Donald L. Horowitz
Duke University
I will pick up on some of Marilynn Brewer’s themes and talk about the transitions from in-group loyalty to violence and back again and about transitions from attitudes to behavior and from behavior to attitudes. I will first talk about ethnic conflict in general and then about a particular species of violence which I call the deadly ethnic riot.
If you look at ethnic conflict around the world, you will find a phenomenon that has common features in the most disparate places. You will find a quite common set of distinctions between putative natives, people who think they belong in the state and have belonged for a long time, and so-called immigrants, people who may have come as recently as fifty years ago or more distantly but who are still categorized as immigrants. What we find is regardless of cultural setting; you can find these distinctions in various states of India, Malaysia, and former Soviet Georgia, where a civil war was fought, in part, over this kind of distinction.
You can find part of a state claiming to be the whole. This is a very common phenomenon in ethnic conflict. You can see this in Romania, in Sri Lanka (the notion is that the Tamils really belong somewhere else), and the Sudan, where the notion is that “if the outsiders don’t belong somewhere else, they at least belong under our thumbs.” You can find ethnically-based political parties almost everywhere there is civilian politics in a divided society. You can find ethnically-divided militaries. You can find intergroup struggles over the symbols of the state and recurrent patterns of hostile attitudes. You see quite similar rhetoric employed in a wide variety of cultural contexts.
There is a tremendous amount of commonality in these conflicts, quite enough to enable us to call this a single phenomenon. Yet, I want to stress some parts of the phenomenon that are not present everywhere, especially some forms of violence that are not present everywhere but have been present in the past—though no longer—in some places. The significance of such events increases if we look at changes over a period of time. Some forms of violence today used to be present in places that we now regard as more or less peaceful.
The specific form of violence I want to talk about is what I call the deadly ethnic riot: that intense, sudden, but not necessarily wholly unplanned, lethal attack by civilian members of one ethnic group on civilian members of another, the victims chosen because of their group membership. This category embraces what are variously called communal, racial, religious, or tribal disturbances, and it embraces the kind of riots seen in India, Indonesia, Nigeria, Sri Lanka, and many other countries. This is a very common and very brutal form of riot. I myself have data on more than 150 such events in more than fifty countries and a good many negative cases, by which I mean non-riots, near-riots, and so on.
The project from which this talk is drawn asks simpleminded questions, like Who, What, How, When, Where, Why, in the hope that the answers to such questions can lead to somewhat more complicated answers. These riots are usually accompanied by quite brutal atrocities. They produce a great many more refugees than they do victims, though they also produce quite a substantial number of victims. They often do some permanent damage to relations with the country to which the refugees fled. The commonness of such events testifies to the importance of ethnic conflicts in the world.
Some countries that used to have this form of violence do not have it anymore. Northern Ireland, for example, in the nineteenth century, was filled with this kind of Protestant-Catholic violence, one riot after another. When the current troubles began in the 1960s, there was a danger that one would see a recurrence of these riots. And there was, actually, one such episode in 1969 in Belfast involving large groups of Catholics and Protestants along the residential boundary between the two communities. There was a good deal of burning, and about ten people died in that riot. But that was almost thirty years ago, and in that event there were relatively few deaths. There was no real recurrence of that form of violence.
Instead, what happened in Northern Ireland was a great deal of intimidation in housing, in the 1970s particularly. People living in so-called mixed neighborhoods were induced by threats to leave. And of course there is the well-known terrorism of the various paramilitaries. But neither the residential intimidation nor the terrorist attacks required mass involvement. The intimidation required only an anonymous threat telling people to move out, and those rare cases of people who did not were subject to individual enforcement actions. The terrorism also can be considered a small-group phenomenon. What was missing, I want to underscore, was mass involvement. What did exist was violence limited, interestingly enough, to what the paramilitaries called legitimate targets, the exact opposite of the target of crowds rioting and choosing victims randomly. Paramilitaries went out of their way to lay down doctrines that defined legitimate targets and differentiated them from illegitimate targets.
