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. 27] CHAPTER 2

Approach, Hypotheses and General Design of the Study [1]

The focal concern of this study is intergroup relations. As an experiment in social psychology, it undertakes to trace over a time period the formation and functioning of negative and positive attitudes of members of one group toward another group and its members as a consequence of experimentally introduced situations. Therefore, the main hypotheses relate to attitudinal and behavioral trends predicted as a result of controlled alterations of the conditions in which experimentally formed in-groups interact.

The general trend of findings from the sociology of small in-groups and their intergroup relations and relevant findings from the work of experimental psychologists led us to the experimental study of the problem of intergroup relations in successive stages. In the present undertaking (Summer, 1954) it will be carried out in 3 successive stages. The main features of these 3 successive stages are the following:

Stage 1: Experimental production of in-groups with a hierarchical structure and set of norms (intra-group relations). In line with our 1949 and 1953 studies, this will be done, not through discussion methods or through lecture or exhortation by resource persons or experts, but through the introduction of goals which arise as integral parts in the situations, which have common appeal value, and which necessitate facing a common problem, discussion, planning and execution in a mutually cooperative way.

Stage 2: Bringing the two experimentally formed groups into functional relations in situations in which the groups find themselves in competition for given goals and in conditions which imply [p. 28] some frustration in relation to one another (intergroup tension).

Stage 3: Introduction of goals which cannot be easily ignored by members of the two antagonistic groups, but the attainment of which is beyond the resources and efforts of one group alone. Such goals will be referred to as superordinate goals throughout this report. Superordinate goals are to be introduced with the aim of studying the reduction of intergroup tension in order to derive realistic leads for the integration of hostile groups. Considerations which led to the selection of this approach rather than other possible alternatives (such as a common enemy, leadership technique or discussion techniques) are stated briefly in the discussion of Stage 3 in the last part of this chapter.

It should be emphasized at the outset that individuals brought into an experimental situation to function as small groups are already members of actual groups in their social settings and thus have internalized values or norms (i. e., attitudes) which are necessarily brought to the situation. With this consideration in mind and in order to give greater weight to experimentally introduced factors in the situation, a special effort will be made in this study not to appeal to internalized values or to prestige symbols coming from the larger setting in the formation and change of positive or negative attitudes in relation to respective in-groups and out-groups.

Background of the Above Summary

The rationale that underlies the above formulation of our approach to the study of intergroup relations stems from relevant findings in both sociology and psychology. They are stated more fully elsewhere (Note 1). Here only a summary statement of these lines of development will be given.

Empirical observations by social scientists and inferences made by psychologists without direct experimental verification present a rather confusing picture at the present time. Therefore it is necessary to state precisely the sense in which the concept "group" and the issue of relations between them (intergroup relations) are used here:

A group may be defined as a social unit (1) which consists of a number of individuals who, at a given time, stand in more or [p. 29] less definite interdependent status and role relationships to one another and (2) which explicitly or implicitly possesses a set of values or norms of its own regulating the behavior of individual members, at least in matters of consequence to the group.

In order that this definition not be unwieldy, common attitudes, common aspirations and goals are omitted. Such shared attitudes, aspirations, and goals are related to and, in fact, are implicit in the concept of common values or norms of a group. From the point of view of the members within the group, these social units may be referred to as in-groups. Again from the point of view of a member within the group, those social units of which he is not a part psychologically or to which he does not relate himself may be referred to as out-groups. It follows that the term intergroup relations refers to the relations between two or more in-groups and their respective members. Whenever individuals belonging to one in-group interact, collectively or individually, with another group or its members in terms of their group identification, we have an instance of intergroup relations.

From a survey of empirical literature it can be stated that intergroup attitudes and behavior regulated by them are produced in the form of social distances and standardized stereotypes as a consequence of functional relations between in-groups. Once these intergroup attitudes and stereotypes are standardized they take their place in the cultural repertory of the group and in many cases, through the vehicle of language, outlast the very functional relations which were responsible for their rise.

These functional relations between groups and their consequences, rather than the study of the deviate individual, constitute the central problem of intergroup relations. Of course, this does not imply a denial of various unique influences in the life history of the individual member (such as personal frustrations, special hardships in the family or other situations). Such personal influences in the life history may have a great deal to do with the individual becoming a non-conformist or deviate in terms of the prevailing scales of attitudes of his group. But such unique or personal influences do not determine the scale themselves. Rather they come in an important way to determine the particular place the individual will occupy within these scales or, in the case of non-conformists or deviates, the acceptance of a position outside of the scale.

