Classics in the History of Psychology

An internet resource developed by
Christopher D. Green
York University, Toronto, Ontario

(Return to index)

Intergroup Conflict and Cooperation: The Robbers Cave Experiment

Muzafer Sherif, O. J. Harvey, B. Jack White, William R. Hood, Carolyn W. Sherif (1954/1961)

[p. 197] CHAPTER 8

Summary and Conclusions

A. The Present Approach

In this book we have presented an experiment on Intergroup relations. The theoretical approach to the problem, the definitions of groups and relations between them, the hypotheses, the selection of subjects, the study design in successive stages, the methods and techniques, and the conclusions to be drawn are closely related. This chapter is a summary statement of these interrelated parts.

The word "group" in the phrase "intergroup relations" is not a superfluous label. If our claim is the study of relations between two or more groups or the investigation of intergroup attitudes, we have to bring into the picture the properties of the groups and the consequences of membership for the individuals in question. Otherwise, whatever we may be studying, we are not, properly speaking, studying intergroup problems.

Accordingly, our first concern was an adequate conception of the key word "group" and clarification of the implications of an individual's membership in groups. A definition of the concept improvised just for the sake of research convenience does not carry us far if we are interested in the validity of our conclusions. The actual properties of groups which brought them to the foreground in the study of serious human problems have to be spelled out.

The task of defining groups and intergroup relations can be carried out only through an interdisciplinary approach. Problems pertaining to groups and their relations are not studied by psychologists alone. They are studied on various levels of analysis by men in different social sciences. In the extensive literature on relations within and between small groups, we found crucial leads for a realistic conception of groups and their relations (Chapter 1).

Abstracting the recurrent properties of actual groups, we attained a definition applicable to small groups of any description. A group is a social unit which consists of a number of individuals who, at a given time, stand in more or less definite

[p. 198] interdependent status and role relationships with one another, and which explicitly or implicitly possesses a set of norms or values regulating the behavior of the individual members, at least in matters of consequence to the group.

Intergroup relations refer to relations between groups thus defined. Intergroup attitudes (such as prejudice) and intergroup behavior (such as discriminatory practice) refer to the attitudes and the behavior manifested by members of groups collectively or individually. The characteristic of an intergroup attitude or an intergroup behavior is that it is related to the individual's membership in a group. In research the relationship between a given attitude and facts pertaining to the individual's role relative to the groups in question has to be made explicit.

Unrepresentative intergroup attitudes and behavior are, to be sure, important psychological facts. But attitude and behavior unrepresentative of a group do not constitute the focal problem of intergroup relations, nor are they the cases which make the study of intergroup relations crucial in human affairs. The central problem of intergroup relations is not primarily the problem of deviate behavior.

In shaping the reciprocal attitudes of members of two groups toward one another, the limiting determinant is the nature of functional relations between the groups. The groups in question may be competing to attain some goal or some vital prize so that the success of one group necessarily means the failure of the other. One group may have claims on another group in the way of managing, controlling or exploiting them, in the way of taking over their actual or assumed rights or possessions. On the other hand, groups may have complementary goals, such that each may attain its goal without hindrance to the achievement of the other and even aiding this achievement.

Even though the nature of relations between groups is the limiting condition, various other factors have to be brought into the picture for an adequate accounting of the resulting intergroup trends and intergroup products (such as norms for positive or negative treatment of the other group, stereotypes of one's own group and the other group, etc.). Among these factors are the kind of leadership, the degree of solidarity, the kind of norms prevailing within each group. Reciprocal intergroup appraisals [p. 199] of their relative strengths and resources, and the intellectual level attained in assessing their worth and rights in relation to others need special mention among these factors. The frustrations, deprivations and the gratifications in the life histories of the individual members also have to be considered.

Theories of intergroup relations which posit single factors (such as the kind of leadership, national character, individual frustrations) as sovereign determinants of intergroup conflict or harmony have, at best, explained only selectively chosen cases.

