After labeling their subjects, the researchers did something that may appear odd. They, in essence, gave each subject a bucket of money and told them to divide it among the other subjects in any way they saw fit. The division was carried out in private. None of the subjects knew any of the other subjects, or could even see them during the course of the experiment. Still, when passing out the money, they favored their in-group, those who shared their group label.
A large body of research replicates the finding that our group-based social identity is so strong that we will discriminate against them and favor us even if the rule that distinguishes them from us is akin to flipping a coin. That’s right: not only do we identify with a group based on the flimsiest of distinctions, we also look at group members differently—even if group membership is unrelated to any relevant or meaningful personal qualities. That’s not just important in our personal lives; it also affects organizations. For example, companies can gain by fostering their employees’ in-group identification, something that can be accomplished by creating and making salient a distinctive corporate culture, as was done very successfully by companies such as Disney, Apple, and Google. On the other hand, it can be dicey when a company’s internal departments or divisions develop a strong group identity, for that can lead to both in-group favoritism and out-group discrimination, and research suggests that hostility erupts more readily between groups than between individuals.14 But regardless of what kind of shared identity does or doesn’t exist within a company, many companies find it effective to use marketing to foster a group identity among their customers. That’s why in-groups based on Mac versus PC ownership, or Mercedes versus BMW versus Cadillac, are more than just computer clubs or car clubs: we treat such categorizations as meaningful in a far broader realm than they have any right to be.
Dog person versus cat person. Rare meat versus medium. Powdered detergent versus liquid. Do we really draw broad inferences from such narrow distinctions as these? The Klee/Kandinsky study, and literally dozens more like it, followed a classic experimental paradigm invented by Henri Tajfel, who conducted the line-length experiment.15 In this paradigm, subjects were assigned to one of two groups. They were told that their group assignments had been made on the basis of something they shared with other members of the group but which, objectively speaking, was really quite meaningless as a way of affiliating with a group—either the Klee/Kandinsky preference or whether they had overestimated or underestimated the number of dots that were quickly flashed on a screen.
As in the study I quoted earlier, Tajfel allowed his subjects to dole out awards to their fellow subjects. To be precise, he had them give out points that could later be cashed in for money. The subjects did not know the identities of the people they were giving points to. But in all cases they knew the group to which the person belonged. In Tajfel’s original study, the handing out of points was a bit complicated, but the crux of the experiment lies in just the way it was done, so it is worth describing.
The experiment consisted of over a dozen stages. At each stage, a subject (“awarder”) had to make a choice regarding how to dole out points to two other subjects (“recipients”), who, as I said, were anonymous. Sometimes the two recipients were both members of the subject’s own group or both members of the other group; sometimes one was a member of the subject’s own group and the other was a member of the other group.
The catch was that the choices offered to awarders did not represent a zero-sum game. That is, they did not entail simply deciding how to divide a fixed number of points. Rather, the options offered added up to varying point totals, as well as differing ways of splitting those points among the two recipients. At each stage, the awarder had to choose from among over a dozen alternative ways to award points. If the awarders felt no in-group favoritism, the logical action would be to choose whichever alternative bestows upon the two recipients the greatest total number of points. But the awarders did that in only one circumstance: when they were dividing points among two members of their in-group. When awarding points to two members of the out-group, they chose options that resulted in awarding far fewer points. And what is really extraordinary is that when the options required awarders to divide points between one in-group member and one out-group member, they tended to make choices that maximized the difference between the rewards they gave to the two group members, even if that action resulted in a lesser reward for their own group member!
That’s right: as a trend, over dozens of individual reward decisions, subjects sought not to maximize their own group’s reward but the difference between the reward their group would receive and that which the other group would be awarded. Remember, this experiment has been replicated many times, with subject pools of all ages and many different nationalities, and all have reached the same conclusion: we are highly invested in feeling different from one another—and superior—no matter how flimsy the grounds for our sense of superiority, and no matter how self-sabotaging that may end up being.
You may find it discouraging to hear that, even when group divisions are anonymous and meaningless, and even at their group’s own personal cost, people unambiguously choose to discriminate in favor of their in-group, rather than acting for the greatest good. But this does not doom us to a world of never-ending social discrimination. Like unconscious stereotyping, unconscious discrimination can be overcome. In fact, though it doesn’t take much to establish grounds for group discrimination, it takes less than we think to erase those grounds. In the Robbers Cave experiment, Sherif noted that mere contact between the Eagles and the Rattlers did not reduce the negative attitude each group had for the other. But another tactic did: he set up a series of difficulties that the groups had to work together to overcome.
In one of those scenarios, Sherif arranged for the camp water supply to be cut off. He announced the problem, said its cause was a mystery, and asked twenty-five volunteers to help check the water system. In reality, the researchers had turned off a key valve and shoved two boulders over it and had also clogged a faucet. The kids worked together for about an hour, found the problems, and fixed them. In another scenario, Sherif arranged for a truck that was supposed to get food for the boys not to start. The staff member who drove the truck “struggled and perspired” and got the truck to make all sorts of noises, as more and more of the boys gathered around to watch. Finally the boys figured out that the driver might be able to start the truck if they could just get it moving. But the truck was on an uphill slope. So twenty of the boys, from both groups, tied a tug-of-war rope to the truck and pulled it until it started.
These and several other scenarios that gave the groups common goals and required cooperative intergroup actions, the researchers noted, sharply reduced the intergroup conflict. Sherif wrote, “The change in behavior patterns of interaction between the groups was striking.”16 The more that people in different traditionally defined in-groups, such as race, ethnicity, class, gender, or religion, find it advantageous to work together, the less they discriminate against one another.17