[p. 30] Considerations determining the approach, plan, and hypotheses: At present there are various and conflicting psychological approaches to the study of intergroup relations. It seems that no amount of argument on an abstract level will prove the advantage of one approach over another. Certain of the empirical considerations which led to the approach to be used in this study will be mentioned briefly in the pages that follow.

The consequential intergroup behavior of individuals (largely revealing friction and tension at the present time) is in terms of their membership in their respective groups. Intergroup behavior of an individual which deviates considerably from the prevailing trends is not a typical case. If the individual's intergroup behavior is too much out of line with the prevailing trend of his respective groups, it is brushed aside or dealt with as deviate by other members.

One approach to intergroup relations is through the study of leadership. Even though leadership undeniably contributes great weight in the shaping of intergroup relations, concentration of research on leadership alone leaves out functional ties to which leadership itself is organically related. Such an approach is in contradiction to the main trend in leadership studies today. These studies are increasingly pointing to the necessity of considering leadership in terms of the whole state of reciprocities within the group.

Another approach in intergroup problems concentrates efforts on in-group relations. Empirical data seem to indicate that the nature of intergroup relations need not be in line with the prevailing character of in-group relations. This approach, which concentrates on improving in-group relations in order to improve intergroup relations, ignores the demonstrated consequences attributable only to the particular character of the interaction process between groups. Solidarity within the group need not be transferred to solidarity between groups, and in fact may contribute to sharpened delineations between groups with all the attendant by-products.

In short, the conception of the present study differs markedly from existing theories which posit one factor or a few factors as sole or primary determinants of the course of intergroup relations. (1) Inherent superiority or inferiority of human groups, [p.31] (2) national character ("war-like people, " "peaceful people"), (3) deep-seated innate instincts of aggression or destruction, (4) frustrations suffered individually, (5) direct economic gain, (6) the character of leadership - - - are variously advanced as sole or primary determinants of intergroup relations. Each of these theories still has its strong supporters.

The present approach does not deny that some such factors may, singly or in combination, be operative as factors in determining the course of intergroup relations (excepting specifically the first and third listed above). "National character," frustrations suffered in common and experienced as a common issue, certain economic gains which become shared goals, or the particular character of the group's leadership may variously become the more weighty determinant of intergroup relations under a given set of circumstances (Note 2).

But conflicting evidence leads us to assert that the weighty factor determining intergroup relations will not be the same for all circumstances. For example, in settled times when in-groups are in a state of greater stability, national character as formed at the time and the existing scale of social distance (or prejudice) will regulate, on the whole, the particular pattern of intergroup relations. But in times of greater flux or crises (due to the impact of technological, cultural, socioeconomic and even military events) some other factor or factors take the upperhand.

One primary point of departure in our approach then, is the principle that various factors are functionally interrelated. In this respect the present approach is opposed to theories which make this or that factor sovereign in its own right; it attempts rather to ascertain the relative weights of all the possible factors that may be operative at the time.

The functional relatedness of various factors leads us to the cardinal psychological principle of our whole plan of study:

In the study of (intra- and inter-) group relations the relative contribution of given external stimulus factors and internal factors pertaining to participating individuals (hunger, sex, status desire, complexes, etc.) have to be analyzed within the framework of the ongoing interaction process among the members in question.

[p. 32] The relative contribution of an external stimulus factor, or an attitude, a drive, or other internal factors, cannot be simply extrapolated from individual situations to interaction situations. Interaction processes are not voids. Whatever drives, motives, or attitudes the individual brings into the situation operate as deflected, modified, and, at times, transformed in the interaction process among the several individuals (who stand or come to stand in time in definite role relations toward one another).

The application of this cardinal principle to the study of group relations is derived from more basic findings in the field of judgment and perception. The judgment of a given weight is not determined solely by its absolute value, but also, within limits, by its relative position in the scale of which it is a part and by the presence or absence of other functionally related anchoring stimuli with values within and without the scale. Likewise placement of attitudinal items on a scale with categories specified by the experimenter or with categories chosen by the subject is determined not only by whatever intrinsic value these items may have when considered singly, but also by their relation to one another and their relation to the stand that the individual has taken on the issue.

