In 1998 and 1999, we went to Rwanda and returned several times in subsequent years for a simple reason: We wanted to discover what had happened there during the 100 days in 1994 when civil war and genocide killed an estimated 1 million individuals. What was the source of our curiosity? Well, our motivations were complex. In part, we felt guilty about ignoring the events when they took place and were largely overshadowed in the U.S. by such “news” as the O.J. Simpson murder case. We felt that at least we could do something to clarify what had occurred in an effort to respect the dead and assist in preventing this kind of mass atrocity in the future. We were both also in need of something new, professionally speaking. Although tenured, our research agendas felt staid. Rwanda was a way out of the rut and into something significant.

Although well-intentioned, we were not at all ready for what we would encounter. Retrospectively, it was naïve of us to think that we would be. As we end the project 10 years later, our views are completely at odds with what we believed at the outset, as well as what passes for conventional wisdom about what took place.

We worked for both the prosecution and the defense at the International Criminal Tribunal for Rwanda, trying to perform the same task — that is, to find data that demonstrate what actually happened during the 100 days of killing. Because of our findings, we have been threatened by members of the Rwandan government and individuals around the world. And we have been labeled “genocide deniers” in both the popular press as well as the Tutsi expatriate community because we refused to say that the only form of political violence that took place in 1994 was genocide. It was not, and understanding what happened is crucial if the international community is to respond properly the next time it becomes aware of such a horrific spasm of mass violence.

Like most people with an unsophisticated understanding of Rwandan history and politics, we began our research believing that what we were dealing with was one of the most straightforward cases of political violence in recent times, and it came in two forms: On the one hand was the much-highlighted genocide, in which the dominant, ruling ethnic group — the Hutu — targeted the minority ethnic group known as the Tutsi. The behavior toward the minority group was extremely violent — taking place all over Rwanda — and the objective of the government’s effort appeared to be the eradication of the Tutsi, so the genocide label was easy to apply. On the other hand, there was the much-neglected international or civil war, which had rebels (the Rwandan Patriotic Front or RPF) invading from Uganda on one side and the Rwandan government (the Armed Forces of Rwanda or FAR) on the other. They fought this war for four years, until the RPF took control of the country.

We also went in believing that the Western community — especially the United States — had dropped the ball in failing to intervene, in large part because the West had failed to classify expeditiously the relevant events as genocide.

Finally, we went in believing that the Rwandan Patriotic Front, then rebels but now the ruling party in Rwanda, had stopped the genocide by ending the civil war and taking control of the country.

At the time, the points identified above stood as the conventional wisdom about the 100 days of slaughter. But the conventional wisdom was only partly correct.

The violence did seem to begin with Hutu extremists, including militia groups such as the Interahamwe, who focused their efforts against the Tutsi. But as our data came to reveal, from there violence spread quickly, with Hutu and Tutsi playing the roles of both attackers and victims, and many people of both ethnic backgrounds systematically using the mass killing to settle political, economic and personal scores.

Against conventional wisdom, we came to believe that the victims of this violence were fairly evenly distributed between Tutsi and Hutu; among other things, it appears that there simply weren’t enough Tutsi in Rwanda at the time to account for all the reported deaths.

We also came to understand just how uncomfortable it can be to question conventional wisdom.

We began our research while working on a U.S. Agency for International Development project that had proposed to deliver some methodological training to Rwandan students completing their graduate theses in the social sciences. While engaged in this effort, we came across a wide variety of nongovernmental organizations that had compiled information about the 100 days. Many of these organizations had records that were detailed, identifying precisely who died where and under what circumstances; the records included information about who had been attacked by whom. The harder we pushed the question of what had happened and who was responsible, the more access we gained to information and data.

There were a number of reasons that we were given wide-ranging access to groups that had data on the 100 days of killing. First, for their part of the USAID program, our hosts at the National University of Rwanda in Butare arranged many public talks, one of which took place at the U.S. embassy in Kigali. Presumably put together to assist Rwandan NGOs with “state-of-the-art” measurement of human rights violations, these talks — the embassy talk, in particular — turned the situation on its head. The Rwandans at the embassy ended up doing the teaching, bringing up any number of events and publications that dealt with the violence. We met with representatives of several of the institutions involved, whose members discussed with us in greater detail the data they had compiled.

