Images of inhumanity and atrocity are burned into our memories. Jewish men, women and children being herded into gas chambers. Entire villages destroyed by rampaging gangs in Rwanda. The systematic use of rape and the destruction of communities as part of ethnic cleansing in the Balkans. The massacre at My Lai in South Vietnam, the abuse of Iraqi prisoners in Abu Ghraib, and most recently, the carnage wrought by suicide bombers in Baghdad, Jerusalem, London and Madrid. Reflecting on these events, we inevitably ask: What makes people so brutal? Are they mentally ill? Are they the products of dysfunctional families or cultures? Or, more disturbingly, is anyone capable of taking part in collective ruthlessness given the right--or rather, the wrong--circumstances? Now the latest research, including possibly the largest social-psychology experiment in three decades, is providing a new window on these conundrums.

Questions about why groups can behave badly have driven some of the most significant developments in social psychology in the 60 years since World War II ended. Starting with the need to understand the psychological processes that made possible the horrors of the Holocaust, scientists have wanted to know how large numbers of apparently civilized and decent people can perpetrate appalling acts.

Initially theorists sought answers to group pathology in individual psychology. In 1961, however, German-born American historian and political philosopher Hannah Arendt witnessed the trial in Jerusalem of Adolf Eichmann, one of the chief architects of the Holocaust. She concluded that far from the defendant demonstrating a perverted and sadistic personality (as psychiatrists for the prosecution claimed), he was utterly unremarkable and disarmingly ordinary. Arendt pronounced Eichmann to be an embodiment of the banality of evil.

Everyday Evil?

First published in 1963 in the New Yorker, Arendt's analysis was considered shocking and heretical. But a series of studies conducted around the same time supported her observations. In experiments at U.S. summer camps during the late 1950s, Muzafer Sherif, a Turkish-born American social psychologist, learned that normal schoolboys became cruel and aggressive toward former friends once they had been placed in different groups that had to compete over scarce resources. Even more striking were obedience studies carried out at Yale University in the early 1960s by Stanley Milgram. Ordinary, well-adjusted males who took part in a bogus memory experiment were told to deliver electric shocks of increasing magnitude to another person who posed as the learner. (In actuality, the learner, an accomplice of the experimenter, received no shocks.) Amazingly, every single teacher was prepared to administer intense shocks of 300 volts, and two thirds obeyed all the experimenter's requests, dispensing what they believed were 450 volts. Participants continued meting out punishments even after hearing the learner complain of a heart condition and yell in apparent agony. Milgram concluded: Arendt's conception of the banality of evil comes closer to the truth than one might dare to imagine.

The vivid culmination of this line of inquiry was the Stanford prison experiment, carried out in 1971 by Stanford University psychologist Philip G. Zimbardo and his colleagues. The researchers randomly assigned college students to be either prisoners or guards in a simulated prison in the basement of the campus psychology building. The goal was to explore the dynamics that developed within and between the groups over a two-week period. The study delivered these dynamics in abundance. Indeed, the guards (with Zimbardo as their superintendent) exerted force with such harshness that the study was halted after only six days.

The experimenters concluded that group members cannot resist the pressure of their assumed stations and that brutality is the natural expression of roles associated with groups who have unequal power. Accordingly, two maxims, which have had immense influence at both a scientific and a cultural level--and which are taught as received knowledge to millions of students around the world every year--are routinely drawn from the Stanford experiment. The first is that individuals lose their capacity for intellectual and moral judgment in groups; hence, groups are inherently dangerous. The second is that there is an inevitable impetus for people to act tyrannically once they are put in groups and given power.

Reexamining Group Power

The weight of the Stanford prison experiment lies in both its dramatic findings and the simple, stark conclusions that have been drawn from it. Over the years, however, social psychologists have developed doubts about the resulting received wisdom.

First, the idea that groups with power automatically become tyrannical ignores the active leadership that the experimenters provided. Zimbardo told his guards: You can create in the prisoners... a sense of fear to some degree, you can create a notion of arbitrariness that their life is totally controlled by us.... Theyll have no freedom of action, they can do nothing, say nothing that we dont permit.... Were going to take away their individuality in various ways.

Second, we know that groups do not perpetrate only antisocial acts. In studies--as in society at large--the group often emerges as a means to resist oppression and the pressure to act destructively. In variants of Milgram's obedience trials, participants were much more likely to defy the experimenter when they were supported by confederates who were also disobedient.

In addition, research after Stanford has confirmed the prosocial and enriching aspects of groups. One particularly influential approach to understanding groups in psychology today is the social identity theory developed in 1979 by social psychologists John Turner, now at the Australian National University, and Henri Tajfel, then at Bristol University in England. This theory holds that it is mostly in groups that people--particularly the individually powerless--can become effective agents who shape their own fate.

