Research article
Shades of truth and lies: Interpreting testimonies of war and violence
Lee Ann Fujii
Department of Political Science, George Washington University
Journal of Peace Research
47(2) 231-241
© The Author(s) 2010
Reprints and permission:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0022343309353097
jpr.sagepub.com
Abstract
How should researchers treat questions of veracity when conducting interviews in settings rent by large-scale violence, such as war and genocide? To what extent should researchers trust narratives that are generated in politically sensitive contexts? The article argues that the value of narrative data does not lie solely in their truthfulness or accuracy; it also lies in the meta-data that accompany these testimonies. Meta-data are informants' spoken and unspoken thoughts and feelings which they do not always articulate in their stories or interview responses, but which emerge in other ways. This article identifies and analyzes five types of meta-data: rumors, inventions, denials, evasions, and silences. The article argues that meta-data are not extraneous to our datasets, they are data and should be viewed as integral to the processes of data collection and analysis. Meta-data indicate how conditions in the present shape what people are willing to say about violence in the past, what they have reason to embellish or minimize, and what they prefer to keep to themselves. Attending to meta-data is important for responding to informants' fears about talking to a researcher and to ensure informants' safety after the researcher leaves the field. It is also crucial for the robustness of researchers' theories and knowledge about political violence and other political phenomena. The article draws from the author's nine months of fieldwork in Rwanda in 2004, as well as the literature on conflict and violence from political science, anthropology, history, and sociology.
Keywords
fieldwork, genocide, narrative data, post-conflict, rumors, silences, war
Introduction
Angelique told a harrowing tale. 'They said I had Tutsi blood', she explained. Her voice was soft, her demeanor somber. It was our first meeting after a long day of multiple interviews. I was in Rwanda to talk to rural people who had lived through or participated in genocidal violence. The year was 2004 - ten years after a civil war that had installed a new regime and a genocide that had cost the lives of half a million people.
We sat side by side on a damp log, the ground still wet from rain. I was eager to hear her story, for here was a woman, I thought, who was Hutu but had nonetheless been targeted because her mother was Tutsi.1 Angelique continued. Some neighbors had dug a hole where she was able to hide with her youngest strapped to her back. Her rescuers covered the hole with leaves, providing adequate camouflage for the night. The next day Angelique and her baby fled to safety with Tutsi from the area.
Angelique's experience was another piece of narrative data I was collecting on mass violence that took place during the Rwandan civil war from 1990 to 1994. Her story was consistent with published testimonies (African Rights, 1995; Des
Forges, 1999) and interviews I had conducted with other genocide survivors. Survival, as Angelique's story illustrated, was often a matter of luck and the life-saving gestures of neighbors, friends, and strangers.
Each time I traveled to the research site where Angelique lived, I looked forward to learning more. As the interviews continued, however, Angelique became less precise. The more I probed, the sketchier her story became.
Angelique seemed to have other things on her mind. Her present life was filled with hardship and struggle. After the war, she had returned with the other refugees who had fled across the border, but the government denied her 'survivor' status. Worse, other Tutsi survivors also denied she was a survivor. This made Angelique ineligible for the benefits the government had promised to genocide survivors, which included housing and assistance with school fees.
By our fourth interview, it started to occur to me that Angelique had made up the entire story of her escape. Her statements were not adding up. When I asked why her former Tutsi friends would have denied she was a survivor like them, she said it was because her husband was Hutu. This seemed
1 Ethnicity in Rwanda is defined through the father, so that a child of a Hutu father is Hutu regardless of the ethnicity of the mother.
Corresponding author:
lafujii@gwu.edu
232
journal of Peace Research 47(2)
odd since other survivors I met had been married to Hutu. When I asked her what had become of her husband, she said she did not know. This, too, struck me as odd - that no news of her husband had ever reached her through other refugees, as was the case with others to whom I had talked. When I asked about her parents' background, she gave similarly vague answers. At one point, she went beyond all credibility and claimed that her father had had 39 wives. Polygamy was common in this part of the country but I had never heard of any man having more than two or three wives. I was beginning to doubt everything she was saying.
How should researchers deal with questions of veracity, especially in post-violence settings when the stakes run particularly high? To what extent should researchers trust personal narratives and local histories that are generated in politically sensitive contexts?
This article argues that the value of oral testimonies researchers collect in places that have recently suffered violence does not lie solely in the truthfulness of their content. It also lies in the meta-data that accompany the testimonies. By meta-data, I mean the spoken and unspoken expressions about people's interior thoughts and feelings, which they do not always articulate in their stories or responses to interview questions. Meta-data can take both spoken and unspoken forms. They include rumors, silences, and invented stories. Meta-data are as valuable as the testimonies themselves because they indicate how the current social and political landscape is shaping what people might say to a researcher. By failing to attend to meta-data, analysts risk misinterpreting ambiguities, overlooking important details, drawing incorrect conclusions, and leaving informants vulnerable to reprisals for having talked to the researcher. Paying attention to meta-data is therefore vital to protecting informants and arriving at robust explanations and theories about violence and its aftermath.
Types of meta-data
When I conducted fieldwork in 2004, it had been ten years since the genocide and civil war in Rwanda. My research design called for multiple interviews with people living in two small, face-to-face communities located well outside the capital. It also called for interviews in two central prisons where former residents of both research sites had been imprisoned on charges of genocide. I asked everyone what they saw or did during the war and genocide. I also asked them about their daily lives before that period.
Ten years after events, memories change. People forget some details and mis—remember others. They rearrange chronologies, confuse sequences, and give greater weight to some moments over others. In addition, institutions of all kinds, from prisons to schools, socialize people to construct the past in certain ways (Payne, 2008: 20). Most of the prisoners my interpreter and I spoke with had been in prison for at least eight years, some close to ten. In that amount of time, it is
likely that prison culture helped to produce one particular way of talking, and perhaps thinking, about the genocide.
Violent episodes may also have much greater salience for some communities than others, regardless of how much time has passed. In the United States, for example, whites and blacks in the same small town recall a past lynching that occurred in their town much differently, with black people retaining the details and whites forgetting them (Wolf, 1992).