Second, the U.S. ambassador at the time, George McDade Staples, helped us gain access to Rwanda government elites —directly and indirectly through staff members.

Third, the Rwandan assigned to assist the USAID project was extremely helpful in identifying potential sources of information. That she was closely related to a member of the former Tutsi royal family was a welcome plus.

Once we returned to the U.S., we began to code events during the 100 days by times, places, perpetrators, victims, weapon type and actions. Essentially, we compiled a listing of who did what to whom, and when and where they did it — what Charles Tilly, the late political sociologist, called an “event catalog.” This catalog would allow us to identify patterns and conduct more rigorous statistical investigations.

Looking at the material across space and time, it became apparent that not all of Rwanda was engulfed in violence at the same time. Rather, the violence spread from one locale to another, and there seemed to be a definite sequence to the spread. But we didn’t understand the sequence.

At National University of Rwanda, we spent a week preparing students to conduct a household survey of the province. As we taught the students how to design a survey instrument, a common question came up repeatedly: “What actually happened in Butare during the summer of 1994?” No one seemed to know; we found this lack of awareness puzzling and guided the students in building a set of questions for their survey, which eventually revealed several interesting pieces of information.

First, and perhaps most important, was confirmation that the vast majority of the population in the Butare province had been on the move between 1993 and 1995, particularly during early 1994. Almost no one stayed put. We also found that the RPF rebels had blocked the border leading south out of the province to Burundi. The numbers of households that provided information consistent with these facts raised significant questions in our minds regarding the culpability of the RPF relative to the FAR for killing in the area.

During this period, we confirmed Human Rights Watch findings that many killings were organized by the Hutu-led FAR, but we also found that many of the killings were spontaneous, the type of violence that we would expect with a complete breakdown of civil order. Our work further revealed that, some nine years later, a great deal of hostility remained. There was little communication between the two ethnic groups. The Tutsi, now under RPF leadership and President Paul Kagame, dominated all aspects of the political, economic and social systems.

Lastly, it became apparent to us that members of the Tutsi diaspora who returned to Rwanda after the conflict were woefully out of touch with the country that they had returned to. Indeed, one Tutsi woman with whom we spent a day in the hills around Butare broke down in tears in our car as we drove back to the university. When asked why, she replied, “I have never seen such poverty and destitution.” We were quite surprised at the degree of disconnect between the elite students drawn from the wealthy strata of the Tutsi diaspora, who were largely English-speaking, and the poorer Rwandans, who spoke Kinyarwanda and perhaps a bit of French. It was not surprising that the poor and the wealthy in the country did not mix; what struck both of us as surprising was the utter lack of empathy and knowledge about each other’s condition. After all, the Tutsi outside the country claimed to have invaded Rwanda from Uganda on behalf of the Tutsi inside — a group that the former seemed to have little awareness of or interest in. Our work has led us to conclude that the invading force had a primary goal of conquest and little regard for the lives of resident Tutsis.

As the students proceeded with the survey, asking questions that were politically awkward for the RPF-led government, we found our position in the country increasingly untenable. One member of our team was detained and held for the better part of a day while being interrogated by a district police chief. The putative reason was a lack of permissions from the local authorities; permissions were required for everything in Rwanda, and we generally had few problems obtaining them in the beginning. The real reason for the interrogation, however, seemed to be that we were asking uncomfortable questions about who the killers were.

A couple of weeks later, two members of our team were on a tourist trip in the northern part of the country when they were again detained and questioned for the better part of a day at an RPF military facility. There the questioners wanted to know why we were asking difficult questions, what we were doing in the country, whether we were working for the American CIA, if we were guests of the Europeans and, in general, why we were trying to cause trouble.

On one of our trips to Rwanda, Alison Des Forges, the pre-eminent scholar of Rwandan politics who has since died in an airplane crash, suggested that we go to the International Criminal Tribunal for Rwanda in Tanzania to seek answers to the questions we were raising. Des Forges even called on our behalf.