When individuals share a sense of identity (for example, we are all American, we are all Catholic), they seek to reach agreement, they like and trust one another more, they are more willing to follow group leaders, and they form more effective organizations. This fact is shown, for example, in extensive studies of cooperation in groups conducted recently by Steven L. Blader and Tom R. Tyler of New York University. As a result, people can pull together to create a social world based on their shared values--instilling a state of collective self-realization, which is good for psychological health. Having the social support to control one's fate results in higher self-esteem, less stress, and lower levels of anxiety and depression.

People who share a sense of identity in a group demonstrate two social features. First, they do not lose the capacity for judgment; instead the basis for their decisions shifts from their individual notions to their commonly held understandings. As field studies by one of us (Reicher) have shown (a summary appears in the Blackwell Handbook of Social Psychology: Group Processes), even the most extreme collective actions, such as a riot, reveal a pattern of behavior that reflects the beliefs, norms and values of the group involved. Second, people's responses vary depending on which group membership is most important to them in any given situation. For instance, the norms and values that we use as employees in the workplace may differ from those that govern us as believers at our place of worship, as activists at a political rally or as patriots at the raising of the flag.

In contrast to the Stanford conclusions, however, social identity theorists have long argued that people do not automatically accept the group memberships that others give them. Quite frequently people distance themselves from groups, especially those that are devalued in society. For instance, in the 1970s Howard Giles and Jennifer Williams, both then at Bristol University, pointed out that many women respond to inequality by downplaying their gender, emphasizing their personal qualities and seeking success as individuals. Only when they believe that they cannot escape--that is, when boundaries between groups are seen to be impermeable, as feminists argued when they identified the glass ceiling--will they identify with the devalued group and act collectively. In addition, they will be prepared to use their collective power to challenge the status quo and try to improve the position of their group only if they view the social system as unstable.

A large body of research, including controlled laboratory studies, extensive surveys and detailed field observations, supports the social identity perspective (for a review, see Social Identity, edited by Naomi Ellemers of Leiden University in the Netherlands and her colleagues). Yet until recently there was no single study in the mold of Sherif, Milgram or Zimbardo that might illustrate and combine the theory's various propositions in a comprehensive and compelling manner. What is more, it seemed impossible to conduct such a study. Despite all the doubts surrounding Stanford, its very severity seemed to put further studies of its type off-limits.

This situation recently changed with the BBC prison experiment. The two of us collaborated with the British Broadcasting Corporation, which funded the research and broadcast the findings in four one-hour documentaries. Our first challenge was to develop ethical procedures that would ensure that, though strenuous, the study would not harm participants. We put into place a raft of safeguards, including on-site clinical psychologists and a round-the-clock independent ethics committee. As the committee's report concludes, we showed that it is possible to conduct dynamic field studies that are also ethical.

The Experiment

Like Stanford, the BBC experiment divided men randomly into guards and prisoners within a custom-built environment. We modeled the setting on a prison, but more broadly we aimed to represent a general class of institutions--such as an office, a barracks, a school--in which one group has more power and privilege than the other. Throughout the study we watched the behavior of participants using unobtrusive cameras. We monitored their psychological states through daily tests. We even checked their well-being by using saliva swabs to measure levels of cortisol--an indicator of stress.

Although our experiment followed the same basic paradigm as that of Stanford, it differed in a number of ways. Unlike Zimbardo, we did not assume any role within the prison, so we could study the group dynamics without directly managing those interactions. Second, we manipulated features of social hierarchy that social identity theory predicts should affect prisoners identification with their group and the forms of behavior in which they subsequently engage. Most significant, we varied the permeability of group boundaries by initially allowing, but then eliminating, opportunities for promotion from prisoner to guard. We expected that given the possibility of advancement, prisoners would try to reject their prisoner identity and work independently to improve their position. We anticipated that this strategy would reinforce the status quo and allow the guards to maintain ascendancy. But after we ruled out promotion (on day three), we thought the prisoners would start collaborating to resist the guards authority.

The results upheld our predictions. At first the prisoners were compliant and worked hard to improve their situation. They started to see themselves as a group and become uncooperative with the guards only when they learned that, no matter how industrious they were, they would remain prisoners. What is more, this shared identity led to improved organization, effectiveness and mental well-being. As the study progressed, the prisoners became more positive and empowered.

The guards, however, surprised us. Several guards were troubled by the idea that groups and power are dangerous, and they were reluctant to exercise control. Uncomfortable with their task, they disagreed with other guards as to how their role should be interpreted and never developed a shared sense of identity. This lack of identity led to a dearth of organization among the guards--which in turn meant that they became increasingly ineffective at maintaining order and increasingly despondent and burned out. As the study progressed, the guards administration became ever more toothless.

