The Commander's Dilemma: Violence and Restraint in Wartime

THE COMMANDER’S DILEMMA

Violence and Restraint in Wartime

Amelia Hoover Green

CORNELL UNIVERSITY PRESS ITHACA AND LONDON

To Jarrod and Henry

Contents

List of Tables and Figures

Acknowledgments

List of Abbreviations

Introduction: Repertoires and Restraint

1. The Commander’s Dilemma

2. Civil War in El Salvador

3. Comparing State and FMLN Institutions and Ideologies

4. Institutions, Ideologies, and Combatant Experiences in FMLN Factions

5. Violence and Restraint in the Salvadoran Civil War, 1980–92

6. The Commander’s Dilemma beyond El Salvador

Conclusion: Policies for Restraint

Appendix

Notes

Works Cited

Index

Tables and Figures

Tables

0.1 Levels and repertoires of wartime violence

1.1 Armed-group institutions and modes of influence

1.2 General empirical implications of the Commander’s Dilemma framework

1.3 Indicators of armed-group and combatant characteristics

3.1 Formal education among FMLN and state recruits

3.2 Recollections of political education among FMLN and state recruits

3.3 Perceptions of discipline among FMLN and state ex-combatants

5.1 Lethal violence as reported by four datasets

5.2 Correlations between raw datasets and MSE estimates, by department (top rows) and year (bottom rows)

5.3 Repertoire elements as a proportion of total violence as reported by four datasets

5.4 Empirical implications of the Commander’s Dilemma and other approaches

6.1 Descriptive statistics: Three Mano River datasets

6.2 Probability of reporting sexual violence, given any report of violence by a group

Figures

0.1 Comparing PTS and SVAC scores for seventy-five civil wars

1.1 Costs of compliance over time, discipline versus political education

1.2 From individual lists to multiple systems estimation

3.1 Pages from the 15 Principles

4.1 Front covers of ERP political-education materials: levels one, two, and three

5.1 Overlap patterns among four Salvadoran data sources

5.2 Custom strata for global MSE estimates

5.3 Comparing MSE and raw data estimates of lethal violence by year

5.4 Comparing reported levels of different repertoire elements among four datasets

5.5 Proportion of survey respondents who saw or heard about rape, by year of recruitment and armed group

Acknowledgments

First and most important, I owe a debt of gratitude to the people who welcomed me to their homes and communities in El Salvador. Their observations are the backbone of this work, although I cannot name them here. I can and very happily do name my colleague Erika Murcia, who served as research assistant, translator, logistics point person, tour guide, and first sounding board for all ideas during each of my trips to El Salvador. Erika’s work made this book possible, and made it better, and I am so grateful for her work and her wisdom. Other colleagues whose research assistance has shaped this book include Quenten Hare, Greta Jusyte, Alexander Nadolishny, Ebony Pleasants, Belén Rodriguez, and Maggie Von Vogt.

This book owes its existence to the steadfast support of my academic mentors, beginning with Carol Nackenoff, Ken Sharpe, and Rick Valelly at Swarthmore. I am grateful also to teachers who managed to give radically different pieces of advice that never tripped over each other or blocked my winding way. Stathis Kalyvas gave project-saving advice at some key, distressful moments. Patrick Ball provided invaluable methodological advice and instruction, unwavering practical support, and an activist’s passion for making work that matters in the world. And I am immeasurably grateful for the mentorship of Elisabeth Wood. Her support has been more important than I have adequate words to describe. Some of the words that come to mind, though: intellectually rich, constant, kind, generous, and deeply ethical. Libby believed in me during the moments when I did not quite believe in myself.

My colleagues, near and far, old and new, junior and senior, long-term and short-, have provided invaluable feedback, conversation, and encouragement. Phillip Ayoub, Julia Azari, Maria Eriksson Baaz, Laia Balcells, Regina Bateson, Kanisha Bond, Zoltan Buzas, Karisa Cloward, Kerry Crawford, Kathleen Cunningham, Will d’Ambruoso, Stephen Engel, Lee Ann Fujii, Scott Gates, Anita Gohdes, Erin Graham, Tamy Guberek, Daniel Gúzman, Paul Kirby, Jeff Klingner, Jule Krüger, Milli Lake, Jason Lyall, Meghan Lynch, Devorah Manekin, Zoe Marks, Michele Leiby, Sarah Parkinson, Rob Person, Megan Price, Emily Ritter, Anastasia Shesterinina, Laura Sjoberg, Paolo Spada, Abbey Steele, and Scott Straus all contributed to this project, and I thank them. I am grateful also to the United States Institute of Peace, Jeff Checkel, Will Moore and Yonatan Lupu, Theo McLauchlin, Helen Kinsella, and Will d’Ambruoso, all of whom invited me to workshops and seminars that shaped this final product. My coauthor and friend Dara Kay Cohen deserves special thanks for many years of discussions, ideas, and support. Several other scholars and former scholars contributed to this book by encouraging me to care for myself and by spending time commiserating about mental health in the academy. Things being what they are, I will not name them here.