With appointments set and with Mount Kilimanjaro in the distance, we arrived in Arusha, Tanzania, for our meeting with Donald Webster, the lead prosecutor for the political trials, Barbara Mulvaney, the lead prosecutor for the military trial, and others from their respective teams. As we began to talk, we initially found that the prosecutors in the two sets of cases — one set of defendants were former members of the FAR military, the other set of trials focused on the members of the Hutu political machine — had great interest in our project.

Eventually, Webster and Mulvaney asked us to help them contextualize the cases that they were investigating. Needless to say, we were thrilled with the possibility. Now, we were working directly with those trying to bring about justice.

The prosecutors showed us a preliminary database that they had compiled from thousands of eyewitness statements associated with the 1994 violence. They did not have the resources to code all of the statements for computer analysis; they wanted us to do the coding and compare the statements against the data we had already compiled. We returned to the U.S. with real enthusiasm; we had access to data that no one else had seen and direct interaction with one of the most important legal bodies of the era.

Interest by and cooperation with the ICTR did not last as long as we thought it would, in no small part because it quickly became clear that our research was going to uncover killings committed not just by the Hutu-led former government, or FAR, but by the Tutsi-led rebel force, the RPF, as well. Until then, we had been trying to identify all deaths that had taken place; beyond confidentiality issues, it did not occur to us that the identity of perpetrators would be problematic (in part because we thought that all or almost all of them would be associated with the Hutu government). But then we tried to obtain detailed maps that contained information on the location of FAR military bases at the beginning of the civil war. We had seen copies of these maps pinned to the wall in Mulvaney’s office. In fact, during our interview with Mulvaney, the prosecutor explained how her office had used these maps. We took detailed notes, even going so far as to write down map grid coordinates and important map grid sheet identifiers.

After the prosecution indicated it was no longer interested in reconstructing a broad conception of what had taken place —prosecutors said they’d changed their legal strategy to focus exclusively on information directly related to people charged with crimes — we asked the court for a copy of the maps. To our great dismay, the prosecution claimed that the maps did not exist. Unfortunately for the prosecutors, we had our notes. After two years of negotiations, a sympathetic Canadian colonel in a Canadian mapping agency produced the maps we requested.

As part of the process of trying to work out the culpability of the various defendants charged with planning to carry out genocidal policies, the ICTR conducted interviews with witnesses to the violence over some five years, beginning in 1996. Ultimately, the court deposed some 12,000 different people. The witness statements represent a highly biased sample; the Kagame administration prevented ICTR investigators from interviewing many who might provide information implicating members of the RPF or who were otherwise deemed by the government to be either unimportant or a threat to the regime.

All the same, the witness statements were important to our project; they could help corroborate information found in CIA documents, other witness statements, academic studies of the violence and other authoritative sources.

As with the maps, however, when we asked for the statements, we were told they did not exist. Eventually, defense attorneys —who were surprised by the statements’ existence, there being no formal discovery process in the ICTR — requested them. After a year or so, we obtained the witness statements, in the form of computer image files that we converted into optically readable computer documents. We then wrote software to search through these 12,000 statements in our attempts to locate violence and killing throughout Rwanda.

The first significant negative publicity associated with our project occurred in November 2003 at an academic conference in Kigali. The National University of Rwanda had invited a select group of academics, including our team, to present the results of research into the 1994 murders. We had been led to believe that the conference would be a private affair, with an audience composed of academics and a small number of policymakers.

As it turned out, the conference was anything but small or private. It was held at a municipal facility in downtown Kigali, and our remarks would be simultaneously translated from English into French and the Rwandan language, Kinyarwanda. There were hundreds of people present, including not just academics but members of the military, the cabinet and other members of the business and political elite.