Following the implications of this general psychological principle, it may be plausible to state that behavior revealing discriminations, perceptions, evaluations of individuals participating in the interaction process as group members will be determined not only by whatever motivational components and personality characteristics each member brings with him, not only by the properties of stimulus conditions specified in an unrelated way, but as influenced, modified, and even transformed interdependently by these and the special properties of the interaction process, in which a developing or established state of reciprocities (roles, statuses) plays no small part. The developing state of reciprocities between individual members can be measured in various differentiated dimensions (e. g., status, popularity, initiative, etc.).

In short, one cannot directly extrapolate from the knowledge of stimulus conditions alone, or motivational components of participating individuals alone, but one has to study behavior in the framework of the actual interaction process with its [p. 33] developing reciprocities.

Carrying this line of conceptualization to the area of inter-group relations, one should start with the recognition that the area of interaction between groups cannot be directly extrapolated from the nature of relations within groups or prevailing practices within them, even though a careful analysis of intra-group relations is an essential prerequisite in any approach to intergroup relations. Numerous instances of intergroup relations in which the pattern (positive or negative) is different from the pattern prevailing within the respective in-groups might be mentioned.

The interaction process between groups and its consequences have to be studied in their own right in addition to studying relations prevailing within the in-groups in question.

The conceptual orientation outlined above determined:

1. the formulation of specific hypotheses,

2. the design of the experiment through 3 successive stages,

3. the choice of criteria in selection of subjects and the choice of setting that will not permit the direct intrusion of influences other than those experimentally introduced,

4. the special considerations related to observational and experimental techniques to be used in the collection of data, and the specific roles staff members will occupy.

Methodological Considerations

The problem of intergroup relations has not been made the domain of experimentation. Literally, there are only a few studies specifically designed to experiment on intergroup relations. Therefore, the present study undertakes to define main functional relations involved in the problem and to point to some unmistakable trends on the basis of data obtained.

In experimental study of intergroup relations it is necessary that various conditions between groups be experimentally [p. 34] introduced and manipulated, the nature of these conditions being defined, and the consequences of their variation predicted.

Recent research in both psychology and sociology and indications of attempts by practitioners in this area are making it increasingly evident that theoretical and practical problems of group relations, including attitudes and change of attitudes regulating behavior of individuals within their respective groups (in-groups) and with out-groups, have to be studied in terms of the interaction processes within and between appropriate group settings.

The usual practice in attitude studies has been to study the effects of already existing attitudes, or to measure attitudes that are already formed. When carried out apart from particular group settings, the study of motives (drives), frustrations, past experience, etc., (which are certainly operative in the formation, functioning, and change of social attitudes pertaining to group relations) has given us items of information whose validity has not been proven in actual issues of group relations. The attempt in this study is to trace the formation, functioning, and change of attitudes towards one's own group, toward its various members, and towards out-groups and their members within the setting of group interaction processes, and as consequences thereof.

In-groups themselves and the attitudes of members towards one another and toward the in-group as a whole are to be experimentally produced. In other words, group attitudes (both intra- and intergroup) will start from scratch and will be produced as a consequence of interaction processes in intra- and intergroup relations through the introduction of specified experimental conditions. The methodological gain from the experimental production of attitudes whose effects or change are to be studied or measured needs no elaboration.

Considerations such as those briefly mentioned above determine the approach taken, the specific hypotheses formulated, and the design of the experiment in 3 successive stages in the present 1954 study. Likewise they determine the choice of particular methods and cautions to be pursued in the collection of data.

[p. 35] To approximate as much as possible the natural process of spontaneous group formation, of in-group and out-group delineation with its consequences so abundantly reported in the literature on small groups, subjects will be kept unaware of the fact that this is an experiment on intergroup relations. (See Subject Selection in the next chapter for information given to teachers and parents concerning the experiment.)

Data concerning in-group formation (Stage 1) and inter-group functioning (in Stages 2 and 3) will be obtained through participant observers who are perceived as part and parcel of the situation by the subjects. All of the staff members directly in contact with the subjects will participate in the role of usual camp personnel, or some role not out-of-ordinary in a camp situation. Moreover, the participant observers should not be detected by the subjects while recording observations contrary to the natural functions of their announced roles. The argument that subjects cease to be mindful that their words and other behavior are being observed and recorded is not in harmony with what we have learned concerning the structuring of perception. The presence of a personage ever observing, ever recording our words and deeds in a situation in which our status and role concerns are at stake, cannot help coming in as an important anchorage in the framework of the interaction process in question. Candid recordings of conversation and moving pictures taken at choice points without the awareness of the subjects will be valuable in addition to other observational data.