The forum in which people recall past violence can also shape the testimonies that people provide. Truth commissions privilege certain types of narrative over others. The Peruvian truth commission, for example, was eager to hear women's stories of victimization but not their stories of heroism (Theidon, 2007). Media can also shape how and when perpetrators confess by providing a platform for perpetrators to depict themselves as heroes even when they admit to atrocities and torture (Payne, 2008).
All these factors enhance the importance of meta-data in our fieldwork and analysis. As the following sections show, meta-data are integral to the research enterprise and constitute valuable data in their own right.
In the following sections, I discuss five forms of meta-data: rumors, inventions, denials, evasions, and silences. None of these meta-data are unique to post-conflict settings; they can emerge in any research that involves face-to-face interactions. The stakes in post-conflict settings, however, are usually higher than in non—violent settings. In the aftermath of war, mass killing, and sexual violence, public accusations and private confessions can lead to reprisals from neighbors, rejection from family members, or repression by the state. In settings of high suspicion, where words can be a matter of life and death, attention to meta-data can help the researcher minimize risks to informants and maximize her chances of finding what she seeks.
Rumors
Ethnographers of war and violence have often noted the prominent role that rumors play in periods of extreme uncertainty. In settings that have recently suffered violence, rumors often reflect insecurities that linger from past violence. Green, for example, talks about the rumors that arose about her and her research assistant when she began going to Mayan women's homes to interview them (after conducting initial interviews in public places). As Green (1999: 75) explains, 'Above all else they had not wanted the gringa to be seen coming to their house. Under the scrutiny of surveillance the women were afraid of what others in the village might say about them and me.' Like Green, I, too, encountered many rumors about my interpreter and me during my fieldwork. Not only did these rumors circulate within my two research sites, they also passed from the sites to the prisons where I was conducting interviews, and from the prisons back to the sites. Attending to these rumors was important because they affected what people were willing to say to me.
Most of the rumors revealed people's suspicions about who I and the Rwandans working with me really were. One rumor
Fujii
233
was particularly troubling because it painted our activities as threatening, hence, worthy of suspicion. The person who told us this rumor was a prisoner we had interviewed twice. With all informants, I made sure to obtain informed consent before we began each interview; and after obtaining consent, I asked people if they had questions for us before we started. On this particular day, the prisoner began by asking if rumors he had heard about us were true.
A woman told me that you passed close by my house. The driver2 called my child. My wife told the driver that the child didn't know anything because he was still too young during the war. I am asking if you went to my house. I don't have any other questions. When we get out of prison, the others are saying that you want to take us to Arusha.
The prisoner actually brings up two rumors - that my driver had approached his child and that my interpreter and I were going to take prisoners to Arusha, presumably to the International Criminal Tribunal for Rwanda which is located there. I denied both rumors in the strongest terms possible. I then used this opportunity to investigate how prisoners vetted information they heard through visitors. I asked this prisoner how he and other prisoners could tell whether news they heard from the outside was true or not. He said it was possible to tell in some instances, when, for example, the rumor contained precise details, such as, in this case, an accurate description of my car. In other instances, the prisoner admitted it was not possible to tell. Since vetting was difficult, rumors stood in for truth and hence, easily enhanced people's suspicions of us.
We heard similar rumors in the research sites. One woman we interviewed several times, whom I call Therese,3 was visibly nervous about talking about the genocide. She articulated her fears more than once. At our third interview, she told us that every time we came, she worried that we would ask her about 'the politics of the genocide'. The rumor she had heard was that my interpreter and I were working for the Rwandan government and that we were talking to people about the impending gacaca, a government initiative to establish community-based courts to try the 120,000 people who had been imprisoned for having allegedly participated in the genocide.4 At the time of my fieldwork, gacaca was scheduled to begin nationwide in the coming months. I tried to use this opportunity to understand why Rwandans traded in rumors. Therese explained that rumors were part of everyday life. 'Even if you stay at home, you always hear rumors about yourself, she stated matter-of-factly. Given the ubiquity of rumors, what Therese feared was not who my interpreter and I said we were,
2 In addition to hiring an interpreter, I hired a driver who was skilled at navigating rough roads and looking after my very old car.
3 All people and place names pertaining to Rwanda are pseudonyms to protect identities.
The number of prisoners has since gone down through government release programs (Tertsakian, 2008; Waldorf, 2006).
but who her neighbors said we were. It was the identities her neighbors had assigned to us that cast suspicion on Therese for having talked with us.
Because suspicions can make people reticent, my interpreter and I took care to act in ways that we hoped would dispel people's fears. We showed up when we said we would and sent word if we could not make our appointment. We strove for consistency so that our appearances in the two research sites became less remarkable and more predictable. Our strategy seemed to work. Over time, Therese's fears subsided as did the suspicions of her neighbors. At a later interview, Therese told us that her neighbors had come to think of us as friends dropping by for a visit. She herself likened our visits to that of a priest - we came for short visits, then left, but always came back again. The shift from threatening government agents to welcome visitors reassured me that our continuing presence was not getting Therese into trouble with her neighbors or authorities. It also indicated that she had begun to trust us.
Rumors such as these illustrate the extent to which field research is a two-way street. As many scholars have pointed out, not only are researchers studying their informants, their informants, in turn, are studying them back - to figure out who the researcher is and whether the researcher is a source of potential threat (Peritore, 1990; Portelli, 1991: 30, 64; Wood, 2003: 41-43; Weinstein, 2007: 357). How informants identify researchers can determine the amount or level of access the researcher can gain. If people suspect that researchers are state agents, informants may invoke a party line in interviews and conversations rather than reveal their deeper thoughts. The task of the researcher is to take people's fears and suspicions seriously and try to allay them as much as possible.
Not all rumors pose barriers, however. Some can lead researchers to new avenues of inquiry or discussion. One woman we interviewed, for example, told us a rumor that was circulating about me. The rumor concerned a local woman who had had a child with a muzungu ('foreigner', usually a white foreigner) man. The rumor was that I was this woman's long-lost daughter come back to her natal hill. The rumor indicated how people were making sense of me and my recurring presence in their community. As Sluka (1990: 121) points out, people will use pre-existing categories to define outsiders. This was clearly what the rumor demonstrated. People were typing me according to their own categories, which, in this case, were umunyarwanda (Rwandan) and umuzungu (foreigner), not 'Black', 'Asian', or 'Hispanic' as I would be typed in the USA.