I want to point briefly to another instance where we think of a relatively high level of ethnic conflict but where mass violence is absent and where terrorism has been the predominant mode of violence: the Basque country of Spain. There is a good deal of group antipathy. There is a large presence of immigrants from other areas who, at least from the standpoint of Basque separatists, constitute a problem to be dealt with. Yet that problem is not dealt with by mass violence but rather by terrorism against the instrumentalities of the state as the principal mode of violence.
As I draw this contrast between deadly ethnic violence and other forms of violence, in order to see what it can tell us about group attitudes, let me turn to another country that has a long history of ethnic riots but would probably come last on our lists: the United States.
The United States has experienced several forms of ethnic violence in the last 150 years or so. In the middle of the nineteenth century, there were nativist riots of exactly the kind that I have been talking about. They were against immigrants, especially against Catholics and Irish and German immigrants. In 1844, there was an ethnic riot in Philadelphia; in 1855, in Louisville, Kentucky, twenty-two Germans were killed in one day on the street.
From about the 1880s to the late 1930s, lynching was the common form of anti-black violence in the South. It was focused on individuals who were said to have violated the mores of Jim Crow, and it was exceedingly brutal. Interestingly enough, the lynch mob went out of its way to show that the violence was not directed at people of African descent who followed the rules. Great care was not always used in selecting the victim, but the alleged offense, that is the violation of the mores of segregated race relations, was said to be necessary to the legitimacy of the violent event. There was often a sham trial before the execution by the mob. The trial was not meaningful except to show that there would be punishment of persons who were violating those rules.
The lynchings were not originally anti-black; the first victims were white when lynching began on the frontier as a substitute for formal justice, long before the Civil War. It moved back to the South at first to fight abolitionists. Only afterward did it become more thoroughly racialized. In the 1890s, lynchings hit a high point and declined steadily in the following decades.
Notice that we have two forms of violence that declined. They overlapped in time but did not coincide within any specific part of the country. These latter ethnic riots were not nativists killing immigrants but whites killing blacks. The first major episode of an anti-black riot was in the draft riot of 1863. This form of violence really got going around the turn of the century and reached a high point in the 1910s. It declined in the 1920s and was pretty much over by the 1940s. White mobs would attack random black victims, largely in northern and border cities like Detroit, St. Louis, and Chicago. As in the case of the earlier immigrant riots, there were quite a lot of deaths in a single episode and a great deal of brutality, and deaths were more random.
Finally, there was a set of non-deadly riots in the form of protest violence in the 1960s, but with roots in the ‘30s and ‘40s, also largely in northern cities. We saw attacks by African-Americans on property owned by whites and on civil and public authorities. They were not principally attacks on people at all. Very few whites were killed; in fact, the majority of those who were killed were rioters killed by police. There were echoes of these riots in Miami in 1980-82 and again in Los Angeles in 1992, though in Los Angeles and Miami there was some emphasis also on interpersonal attacks, on Hispanics and Koreans, respectively.
Notice that each form of violence reached a peak, declined suddenly, and trailed off slowly thereafter. The precise form, however, makes a difference—or rather two differences. The first concerns the death count, and the second concerns what the form tells us about group relations.
The violence of the 1960s produced remarkably few deaths for the number of episodes. There were more than 500 identifiable disturbances, but fewer than 300 people were killed overall over seven or eight years. In fact, in only six percent of those disturbances was anyone killed at all. Yet the property damage was enormous, in the hundreds of millions of dollars, which tells us that there was rather intense sentiment at these events. Still, the deaths in all the hostilities of the 1960s equaled the number of people lynched in only one year of the 1890s. This number of deaths is about the same as in any single ethnic riot today in Asia, Africa, or the former Soviet Union. And this is a much smaller number than were killed in the worst riots of northern Nigeria in 1966. Probably about ten times that number were killed in those riots.