Of course leadership counts in shaping intergroup behavior, the prevailing norms of social distance count, so do the structure and practices within the groups, and so do the personal frustrations of individual members. But none of these singly determines the trend of intergroup behavior at a given time. They all contribute to the structuring of intergroup behavior, but with different relative weights at different times. Intergroup behavior at a given time can be explained only in terms of the entire frame of reference in which all these various factors function interdependently. This approach, here stated briefly, constituted the starting point of our experiments on intergroup relations. The approach was elaborated fully in our previous work, Groups in Harmony and Tension.

The relative weights of various factors contributing to intergroup trends and practices are not fixed quantities. Their relative importance varies according to the particular set of conditions prevailing at the time. For example, in more or less closed, homogeneous or highly organized groups, and in times of greater stability and little change, the prevailing social distance scale and established practices toward out-group which have been standardized in the past for group members will have greater weight in determining the intergroup behavior of individual members. But when groups are in greater functional interdependence with each other and during periods of transition and flux, other factors contribute more heavily. In these latter cases, there is a greater discrepancy between expressed attitude and intergroup behavior in different situations, attributable to situational factors, as insistently noted by some leading investigators in this area of research. Alliances and combinations among groups which seem strange bedfellows are not infrequent in the present world of flux and tension.

[p. 200] Because of their influence in social psychology today, two other approaches to intergroup behavior deserve explicit mention. A brief discussion of them will help clarify the conception of the experiment reported in this book.

One of these approaches advances frustration suffered in the life history of the individual as the main causal factor and constructs a whole explanatory edifice for intergroup aggression on this basis. Certainly aggression is one of the possible consequences of frustration experienced by the individual. But, in order that individual frustration may appreciably affect the course of intergroup trends and be conducive to standardization of negative attitudes toward an out-group, the frustration has to be shared by other group members and perceived as an issue in group interaction. Whether interaction focusses on matters within a group or between groups, group trends and attitudes of members are not crystallized from thin air. The problem of intergroup behavior, we repeat, is not primarily the problem of the behavior of one or a few deviate individuals. The realistic contribution of frustration as a factor can be studied only within the framework of in-group and intergroup relations.

The other important approach to intergroup relations concentrates primarily on processes within the groups in question. It is assumed that measures introduced to increase cooperativeness and harmony within the groups will be conducive to cooperativeness and harmony in intergroup relations. This assumption amounts to extrapolating the properties of in-group relations to intergroup relations, as if in-group norms and practices were commodities easily transferable. Probably, when friendly relations already prevail between groups, cooperative and harmonious in-group relations do contribute to solutions of joint problems among groups. However, there are numerous cases showing that in-group cooperativeness and harmony may contribute effectively to intergroup competitiveness and conflict when interaction between groups is negative and incompatible.

The important generalization to be drawn is that the properties of intergroup relations cannot be extrapolated either (1) from individual experiences and behavior or (2) from the properties of interaction within groups. The limiting factor bounding intergroup attitudes and behavior is the nature of relations between groups. Demonstration of these generalizations has been one of the primary objectives of our experiment.

[p.201] B. The Experiment

The Design in Successive Stages

Experimental Formation of Groups: In order to deal with the essential characteristics of intergroup relations, one prerequisite was the production of two distinct groups, each with a definite hierarchical structure and a set of norms. The formation of groups whose natural histories could thus be ascertained has a decided advantage for experimental control and exclusion of other influences. Accordingly, Stage l of the experiment was devoted to the formation of autonomous groups under specified conditions. A major precaution during this initial stage was that group formation proceed independently in each group without contacts between them. This separation was necessary to insure that the specified conditions introduced, and not intergroup relations, were the determining factors in group formation. Independent formation of distinct groups permitted conclusions to be drawn later from observations on the effects of intergroup encounters and engagements upon the group structure.

The distinctive features of our study are Stages 2 and 3 pertaining to intergroup relations. The main objective of the study was to find effective measures for reducing friction between groups and to discover realistic steps toward harmonious relations between them. If we had attempted to get two groups to cooperate without first bringing about a state of friction between them, there would have been no serious problem to be solved. The great task that social scientists, practitioners and policy-makers face today is the reduction of prevailing intergroup frictions.

Intergroup Conflict: After formation of definite in-groups, we introduced a period of intergroup relations as Stage 2 of the experiment. During this stage, the two experimentally formed groups came into contact under conditions which were competitive, so that the victory of one group meant loss for the other. This series of encounters was conducive to successive frustrations whose causes were experienced as coming from the other group.