That people were defining me as part Rwandan provided a useful entry point for talking about ethnicity. This was invaluable since the government had banned talk of'Hutu', 'Tutsi', and 'Twa', the three ethnic groups that make up the population. Focusing on how people were constructing my ethnicity allowed me to probe a subject that was critical to my research, without appearing to violate the government's ban. By talking
234
journal of Peace Research 47(2)
about my ethnicity, I believed we were on safe ground. As this example shows, rumors can point to innovative ways to talk about sensitive subjects. Rumors can therefore present access points and not just obstructions.
Attending to rumors is critical in settings rent by violence, when social relations are fragile or fractured. Rumors can indicate the source of people's fears about talking to a researcher. The greater the suspicions, the less likely people will talk openly about past violence or related subjects. Tending to rumors can help researchers gauge levels of trust. It can mean the difference between accessing meaningful stories and listening to empty talk. It can also mean the difference between research that makes people more vulnerable to reprisals and that which minimizes potential harms.
Inventions
Of equal concern to researchers is the possibility of embellished, distorted, or made up stories. Researchers usually try to do their best to distinguish 'fact' from 'fiction' as a way to minimize 'errors' in their datasets. Yet, when it comes to data gathered in the field, 'fact' and 'fiction' may be a false and misleading binary. Even the most earnest and honest informants can make mistakes when recounting past events. Not every story, moreover, lends itself to determinations of truth. People's beliefs about how the world works, for example, cannot be subject to a truth test.5 Similarly, the value of people's narrations about their experiences of violence - what they saw, did, felt, or heard - does not necessarily lie in their 'accuracy' or 'truthfulness'. Their value might lie in the meaning with which the narrator endows the events or moments she narrates. Stories situate the narrator in a larger context; the importance of context makes even inaccurate details revealing. As Portelli (1991: 51) explains:
Oral sources are credible but with a different credibility. The importance of oral testimony may lie not in its adherence to fact, but rather in its departure from it, as imagination, symbolism, and desire emerge. Therefore, there are no 'false' oral sources. .. . 'wrong' statements are still psychologically 'true,' and . .. this truth may be equally as important as factually reliable accounts, (emphasis in original)
As Portelli teaches us, the stories people tell - inventions and all - are valuable because they reflect the speaker's state of mind, aspirations, and desires. All these elements tie the speaker to her larger community and reveal different kinds of truths, such as the psychological truths Portelli points to or the emotional truths that Payne (2008: ch. 7) describes. As Payne shows, even literary fiction can reveal previously undisclosed truths about state violence. A novel written by former Brazilian Air Force officer Pedro Correa Cabral provided the first insider's view of the atrocities that the Brazilian
I thank one of the anonymous reviewers for this point.
security apparatus had committed. Not only was Cabral's novel pathbreaking for its insider perspective, it also remains the primary way that Brazilians have learned about the state-sponsored violence committed in their name. Similarly, the confession of a former Vlakplass death squad worker in South Africa, which was riddled with false identity claims and lurid details about the man's supposed violent activities, nevertheless 'remained a vivid account of the kinds of atrocities the regime committed' (Payne, 2008: 225).
Chandler's (1999) analysis of documents found at the S-21 prison in Cambodia further underscores how known lies can reveal new insights. S-21 was the highly secretive institution that the Pol Pot regime used to extract lengthy and detailed confessions from political prisoners by way of torture. These written confessions, some of which run into hundreds of pages, are filled with imagined and fantastical claims, yet they shed valuable light on the psychology of the Pol Pot regime and how it pursued its ideology of permanent revolution inside the walls of S-21.
In my own work, I pondered the value of my interviews with Angelique after I began to suspect she may have made up her story of being targeted for having 'Tutsi blood'. The point at which I doubted everything was when she claimed her father had had 39 wives. Thirty-nine seemed implausible, and as Dean & Whyte (1970: 126) point out, when a story appears implausible, there is reason to question it.
At that moment, I believed that Angelique had been spinning tales but was her purpose to deceive me? At first, I became angry thinking that she had lied to me. Over time, however, I came to believe that her purpose in telling the story was not to deceive, but to make sense of her current situation. Angeli-que's present was difficult for her to bear. The war had left her a widow with many children to feed; her house was meager even by local standards. She felt marginalized and rejected by the other Tutsi who had denied she was a survivor.
As I reflected further on Angelique's story, I realized that it might have contained elements of truth. It is quite possible that Angelique's mother was Tutsi. It is also possible that her father had had multiple wives, indicating that he been a man of some means and that Angelique had once been associated with a person of power and prestige. Like the fictionalized confessions that Payne analyzes, Angelique may have made up the specific details of what happened to her, but her story may have still accurately depicted how some Tutsi did survive the mass violence of 1991 in her region.
Angelique was not the only person to exaggerate or embellish. Sophie was close to 80 when we first met her in 2004. During the genocide, she managed to rescue many people by hiding them in her small house. Despite her advanced age, Sophie was vivacious and loved to tell stories. Dates and chronologies, however, were usually a blur to her. 'It's been ten years', she explained when she could not remember the specific year in which a certain event had taken place. Sophie did not claim to remember everything, but like most good storytellers, she was committed to whatever story she did tell. Did Sophie
Fujii
235
embellish at times? Undoubtedly. Did these embellishments turn her stories into lies? I would argue not. Sophie's stories were not litanies of facts but narratives with characters and plots whose purpose was to bring to light forgotten histories. It was not dates that were important to her, but who did or said what to whom and why.
