulnerable civilians, often with the intention to ruin them for life.
NOTES

1. This paper joins other recent works that analyze variation in wartime sexual violence but focuses on a wider range of variation. See Lisa Boswell Sharlach, Sexual Violence as Political Terror (Ph.D. diss., University of California, Davis, 2001); J. Robert Lilly and Pam Marshall, Rape in Wartime, in The Encyclopedia of Criminology and Deviant Behavior, ed. Clifton D. Bryant (London: Brunner-Routledge, 2000), 318-22; and Mia Bloom, War and the Politics of Rape: Ethnic Versus Non-ethnic Conflicts, unpublished manuscript. 2. See UNESCO, Contemporary Forms of Slavery: Systematic Rape, Sexual Slavery and Slavery-like Practices during Armed Conflict, E/CN.4/Sub.2/1998/13 (New York: United Nations, 1998); and Human Rights Watch, Well Kill You if You Cry: Sexual Violence in the Sierra Leone Conflict (New York: Human Rights Watch, 2003). 3. Sexual violence differs from the broader category of gender violence in that the latter category includes violence that occurs because of the victims gender without the kinds of sexual contact included in sexual violence. 4. Norman M. Naimark, The Russians in Germany. A History of the Soviet Zone of Occupation, 1945-1949 (Boston: Belknap Press of Harvard University Press, 1995), 69-140. 5. Ibid., 72; see also 74. 6. Ibid., 106-7. 7. Lilly and Marshall, Rape in Wartime. 8. Antony Beevor, The Fall of Berlin 1945 (New York: Penguin, 2002), 300, 326, 413. 9. Naimark, The Russians in Germany, 88-90. 10. The records of women requesting abortions confirm a high prevalence of rape in Berlin. While abortion was technically illegal, authorities suspended the law in the case of rape by foreigners; permission was granted to nearly all cases in the district whose records were analyzed by Atina Grossman. See Atina Grossman, A Question of Silence:

336

POLITICS & SOCIETY

The Rape of German Women by Occupation Soldiers, in West Germany under Construction: Politics, Society, and Culture in the Adenauer Era, ed. Robert G. Moeller (Ann Arbor: University of Michigan Press), 33-51. 11. Beevor, The Fall of Berlin 1945, 410. 12. The incidence of rape (incidents/population) would be much higher than the prevalence (victims/population) given the pattern of gang rapes and multiple incidents suffered by the same person. 13. Quoted in Naimark, The Russians in Germany, 72. 14. Ibid., 71. 15. Jeffrey Burds, personal communication, December 3, 2006. 16. Wendy Jo Gertjejanssen, Victims, Heroes, Survivors: Sexual Violence on the Eastern Front during World War II (Ph.D. diss., University of Minnesota, 2004). 17. Jonathan C. Friedman, Speaking the Unspeakable: Essays on Sexuality, Gender, and Holocaust Survivor Memory (Lanham, Md.: University Press of America, 2002), ch. 2. The German military treated rape of civilians by German soldiers on the eastern front much more leniently than on the western front, where military courts imposed significantly more severe punishment. See Birgit Beck, Rape: The Military Trials of Sexual Crimes Committed by Soldiers in the Wehrmach, 1939-1944, in Home/Front: The Military, War and Gender in 20th Century Germany, ed. Karen Hagerman and Stefanie Schuler-Springorum (New York: Berg, 2002), 255-73. 18. Gertjejanssen, Victims, Heroes, Survivors, 220. 19. Based on Gertjejanssens description of the camp brothels, I estimate the number of women in camp brothels to have been between 1,000 and 10,000. 20. Iris Chang, The Rape of Nanking: The Forgotten Holocaust of World War II (New York: Penguin, 1997). Chang draws on a wide range of documents, including the diaries and reports of international observers who remained in Nanjing throughout the violence, as well as some interviews. It is not clear how Chang arrives at this estimate. 21. Joshua A. Goldstein, War and Gender: How Gender Shapes the War System and Vice Versa (Cambridge: Cambridge University Press, 2001), 367. The system was begun in 1932 but expanded extensively in the aftermath of the violence in Nanjing. 22. Japanese Cabinet Councillors Office on External Affairs, On the Issue of Wartime Comfort Women, cited in UNESCO, Contemporary Forms of Slavery, appendix, 9(a). The precise number of women forced to serve as military sexual slaves is not well documented, as the Japanese destroyed much of the documentation in 1945. 23. Cited in Chung Hyun-Kyung, Your Comfort versus My Death: Korean Comfort Women, in Wars Dirty Secret: Rape, Prostitution, and Other Crimes against Women, ed. Anne Llewellyn Barstow (Cleveland, Ohio: Pilgrim Press, 2000), 17-19. 24. Cited in Goldstein, War and Gender, 363; and Cynthia Enloe, Maneuvers: The International Politics of Militarizing Womens Lives (Berkeley: University of California Press, 2000), 140. Twenty thousand girls and women comprise 2.1 percent of female Muslims in pre-war Bosnia-Herzegovina of all ages. The UN Special Rapporteur of the Commission on Human Rights initially made a lower estimate of 11,900 rapes, based on 119 pregnancies resulting from rape that were aborted in six major medical centers (the rapporteur assumed a rate of pregnancy after rape of about 1 percent); cited in Todd Salzman, Rape Camps, Forced Impregnation, and Ethnic Cleansing, in Barstow, ed., Wars Dirty Secret, 76. However, as Salzman points out, on the one hand many women were raped more than once and others had no access to medical facilities and induced abortion themselves, abandoned the child, or kept the child. On the other hand, many