My editor at Cornell, Roger Haydon, has been patient and encouraging throughout the project, from a lengthy crack-of-dawn breakfast at the International Studies Association meeting in 2015, to the inevitable delays, to the last-minute questions of a first-time author. Thank you.

I received funding at various stages of this project, without which it could never have been completed. I am grateful to the United States Institute of Peace Jennings Randolph Junior Fellows program; the Whitney and Betty MacMillan Center for International and Area Studies at Yale University; the Institute for Social and Policy Studies, also at Yale University; the Human Rights Data Analysis Group (then in incubation at the Benetech Initiative); and Drexel University. An Antelo Devereux Award for Excellence in Research allowed me to complete a final field visit to Morazán in 2015. I also thank the Museo de la Palabra y la Imágen, in San Salvador, for unfettered access to its archives.

I spent most of the years I worked on this book living among housemates and friends. Everyone who shared a home with me during the last decade deserves my thanks for food, drink, shoulders to cry on, gossip, hootenannies, and enforced downtime. Louisa Egan Brad, Stephen Engel, Joe Blodgett, Joey Kotfica, Jamie Timmons, Elizabeth Eager, and Angelo Hernandez: I’m in your debt. Neighborhood friends who have made Philadelphia our forever home (and made parenting feel possible) include Jenny Lunstead, John Pappas, Carrie Fafarman, and Aaron Fafarman. The Scones had my back. Laurel Eckhouse and Emily Clough have taught me everything about friendships of character and surviving in the academy. Anne Bakken, my friend since childhood, always has Diet Coke and encouragement on hand when I need them.

Finally, I could not have done this without my family. John and Betsy Hoover, Ray and Donny Hoover, you gave me my love of learning and helped me be my best nerd-self. Laury Fischer, Sue Schweik, Sara Dickey, Rob Jackson, Zach Fischer, Marria Kee, Solon Smith, and all: thank you for welcoming me and sharing my excitement about this work. Finally, to Jarrod, whom I met as I started this project, and Henry, who joined us as I raced toward its finish, thank you for making my life awesome every day: for devotion, conversation, hilarious antics, domestic and emotional labor, guitar tunes, big jumps, blueberry smiles, resilience, patience, and love. This book is for the two of you.

Abbreviations

Introduction

REPERTOIRES AND RESTRAINT

Repertoires are learned cultural creations. … They emerge from struggle.

—Charles Tilly, Contentious Repertoires in Great Britain, 1758–1834

Over about three days in December 1981, an elite, American-trained battalion of the Armed Forces of El Salvador murdered approximately one thousand people in and around the village of El Mozote, in the department of Morazán. Rufina Amaya was one of the few eyewitness survivors to these killings. Her testimony, a version of which was published in 1996 as Luciérnagas en El Mozote, describes both the killings and the crimes that accompanied them:

[We] were taken out of the houses and had to lay face down in the street, including children, and [the soldiers] took everything: necklaces, money. … [The next morning the soldiers killed some of the men.] These they beheaded and threw into the convent. At noon, they finished killing all the men and went to get girls to take them to the hills [where many were raped]. … At five in the afternoon they took me with a group of 22 women. I was the last in line. I was breastfeeding my daughter. They took her from my arms. … The soldiers killed that group of women without realizing that I had hidden, and went to bring another group. About seven o’clock in the evening they finished. …

After joking about the witches burning in the sacristy, Amaya writes, soldiers looted snacks and drinks from the local store. Amaya crawled away and hid until the soldiers moved on, then took shelter in a cottage near the Rio Sapo. Eight days later, a local family found Amaya as they were bringing corn for storage; by Christmas, rebel forces had begun to inspect El Mozote and the surrounding hamlets, and to bury the dead. American reporters reached the area in early January. Both Raymond Bonner, in the New York Times, and Alma Guillermoprieto, in the Washington Post, interviewed Amaya for their stories about the massacre, published on January 27, 1982.¹