We presented two main findings, the first derived from spatial and temporal maps of data obtained from the different sources already mentioned. The maps showed that, while killing took place in different parts of the country, it did so at different rates and magnitudes — begging for an explanation we did not yet have. The second finding came out of a comparison of official census data from 1991 to the violence data we had collected. According to the census, there were approximately 600,000 Tutsi in the country in 1991; according to the survival organization Ibuka, about 300,000 survived the 1994 slaughter. This suggested that out of the 800,000 to 1 million believed to have been killed then, more than half were Hutu. The finding was significant; it suggested that the majority of the victims of 1994 were of the same ethnicity as the government in power. It also suggested that genocide — that is, a government’s attempts to exterminate an ethnic group — was hardly the only motive for some, and perhaps most, of the killing that occurred in the 100 days of 1994.

Halfway into our presentation, a military man in a green uniform stood up and interrupted. The Minister of Internal Affairs, he announced, took great exception to our findings. We were told that our passport numbers had been documented, that we were expected to leave the country the next day and that we would not be welcomed back into Rwanda — ever. Abruptly, our presentation was over, as was, it seemed, our fieldwork in Rwanda.

The results of our initial paper and media interviews became widely known throughout the community of those who study genocides in general and the Rwandan genocide in particular. The main offshoot was that we became labeled, paradoxically, as genocide “deniers,” even though our research documents that genocide had occurred. Both of us have received significant quantities of hate mail and hostile e-mail. In the Tutsi community and diaspora, our work is anathema. Over the past several years, as we have refined our results, becoming more confident about our findings, our critics’ voices have become louder and increasingly strident.

Of course, we have never denied that a genocide took place; we just noted that genocide was only one among several forms of violence that occured at the time. In the context of post-genocide Rwandan politics, however, the divergence from common wisdom was considered political heresy.

Following the debacle at the Kigali conference, the ICTR prosecution teams of Webster and Mulvaney let us know in no uncertain terms that they had no further use of our services. The reasons for our dismissal struck us as somewhat outrageous. From the outset, the prosecution claimed it was not interested in anything that would prove or disprove the culpability of any individuals in the mass killings. Now, they said, the findings we’d announced in the Kigali conference made our future efforts superfluous.

Shortly after our dismissal, however, Peter Erlinder, a defense attorney for former members of the FAR military who were to be tried, contacted us. This was after several others from the defense had also attempted to contact us, with no success.

We had misgivings about cooperating or working with the defense, the gravest being that such work might be seen as supporting the claim we were genocide deniers. After months of negotiating, we finally met Erlinder at a Starbucks in Philadelphia, Pa. The defense could have made a better choice for roping us in. Erlinder, a professor at the William Mitchell College of Law, was an academic turned defender for the least likable suspects.

After we obtained lattes and quiet seats in the back of the coffee shop, Erlinder came straight to the point: He was, of course, interested in establishing his client’s innocence, but he felt it would help the defense to establish a baseline history of what had taken place in the war in 1994. As he explained, “My client may be guilty of some things, but he is not guilty of all the things that any in the Rwandan government and military during 1994 is accused of. They have all been made out to be devils.”

What he asked was reasonable. In fact, he made the same essential offer the prosecution had: In exchange for our efforts at contextualizing the events of 1994, Erlinder would do the best he could to assist us in getting data on what took place. With Erlinder’s assistance, we were able to obtain the maps we’d seen in Mulvaney’s office and the 12,000 witness statements. With this information, we were able to better establish the true positions of both the FAR and RPF during the civil war. This greater confidence of the location of the two sides’ militaries made — and makes — us more certain about the culpability of the FAR for the majority of the killings during the 100 days of 1994. At the same time, however, we also began to develop a stronger understanding of the not insignificant role played by the RPF in the mass murders.

About this time, we were approached by an individual associated with Arcview-GIS, a spatial mapping software firm that wanted to take the rather simplistic maps that we had developed and improve them, thereby showing what the company’s program was capable of. Our consultant at Arcview-GIS said the software could layer information on the map, providing, among other things, a line that showed, day by day, where the battlefront of the civil war was located, relative to the killings we had already documented.

This was a major step. In line with the conventional wisdom, we had assumed that the government was responsible for most all of the people killed in Rwanda during 1994; we initially paid no attention to where RPF forces were located. But it soon became clear that the killings occurred not just in territory controlled by the government’s FAR but also in RPF-captured territory, as well as along the front between the two forces. It seemed possible to us that the three zones of engagement (the FAR-controlled area, the RPF-controlled area and the battlefront between the two) somehow influenced one another.