On the second point, the implications of these forms of violence, the reason nativist violence against immigrants ceased was that a group formerly seen to be alien eventually came to be incorporated in the definition of one of the contending categories in society. The boundary changed; there was amalgamation. A great defect of crossnational analysis is the frequently cited one of taking multiple snapshots. In point of fact, the rather depressing picture we have of deadly riots in one country after another can change. We have seen changes twice in the United States as nativist riots and anti-black riots and lynchings passed with a redefinition of group identity, an attitudinal change within the country.
This brings me to the underlying question I have come with: what are the underlying supports for deadly riots? Certainly intergroup antipathy is not enough. Let me deal first with what looks to be absent when violence does not take the form of deadly riots. I will turn briefly to the protest violence in the 1960s in the United States, and then to Northern Ireland terrorism, before coming to deadly riots by way of contrast.
During the 1960s violence, there were surveys of black opinion. Long after the violence subsided, it was quite clear that a substantial fraction of ghetto residents believed the violence to be justified. In post-riot surveys, many respondents showed some sympathy with the rioters. In Detroit, 46% felt at least some sympathy. In a fifteen-city survey, 54% of black respondents felt some sympathy with the rioters. In the late 1960s, anywhere between 12% and 31% (depending on the survey) of black respondents were willing to advocate violence either as the best or as a necessary means to equality. But it is interesting that killing had no support in the survey. Black respondents found the violence useful to call attention to grievances, but there was little approval given to the sniping and firebombing that took place. Extreme violence was not regarded as justifiable, and survey responses are devoid of the theme of revenge or violence as an end in itself. On this dimension, deadly riots are as different as they can be.
In Northern Ireland, a 1972 survey asked whether people supported wrongs in order to advance an ideal. Nineteen seventy-two was not a very good year in Northern Ireland, but when they were asked, “Is the use of force wrong to advance an ideal?” 52% of respondents said “very wrong”; 27% more said “usually wrong”; 18% said it was “sometimes the only way,” and only two percent said it was “the only way.” The Irish Republican Army and the other paramilitaries have been constrained, I want to argue, largely on the basis of this configuration of opinions, to develop a restrictive doctrine of legitimate targets because of the need to retain at least the minimal support of their communities. The IRA has eschewed attacks on Protestant civilians except as part of its war against so-called economic targets. Even then, the IRA has had a terrible time justifying those attacks. The IRA has justified those attacks with reference to the targets’ support for the Unionist cause, and this has not been a very popular form of justification.
To go further, in 1968-74 there were other surveys of attitudes in Northern Ireland. Each side said that, as people, the others were “about the same as our people.” In another survey, they called each other “ordinary people.” In 1979, 61% of Catholics thought the Irish Republic ought to get tougher with the IRA. Sixty-four percent thought that the Irish Republic ought to extra-dite IRA members for political crimes to the British-controlled north. Most people favored increasing contact, and in the period we are talking about (1969-91) intermarriage just about doubled in Northern Ireland, from about six to eleven percent. In Belfast, 20% of all marriages were Protestant-Catholic marriages. Contrast this with the usual refrain in countries that have severe divisions and deadly riots; one slogan in Assam, in India, was, “Drive out the Bengalis.” Those slogans are not really seen in Northern Ireland anymore, although they were seen there in the nineteenth century.
There are a lot of other data on the support for the IRA campaign of violence: under ten percent, even under five percent. Terrorism is clearly not the weapon of the politically strong, and quite obviously there would not be any support at this point for deadly ethnic riots.