Only after an unmistakable state of friction between the two groups was manifested in hostile acts and derogatory stereotypes was the stage of reducing intergroup friction introduced.

[p. 202] Reduction of Intergroup Hostility: Various measures could have been tried in this experimental attempt toward the reduction of intergroup friction. One possible measure is the introduction of a "common enemy." Exposed to a common enemy, groups may join hands to do away with the common threat. This measure was not resorted to because it implies intergroup conflict on a larger scale.

Another possible approach is through dissemination of specific information designed to correct prevailing group stereotypes. This measure was not seriously considered because of the large body of research showing that discrete information, unrelated to central concerns of a group, is relatively ineffective in changing attitudes. Stereotypes crystallized during the eventful course of competition and conflict with the out-group are usually more real in the experience of the group members than bits of information handed down to them.

The alternative of channeling competition for highly valued rewards and prizes along individualized directions may be effective in reducing intergroup friction by breaking down group action to individual action. This measure may be practicable for small groups and is attempted at times by supervisors in classroom and recreational situations. However, frictions and conflicts of significant consequence in life and the problem of their resolution are in terms of group demarcations and alignments.

The initial phase of Stage 3 was devoted to testing the effects of intergroup contact involving close physical proximity in activities that were, satisfying in themselves, such as eating meals or seeing a movie. This initial phase was introduced with the objective of clarifying the blanket term "contact" as applied to intergroup relations.

The alternative chosen as the most effective measure for reducing intergroup friction was the introduction of a series of superordinate goals, in line with the hypothesis stated prior to the experiment. Superordinate goals are goals of high appeal value for both groups, which cannot be ignored by the groups in question, but whose attainment is beyond the resources and efforts of any one group alone.

[p. 203] Research Methods

The methods used in this experiment to bring about the formation and subsequent change of attitude and behavior ill directions predicted by the hypotheses were neither lecture method nor discussion method. Instead, the procedure was to place the members of respective groups in demanding problem situations, the specifications of which met the criteria established for the experimental stage in question. The problem situations concerned activities, objects or materials which we knew, on the basis of the expressed preferences of the individuals or the state of their organisms, were highly appealing to them at the time. Facing a problem situation which is immediate, which must be attended to, which embodies a goal that cannot be ignored, group members do initiate discussion, do plan, do make decisions and do carry through the plans by word and deed until the objective is achieved. In this process, the discussion becomes their discussion, the plan becomes their plan, the decision becomes their decision, the action becomes their action. in this process, discussion has its place, planning has its place, action has its place, and when occasion arises, lecture has its place, too. The sequence of these related activities need not be the same in all cases. In many instances, we observed two or three of them carried on simultaneously.

Thus, problem situations introduced in Stage 1 embodied goals of immediate appeal value to all members-within a group, and the goals required their concerted activity or coordinated division of labor for attainment. The problem situations of Stage 2 offered goals whose attainment by one group necessarily meant failure for the other group. Intergroup conflict was generated in the course of such engagements. The main part of Stage 3 consisted of introducing a series of situations conducive to super-ordinate goals requiring joint action by both groups towards common ends. In every stage, changes in attitudes and action were not attempted through a single problem situation, but through the cumulative effect of a series of varied activities which, during each stage, had the distinctive characteristics summarized here.

All problem situations were introduced in a naturalistic setting and were in harmony with activities usually carried out in such a setting. The individuals participating in the study were not aware that each step was especially designed to study a particular phase of group relations. Once the problem situation was [p. 204] introduced under specified conditions and at a specified time, the initiative, discussion and planning were theirs, of course within  bounds insuring health, security and well-being of the individuals studied.

Every effort was made that the activities and the flow of interaction in these activities appear natural to the subjects. Yet these activities and the interaction in them were experimental: Problem situations were chosen for each stage according to specified criteria (Chapter 2) and were introduced under specified conditions (including the place, terrain, time, arrangement of facilities, stimulus materials available, etc.). The choice of an isolated site made it possible to restrict interaction situations and the individuals involved in them to those appropriate during each experimental stage.