Amaya quickly became the victim of a vicious smear campaign by US and Salvadoran officials, who claimed alternately that no massacre had occurred, or that civilians in northern Morazán were guerrilla sympathizers (Binford 2016, chap. 4; Bonner 1984; Guillermoprieto 2007). Yet Amaya lived to see her story vindicated by the Truth Commission at war’s end, played a key role in the exhumation and reburial of victims, and remained in Morazán until her death, in 2007. The El Mozote massacre, its truth no longer in doubt, has become the best-known episode of violence against civilians during the Salvadoran civil war, and a stand-in for the brutality of state forces. Yet it is not a representative episode: when they killed, Salvadoran forces typically did so in ones and twos, or from a distance. Death squads seized individual victims from their homes and dumped their mutilated bodies at San Salvador’s giant municipal garbage dump; planes and helicopters dropped bombs that obliterated isolated villages. But nonlethal violence was at least as common: paramilitary security forces detained thousands of civilians, torturing them for information, punishment, and fun. Soldiers stole food, sexually harassed and raped women during operations in villages, and otherwise conducted an indiscriminate and deeply alienating campaign of terror against the civilian population.

State violence in El Salvador contrasts starkly with violence committed against civilians by the opposing guerrilla forces, the Farabundo Martí National Liberation Front (Frente Farabundo Martí para la Liberación Nacional [FMLN]). Prototypical violence by the FMLN included the targeted assassinations of individuals thought to be collaborating with the government, including a number of mayors (Betancur, Figueredo Planchart, and Buergenthal 1993). Civilian massacres by FMLN cadres were nearly unheard of. Indeed, the most sustained campaign of FMLN murders appears to have occurred when the political-military commander of the paracentral region (the departments of Cabañas, San Vicente, and La Paz), known as Mayo Sibrian, spiraled into paranoia following a period of imprisonment in 1984–85. Between assuming command of the region in 1986 and his execution in 1991, Mayo Sibrian ordered the torture and killing of hundreds of guerrilla fighters and civilian collaborators on charges of infiltration (Galeas and Ayala 2008). Here too, a common pattern of violence emerges. A witness to the violence (quoted in Galeas and Ayala 2008, Prologue, part 1) testified:

Ethel [Pocasangre Campos, a guerrillera known as Crucita] was accused of treason. … On the 22nd of September of 1986, at a place in the village of San Bartolo, near a hill, her own guerrilla commanders tied her up and threw her, half-naked, into the mud. During several hours they tortured her, beating her with a club made of guava wood, while they demanded that she confess and give up her supposed accomplices. After that, she was executed and buried in a common grave, together with fifteen other combatants killed in the same way that day.

Yet despite the climate of fear created by Mayo Sibrian’s actions, and despite what observers describe as low morale among his troops, it seems clear that the violence, while horrific, was also tightly controlled. Unlike the laundry list of state atrocities, there appeared to be little violence against nonaligned civilians. No clear instances of sexual violence, extortion, or theft are discussed, even by the most critical sources; the closest we come is the reference to Crucita’s having been half-naked as she faced torture. Across the span of the war, FMLN cadres committed low levels of violence overall, and extremely low levels of sexual violence and property crimes. The story of why and, more important, how the FMLN implemented this policy makes up an important part of this volume.

A partial contrast to the Salvadoran war can be found in Perú, where Shining Path guerrillas were responsible for a number of brutal retributive massacres. Perhaps the most famous of these occurred in the village of Lucanamarca, where in 1983 Shining Path cadres murdered over sixty men, women, and children, using extraordinarily brutal, personal, performative violence. Antonio Quincho testified to the Peruvian Truth and Reconciliation Commission that he returned to his home a few days after the massacre to find victims with their hands and feet tied up, the braids cut off the young women, who had been cut with axes, knives, and picks; they had even thrown hot water on them. … We found children with their little hands burned, their little faces. … They had pulled out the guts of newborn babies and stomped their heads until the brains came out (testimony to the Truth and Reconciliation Commission for Perú [Comisión de la Verdad y la Reconciliación de Perú (CVR)], vol. 7, chap. 2, sec. 6, 46; my translation).