All the goals in the in-group stage and in the negative and positive intergroup stages will be introduced through conditions inherent in immediate situations (such as eating, overnight camping or some activity expressly desired by the subjects), and not in the form of abstract incentives distantly related to the immediate goals of ongoing activities and situations. For example, attainment of food will be introduced, not as a hypothetical problem or discussion situation, but through arranging conditions at a time when group members are getting hungry in a place where no other food is available so that members have to cooperate with one another to prepare available ingredients with facilities in the situation. (After subjects take the initiative along some plan, all necessary help and skill can be extended to carry out their plan more effectively.)

[p. 36] The technique of problem solving, that is, attainment of goals introduced in the manner described above, will not be through methods introduced by the experimenter, such as discussion method or lecture method. One of the guiding principles in the present study is that an actual problem situation faced by group members, as a common goal to be attained or a common deprivation to be taken care of, will necessarily lead to various suggestions, counter-suggestions, proposals and their weighing - - - in short, to discussion by group members.

When the group is faced with a situation involving common goals or deprivations, group activity will arise. This group activity may be in the form of suggestions from various members, leading to discussion, decisions, planning and execution. When group activity in relation to common goals is initiated, effective ways of dealing with the situation may involve group discussion, or analysis of the situation by a member who is conceded to know more about the topic than others, or (especially if the group is well-structured or the situation and available means sufficiently compelling) more direct action by higher status members or by the whole group may be taken. Those familiar with sociological findings on informally organized small groups, know well that such groups, facing plans to be executed or problems to be solved, do discuss, do plan, and do execute plans. In this interaction process involving an actual problem or goal situation perceived as common to the group, discussion of alternatives has its place, at times exhortations (lectures) and skills of particular members in verbal and non-verbal ways have their places. The various activities involved in the interaction process, viz., discussion, exhortation, planning, and execution, may be carried out in sequence, or in rapid succession, or the common decision may be implicit in the action itself, if the goal and means stand out clearly. The sequence followed and methods used will be determined in part by the nature of the problem, in part by the particular character of group structure (in which leadership, as part and parcel of the hierarchical structure of the group, plays no small part), in part by the particular set of values or norms prevailing in the group, and also by the character and norms of the general sociocultural setting of which the group in question is a part.

Emphasis on studying the interaction process in a natural setting, while approximating experimental control and techniques, does not eliminate the possibility of checking the validity of observed [p. 37] trends by precise laboratory techniques at "choice" points. If there is any validity in the recent generalizations concerning perceptual and judgmental variations ("distortions") as a function of attitude or motive, relevant perceptual or judgmental tasks of the type used in the laboratory can very well be introduced at a few choice points. The stimulus materials used in these experimental units are of an indirect and unstructured type not involving direct questions about developing group attitudes. The procedures are perceived by the S' s as part of the camp activities, and not as experiments which clutter the flow of their interaction process.

In fact, on the methodological side, the plan of the study aims at two additional objectives:

The first involves the introduction of laboratory-type experimental procedures as supplements for obtaining data concerning the effects of group interaction with the aim of establishing short-cut methods for tapping behavioral trends to supplant laborious, gross behavior observations (see experimental units at the end of Stages 1, 2, 3 later in this chapter).

The second is to secure personal data (e. g., intelligence, personal characteristics) through available testing procedures which can be related to various dimensions of behavior manifested in the interaction process in various stages. This aspect is not to be carried out in the present 1954 study owing to lack of facilities. As this line of research develops it can be brought to the foreground as one of the important problems.


Subjects will be 24 twelve-year-old boys from established Protestant families of middle-class socioeconomic standing, who are normal (no "problem" cases), who have not experienced any unusual degrees of frustration in their homes or other situations, who are not school or social failures (no isolates), and who have a similar educational level. (See section on subject selection, Chapter 3.)

[p. 38] A nominal fee of $25 or less will be charged. This nominal fee will give us the privilege of asking parents not to visit their boys during the experiment. Staff members will have no visitors.

Three Successive Stages and the Hypotheses

The hypotheses will be listed under their appropriate stages, since the account of these stages specifies in outline the conditions under which the particular hypothesis holds true.

Our general hypothesis in regard to intergroup relations (which is the main concern of the present study) is that intergroup attitudes and behavior are determined primarily by the nature of functional relations between groups in question (and not primarily by the pattern of relations and attitudes prevailing within groups themselves, nor primarily by the deviate or neurotic behavior of particular individual members who have suffered more than the usual degree of frustration in their personal life histories).