This brings me to the subject of deadly riots generally. They are characterized by authoritative social support, that is, conduct by political authorities or social superiors that lends approval to violent behavior. It becomes a kind of green light for violence. Sometimes the rioters misread the signals, as they did in 1930 in Burma, when the Burmese thought the British would not mind if they killed Indians because after all the British had made clear their objections to the Indian nationalist movement. But sometimes governments do foster violence. I am now not talking about genocide but deadly riots. In Cambodia, for example, in 1970, when there was a powerful campaign against the Viet Cong, the Viet Cong came for many purposes to be viewed synonymously with the Vietnamese, and there were a good many Vietnamese in Cambodia; the deadly result, predictably, followed.
In Delhi in 1984, after Mrs. Gandhi’s assassination, television coverage showed her lying in state, and one could hear on camera the refrain “blood for blood,” which seemed to signal to a good many people in Delhi that Sikhs were fair game for attack. Permission is important to rioters. Of course, it creates an air of impunity, but equally important it legitimizes violence. The same goes for police inactivity or ineffectiveness or even violence, which is very common in deadly riots. Occasionally you will see police participate in violence.
Sometimes rioters will engage in violence even in the face of police opposition, provided they see their situation as sufficiently desperate. But more often, police intolerance will inhibit riots, and, if the ethnic hostility is still great, the intolerance may convert the hostility into some other form of violence, such as terrorism or other forms of retail rather than wholesale violence. One example of how this conversion can happen is in a city of the Tuva autonomous republic in the former Soviet Union. In this city, there is active anti-Russian hostility. It got very violent in 1990, but the police made it clear that it was not going to tolerate violence, and so those who had an interest in violence against Russians decided to refrain from wholesale violence and turned instead to retail violence, by individual murders of Russians. The police component to this situation is far from insignificant. But the more common pattern in general is police indifference, which encourages the rioters.
The same goes for the rarity of punishment. Most of the time, there are no prosecutions. When there are prosecutions, they are typically diminished. The few who are convicted are usually convicted for lesser crimes. Once again, the message to the rioters is that they are not in much danger. They also get this message, by the way, from the targets, who are rarely in a position to defend themselves or to retaliate later. Revenge riots by the target are exceedingly rare. But, more importantly, the absence of punishment also signals permission for the rioters.
This brings me to my main point, a point usually overlooked: the killings are actually approved in the wider society in which the rioters form a part. I want to show that this is so, why it is so, and how it is so.
First of all, contrary to what we might think, rioters are not merely marginals or deviants. It is true that, for the most part, the middle class does not participate in the killing, and often criminals and others who are imperfectly integrated into the society do latch on to the riots and participate. But they are not typically the central actors. What stands out most is the ordinariness of most participants: textile workers and manual laborers in 1969; Singhalese hospital workers attacking Tamil hospital staff in 1977; and so on. We see something like a random sample of mainly employed, young (between 18 and 30, usually), working-class men (not women), in most (but not all) cases, with a bias toward unskilled laborers. In other words, there is nothing specially marginal, nothing unusual, nothing otherwise pathological about the composition of the crowd, except, of course, that it does horrible things. That ordinary people are drawn in suggests that the violence has legitimacy and social support. If it did not, otherwise respectable people could not resume ordinary life, free of social sanction, after the fact. The ordinariness of the killing crowd testifies to its reflection of the norms and feelings of the group from which it springs.
Now there is also evidence of the legitimizing of violence. Surveys generally do not ask whether it is acceptable to kill other people. We have very few surveys in which that is part of the protocol, unfortunately for my work. We do not ask questions that investigators think will lack a significant diversity of responses. At present we are reduced to inferring attitudes from what people say and do after a riot. But it turns out to be easy to draw the inference.
There is an utter absence of remorse after the deadly ethnic violence. The very best you get is the suggestion that the violence was a blessing in disguise, by which the respondent means to impart only a very mild form of condemnation. Respondents basically say that the riot showed what needed to be shown about underlying tensions or about the need for better policies, usually skewed to benefit the attacking group. Much more often you get wholehearted approval of the killings. Here are the usual refrains: “We taught them a lesson.” “They were arrogant and deserved what they got.” “They brought it on themselves.” People who would not themselves engage in violence nevertheless approve it and are willing to say what I have just quoted you. I must say that when I began to realize that this was the nearly universal response to riots, I was at first quite chilled by it.