Sophie, like Angelique, felt marginalized in her community. Unlike Angelique, however, for Sophie being marginalized meant having greater freedom to say and do what she wanted. It made her more forthcoming, not less. Being more forthcoming did not necessarily make Sophie's stories more reliable in terms of factual accuracy, but it did make them more available to cross-checking with her at subsequent interviews and through other sources. By contrast, when the details of Angelique's story became less precise and more fantastical, it became harder to confirm any part of the story, even with her. Indeed, it was my inability to gain greater clarity from Angelique that finally prompted me to ask different questions of the data. Why would someone make up a story like that? What is the story behind the story? Angelique's testimonies hinted at the value she placed on being recognized as a 'survivor' and a 'victim'. The importance she placed on these categories did not seem to reflect purely material motives - what she could obtain from the government or a foreign researcher by being classified as a 'survivor'. It had more to do with a type of social hierarchy that was in place at the time, a hierarchy that placed survivors near or at the top. This hierarchy had left Angelique at the bottom with no obvious way to advance. Claiming that others had denied her survivor status and, perhaps, a prestigious pedigree through her father were ways to mitigate the disappointment she felt with her current situation. Angelique's invented narrative seemed to have been an example of a 'uchronic' story. It was not a story of what was, but rather, what should have been (Portelli, 1991: Ch. 6). This alternative version restored Angelique's dignity and social importance.
I interpreted Sophie's embellishments, by contrast, as coming from a place of critique. Sophie had strong opinions and enjoyed expressing them. Indeed, Sophie seemed to relish any opportunity to show that she was not like her neighbors and that she was unafraid to criticize others. For Sophie, the goal of relating stories was not to invent a new life history that made it easier to cope with her present reality, as seemed to be the case with Angelique. It was to demonstrate her willingness to speak openly about anything, in violation of cultural norms that proscribed such behavior.
One might imagine other possible explanations for Angelique's inventions and Sophie's embellishments, such as trauma or age. Over the course of multiple interviews, however, I never saw any hint of mental or physical impairment. Instead, I came to see that elements of implausibility or inaccuracy did not render their stories unusable - to the contrary. They told me quite a bit about the social terrain of my two research sites and where each woman located herself in that terrain. Angelique's stories embodied her aspirations for a better life and disappointment with her current situation. Sophie's stories shed
light on her moral code and the value she placed on speaking out rather than hiding her true thoughts and feelings. Their stories also revealed the causal logics the women used to understand how social status, ethnicity, and power worked. For Angelique, the three were intertwined and left her feeling a 'victim' of cruel circumstances beyond her control. For Sophie, the fact that social location and power were so closely tied made it all the more important for her to speak out against the prevailing beliefs and practices that kept those links in place.
As these examples show, narratives of violence, even those filled with inaccuracies, fictions, and lies, can embody all sorts of truths - emotional, psychological, and moral. These truths are as crucial to researchers' analyses of past violence and prospects for future peace as any 'fact-based' reports.
Denials
A third discursive strategy I encountered in the field was denial. This was most pronounced among survivors of violence that occurred after the genocide.
I met one of these 'survivors' unexpectedly. I had asked the local authority if there were any women 'rescape'es who might be willing to talk with me. In using the word 'rescape", I had assumed that my meaning was clear - that I wanted to talk with survivors of the violence targeted at Tutsi civilians, which had occurred in the region in 1991 after an attack by the RPF on a nearby town.6 When the rescapee arrived for her initial interview, I had certain expectations about how it would go. Genocide survivors were usually quite willing to talk about their experiences of violence. So when I began with questions about the period of 1990-94, I was taken aback when she answered by focusing on the period after 1994. When I asked what happened to her during 1990 to 1994, she responded: 'I encountered some problems from the war. Especially the war of 1997 when they struck me with knives and machetes.' A few questions later, I tried to steer her back to the period I was interested in and asked if she had encountered violence before 1997. Her reply again focused on events in 1997. I made one last try and asked whether there had been any violence in her area before 1997. She replied 'no' and once again emphasized that problems started after the arrival of the RPF, which she dated to 1997.
This woman's denials that any violence had occurred before 1997 were not unique. Other women who lost husbands after the war and genocide also maintained that there had been no violence between 1990 and 1994. While I did not find their denials credible, I did not think of them as simply lies or deceptions. Rather, I concluded that what was paramount for these women was their own victimization which had occurred after the genocide. This realization pushed me to inquire into their experiences as victims first, before trying to talk to them
6 The Rwandan Patriotic Front (RPF) was a rebel group that had invaded Rwanda from Uganda in 1990 and took over the country after claiming victory in 1994.
236
journal of Peace Research 47(2)
about other forms of victimhood. Only by allowing these informants to speak about their own experiences of violence could I get them to acknowledge, however grudgingly, that others were targeted for violence during an earlier period. It was as if acknowledging the violence perpetrated against other victims took away from their own status as victims. In a way, they were right. In its gacaca initiative, the RPF-led government had reserved the term 'survivor' (and by extension, the category of 'victim') for those who had been targeted during the genocide. By doing so, the government was effectively denying redress to victims of other forms of violence, most notably violence that the RPF had committed during and after the civil war. Furthermore, by restricting my questions to the period of the 1990-94, I, too, was designating these women's experiences as irrelevant, since the stories they were providing fell outside the temporal boundaries I had established for my research.
What these women's denials taught me was that informants do not experience violence in the same neat, analytic packages that we researchers use in our fieldwork Rather, people experience, remember, and recount violence through the lens of their own victimization. This meant that I could not pre-specify a designated time period and expect that informants would go along with my timelines, especially when my demarcations did not match their experience, or worse, silenced them. This was a critical realization, since I was trying to gather as many different vantage points on the civil war and genocide as possible. If those who suffered violence after the genocide refused to talk about their experiences during the genocide, then my data would have been missing an important set of perspectives.
Like rumors and inventions, denials can be informative, not just obfuscating. People deny some stories to focus on others. This woman knew I would hear about the war and genocide from others. What she insisted I hear from her was her personal story of violence, which occurred after the war and genocide. This was not an insignificant move on this woman's part, since the current regime brooked no criticism of its actions during the civil war. Volunteering this information could get her in trouble with the authorities, yet she volunteered the information anyway. Her denial of one form of violence made her quite vocal about another. Her denials did not shut down dialogue; they merely redirected the interview to another set of events, about which she was more than willing to talk despite the attendant risks.