Both the 1949 and 1953 experiments started with a stage of spontaneous friendship choices (Note 3). This stage, to which the first days of the experiments were devoted, was introduced to rule out the possibility of attributing the experimental in-group formation to personal affinities that subjects develop for one another. This alternative explanation was ruled out on the basis of reversals of friendship choices away from interpersonal preferences and in the direction of the experimentally produced in-groups in our 1949 and 1953 experiments. The stage of interpersonal friendship choices, therefore, is eliminated from this 1954 undertaking, and the study is designed in 3 stages instead of the more complex 4 stage design of the 1953 attempt.

In the two previous studies, the assignment of the subjects to two experimental groups was done towards the end of the first stage, that of spontaneous friendship choices. The basis for this division was not only the splitting of spontaneous friendship choices but also matching the groups as much as possible in terms of observed skills, athletic ability, etc., as well as in terms of data collected during the period of subject selection. Since dropping the period of spontaneous friendship choices [p. 39] eliminates the possibility of actual observation at the camp prior to assignment of subjects to two groups, we have to rely exclusively on the data from the observations at schools, teacher evaluations, school ratings, and data from interviews in actual home situations during the subject selection period. Utmost care will be exhibited by staff members to obtain two groups matched in as many dimensions as possible relevant to the activities that will be introduced, especially those to be utilized in the intergroup stages.

Stage 1: (5-6 days) Experimental in-group formation

The chief aim of Stage 1 is the production of in-groups through manipulation of conditions in which interaction takes place. This step is necessary in order that intergroup relations may be studied between in-groups, whose formation and functioning can be specified.

With the aim of specifying the formation and structure of the experimental in-groups, the two groups will be kept apart and their activities separated as much as possible, especially during the first days of this stage. Otherwise any functional contacts between the two groups would certainly have some consequence both for in-group formation and for the later stages of intergroup relations.

Conditions conducive to bringing about in-group formation (with hierarchical statuses and roles which will be clear-cut at the upper and bottom ends of the hierarchy) will consist of a series of common and interdependent activities prompted by goals integral to the actual situations in which the subjects find themselves (e. g., getting a meal when they are hungry or water when thirsty). The attainment of the goal will necessarily require cooperation and reciprocal relations. As a result, the initial discussion and the activities that follow will be real to the subjects, unlike discussion topics introduced or hinted by experimenters (or leaders) which are not immediately inherent in the situation. (Topics used in many discussion group studies are often conducive to individual 'shining' in verbal skills or debating.)

The effects of the series of activities conducive to group formation will be studied in terms of:

[p. 40] (a) behavioral observations - - - verbal and non-verbal,

(b) ratings of emerging relationships by the participant observers (looking from outside),

(c) sociometric ratings in several relevant dimensions (looking from inside),

(d) experimental indices in terms of judgmental and perceptual variations reflecting the reciprocal role and status attitudes that emerge among group members toward each other. Before these indices are obtained, we can make predictions of the direction and degree of such variations.

As emphasized in the introductory theoretical and methodological considerations, the focal point is to maintain the natural flow of the interaction process within groups and, later, between groups under conditions which appear life-like to the subjects. Any observational procedure, or laboratory-type experiment or repetition of sociometric tapping which clutters the flow of interaction is antithetical to the main conception of this study. Therefore, only one judgmental experiment will be used during the stage of in-group formation. It is perfectly feasible to design an experiment primarily to study in-group formation and related problems and to devote the entire time to it. In that case, of course, it would be possible to introduce various experiments studying the progressive development of in-group structure and its effects on in-group members.

Hypothesis 1 (Stage 1)
A definite group structure consisting of differentiated status positions and reciprocal roles will be produced when a number of individuals (without previously established interpersonal relations) interact with one another under conditions (a) which situationally embody goals that have common appeal value to the individuals, and (b) which require interdependent activities for their attainment.

The hypothesis above is formulated on the basis of empirical findings by sociologists like F. Thrasher, Clifford Shaw, and William Whyte. These and other authors stated generalizations [p. 41] in line with it. Our findings in this respect will serve as experimental verification. This hypothesis was supported by the results of both our 1949 and 1953 experiments cited previously.

The hypothesis will be considered to be verified if the individuals can be placed on a pyramidal hierarchy (the leader being at the apex) on the basis of (a) observational data, (b) status ratings of subjects in the respective groups by participant observers, and (c) sociometric indices.