The Shining Path’s ideology was ostensibly similar to that of the Salvadoran FMLN, yet Shining Path cadres were responsible for vast and brutal lethal violence—approximately half of all killings during the Peruvian civil war (Ball et al., 2003). Much of the violence committed by Shining Path cadres could be described as overkill—violence in excess of that required to kill, including performatively gruesome acts such as those described above. Indeed, the Shining Path has been described as a cult of violence, in which violence became not only a means to exert control but a purifying force in itself. Yet in most times and in most areas, Shining Path cadres committed very few acts of sexual violence relative to government forces; in this regard, the Lucanamarca massacre looked very little like El Mozote.

The conflicts in El Salvador and Perú featured leftist guerrilla forces fighting large, conventionally oriented state militaries. Post–Cold War conflicts, on the other hand, often pit weak state militaries, no longer propped up by Cold War superpowers, against nonstate actors whose political motives may be outweighed by their profit motives (cf. Collier 2000). S. J., a wealthy woman from Bombali District in northern Sierra Leone, was one of thousands caught in brutal fighting between the Revolutionary United Front (RUF) and the Sierra Leone Army (SLA) in 1999. In a 2003 Human Rights Watch report, she described murder, rape, torture, and pillage by RUF combatants:

I saw them kill three people and were it not for God, I would have been the fourth. Then they burned thirteen houses and looted all our things. I ran with my four children to the house in the bush where we tend to the cows … Seven rebels surprised us there. … They started stealing what few possessions I had and then [the commanding officer of the unit] said that I should be raped. … Four raped me and the last one to rape me … put a knife to my throat and said he was going to kill me but the C.O. said I shouldn’t be killed. Then they tied my hands behind me and [the C.O.] burnt me. He scooped up hot charcoal from the fire we had been cooking with and tried to burn my face with it. I struggled and turned my face so he burned my chest instead. (Human Rights Watch 2003, 37)

Across the country in Bonthe District, J. K. suffered violence at the hands of Community Defense Forces (CDFs), also known as Kamajors:

One of the Kamajors called Kinie said that they had been told that my brother was in the village and was planning to attack them. I assured them no one knew where he was. During this argument, the other civilians in [the] village became afraid and fled into the bush. As soon as the Kamajors forced their way into my bedroom, I followed them to check up on what they were doing. Kinie and another Kamajor whose name I did not know pushed me to the ground, tearing off my clothes. I screamed for help but no one came to my rescue. Even my father who was in the house was unable to help me. They both raped me while the others stood around laughing. When they left the village, they looted some goats and chickens. There was no one to report the incident to and I had no money to pay for a hospital visit. I decided to leave everything to the Almighty God. (Human Rights Watch 2003, 47)

J. K.’s story is superficially similar to S. J.’s. Indeed, Human Rights Watch published both testimonies as part of a report documenting sexual violence by all parties to the conflict. But it is important to know that sexual violence by Kamajors was relatively rare, while sexual violence by RUF forces was common. While CDF violence was significant, it seldom featured sexual violence, amputation, or the other crimes for which RUF cadres were feared and reviled. Why? Why did Salvadoran troops massacre the village of El Mozote and terrorize the broader civilian population, while its rebel enemies carefully controlled violence against civilians?

Consider for a moment the full range of abuses described in the firsthand narratives above. Soldiers steal animals, poison wells, and set fires. They sexually harass women and cut their hair. They rape and torture. They discuss, argue, accuse, joke, drink looted soda pop, and grab chickens on their way out of town. The nitty-gritty details of violence play an important role in eyewitness narratives, and to nonexpert readers they have a clear feel— regimented or berserk, ritualistic or oddly workmanlike. Yet many or most would be erased from systematic studies of wartime violence. Social scientists often strip away gore and complexity as we search for the logic of violence (cf. Kalyvas 2006). In particular, we often focus on lethal violence, excluding other forms from our analyses and our attention. The stories above are reduced to a list of body counts; stories that include horrific violence other than death are often erased completely. Unfortunately, the simple, compelling theories of (lethal) violence that are social scientists’ stock-in-trade offer few explanations for the abduction, destruction, detention, disappearance, disfigurement, displacement, sexual violence, slavery, theft, and torture that often accompany killings.

But the fact remains that, despite an intense narrative and historical focus on death, most victims of atrocities during war survive. Disregard for living victims, I argue, is an analytical problem as much as a moral one. The main question of this book is, Why do some combatants and groups commit many types of violence, seemingly indiscriminately, while others carefully limit the forms or levels of violence they employ?