Techniques of data collection were also determined by the theoretical approach and methodological considerations briefly stated above. The subjects were not aware that behavioral trends reflecting favorable or unfavorable, friendly or hostile intergroup attitudes were being studied. Knowing that one is under constant observation cannot help becoming a factor in structuring experience and behavior, particularly when the observation is related to our status concerns, our acceptance or rejection by others, our good or bad intentions toward others.

To the subjects, the participant observers appeared to be personnel of a usual camp situation. They were introduced as senior counselors. In this capacity they were close to their respective groups in a continuing way. True to their announced roles, the participant observers jotted down relevant observations out of the subjects' sight, and then expended their notes later each day.

When the technique of observation is adapted to the flow of interaction, there is danger of being selective in the choice of events to be recorded. The effective remedy against possible selectivity is using a combination of methods to check findings obtained with one method against those obtained by other methods.

The events which revealed stabilization and shifts in statuses, and crystallization of negative and then positive intergroup attitudes were recurrent and so striking that one could not help observing them. However, in testing our main hypotheses, we [p. 205] supplemented the observational method with sociometric and laboratory-like methods. One distinctive feature of this study was introducing, at choice points, laboratory-like techniques to assess emerging attitudes through indirect, yet precise indices. Such laboratory-like assessment of attitudes is based on the finding that under relevant conditions, simple judgments or perceptions reflect major concerns, attitudes and other motives of man.

Reliability of observation and observer ratings was checked by comparing those of the participant observer with independent observations by others in crucial test situations. One such test situation illustrates the technique. When the status hierarchy in one group became stabilized toward the end of Stage 1, a problem situation was introduced which, like other problem situations of this stage, required initiative and coordination of the membership. A staff member who was not with the group regularly and who had not rated the status positions from day-to-day, observed the group interaction in this situation. On this basis he made independent ratings of the status hierarchy, which were significantly correlated with those of the participant observer of that group.

C. Main Conclusions

Individual Characteristics and Intergroup Behavior

In this experiment, the rigorous criteria and painstaking procedures for selecting subjects ruled out explanations of hostile or friendly intergroup attitudes in terms of differences in socio-economic, ethnic, religious, or family backgrounds. Similarly, the criteria for subject selection insured against explanations on the basis of unusual individual frustrations, failures, maladjustment or instability.

The subjects came from families who were established residents of the same city. They were stable families composed of natural parents and siblings. No subjects came from broken homes. Their religious affiliations were similar. They were from the middle socio-economic class. They were of the same chronological and educational level. They had all made satisfactory progress academically; none had failed in school. In school and neighborhood, their social adjustment was above average. None was a behavior problem in home, neighborhood or school.

[p. 206] In short, they were normal, healthy, socially well-adjusted boys who came from families with the same or closely similar socio-economic, ethnic, and religious backgrounds.

Since none of the individuals was personally acquainted with others prior to the experiment, pre-existing positive or negative interpersonal relations did not enter into the rise of intergroup attitudes.

The conclusion that explanations of the intergroup trends and attitudes on the basis of individual characteristics are ruled out in this experiment should not be construed to mean that the relative contributions of individuals within their own groups and in intergroup relationships are unimportant. Individuals do contribute differentially both in shaping and carrying on the trend of group relationships. This experiment does indicate, however, that intergroup attitudes are not merely products of severe individual frustrations or background differences brought to the situation.

Formation of Group Organization and Norms

When the individuals interacted in a series of situations toward goals with common appeal value which required interdependent activity for their attainment, definite group structures arose. These groups developed stable, but by no means immutable status hierarchies and group norms regulating experience and behavior of individual members.

More concretely, a pattern of leader-follower relations evolved within each group as members faced compelling problem situations and attained goals through coordinated action. As group structure was stabilized, it was unmistakably delineated as an "in-group." Certain places and objects important in group activities were incorporated as "ours." Ways of doing things, of meeting problems, of behaving under certain conditions were standardized, permitting variation only within limits. Beyond the limits of the group norms, behavior was subject to group sanctions, which ranged from ridicule, through ignoring the offender and his behavior, to threats, and occasionally to physical chastisement.