ELISABETH JEAN WOOD

337

pregnant women who sought abortions did not indicate that pregnancy originated in rape. On balance, Salzman argues that the number of pregnancies was likely significantly higher than 119, and he concurs with the 20,000 estimate (Salzman, Rape Camps, 76-77, 63). 25. United Nations Security Council (UNSC), Rape and Sexual Assault, in Final Report of the United Nations Commission of Experts Established Pursuant to Security Council Resolution 780 (1992), S/1994/674/Add.2, vol. V (New York: United Nations, 1994), annex IX.I.C. 26. This history of Foa draws on UNSC, Rape and Sexual Assault; and Joanne Barkan, As Old as War Itself: Rape in Foa, Dissent (Winter 2002): 60-66. 27. UNSC, Rape and Sexual Assault, annex IX.2.A.20. 28. Eight men from Foa were indicted by the International Criminal Tribunal for the former Yugoslavia (ICTY) on sixty-two counts of sexual assault and rape as crimes against humanity and grave breaches and violations of the laws and customs of war (see Barkan, As Old as War Itself, 65). The three who were tried received sentences of twenty-eight, twenty, and twelve years. 29. UNSC, Rape and Sexual Assault, esp. annex IX. 30. Ibid., annex IX.I.A. 31. Ibid., annex IX.A. 32. Ibid., annex IX.I.C. 33. Ibid., annex IX.A III.A.2. 34. Ibid., annex IX, Conclusions. 35. Of course, one reason the ethnic violence in the former Yugoslavia seemed troubling to many observers was the fact of significant intermarriage before the war: from 1981 to 1991, 18.6 percent of new marriages in Bosnia-Herzegovina were inter-ethnic (1991 census figures cited in Enloe, Maneuvers, 142). 36. See Amnesty International, Sri Lanka: Torture in Custody, 1999, http://www .amnestyusa.org/countries/sri_lanka/reports.do; and United Nations Development Fund for Women, Gender Profile of the Conflict in Sri Lanka, rev. October 31, 2005, http://www.womenwarpeace.org/sri_lanka/sri_lanka.htm. 37. Amnesty International documented sexual violence by government forces against eleven women between 1999 and 2001. See Amnesty International, Sri Lanka: Rape in Custody, 2002, http://www.amnestyusa.org/countries/sri_lanka/reports.do. 38. Miranda Alison, Cogs in a Wheel? Women in the Liberation Tigers of Tamil Eelam, Civil Wars 6, no. 4 (2003): 34-54. 39. Human Rights Watch, Living in Fear: Child Soldiers and the Tamil Tigers in Sri Lanka, vol. 16, no. 13 (C) (New York: Human Rights Watch, 2004); and United Nations Development Fund for Women, Gender Profile. 40. Benny Morris, The Birth of the Palestinian Problem Revisited (Cambridge: Cambridge University Press, 2004). 41. In the testimonies compiled by Human Rights Watch and Physicians for Human Rights, victims reported perpetrators wanting sex with a virgin, wanting a new wife, wanting to send a message to the government, and so on, but do not report perpetrators stating a wish to have sex with or to punish a person of particular ethnicity or religion. See Human Rights Watch, Well Kill You if You Cry; and Physicians for Human Rights (PHR), War-Related Sexual Violence in Sierra Leone (Boston: PHR, 2002). 42. Truth and Reconciliation Commission of Sierra Leone, Final Report, 2005, http://trcsierraleone.org/drwebsite/publish/index.shtml, ch. 3b, para. 282. 43. Human Rights Watch, Well Kill You if You Cry.