After six days, the prisoners collaborated to challenge the fragmented guards, leading to an organized breakout and the collapse of the prisoner-guard structure. Then, on the ruins of the old system, both prisoners and guards spontaneously established a more equal system--in their words, a self-governing, self-disciplining commune. Once again, though, some members were bothered by the idea of using power. They did not discipline individuals who neglected to perform assigned tasks and broke the commune's rules.

At this point, we got a second surprise. Supporters lost faith in their ability to make the commune work, leaving its members in disarray. In response, a number of former prisoners and former guards proposed a coup in which they would become the new guards. They asked for black berets and black sunglasses as symbols of a new authoritarian management that they wanted to impose. They talked of re-creating the guard-prisoner divide but this time ensuring that the prisoners toed the line--using force if necessary. We expected those who had supported the commune to defend the democratic arrangement they had put in place. But nothing of the kind happened. Instead they lacked the individual and collective will to defy the new regime. Psychometric data also indicated that they had become more authoritarian-minded and more willing to accept strict leaders.

In any event, the coup never occurred. For ethical reasons, we could not risk the type of force witnessed in the Stanford study, and so we brought the study to a premature close on the eighth day. But whereas the outcome resembled that of Stanford, the path our participants took to reach that point was very different. In particular, the specter of tyranny was very clearly not a product of people acting naturally in terms of the groups to which they had been assigned. Instead it arose from the failure of those groups: for the guards, the inability to develop a cohesive bond and, in the case of the commune, the breakdown in the attempt to turn collective beliefs into reality.

Lessons for Society

Why did participants who had rejected mild inequalities imposed on them at the start of the study, and who had fought so hard to establish a democratic system, end up shifting toward a self-supported tyranny? The answers lie in a basic corollary to our arguments. Groups, we have discussed, are ultimately about collective self-realization. They use social power to make an existence in the image of their shared beliefs and values. But when groups cannot produce such a working order, their members become more willing to accept other social structures--even if these new systems violate their existing way of life. Thus, when the guards could not impose their authority, they became more willing to agree to democracy. More ominously, though, when the commune fell apart, its members became less motivated to defend democracy against tyranny.

From this study and other research into social identity processes, we can draw conclusions that have important implications for academia and society at large. In general terms, we concur with Sherif, Milgram, Zimbardo and others that tyranny is a product of group processes, not individual pathology. Yet we disagree about the nature of these processes. From our standpoint people do not lose their minds in groups, do not helplessly succumb to the requirements of their roles and do not automatically abuse collective power. Instead they identify with groups only when it makes sense to do so. And when they do, they actively and knowingly attempt to implement collective values--the way in which they exercise power depends on these values. In short, groups do not deny people choice but rather provide them with both the grounds and the means to exercise choice.

Of course, this argument does not deny that people can do terrible things in groups. But not all groups in charge and certainly not all prison guards are brutal. To propose that there is something inherent in group psychology that mandates excessive cruelty is to take the spotlight off the specific factors that lead particular groups to become vicious, brutish and tyrannical.

Two interrelated sets of circumstances can lead to a tyrannical group dynamic. The first arises from the success of groups that have oppressive social values. It has been pointed out, for instance, that the worst atrocities occur when people believe they are acting nobly to defend against a threatening enemy. One might wonder: How do they come to hold those beliefs? In turn we ask: What is the role of national leaders in demonizing out-groups--such as Jews, Tutsis or Muslims? What about immediate superiors of military units who actively encourage brutality or passively condone it? What part do ordinary men and women play when they laugh or turn a blind eye to an out-group member who is being humiliated? As our questions imply, we believe that people at every level of the group help to foster a collective culture of hate and are responsible for its consequences.

Less straightforwardly, the second constellation of factors that can give birth to tyranny occurs when groups who seek to instill democratic and humane social values do not succeed. When a social system collapses, people will be more open to alternatives, even those that previously seemed unattractive. Moreover, when the collapse of a system wreaks such havoc that a regular and predictable social life becomes impossible, the promise of a rigid and hierarchical order becomes more alluring. Thus, the chaotic failure of the democratic Weimar led to the rise of Nazism; the deliberate divisions imposed by ruling powers facilitated the ascendancy of brutalizing regimes in postcolonial Africa and in the post-Soviet Balkans; and the suppression of postwar organization paved the way to the reemergence of antidemocratic forces in Iraq. In each case, the rejection of democracy can be traced back to political strategies that deliberately sought to break down groups and strip them of power. Rather than striving to make people fearful of groups and power, we suggest, people should be encouraged to work together to use their power responsibly.

To the extent that received wisdom encourages policymakers to foment the very conditions that can promote oppressive regimes, that thinking may be not only intellectually limited but downright dangerous. It was certainly perilous for the participants in the BBC prison experiment. Their tragedy was to neglect the reasonable exercise of power for fear of tyranny. As a bitterly ironic consequence, they set up the very conditions whereby the tyranny they feared came back to haunt them.