[p. 207] In-Group Cooperativeness Is Not Directly Transferable

When two groups met in competitive and reciprocally frustrating engagements, in-group solidarity and cooperativeness increased. Toward the end of intergroup friction (Stage 2), in-group solidarity became so strong that when the groups were taken to a public beach crowded with outsiders and affording various distractions, our groups stuck almost exclusively to activities within their respective in-groups. Psychologically, other people did not count as far as they were concerned. In the presence of so many people and distractions, this intensive concentration of interests and activities within the group atmosphere would have been impossible had the groups gone there before attaining such a high degree of solidarity.

This heightened in-group solidarity and cooperativeness were observed at the very time when intergroup hostility was at its peak, during the period when the groups asserted emphatically that they would not have anything more to do with each other. This can only mean that the nature of intergroup relations cannot be extrapolated from the nature of in-group relations. In-group solidarity, in-group cooperativeness and democratic procedures need not necessarily be transferred to the out-group and its members. Intergroup relations cannot be improved simply by developing cooperative and friendly attitudes and habits within groups.

Consequential Intergroup Relations Affect In-group Relations

Special note should be made of a related finding, namely that consequential intergroup relations have an impact on the in-group organization.

When it became evident that certain members of one group, including the leader, were not living up to the responsibilities expected of them by other members during the eventful course of intergroup competition, leadership changed hands. Those individuals who distinguished themselves by giving a good account for their group rose in the status hierarchy. Internal shifts in status were observed again during the cooperative intergroup activities of Stage 3. Functional relations between groups which are of consequence tend to bring about changes in the pattern of in-group relations.

[p. 208] Limiting Conditions for Intergroup Attitude and Behavior

We have seen that the individuals studied in this experiment were selected in ways which rule out explanations for the direction of intergroup behavior on the basis of differences in their backgrounds or on the basis of their individual frustrations, instabilities and the like. In the preceding sections, we have seen evidence that in-group properties were affected by consequential intergroup relations. Thus the intergroup hostility and its reduction cannot be explained merely by the nature of relationships within the groups.

Our findings indicate that the limiting condition determining friendly or hostile attitudes between groups is the nature of functional relations between them, as defined by analysis of their goals. When the groups competed for goals which could be attained by only one group, to the dismay and disappointment of the other, hostile deeds and unflattering labels developed in relation to one another. In time, derogatory stereotypes and negative attitudes toward the out-group were crystallized. These conclusions are based on observations made independently by observers of both groups and other staff members. Sociometric indices pointed to the overwhelming preponderance of friendship choices for in-group members. Experimental assessment of intergroup attitudes showed unmistakable attribution of derogatory stereotypes to the villainous out-group and of favorable qualities to the in-group. Laboratory-type judgments of performance showed the tendency to overestimate the performance attributed to fellow group members and to minimize the performance of members of the out-group.

What Kind of Contact Between Groups is Effective?

The novel step in this experiment was Stage 3, in which intergroup friction was reduced. We have already stated why we discarded certain procedures in this stage, such as introducing a "common enemy" or disseminating information. In order to clarify the term "contact," we tried the method of bringing the groups into close proximity in a series of activities. Most of these contact situations involved activities which were satisfying in themselves, such as eating good food in the same room, attending a movie together, or engaging in an exciting activity like [p. 209] shooting fireworks. But none of them created a state of interdependence between the groups. Such contact situations did not prove effective in reducing friction. Instead contact situations not conducive to interdependence were used by our groups for overt acts of hostility and further exchanges of unflattering invectives.

The ineffectiveness of contacts during which hostile groups engaged, while in close physical contiguity, in activities which were themselves satisfying to each individual has obvious implications for psychological theorizing.

The Introduction of Superordinate Goals

During the final period of the experiment, the prevailing friction between groups was reduced. Reduction of the conflict and hostility was observed in reciprocally cooperative and helpful intergroup actions, in friendly exchanges of tools, in developing standard procedures for alternating responsibilities and in meeting problems. The change in behavior and patterns of interaction between the groups was striking to all observers. The reliability of these observations is established by sociometric indices which showed increases of friendship choices for the erstwhile antagonists and also in the sharp decrease of unfavorable stereotypes toward the out-group. Favorable conceptions of the out-group developed, so that ratings of the in-group and out-group were no longer a set of contrasted polarities.