338

POLITICS & SOCIETY

44. Truth and Reconciliation Commission of Sierra Leone, Final Report, paras. 292-96; and Human Rights Watch, Well Kill You if You Cry, 35-42. 45. Lynn L. Amowitz, Chen Reis, Kristina Hare Lyons, Beth Vann, Binta Mansaray, Adyinka M. Akinsulure-Smith, Louise Taylor, and Vincent Iacopino, Prevalence of War-Related Sexual Violence and Other Human Rights Abuses among Internally Displaced Persons in Sierra Leone, Journal of the American Medical Association 287, no. 4 (2002): 513-21. The survey design combined systematic random sampling and cluster sampling in four locales representing 91 percent of the internally displaced population. The estimated prevalence rate appears to be several times higher than the peacetime rate (the estimated lifetime prevalence of non-war-related sexual violence is 9.0 percent; Amowitz et al., Prevalence of War-Related Sexual Violence, 518). 46. Ibid., table 3. 47. Ibid., table 2. 48. Human Rights Watch, Well Kill You if You Cry. 49. See Physicians for Human Rights, War-Related Sexual Violence in Sierra Leone, ch. 4; and Human Rights Watch, Well Kill You if You Cry, ch. V. According to Physicians for Human Rights, girls and women who undergo female genital cutting are at increased risk for genital trauma and related complications after rape (War-Related Sexual Violence, 49). Human Rights Watch reports that 90 percent of females in Sierra Leone undergo female genital cutting (Well Kill You if You Cry, 24). 50. Truth and Reconciliation Commission of Sierra Leone, Final Report, ch. 3b, 299-311. 51. Amowitz et al., Prevalence of War-Related Sexual Violence, table 3. 52. Forced marriages in the sense of marriages of girls without their consent, often at a very young age, was common in Sierra Leone before the war but required permission of the girls family (Human Rights Watch, Well Kill You if You Cry, 17, 23-24). 53. Truth and Reconciliation Commission of Sierra Leone, Final Report, para. 298. 54. This summary of events at My Lai draws on Michael Bilton, Four Hours in My Lai (New York: Penguin, 1992); Susan Brownmiller, Against Our Will (New York: Bantam, 1975); Seymour M. Hersh, My Lai 4: A Report on the Massacre and Its Aftermath (New York: Random House, 1970); Herbert C. Kelman and V. Lee Hamilton, Crimes of Obedience (New Haven, Conn.: Yale University Press, 1989); William R. Peers, The My Lai Inquiry (New York: Norton, 1979); and James Olson, My Lai: A Brief History with Documents (New York: Bedford, 1998). 55. Mitch Weiss, Rogue GIs Unleashed Wave of Terror in Central Highlands, Toledo (Ohio) Blade, October 22, 2003. 56. Mark Danner, The Massacre at El Mozote (New York: Vintage, 1994). 57. Truth Commission for El Salvador, From Madness to Hope: The 12 Year War in El Salvador. Report of the Truth Commission for El Salvador, reprinted in The United Nations and El Salvador, 1990-1995, United Nations Blue Books Series, vol. IV (New York: United Nations, 1993). 58. Truth Commission for El Salvador, From Madness to Hope, unpublished annex (Tomo II: 8-10, 15). 59. Elisabeth Jean Wood, Insurgent Collective Action and Civil War in El Salvador (Cambridge: Cambridge University Press, 2003), ch. 4. 60. See the various maps and graphs compiled in Truth and Reconciliation Commission, Peru, Final Report (2003), http://www.cverdad.org.pe/ingles/ifinal/index.php, Statistical Annex. 61. In her analysis of testimonies to the commission, Michele Leiby found that Sendero Luminoso was responsible for nearly 20 percent of reported cases of sexual violence. Her