Repertoires of Violence

Answering this book’s central question requires sustained, systematic attention to the repertoire of violence, which I define as the forms of violence frequently used by an actor, and their relative proportions. By violence, I mean direct physical aggression, either against persons or against immediate necessities such as food or shelter; or specific threats of imminent, direct physical violence. Murder, sexual violence, arbitrary detention, torture, amputation, and other physical violence are included, but so—for the purposes of this volume—are looting, livestock killings, and the burning of homes. While I recognize institutional, structural, and psychological violence as elements of any repertoire, in this volume I generally focus on direct physical violence. (I understand, of course, that this means my own work is subject to the same concerns about ignoring analytically and morally consequential forms of violence that I raise about other works. My hope is that this book will be the first of many to systematically examine multiple forms of violence in conflict.)

While some political scientists have examined repertoires of violence during individual wars (notably James Ron’s [1997] work on the Israeli Defense Forces; and Francisco Gutierrez Sanín’s [2004, 2008] work on the Revolutionary Armed Forces of Colombia), none has theorized repertoires per se, and most (again, with notable exceptions: see Wood 2006b, 2009; Cohen 2013, 2016) have essentially ignored nonlethal violence. Among the most important goals of this book is to consider how, and how systematically, repertoires of violence differ from time to time, place to place, and group to group.

As a concept, the repertoire of violence is a direct descendant of Charles Tilly’s (1993) concept of repertoires of contention, the definition of which serves as the epigraph for this introduction. Tilly was thinking about the nonelectoral ways that people, often organized in movements, push for political goals: boycotts, letter writing, demonstrations, sit-ins, riots, and so on. I am thinking about the violent acts that armed groups and fighters use (primarily, in this book, against civilians) for any reason or—and this will become important—for no apparent reason at all. But the analogies connecting repertoires in art, repertoires of contention, and repertoires of violence are complicated and incomplete. Tilly himself distinguishes, usefully yet confusingly, between the repertoire of improvisations available to jazz musicians and the tightly defined rotation of a repertory theatrical company. A sixteen-bar solo is a very different endeavor from a decision about Lear versus Godot.

Tilly stresses that repertoires are learned cultural creations. … At any point in history, however, [people] learn only a rather small number of alternative ways to act collectively (1993, 264). For Tilly, what is interesting about repertoires of contention is their boundedness. He stresses improvisation, but also learning, deliberation, choice making, and strategy. One thing that sets repertoires of wartime violence apart from repertoires of peacetime contention, I argue, is the extent to which desperation or lack of information can destroy deliberation, choice making, and strategy. Repertoires of violence may change suddenly; changes may diffuse through larger groups quickly, or remain isolated. While repertoires of violence are bounded, they are less constrained by external forces than repertoires of contention are.

Repertoires of violence may be narrow, encompassing only a few tightly controlled forms of violence, or broad, comprising many forms. An armed group’s repertoire of violence may be uniform across individuals or subgroups, or nonuniform. Qualitatively, by broad repertoire of violence I mean a repertoire of violence in which many different forms of violence are represented, including both forms of violence that often require planning or infrastructure and are often ordered by commanders (killing, torture), and forms of violence that often happen in the absence of orders (sexual violence, looting, extortion). The FMLN, Shining Path, and Kamajors used narrow repertoires of violence in most times and places, contrasting with the Salvadoran Army, the Peruvian Army, and the Revolutionary United Front.

In addition to describing repertoires of violence qualitatively, I often describe the breadth of a repertoire of violence using one of several rough numerical proxies: the number of forms of violence regularly used (although that measure requires defining regularly), the proportion of episodes of violence comprising a particular form of violence (often, as I describe in the next chapter, looting or sexual violence), or the ratio of lethal to nonlethal violence, for example. I refer to an actor’s violence as restrained when it uses low levels of violence (as measured by total incidents of violence), narrow repertoires of violence (specializing in one or a few types of violence), or both. Often, though, I eschew quantitative measures altogether, simply because accurate measurement of wartime violence is extraordinarily difficult. Where I do use quantitative measures, I use a variety of strategies to create repertoire estimates, and to check these estimates against qualitative and historical understandings.

I have argued (and will argue at greater length in chapter 1) that repertoires and levels of violence are analytically separable; that is, it is both possible and fruitful to think of them separately. At the same time, it is important to acknowledge how repertoires and levels of violence affect one another. For example, in El Salvador—the site of much of the research for this book—the overall level of violence against civilians by state forces dropped significantly after Vice President George H. W. Bush, speaking at a gathering of Salvadoran business leaders, issued a warning to elites in late 1983 (quoted in Apodaca 2006, 89): Providing assistance to you [the Salvadoran state] is not a popular cause in the United States. Publicity about death squads, great inequalities of income, the killing of American citizens, and military setbacks make it a very unpopular proposition in my country. … Without actions in these areas there is no point in trying to obtain additional funds for El Salvador, and to be honest, we will not even make the effort, because it will be fruitless.