The end result was obtained through introduction of a series of superordinate goals which had compelling appeal value for both groups but which could not be achieved by the efforts and resources of one group alone. When a state of interdependence between groups was produced for the attainment of superordinate goals, the groups realistically faced common problems. They took them up as common problems, jointly moving toward their solution, preceding to plan and to execute the plans which they had jointly envisaged.

In this experiment, the setting and circumstances for the introduction of superordinate goals were elaborately prepared by the experimenters. But beyond setting the scene, the methods followed, the discussion necessary for the solution, the plans to [p. 210] be made and executed were left to the groups themselves. Faced with superordinate goals, the groups carried on discussion when necessary, listened to the advice and suggestions of members of both groups who were resourceful, made decisions, and even combined discussion, decision and deeds simultaneously when the goal was attained more effectively this way.

Cumulative Effects of Superordinate Goals

If the hostile attitudes generated during intergroup friction had any stability, it could not be expected that one or two situations embodying superordinate goals could wipe them out. Indeed intergroup antagonisms did not disappear in one stroke. At first, cooperative interaction involving both groups developed in specific situations in response to common problems and goals, only to be followed by a renewal of sharply drawn group lines and intergroup friction after the challenge had been met. Patterns and procedures for intergroup cooperation were laid down at first on a small scale in specific activities. Only during interaction in a series of situations involving superordinate goals did intergroup friction begin to disappear and the procedures for intergroup reciprocity developed in specific situations extend spontaneously to widening areas of activity.

In the sequential events of Stage 3 (Chapter 7), it was abundantly evident that the series of activities conducive to superordinate goals provided opportunities for members of the two groups to work out and develop procedures for cooperation in various spheres of action. Once a cooperative pattern was effective in a specific activity, it was extended by members of both groups to related actions. In the face of successful functioning of such procedures, the occasional dissident member who preferred the old days of intergroup strife or self-imposed separation found it more difficult to make his voice count in his own group.

Some procedures successful in intergroup interaction had previously been used by the groups in meeting problems within their own groups. But their transfer to intergroup interaction involved a significant step: the tacit recognition that the procedures now involved groups of individuals and not merely so many individual members within a group. Each individual within his group had been expected and encouraged by others to contribute to group efforts to the best of his abilities. Now, each [p. 211] group expected the other to contribute its share to meeting intergroup problems. While previously solutions were experienced as equitable or not relative to the individual's expectations and contributions within his group, now justice was also evaluated relative to equitable participation and opportunity for the groups as well.

The Same Tools May Serve Intergroup Conflict or Cooperation

In planning and working towards superordinate goals, there were times when the groups used jointly the tools and techniques which had been used by one or both groups separately in the service of tights during the intergroup conflict. Tools and techniques can be put to the service of harmony and integration as well as of deadly competition and conflict. Tools, in themselves, are not opposed to cooperation among individuals using them. It is the individuals as group members who put the tool to use in their opposition to other groups.

Even the proprietary pride that a place, a technique, a tool is "ours" takes on a different significance when the trend in intergroup relations is cooperation toward superordinate goals. Use of the technique or the tool in intergroup activities now implies a contribution toward a goal common to both groups - a contribution by the group in which members may take personal pride and which can be reciprocated by the other group equally enjoying its benefits through its own contributions at that or future occasions.

Superordinate Goals Alter the Significance of Other Influences

Contacts between groups in the course of actions towards superordinate goals are effective. They are used for developing plans, making decisions, and for pleasant personal exchanges. Information about the out-group becomes a matter of interest to group members and is actually sought in the course of interactions between members of the two groups. Leaders find that the trend toward intergroup cooperation in activities involving superordinate goals widens the spheres in which they may take positive steps toward working out procedures for joint endeavors and planning future contacts. Mingling with members of the other group and sharing in activities with them is no longer perceived.

[p. 212] by in-group members as "betrayal" or "treason". Similarly, the out-group member who engages in activities with the in-group is no longer seen by them as a strange and threatening figure in "our midst." On the contrary, intermingling of persons from different groups becomes a joint opportunity to work towards goals shared by both groups.

These are products of interaction toward goals superordinate to all groups, which are genuinely appealing to all, whose attainment requires equitable participation and contributions from all groups in interdependent activities.