In his book, The Limits of Humanitarian Intervention, Alan Kuperman argued that given the logistical challenges of mounting a military operation in deep central Africa, there was little the U.S. or Europe could have done to limit the 1994 killings. To support his position, Kuperman used U.S. Defense Intelligence Agency information to document approximate positions of the RPF units over the course of the war. We updated this information on troop locations with data from CIA national intelligence estimates that others had obtained through the Freedom of Information Act and then updated it again, incorporating interviews with former RPF members, whose recollections we corroborated with information from the FAR.

Our research showed the vast majority of the 1994 killing had been conducted by the FAR, the Interahamwe and their associates. Another significant proportion of the killing was committed not by government forces but by citizens engaged in opportunistic killing as part of the breakdown of civil order associated with the civil war. But the RPF was clearly responsible for another significant portion of the killings.

In some instances, the RPF killings were, very likely, spontaneous retribution. In other cases, though, the RPF has been directly implicated in large-scale killings associated with refugee camps, as well as individual households. Large numbers of individuals died at roadblocks and in municipal centers, households, swamps and fields, many of them trying to make their way to borders.

Perhaps the most shocking result of our combination of information on troop locations involved the invasion itself: The killings in the zone controlled by the FAR seemed to escalate as the RPF moved into the country and acquired more territory. When the RPF advanced, large-scale killings escalated. When the RPF stopped, large-scale killings largely decreased. The data revealed in our maps was consistent with FAR claims that it would have stopped much of the killing if the RPF had simply called a halt to its invasion. This conclusion runs counter to the Kagame administration’s claims that the RPF continued its invasion to bring a halt to the killings.

In terms of ethnicity, the short answer to the question, “Who died?” is, “We’ll probably never know.” By and large, the Hutu and the Tutsi are physically indistinct from one another. They share a common language. They have no identifiable accent. They have had significant levels of intermarriage through their histories, and they have lived in similar locations for the past several hundred years. In the 1920s and 1930s, the Belgians, in their role as occupying power, put together a national program to try to identify individuals’ ethnic identity through phrenology, an abortive attempt to create an ethnicity scale based on measurable physical features such as height, nose width and weight, with the hope that colonial administrators would not have to rely on identity cards.

One result of the Belgian efforts was to show — convincingly — that there is no observable difference on average between the typical Hutu Rwandan and the typical Tutsi Rwandan. Some clans — such as those of the current president, Paul Kagame, or the earlier Hutu president, Juvenal Habyarimana — do share distinctive physical traits. But the typical Rwandan shares a mix of such archetypal traits, making ethnic identity outside of local knowledge about an individual household’s identity difficult if not impossible to ascertain — especially in mass graves containing no identifying information. (For example, Physicians for Human Rights exhumed a mass grave in western Rwanda and found the remains of more than 450 people, but only six identity cards.)
In court transcripts for multiple trials at the ICTR, witnesses described surviving the killings that took place around them by simply hiding among members of the opposite ethnic group. It is clear that in 1994, killers would have had a difficult time ascertaining the ethnic identity of their putative victims, unless they were targeting neighbors.

Complicating matters is the displacement that accompanied the RPF invasion. During 1994, some 2 million Rwandan citizens became external refugees, 1 million to 2 million became internal refugees, and about 1 million eventually became victims of civil war and genocide.

Ethnic identity in Rwanda is local knowledge, in much the same way that caste is local knowledge in India. With the majority of the population on the move, local knowledge and ethnic identity disappeared. This is not to say that the indigenous Tutsi were not sought out deliberately for extermination. But in their killing rampages, FAR, the Interahamwe and private citizens engaged in killing victims of both ethnic groups. And people from both ethnic groups were on the move, trying to stay out in front of the fighting as the RPF advanced.