The approval is crucial for a reason that is usually misunderstood or overlooked. The rioters want to do the right thing, and they are convinced they are doing the right thing. They see the riots as having been brought on by the target group and its behavior. Justification is inevitably cast in terms of target group behavior. That is why precipitating events are necessary to the riot. The riot does not just happen out of nowhere; it follows a precipitating event or chain of events. Contrary to conventional wisdom, not just any precipitating event will do. It needs to be one that crystallizes the unsatisfactory state of group relations or threatens the position of the group that then initiates the violence, and the precipitant must be significant in that it constitutes both a threat and a justification. Likewise, authoritative social support is necessary not just to assure impunity, which I have said is important, but also the legitimacy of riots.
How does this work? What are the cognitive mechanisms that allow people to kill each other but feel no remorse? If you look at pre-riot events, the last precipitant before the violence breaks out is very often a set of powerful rumors. Usually these are rumors of aggression. The target group has “poisoned the water supply” (you will be surprised at how often that one goes around), or the target group has raped and killed and cut the breasts off of women, or the target group has organized an army that is “on the march.” I mean literally an army that is on the march or has already attacked and killed large numbers of “our” people. These are the most common rumors. These rumors are almost always totally false. No such atrocities have occurred; no armies are on the march. If there has been a fight, an injury will be reported as a death, and, if there has been a death, it will be reported as many deaths.
The rumors, however, have a crucial function. I need to digress for a moment to say that those who analyze rumors these days, in contrast to those who used to analyze rumors, often stress the manipulative character of the rumor, saying that one should focus on the person who started the rumor to find out who is fomenting the violence. Actually, I want to turn this around and go back to the earlier analysis of rumors, which was to ask, why does a rumor become so widely believed? I think it is much more important to ask why ordinary people would kill than it would be to ask why people would want to manipulate other people into killing. Once people are unwilling to kill, the manipulators will go their own way and look for other kinds of opportunities. There is already a demand for violence.
The rumors have a crucial and rather simple function; they legitimate violence as self-defense: “If we don’t stop them, they will kill us.” The rumors do something else, too. If we are talking about the use of violence in repelling mass aggression, we are talking, after all, about warfare. It is very clear, from pre-riot rituals, from the traditional martial motifs in the course of the riots, and from the rumors of aggression, that the attackers often see themselves as participating in a form of warfare. Whatever else one may say about warfare, it is regarded not as a series of separate transactions, each to be reckoned and justified separately, but as a type of extended transaction in which one killing is not considered sui generis, and it is certainly not subject to judgment apart from the totality of the event. The killing is seen as not only justified but necessary. The precipitators display the malevolent intentions of the target group and are an integral link in the chain of events.
Furthermore, in this particular form of warfare, as in much other warfare, there is no moral community between aggressors and victims, no empathy, and so it is possible to consider that lives on the other side of the boundary are of trivial value. It is in this sense, to return to the themes of this conference’s proceedings, that group loyalty is seen as a cause, if you want to use the word cause, of violent conflict.
At the outset of this study, I read a lot of riot reports, as you can imagine, and I was quickly faced with a problem of deciding what a riot was. This was not just a definitional problem but an operational one. When am I dealing with one riot, and when am I dealing with more than one riot? Sometimes you find that you have more than one crowd chasing more than one ethnic category of victim; this is not usual, but it happens. So is that just one riot or more than one riot? I will not bore you with the solution that I have devised, but I assure you that it is no easy matter to solve this problem. But in the course of my work I read a lot of literature on ontology, specifically on the individuation of acts and events.