As this example illustrates, denials, like rumors, are not necessarily barriers, they can also be openings. They can lead to questions and information the analyst was not originally seeking. In my case, the women's denials led me to other stories of victimization which I had not intended to collect. In other cases, denials might lead the researcher away from stories of victimhood altogether to stories of heroism, as Theidon (2007: 465-471) found when she interviewed women who had suffered sexual violence during the civil war in Peru. Contrary to accounts that cast these women as victims only, these
women portrayed themselves as agents, who developed inventive strategies to save themselves and their families from being raped or killed. These strategies included getting pregnant by local men or trading sex for the lives of family members.
Denials can also provide insight into current divisions which people may not want to talk about because of government policies that preclude such talk or nosy neighbors who might take advantage of such information. In my case, for example, informants' denials seemed to indicate that social divisions between genocide survivors and those who suffered violence at the hands of the RPF did indeed exist, contrary to the state's claim that its policies were moving the country toward unity. These divisions, moreover, seemed to be the product, not producer, of the genocidal violence, which was a key finding of my project (Fujii, 2009).
These women's denials also taught me how over-reliance on analytic categories can lead to systematic holes in the data. By denying these women the opportunity to tell stories that did not fit my analytic grid, I risked losing out on the stories I was seeking.
Evasions
In addition to denials, I also encountered strategies of evasion on the part of some informants. Some people avoided answering particular questions; others avoided being interviewed altogether. Avoiding interviews did not always mean the person had something to hide. During the course of my fieldwork, I came across people who seemed bored by the interview process or simply preferred doing something else with their time. Conversely, agreeing to interviews did not imply any promise of openness on the part of the informant. Some people I interviewed, for example, readily agreed to talk but said very little.
The most blatant example of the latter was a man I call Robert. The local authority suggested we talk to Robert because he had been the local authority after the war. Robert was also a genocide survivor. Since we were having trouble finding people that day (because it was market day, most people were out), I welcomed the opportunity to interview him.
Entering his house, I noticed that Robert seemed to be fairly well off by local standards. There was a scale on a table near the front door, indicating that he was a merchant of some kind. His teenage daughter, dressed in a school uniform, greeted us in French. As we sat down, he made a big show of welcoming us into his home. Throughout the interview, he played the part of cordial host, all the while clutching a wad of 100 Rwandan franc bills in his hand.
I began the interview as I had with other genocide survivors - with the expectation that this man would talk openly about his experience during the genocide. As the interview progressed, however, I noticed that rather than getting more detailed (as was usually the case with survivors), his answers became more general. He began the interview saying that he had seen everything. He explained that in 1994, Hutu were being trained to kill Tutsi. He also named the local person who
Fujii
237
was in power during the genocide. As the interview continued, however, he began saying he did not know the answer to my questions.
Q. How did the attackers know who the Tutsi were?
I don't know how they knew that, but they had made lists well in advance and used those at the time of the killing.
Q. They made the lists before the shooting down of the president's plane?
I don't know when they did it.
Q. Did the war that started in 1990 change anything here in Ngali?
We heard that there was a war at the border, that the Inyenzi-Inkotanyi [the RPF] were attacking Rwanda. Friendships between people began to erode and before you know it, there was a conflict between the ethnic groups, saying that the Inyenzi were Tutsi.
Q. After having arrived at [the secteur where you fled], what did you do next?
We stayed at my father-in-law's until the arrival of the Inkotanyi.
Q. When did the Inkotanyi arrive at Ngali? I don't know because I wasn't here at Ngali. Q. [When did they arrive] at [the secteur where you fled] ? I don't remember. I was in the house. I didn't go out at that time.
This entire exchange struck me as odd for a survivor. Many prisoners claimed to have seen everything and then became vague when I asked them for details of what they saw or did. This man was showing the same tendency. With prisoners, even those who had confessed their participation in the genocide, I understood this vagueness to be a way to deflect guilt or minimize responsibility for their deeds. This man was a genocide survivor so his evasions perplexed me. He stated that there were lists circulating, but claimed not to know who had drawn them up or when. Yet, if such lists had existed (and other testimony corroborated this point), those who drew them up must have been locals since only locals would have known which households were Tutsi. Robert's claimed ignorance seemed at odds with his claim that he saw everything.
Robert's response that he did not know when the RPF arrived in his own secteur did not seem believable either since his father-in-law's secteur7 adjoined his own. Later in the same interview, when I asked Robert about his own experience being targeted, he avoided specifying which Hutu were targeting him. Were they outsiders? Neighbors? Militia? Instead, he resorted to the unassailably general statement that 'the Hutu were hunting the Tutsi'.
All these evasions puzzled me. Then, just after leaving the man's house, my interpreter pointed out that Robert must be the brother of a prisoner we had recently interviewed. This prisoner was also a Tutsi genocide survivor and had told us a
In 1994, a secteur was comprised of roughly 3,000 to 5,000 people.
complicated story of how he came to be imprisoned after the genocide. While the details were a blur to me at the time, what I did recall was that it was family problems that had landed this man in prison. My interpreter had figured out the link when she recognized the names Robert had given for his parents (in response to a standard question I asked at initial interviews).
Another clear indication that Robert was hiding information was his claim that he had only one brother, who had died in the genocide. What he neglected to mention was that he had another brother who was languishing in prison. That omission was clearly telling but telling of what? After our initial interview, we asked Robert if we could come back another day. He readily agreed. We made an appointment to talk with him again in two weeks' time. When we returned on the appointed day, he was nowhere to be found. We left a message with his wife that we would return on another specified day. When that day came, he was absent again. We then asked the local authority, who made his own inquiries, but said he could not locate him either. It was obvious that Robert was trying to avoid us, but why would a genocide survivor have reason to hide from a researcher conducting research on the genocide? He had been among the victims of the genocide, after all, not one of its perpetrators.
Because he was avoiding us, I decided to inquire of others about Robert. One of the people I asked was Sophie, whom I knew would talk honestly. Another was a genocide survivor who, with no prompting, told us that Robert had threatened to kill her if she did not falsely accuse a man Robert wanted to threaten. After talking with these women and others, I developed a picture of a man who was not well liked before the genocide. After the genocide, dislike turned to fear when Robert became the new local authority and used his new found power to pursue his private interests through violence. He had people imprisoned (including the brother mentioned above) and had others killed outright. I asked people about his motives. Was he simply after revenge? According to those I talked to, Robert's motive was not revenge, but greed. He had people imprisoned and killed so he could take their property.
