| "A group of Hutu men, upon hearing the news of the president's assassination, armed themselves with machetes and rounded up the Tutsis in our area." This beginning, told to me many times during my eighteen months of fieldwork in Burundi, had different endings. In some places, these men went on to attack and kill their Tutsi neighbors. In others, respected Hutu members of the community talked the men down. In still others, a group of potential attackers never formed.;We are faced with an empirical puzzle: how can we explain these varied local reactions to the same national tragedy? And why, during the sixteen years of civil war that followed, did different areas in this tiny country have such different experiences, ranging from wide-scale violence to relative peace?;When armies, militias, rebels, or guerrillas fight for political control, civilians often bear the brunt of the costs, both human and material. But civilians are not just acted upon; they, too, are actors, whose beliefs and decisions influence outcomes in and beyond their communities. Drawing on oral history and archival data collected during eighteen months of fieldwork, this dissertation examines the causes of local variation in civilian-on-civilian violence.;In the first paper, I demonstrate that different local experiences of past violence led to the formation of different community-level distributions of beliefs about relative strength and the likely behavior of the opposing group. These different beliefs, in turn, led to different decisions about whether to participate in civilian-on-civilian violence. I provide empirical support for these arguments by tracing the formation of these different beliefs in eight different communes in Burundi, over the period 1972-1992. Using interview and archival evidence, I demonstrate that my account best explains the observed empirical patterns, whereas other plausibly relevant preferences, beliefs, and structures do not explain the spatial and temporal variation in violence at the local level.;In the second paper, I use oral histories and contemporaneous archival documents to re-create the aftermath of the assassination of Hutu president Melchior Ndadaye in 1993. I note that even in areas where the beliefs of the majority of Hutus led them to choose peace, some extremists with high risk tolerance decided to engage in violence. However, violence was not inevitably the outcome. Local leaders could intervene at this point to change decisions. I show that when peace was threatened, local leaders were successful at maintaining peace via six main methods: persuasion, obedience, coercion, deception, prejudice reduction, and collaboration with the Tutsi military. Local peace, therefore, was highly contingent and threatened on multiple occasions.;In the third paper, I turn my focus to methods of studying civil war violence. Scholars often see the trade-off between qualitative and quantitative analysis as depth of understanding versus breadth and representativeness. But when analyses rely on information provided by human subjects, scholars collecting large quantities of data often neglect to examine a fundamental assumption: that most people will answer questions either truthfully or in a predictably biased way. This assumption ignores important lessons from anthropology and psychology about the complex ways in which people relate their stories and the ways that this story-telling varies from culture to culture. If this key assumption does not hold -- if respondents provide information that systematically deviates from the truth in unpredictable directions -- our quantitative information will end up representative, statistically significant, and wrong. This paper proceeds in three sections. First, using evidence from my fieldwork in Burundi, I provide examples of how respondents answer questions about sensitive topics in unexpected ways. Second, I suggest a three-part classification of reasons why respondents may modify the responses they provide: intentional modifications to protect self or others, intentional modifications to promote self or others, and unintentional modifications. Finally, I argue that there are strong reasons to believe that even the best survey practices may be missing systematic lies that ethnographic methods can often detect. To obtain accurate information about sensitive issues, ethnography does better, because it is uniquely equipped to deal with issues of culture, trust, and contradiction. |