Much of the drop in the government’s level of violence against civilians occurred because state forces lowered their level of lethal violence—the form of violence that, state leaders knew, most concerned their American patrons. In chapter 7, I find no evidence that levels of other forms of violence changed appreciably in either direction. There are two ways to evaluate this change. We could say that repertoires broadened: as killings declined, nonlethal violence necessarily formed a larger proportion of the overall repertoire. By another measure, the repertoire changed very little: killings declined, but they remained a significant part of the state’s repertoire, and consequently the most common forms of violence remained the same. By any measure, the repertoire was, and remained, broad.

What can we infer from this pattern? Existing theories don’t provide full explanations. We know that the change in violence was not the result of rational calculation among enlisted people or low-level officers, because the decline occurred rather suddenly across many state forces. We know that it was not a response to changing rebel strategies, which were essentially static during this period. The state’s recruiting pool remained constant; it wasn’t suddenly recruiting people who would torture but not kill. The relative timing of the change visà-vis the Bush speech suggests that top Salvadoran leaders responded to threats and incentives from their American patrons, a dynamic more often associated with external sponsorship of rebels (e.g., Salehyan, Siroky, and Wood 2014). But even this doesn’t fully explain the change in some, but not all, forms of violence. How much control did Salvadoran leaders actually exercise over low-level officers and soldiers? The high command may have demanded explicitly that individual units decrease killings but continue other forms of violence. Or it may have instructed units only about killings, leaving lower-level personnel to determine a course of action regarding other forms of violence. Or perhaps the high command demanded limits on all forms of violence, but lower-level personnel couldn’t, or wouldn’t, fully comply.

The evidence I present in chapter 4 suggests that top Salvadoran military leaders lacked both the will and the capacity to control violence against civilians. Orders from the top curtailed killings after late 1983, but the situation with other forms of violence was vastly more complicated. Some violence continued, with commanders’ knowledge, because it was less visible to American patrons than the death squad murders and massacres of 1979–83. Paramilitary security forces, in particular, maintained a relatively well-organized system of arbitrary detention and torture in jails and prisons around the country. Within regular-army units, on the other hand, much of the continuing violence was opportunistic, committed either in the absence of orders or against them.

Why study repertoires?

Repertoires are multidimensional; they are difficult to describe succinctly or measure systematically. My colleagues often, and rightly, ask, Why add complexity, when we already have enough difficulty understanding variation in single forms of violence? First, and most important, developing repertoire measures allows the researcher—particularly the quantitative researcher—to examine variation in violence on dimensions other than the number of killings. As Cohen (2016) has persuasively argued, to analyze killings exclusively is to ignore many women victims of violence, because killings are much more likely to target military-age men (cf. Carpenter 2006). Theidon (2004), examining the construction of memory in the highlands of Perú, describes how men steered public meetings toward episodes of slaughter, noting particularly a case in which a local man was burned alive. In a women-only setting, equally impassioned accounts emerged of soldiers destroying food stores, stealing prized livestock, and disappearing breadwinners. Women also discussed both the heroism of and the stigma attached to women who agreed to have sex with soldiers in exchange for the freedom of their spouses or other loved ones. For many of Theidon’s respondents, soldiers shit[ting] in the wheat [stores] was as threatening a demonstration of impunity as a murder or rape. Considering repertoires helps to retain some fidelity to these experiences and priorities, which have often been swept aside in the dramatic narratives of historical memory or social-scientific analysis.

Analytically, examining repertoires overturns some common misunderstandings about violence, as a general phenomenon, that can creep into social-scientific accounts. Until relatively recently, nearly all quantitative research on wartime violence against civilians focused on killings as a stand-in for violence. Most scholars would describe this as an attempt to count systematically one war crime that can often be counted; others have more explicitly chosen to measure killing because of its presumed status as the ultimate atrocity—the absolute violence (Kalyvas 2006, 20, quoting Sofsky 1998, 53). This analytical move assumes, implicitly or explicitly, either that all types of violence have the same cause, or that nonlethal violence is purely anomic and random. Research on sexual violence demonstrates that at least this one category of nonlethal atrocities