Robert's story indicates how closely tied power and violence could be at the local level. As conseiller, Robert wielded almost absolute power over the lives of local residents. This type of power sounded similar to the stories I had heard about the person who led the genocide in Robert's community. Robert's case thus illustrated how one person could exercise complete control in a single community.
More generally, the case of Robert cautions against viewing victims as uniformly innocent and perpetrators as the only actors capable of violence. It speaks to the need for stripping these categories of their normative assumptions. It also suggests that analysts should not treat the stories of victims as necessarily more accurate than other actors' testimonies, since victims, too, may have their own reasons to withhold information.
Silences
Like evasions, silences, too, can be polyvalent. Their meanings can be multiple and contradictory. They can both hide and reveal.
238
journal of Peace Research 47(2)
I had expected people to be silent on one subject - sexual violence. I had read about how widespread mass rape was during the genocide8 but did not expect anyone to talk about it because of the social stigma attached to rape. I also chose not to pose direct questions about this aspect of the genocide as I felt unprepared to broach what I believed to be a sensitive and potentially traumatizing subject. Instead, I waited for informants to bring up the topic themselves. Only three people did so. One was a man who said he had heard that former Rwandan army soldiers had committed sexual violence against Tutsi women and girls in a neighboring secteur. Another was a female genocide survivor who, while fleeing, ran into two local killers, one of whom threatened to rape her, but she managed to escape. The third was a prisoner who had not confessed to participating in the genocide (thereby relinquishing the possibility of a reduced sentence). He related a story of how his cousin and an accomplice had killed multiple members of the prisoner's family and raped a girl who was hiding at his mother's house.
These cases were the exceptions. People were, for the most part, silent on this subject. That silence reflected not only the sensitivity of the topic, but also a general disbelief that sexual violence could occur in one's own community. One (male) genocide survivor insisted, for example, that there had been no rape in his community during the genocide because he had never heard anyone talk about it. When I asked him if a woman who had been raped would feel free to talk about it, he conceded she would not.
People's silence about sexual violence also seemed to reflect a common shame around this form of violation. As Olujic (1995) points out, victims who talk about their rape generally bring more, not less, shame to themselves and their families. For this reason, they often choose to remain silent as a way to protect their families.
In addition to silence on sexual violence, people were also largely silent on another subject - pillaging. This silence was quite unexpected, particularly since many people had already admitted to participating in mass murder. A typical response to questions about pillaging during the genocide was to acknowledge that pillaging had occurred, but not to implicate oneself or specific individuals. Some of the confessed killers I spoke with talked about pillaging as one of several tasks that local authorities ordered them to do. Only one admitted to pillaging on his own.
In addition to pillaging and sexual violence, people were also largely silent on the subject of atrocities committed during the genocide. When people talked of killing, they did so with an economy of words: 'We cut him' or 'they killed him'. I often inquired as to the instruments of death. The answers were consistent but perfunctory: hoes, clubs, axes, and machetes. People rarely volunteered details about specific atrocities.
See, for example, Landesman (2002).
Unlike silence around rape or sexual violence, silence on atrocities is not common across societies. The Mayan widows that Green interviewed, for example, readily described the atrocities committed against them and their families during la violencia. As Green (1999: 75) explains: 'The women, without prompting, took turns recounting their stories of horror. Using vivid detail, they would tell of the events surrounding the deaths or disappearances of their husbands, fathers, sons, brothers,
French journalist Jean Hatzfeld (2000) elicited similarly detailed accounts of the atrocities that a group of Rwandan genocide survivors lived through and witnessed. Why was I unable to elicit similar details? Perhaps I did not ask the right questions or establish the requisite level of rapport with informants. I chose not to ask direct questions about the most intimate details of the violence because such questions felt too invasive. I had no obvious entry point as I did with the subject of ethnicity. Perhaps it was my own norms that kept me from broaching this subject, despite being keenly interested in expressive forms of violence. Or perhaps it was the realization that 'interviewing is also interrogation, and many subjects will not allow it to penetrate beyond a certain level of generality' (Peritore, 1990: 360). In other words, perhaps I knew not to ask.
Silences require careful handling since one explanation does not fit all. In the case of sexual violence, I interpreted the silence to mean this was a topic not to be broached - by me or anyone else. In the case of atrocities, I was uncertain as to the reason for the silence but did not think to discuss it with my interpreter or other colleagues and friends at the time.
On the subject of pillaging, I was at a similar loss but did consult two Rwandan friends. They provided two different explanations. One colleague reasoned that admitting to pillaging made a person liable for paying restitution to the victim or victim's family. The other conjectured that by admitting to pillaging, the person would also be admitting to coveting what another had, a shameful admission in Rwandan society. Both explanations made sense to me, for people were not silent on pillaging in general, but on their own, individual involvement.
The example of pillaging shows clearly how conditions in the present shape testimonies of the past, to the point of denying researchers access to the data they seek. The silence on pillaging, for example, could make it difficult, if not impossible, to investigate the role that pillaging played in motivating people to participate in the genocide. One common theoretical claim, for example, is that people participate in collective action when selective incentives exist (Olson, 1965). In the case of the Rwandan genocide, collective action theory suggests that many would have participated in the opportunity to loot victims' belongings. Silence on this subject, however, would make it difficult to test this hypothesis or to reconstruct actors' agency in this area.
Silences are not always collective, however. Individuals can also be silent on certain subjects. Their silence does not necessarily mean they are less truthful or forthright than those who
Fujii
239
are more talkative. Therese, the woman who likened our visits to that of a priest, said very little about the genocide during our many interviews, but was quite open about all other topics. Robert, by contrast, claimed to have seen everything but told us very little. Like 'fact' and 'fiction', talking and silence may also be a false binary.
Silences can also be a collaborative effort between researcher and participant. As Warren (1998) points out, 'strategic ambiguities' always arise in narratives about war and violence. Often, such ambiguities are not invitations to probe more deeply, but rather subtle admonishments to the researcher to respect certain topics as 'off limits'. As Malkki (1995: 51) remarks about her own fieldwork experience in the 1980s interviewing Burundi refugees living in Tanzania:
the success of the fieldwork hinged not so much on a determination to ferret out 'the facts' as on a willingness to leave some stones unturned, to listen to what my informants deemed important, and to demonstrate my trustworthiness by not prying where I was not wanted.