ELISABETH JEAN WOOD

339

finding is based on a wider definition of sexual violence and her coding of reported instances of homicides, torture, or other human rights violations as also sexual violence if they included sexual violence. See Michele L. Leiby, Sexual Violence as a Strategic Weapon of War: Latin America, unpublished manuscript. 62. Ethel Tobach and Rachel Reed, Understanding Rape, in Evolution Gender and Rape, ed. Cheryl Brown Travis (Cambridge, Mass.: MIT Press, 2003), tables 5.1 and 5.2. 63. Patricia D. Rozee, Forbidden or Forgiven? Rape in Cross-Cultural Perspective, Psychology of Women Quarterly 17 (1993): 499-514. For example, in some societies, sexual access to women is granted to guests, brothers, or other associates of the husband (and the women are beaten or killed if they refuse). And in some societies, female transgression of social norms (such as women seeing ceremonial artifacts strictly reserved for males) is punished by rape, sometimes group rape in a public place. 64. Even if all females who were pregnant as a result of rape reported the incident to health authorities and were in fact pregnant as a result of rape, that would not be enough to infer the incidence of rape given the fact that multiple rapes appear to be frequent in wartime settings. 65. World Health Organization, Sexual Violence, in World Report on Violence and Health (Geneva: World Health Organization, 2002), table 6.1, 151. For developing countries, the data is compiled from face-to-face surveys in the capital city; there is apparently no correction for possible rural-urban differences other than for variation in household size. Given the challenges to compiling comparable sexual violence data, I average across the lowest and highest three cities. 66. J. N. van Kesteren, P. Mayhew, and P. Nieuwbeerta, Criminal Victimization in Seventeen Industrialised Countries: Key Findings from the 2000 International Crime Victims Survey (The Hague: Ministry of Justice, Research and Documentation Center [WODC], 2000), app. 4, table 6, 188-89. For industrialized countries, the surveys are national samples and done by phone (with the exception of Malta). The high reported rates in Sweden and Finland probably reflect high rates of binge alcoholism, with attendant violence, or higher rates of reporting sexual assault, and a more inclusive definition of assault. 67. See Rozee, Forbidden or Forgiven? 504. 68. Peggy Sanday, Socio-cultural Context of Rape: Cross Cultural Study, Journal of Social Issues 37, no. 4 (1981): 27. 69. Rozee, Forbidden or Forgiven? 70. Ibid., table 1. 71. Ibid., table 1. 72. Neil Mitchell and Tali Gluch, The Principals and Agents of Political Violence and the Strategic and Private Benefits of Rape (paper presented at the annual meeting of the American Political Science Association, Chicago, September 2004). 73. The United Nations Interregional Crime and Justice Research Institutes (INICRI) crime victimization data for most developing countries goes back only to 1996 or 1997 (and in a few cases to 1992), and only a few countries with recent civil wars are included. Of the cases discussed above, El Salvador, Sri Lanka, Israel, and Sierra Leone are not included; while a survey was carried out in Yugoslavia in 1996, it is difficult to see its relevance for Bosnia-Herzegovina in 1992. 74. Madeline Morris, By Force of Arms: Rape, War, and Military Culture, Duke Law Journal 45, no. 4 (1996): 651-781. 75. Ibid. 76. Randy Thornhill and Craig Palmer, The Natural History of Rape: Biological Basis of Social Coercion (Cambridge, Mass.: MIT Press, 2002). 77. See the essays in Brown Travis, Evolution Gender and Rape.

340

POLITICS & SOCIETY

78. Richard Borshay Lee, The Kung San: Men, Women and Work in a Foraging Society (Cambridge: Cambridge University Press, 1979). 79. Mary P. Koss, Evolutionary Models of Why Men Rape: Acknowledging the Complexities, in Brown Travis, Evolution Gender and Rape, 191-205. 80. Jerry A. Coyne, Of Vice and Men: A Case Study in Evolutionary Psychology, in Brown Travis, Evolution Gender and Rape, 171-89. However, see Jonathan A. Gottschall and Tiffani A. Gottschall, Are Per-Incident Rape-Pregnancy Rates Higher than Per-Incident Consensual Pregnancy Rates? Human Nature 14, no. 1 (2003): 1-20, for a discussion of the possibility that a conditional rape strategy that accounts for such targeting objections could have evolved. 81. For example, Mitchell and Gluch, in Principals and Agents, argue that the principalagent problem confronting armies is the tendency of combatants to seek to engage in more sexual violence than the leadership deems optimal. 82. Inger Skjelsbaek, Sexual Violence and War: Mapping out a Complex Relationship, European Journal of International Relations 7, no. 2 (2001): 218. 83. Allan Mazur and Alan Booth, Testosterone and Dominance in Men, Behavioral and Brain Sciences 21 (1998): 353-97. 84. John Archer, Nicola Graham-Kevan, and Michelle Davies, Testosterone and Aggression: A Reanalysis of Book, Starzyk, and Quinseys 2001 Study, Aggression and Violent Behavior 10 (2005): 241-61. 85. Mazur and Booth, Testosterone and Dominance. 86. Richard E. Nisbett and Dov Cohen, Culture of Honor: The Psychology of Violence in the South (Boulder, Colo.: Westview, 1996). 87. Mazur and Booth, Testosterone and Dominance. 88. Theodore D. Kemper, Social Structure and Testosterone (New Brunswick, N.J.: Rutgers University Press, 1990); and Theodore D. Kemper, Fantasy, Females, Sexuality and Testosterone, Behavioral and Brain Sciences 21 (1998): 378-79; and Jeremy Freese, Jui-Chung Allen Li, and Lisa D. Wade, The Potential Relevances of Biology to Social Inquiry, Annual Review of Sociology 29 (2003): 233-56. 89. Lea H. Studer, A. Scott Aylwin, and John R. Reddon, Testosterone, Sexual Offense Recidivism and Treatment Effect among Adult Male Sex Offenders, Sexual Abuse: A Journal of Research and Treatment 17, no. 2 (2005): 171-81. 90. Versions of the argument can be found in many feminist works; the classic work is Brownmiller, Against Our Will. 91. Cynthia Enloe, Does Khaki Become You? The Militarization of Womens Lives (Berkeley: University of California Press, 1983). 92. Goldstein, War and Gender, 253-300. 93. Ibid., 356-60. 94. Morris, By Force of Arms, 692. 95. Ibid., 707, 701-6. There appears to be significant consensus among researchers about the factors at the individual level that are associated with increased likelihood to suffer sexual violence or to perpetrate sexual violence in peacetime. In addition to Morriss summary, see World Health Organization, Sexual Violence, ch. 6. 96. Lisa Sharlach, Gender and Genocide in Rwanda: Women as Agents and Objects of Genocide, Journal of Genocide Research 1, no. 3 (1999): 387-99. 97. Stathis Kalyvas, The Logic of Violence in Civil Wars (Cambridge: Cambridge University Press, 2006). 98. Lisa Sharlach, Rape as Genocide: Bangladesh, the former Yugoslavia, and Rwanda, New Political Science 22, no. 1 (2000): 89-102; and Bloom, War and the Politics of Rape.