In the end, our best estimate of who died during the 1994 massacre was, really, an educated guess based on an estimate of the number of Tutsi in the country at the outset of the war and the number who survived the war. Using a simple method —subtracting the survivors from the number of Tutsi residents at the outset of the violence — we arrived at an estimated total of somewhere between 300,000 and 500,000 Tutsi victims. If we believe the estimate of close to 1 million total civilian deaths in the war and genocide, we are then left with between 500,000 and 700,000 Hutu deaths, and a best guess that the majority of victims were in fact Hutu, not Tutsi.

This conclusion — which has drawn criticism from the Kagame regime and its supporters — is buttressed by the maps that we painstakingly constructed from the best available data and that show significant numbers of people killed in areas under control of the Tutsi-led RPF.

One fact is now becoming increasingly well understood: During the genocide and civil war that took place in Rwanda in 1994, multiple processes of violence took place simultaneously. Clearly there was a genocidal campaign, directed to some degree by the Hutu government, resulting directly in the deaths of some 100,000 or more Tutsi. At the same time, a civil war raged — a war that began in 1990, if the focus is on only the most recent and intense violence, but had roots that extend all the way back to the 1950s. Clearly, there was also random, wanton violence associated with the breakdown of order during the civil war. There’s also no question that large-scale retribution killings took place throughout the country — retribution killings by Hutu of Tutsi, and vice versa.

From the beginning, the ICTR’s investigation into the mass killings and crimes against humanity in Rwanda in 1994 has focused myopically on the culpability of Hutu leaders and other presumed participants. The Kagame administration has worked assiduously to prevent any investigation into RPF culpability for either mass killings or the random violence associated with the civil war. By raising the possibility that in addition to Hutu/FAR wrongdoing, the RPF was involved, either directly or indirectly, in many deaths, we became in effect persona non grata in Rwanda and at the ICTR.

The most commonly invoked metaphor for the 1994 Rwandan violence is the Holocaust. Elsewhere, we have suggested that perhaps the English civil war, the Greek civil war, the Chinese civil war or the Russian civil war might be more apt comparisons because they all involved some combination of ethnic-based violence and the random slaughter and retribution that can occur when civil society breaks down altogether.

Actually, though, it is difficult to make authoritative comparisons when it remains unclear exactly what happened in the Rwandan civil war and genocide.

Contemporary observers — including Romeo Dallaire, the commander of the ineffective U.N. peacekeeping force for Rwanda in 1993 and 1994 — claim that much of the genocidal killing had been planned by the Hutu government as early as two years in advance of the actual RPF invasion. Unfortunately, we have not been able to gain access to the individuals who have information on that score to either corroborate or to refute the hypothesis. The reason? Convicted genocidaires who have been implicated in the planning of the slaughter now reside out of contact with potential interviewers in a U.N.-sponsored prison in Mali.

We wanted to put questions to these planners, specifically to ask them what their goals were. Was the genocide plan an attempt at deterrence, an effort that the FAR leadership thought might keep the RPF at bay in Uganda and elsewhere? Did the FAR government actually hope for war, believing — incorrectly as it turned out — that it would win? Was the scale of the killing beyond its expectations? If so, why do FAR leaders believe events spun so badly out of control, compared to previous spasms of violence in the 1960s, ’70s and ’80s?

Unfortunately, the U.N. prosecutors in Tanzania told us they could not arrange a meeting with the convicted planners and killers, but we were free to go to Mali on our own. We were told we would probably get in to see the prisoners, but the prison is in the middle of nowhere, in a country where we had no contacts. We had to let go.

Even without access to convicted genocidaires, we continued to piece together what had happened in 1994 with the help of a grant from the National Science Foundation. The grant allowed us to be more ambitious in our pursuit of diverse informants who started popping up all over the globe, to refine our mapping and to explore alternative ways of generating estimates about what had taken place. While our understanding has advanced a great deal since our first days in Kigali, it is hard not to see irony in a current reality: Some of the most important information about what occurred in Rwanda in 1994 has been sent — by the very authorities responsible for investigating the violence and preventing its recurrence, in Rwanda and elsewhere — to an isolated prison, where it sits unexamined, like some artifact in the final scene of an Indiana Jones movie.