Like Warren, Malkki, and others who have studied war and violence up close, I, too, never pressed anyone to talk about anything he or she did not want to discuss. When I encountered hesitation or resistance, I switched to entirely different topics to demonstrate my willingness to respect the informant's boundaries. 'Not asking' was one way I could demonstrate my trustworthiness.
Implications
What differences do meta-data make in terms of our knowledge and theories of political violence? The answer depends in part on disciplinary norms. Most writing on conducting fieldwork in conflict or post-conflict zones comes from anthropologists and oral historians. Political scientists have reflected less on the strategies they use to collect data in these settings.9 Recently, however, a growing micro-level research program in political science has emerged that has prompted closer scrutiny of the most suitable methods for investigating questions of social violence. The best of these scholars are extremely conscientious about their methods.10 Indeed, their concern with rigor is a key strength of their work, for they know all too well that the robustness of their findings and theories depend on the quality of the underlying data.
More generally, however, political scientists tend to reflect less on the many decisions they make 'backstage', such as how they choose their interlocutors, how they train enumerators, what languages they use for interviews, and how they adjust for rumors or silences during fieldwork. Yet, it is these backstage decisions that bear directly on the quality of the data,
9 Wood (2003, 2006), a political scientist, is one exception.
10 Exemplars include Wood (2003), Kalyvas (2006), Posner (2005), Straus (2006), and Wilkinson (2004).
no matter the methods employed. Surveys or other forms of one-shot interviews are no less vulnerable to systematic denials or silences than multiple interviews. People are no less apt to embellish or edit their stories to a native-born enumerator than a foreign researcher. For these reasons, meta-data should be a concern to all researchers who venture into the field.
Failing to attend to meta-data can have clear consequences for a study's findings. Rumors about who the researcher is can create barriers before the research has begun, making the very people in whom the researcher is interested less likely to talk honestly, openly, or at all. Systematic silences, evasions, and denials can obscure the identities of perpetrators, lead to under- and overestimations of popular participation in violence, and foreground certain actors, such as thugs, while downplaying the role of other actors, such as neighbors.
Failing to attend to meta-data can also lead to faulty conclusions. We might imagine, for example, what would have happened if Elisabeth Wood (2003) had not attended to the meta-data that emerged during her extended fieldwork in El Salvador. In trying to explain patterns of insurgent collective action, she might have missed the expressive motivations that underscored informants' accounts of why they participated in the insurgency at great risk to themselves. She might have missed, for example, the moral commitment and pride that campesinos expressed when they talked about their role in bringing about social and political change in their country or discounted them as post hoc rationalizations. She might have missed the importance of new found collective beliefs about social justice which invested land occupation with moral, not just material, significance. Given that her data refuted alternative explanations, such as those emphasizing class differences or selective incentives, she would have been left with no explanation for the patterns of mobilization she observed.
Looking at my own experience, I would have drawn very different conclusions as well. I might have read Therese's silence on the genocide as complicity, dismissed Sophie's stories as too riddled with inaccuracies, and accepted Angelique's story of victimization at face value. I might have concluded prematurely that anyone closely situated to a local leader of the genocide (as Therese was) must have also participated in the genocide; or overestimated the extent to which genocidaires went after Hutu with Tutsi mothers. These mis-readings would have left me with unexplained patterns as well. I would have been unable to account for why participants in the genocide tried at times to save Tutsi instead of kill them, why perpetrators sometimes targeted Hutu (with no Tutsi parent), and why family members of killers could become targets themselves.
Those attending to meta-data know that words can hide11 just as silences can reveal. Hatzfeld (2000, 2003), a journalist, may have succeeded in eliciting graphic details from both
11 See, for example, Robben (1995) on 'ethnographic seduction'.
240
journal of Peace Research 47(2)
genocide perpetrators and survivors, but the highly articulate and chillingly detailed language in which his informants speak - in stark contrast to the way the perpetrators with whom Straus (2006), Mironko (2004), and I spoke - does not necessarily mean his informants are telling more 'truth'.
Conversely, silences and denials on the part of perpetrators can open up new forms of dialogue and public debates, which can lead to disclosures of new truths and knowledge (Payne, 2008). Silences can also enable speakers to transform old identities and construct new ones by adding and subtracting key elements of stories about collective pasts (Smith, 2007). As these examples show, meta-data are not ancillary or extraneous parts of the dataset, they are data.
There are multiple ways to access and identify meta-data. One strategy is to interview people multiple times over an extended period. Multiple interviews enable the researcher to respond to the fears and suspicions that are likely to exist in post-conflict settings. Knowing that informants can face reprisals from many quarters - neighbors, family, state agents -should prompt researchers to find ways to ensure people's safety, not only during the period of research but just as importantly, long after the researcher has left the field.
A second strategy that researchers might use to probe for meta-data is to invite informants to pose questions of their own. The questions informants raise may speak directly to their reasons for and concerns about agreeing to interviews, their aspirations for establishing a relationship with an outsider, and their assumptions about what they might get in return for their time. By allowing informants to ask questions (and not just answer them), the researcher might learn of people's concerns directly without having to guess or assume. The standard one-way interview will not always bring these issues out in the open.
A third practice that scholars might incorporate is sustained self-reflection both during and after fieldwork. This process might involve regular dialogue with research assistants and colleagues in the country about rumors and gossip that commonly arise about outsiders and those that arise specifically about the researcher. My interpreter did not always relay this information to me so I had to ask. Researchers might also take note of puzzles that arise from everyday conversations or interactions and revisit those puzzles to see if time in the field has helped to resolve them. Researchers should then reflect on the source of that clarity - why previous puzzles suddenly make sense.