What is most important is that the rioters do not define the riot episode as I have, as beginning with the activation by the precipitant and then following with the outbreak of violence. As amateur ontologists, or perhaps, we should say, as the truly professional ontologists that they are, the rioters see everything as related to everything else. They do not individuate the sequence of events that I call the riot. That is part of what makes it easy for them to justify the killing; that is, the event for them begins long before the riot and has to do with the threat posed by the target. And that is the mechanism by which ordinary people can justify killing: by considering it not as a separate event but as a link in a long, long chain, and a very threatening one at that.
All of this brings me back to the cases in which I began, those in which deadly riots are no longer present and have been superseded by terrorism or protest violence. In those cases, the justification for killing is much, much rarer, and that makes it hard to induce people to kill or at least to kill civilians en masse. It is probably also the case that deaths from collective violence in general, interestingly, not just in the two cases of the United States and Northern Ireland in the West, but also overall in the West, are down. Oddly enough, when I went looking for the answer to that question, it proved very difficult to pin it down, but it certainly seemed to be the case. Britain has been well studied; there it is very clear that labor violence, community violence, gang violence, and political violence all declined sharply from 1900 to 1975. (Only football violence increased during that period.) This is almost surely true in the United States as well.
Perhaps the professionalization of police has had something to do with it, but one also wants to go to broader explanations that relate to liberal states that do not find it easy to permit their citizens to kill each other face-to-face any more. People have become much more equality-minded. But is this a sufficient explanation? After all, people would not have become equality-minded if intergroup antipathies had not declined at the same time or before.
Underlying these changes in patterns of violence is change in social support for violence. Whereas in Northern Ireland in the nineteenth century, there was a clear and almost unanimous support for the violence, this has just simply stopped being true. The same goes for interpersonal black-white violence in the United States. Interestingly enough, in the United States, the changes have gone hand in hand with larger changes in attitudes regarding race relations. But, in Northern Ireland, the growing intolerance for interpersonal violence, while it has been accompanied by some of those same underlying changes in attitudes, has not been accompanied by anything like the growth of an interethnic political center dedicated to accommodation. In fact, that political center is exactly what Britain and the Irish Republic government have been determined to create at all costs, and so far quite unsuccessfully. In some ways, that makes Northern Ireland the more interesting case because it suggests that the conflict does not need to abate altogether in order for the deadly ethnic violence to become obsolete.
For some divided societies, this is very good news indeed because we have quite a lot of societies with a lot of conflicts, and we would not want to impose on them maximal attitudinal change as a prerequisite for a decline in violence. Most probably will not experience powerful changes in interethnic attitudes anytime soon. But it is still completely unclear what produces the sort of change that Northern Ireland and a handful of other countries have undergone as their relatively severe attitudes have been moderated over time.
Much that occurs in the buildup to riots—precipitants, rumors of aggression, a sense of warfare—goes to support the need and justification for a violent response. The lack of moral community between groups makes possible the judgments about the legitimacy of killing. Those judgments can change over time, even in severe conflict cases, but the process by which they change, and the role of deliberate intervention in that process, has thus far been largely hidden from view.
To conclude, I come back to the theme of group loyalty and violent conflict. It seems to me that the concept of group loyalty requires some unpacking. One can talk, at one level, of group loyalties sufficient to create a conflict of interest between groups. I do not mean to be materialist about it. I mean to include all of those things, evaluative and symbolic, that are dimensions along which conflict of interest can proceed. At another level, one can speak of group loyalties sufficient to create severe conflicts over the nature of the state and the place of groups within it, that is, sufficient to create exclusionary political institutions in which, for example, immigrants are excluded. And one can speak at yet another level of loyalties sufficient to produce an absence of moral community and therefore sufficient to provide a soil in which one can ground sanctions for widespread violence.
What is interesting is that groups can move among these three levels, that is, from lower levels of loyalty to higher, more intense levels of loyalty or vice versa. Groups can move from a lack of a moral community to really severe conflict or to something in between. But thus far, the truth of the matter is, nobody can tell us why.
 |
 |
 |