ELISABETH JEAN WOOD

341

99. Enloe, Maneuvers, 134. 100. Ibid., 124. 101. Stathis Kalyvas, The Ontology of Political Violence: Action and Identity in Civil Wars, Perspectives on Politics 1, no. 3 (2003): 475-94. 102. Jeremy Weinstein, Inside Rebellion: The Politics of Insurgent Violence (Cambridge: Cambridge University Press, 2006). 103. Human Rights Watch, Youll Learn Not to Cry: Child Combatants in Colombia (New York: Human Rights Watch, 2003), 10. 104. Kalvyas, The Ontology of Political Violence. 105. Wood, Insurgent Collective Action. 106. See Amelia Hoover, Disaggregating Violence during Armed Conflict: Why and How, unpublished manuscript, for an analysis of how different repertoires of violence can be understood via principal agent models in which elites (the principals) have distinct preferences than do combatants (the agents) for different types of violence. 107. Jennifer L. Green, Uncovering Collective Rape: A Comparative Study of Political Sexual Violence, International Journal of Sociology 34, no. 1 (2004): 97-116. 108. Green, in Uncovering Collective Rape, takes a distinct approach, identifying cases of collective rape via online searches of the New York Times archive. The strength of this approach is that she has tentative findings about the correlates of collective rape; the weakness is the reliance on a single source and the difficulty in coding relative prevalence of rape based on that source. 109. Elisabeth Jean Wood, The Ethical Challenges of Field Research in Conflict Zones, Qualitative Sociology 29, no. 3 (2006). 110. Francisco Gutirrez Sann, personal communication, July 2005. 111. Kalyvas, The Logic of Violence. 112. Scott Straus, The Order of Genocide: Race, Power, and War in Rwanda (Ithaca, N.Y.: Cornell University Press, forthcoming). 113. Weinstein, Inside Rebellion. 114. Wood, Insurgent Collective Action. 115. Macartan Humphreys and Jeremy Weinstein intend to carry out such a disaggregated analysis of patterns of violence in the Sierra Leone conflict, drawing on a survey they carried out with ex-combatants to document (among other things) patterns of command-and-control and discipline among particular units. However, as yet they do not have adequately disaggregated data on human rights violations. If such data becomes available, they will be able to test how well unit discipline predicts sexual and other violence. See Macartan Humphreys and Jeremy Weinstein, Handling and Manhandling Civilians in Civil War: Determinants of the Strategies of Armed Factions, unpublished paper, 2004. 116. Straus, The Order of Genocide. Elisabeth Jean Wood (elisabeth.wood@yale.edu) is professor of political science at Yale University and research professor at the Santa Fe Institute. She is the author of Forging Democracy from Below: Insurgent Transitions in South Africa and El Salvador (Cambridge University Press, 2000), and Insurgent Collective Action and Civil War in El Salvador (Cambridge University Press, 2003).