Researchers usually acquire a certain amount of local knowledge while in the field; it is the immersion into local cultures and perspectives that slowly shifts and transforms the researcher's own sense of what is normal and credible. Sustained fieldwork creates the opportunity to build trust and rapport through active and attentive listening12 - with informants certainly, but also acquaintances, friends, colleagues,
12 I thank one of the anonymous reviewers for this point.
and research assistants. This trust becomes the main avenue for identifying, interpreting, and responding to meta-data. Some scholars deny that trust and rapport come with time. As Belou-sov et al. (2007: 156) argue, in 'crisis-ridden research settings', rapport between researcher and researched may actually diminish over time. As Lofving (2005: 89), too, observes: 'Lying, misinformation and direct silence adhere to the communicative tool kit of people in politically unstable circumstances.' While strategies of dissimulation may indeed constitute modes of survival in conflict and post-conflict settings, meta-data can help researchers make sense of the ambiguities and complexities such strategies generate. By giving meta-data systematic attention, researchers can learn to read the different shades of'truth' and 'lies' they encounter in the field and find answers to questions we never knew we had.
Acknowledgments
The author wishes to thank Elisabeth King, Michelle Moyd, Robin Turner, Tamara Walker, Elisabeth Wood, Dvora Yanow, and three anonymous reviewers for their extremely helpful comments on this article.
References
African Rights (1995) Death, Despair, and Defiance. London: African Rights.
Belousov, Konstantin; Tom Horlick-Jones, Michael Bloor, Yakov
Gilinkskiy, Valentin Golbert, Yakov Kostikovsky, Michael Levi
& Dmitri Pentsov (2007) Any port in the storm: Fieldwork
difficulties in dangerous and crisis-ridden settings. Qualitative
Research 7(2): 155-175. Chandler, David (1999) Voices from S-21: Terror and History in Pol
Pot's Secret Prison. Berkeley, CA: University of California Press. Dean, John P & William Foote Whyte (1970) How do you know if
the informant is telling the truth? In: Lewis Anthony Dexter (ed.)
Elite and Specialized Interviewing. Evanston, IL: Northwestern
University Press, 119-31. Des Forges, Alison (1999) Leave None To Tell the Story. New York:
Human Rights Watch & Federation Internationale des ligues des
droits de l'homme. Fujii, Lee Ann (2009) Killing Neighbors: Webs of Violence in Rwanda.
Ithaca, NY: Cornell University Press. Green, Linda (1999) Fear as a Way of Life. New York: Columbia
University Press.
Hatzfeld, Jean (2000) Dans le Nu de la Vie: Recits desMarais Rwandais [Life Laid Bare: The Survivors in Rwanda Speak]. Paris: Seuil.
Hatzfeld, Jean (2003) Une Saison de Machettes [Machete Season]. Paris: Seuil.
Kalyvas, Stathis N (2006) The Logic of Violence in Civil War.
Cambridge: Cambridge University Press. Landesman, Peter (2002) A Woman's Work. New York Times
Magazine, 15 September, 82-134. Lofving, Staffan (2005) Silence and the politics of representing
rebellion: On the emergence of the neutral Maya in Guatemala.
In: Paul Richards (ed.) No Peace No War: An Anthropology of
Fujii
241
Contemporary Armed Conflicts. Athens, OH: Ohio University Press, 77-97.
Malkki, Liisa (1995) Purity and Exile. Chicago, IL: University of Chicago Press.
Mironko, Charles K (2004) Social and political mechanisms of mass murder: An analysis of perpetrators in the Rwandan genocide. PhD dissertation, Anthropology, Yale University, New Haven, CT.
Olson, Mancur (1965) The Logic of Collective Action: Public Goods and the Theory of Groups. Cambridge, MA: Harvard University Press.
Olujic, Maria B (1995) The Croatian war experience. In: Carolyn Nordstrom & Antonius C G M Robben (eds) Fieldwork Under Fire. Berkeley, CA: University of California Press, 186-204.
Payne, Leigh (2008) Unsettling Accounts. Durham, NC: Duke University Press.
Peritore, N Patrick (1990) Reflections on dangerous fieldwork. American Sociologist Winter. 359—372.
Porteiii, Alessandro (1991) The Death of Luigi Trastulli and Other Stories. Albany, NY: State University of New York Press.
Posner, Daniel N (2005) Institutions and Ethnic Politics in Africa. New York: Cambridge University Press.
Robben, Antonius C G M (1995) The politics of truth and emotion among victims and perpetrators of violence. In: Carolyn Nordstrom & Antonius C G M Robben (eds) Fieldwork Under Fire. Berkeley, CA: University of California Press, 81-103.
Sluka, Jeffrey A (1990) Participant observation in violent social contexts. Human Organization 49(2): 114—126.
Smith, Tammy (2007) The structural dynamics of narrative silence.
Paper presented at Columbia University's ISERP Contentious
Politics Workshop, New York, 1 October. Straus, Scott (2006) The Order of Genocide: Race, Power and War in
Rwanda. Ithaca, NY: Cornell University Press. Tertsakian, Carina (2008) Le Chateau: The Lives of Prisoners in
Rwanda. London: Arves. Theidon, Kimberly (2007) Gender in transition: Common sense,
women, and war. Journal of Human Rights 6(4): 453—478. Waldorf, Lars (2006) Mass justice for mass atrocity: Rethinking local
justice as transitional justice. Temple Law Review 79(1): 1—87. Warren, Kay B (1998) Indigenous Movements and Their Critics:
Pan-Maya Activism in Guatemala. Princeton, NJ: Princeton
University Press.
Weinstein, Jeremy M (2007) Lnside Rebellion: The Politics of Insurgent
Violence. Cambridge: Cambridge University Press. Wilkinson,  Steven  I  (2004)   Votes and Violence. Cambridge:
Cambridge University Press. Wolf, Charlotte (1992) Constructions of a lynching. Sociological
Inquiry 62(1): 83-97. Wood, Elisabeth Jean (2003) Insurgent Collective Action and Civil
War in El Salvador. Cambridge: Cambridge University Press. Wood, Elisabeth (2006) The ethical challenges of field research in
conflict zones. Qualitative Sociology 29(3): 373—386.
LEE ANN FUJII, b. 1962, PhD in Political Science (George Washington University, 2006); Assistant Professor, George Washington University (2007). Most recent book: Killing Neighbors: Webs of Violence in Rwanda (Cornell University Press, 2009).