(a, b) Observational data: The ratings of emerging status relations will be a part of the daily observational reports of the participant observers. Thus, the ratings will serve as a day-to-day index of the trend from mere togetherness situations (in which unstable, transitory differential effects are manifested) to various degrees of stabilization of established reciprocities which constitute the group structure at a given time. When three consecutive ratings (especially of positions at the top and bottom of the status hierarchy) by participant observers of their respective groups show a high degree of correspondence, we can say a definite in-group structure has formed. At this point the similar ratings independently made by junior counselors and other staff members who have had sufficient contact with the groups may be used as further checks. At that time, sociometric ratings and the judgmental experiment with the target board will be introduced (see c and d below).

Observational data consisting of the frequencies of suggestions for activities made by various members and the proportion of acceptance and observance of these suggestions will be obtained. The latter measure might be termed the initiative ratio.

Other observational data along various dimensions will be desirable. Observers will make their ratings of group structure along these dimensions.

Frequency of suggestions (for engaging in this or that activity, etc.) addressed to various group members is one such dimension. It is a plausible hunch that the number of suggestions for group activities which are received by various members will be proportional to the status each achieves in the group. When members are placed according to the frequencies of suggestions addressed to them, we may be getting a placement of members [p. 42] pyramidal in shape very much like the one mentioned above. It is plausible to state this tendency in the form of an auxiliary hypothesis:

Hypothesis 1 a (Stage 1)
If a definite group structure develops, it will be reflected in a consistent pattern in directions of communication. The specific pattern in direction of communication will be as follows: The higher the status of a group member the greater the frequency of suggestions (for group activities) addressed to him.

It seems feasible to represent the pattern in directions of communication visually in the form of a chart. We should think that through the course of a study such as this, variations in such charts would be obtained. The chart of directions of communication at a given time will correspond closely to the chart of initiative ratios and the pattern of judgmental variations in the way of overestimations and underestimations of performance. A suggestion for activities coming from any member may be kicked around among the group. Even if it is not initially addressed to the top position (leader), but to middle position members or lieutenants, it will be kicked around until a nod expressing approval or, at least, no disapproval from the top position member (leader) is perceived.

(c) Sociometric data (Note 4): Sociometric data obtained from the subjects themselves along various dimensions (popularity, initiative, degree of service for the well being of the group, etc.) will be significant indices in terms of relations perceived by the group members themselves. The sociometric indices (looking from within) should give very much the same trend as those represented in the ratings, frequencies, and charts obtained through observational data mentioned above. We shall consider this hypothesis verified only in cases in which there is a high degree of correspondence between (a) observational, (b) sociometric, and (c) experimental indices.

(d) Experimental indices to be obtained through laboratory-type judgmental experiments introduced at this point: Recent findings which indicate the feasibility of measuring attitudes and other motivational components through perceptual and judgmental [p. 43] indices suggest that the reciprocities developing among members of a group as status and role relations will be reflected in the differential ways group members perceive and judge one another. One index of these differential judgments as a function of relative statuses or roles will be based on the tendency to expect higher or lower performance in activities engaged in by members occupying various status positions. (Differential expectations proportional to status positions occupied.) Relative over- and underestimates of performance in experimentally introduced tasks may be utilized to measure indirectly the status hierarchy of group members. If this proves to be the case, such experimental indices can be developed to check the validity of gross observational findings, and eventually to supplant them. Such an attempt will be made in this study with the following hypotheses:

If Hypothesis 1 holds, it can be predicted that:

Hypothesis 1 b (Stage 1)
(a) The higher the status of a member in the group, the greater his tendency to overestimate his performance in an activity the group engages in.
(b) The higher the status of a member in the group, the greater the tendency of other group members to overestimate his performance.
(c) The lower the status of a member in the group, the less his tendency to overestimate his performance in an activity the group engages in.
(d) The lower the status of a member in the group, the less the tendency of other members to overestimate his performance, even to the point of underestimating it.

This psychological tendency was demonstrated in established informal cliques in an experiment at the University of Oklahoma carried out as one unit of a research project supported by the Office of Naval Research (Note 5). However, in that study indices used were estimates of future performance, whereas in the 1953 study mentioned above direct judgments of performance were used (Note 6). The experiment to be introduced here follows the [p. 44] procedures used in 1953 utilizing direct judgmental indices.

Hypothesis 2 (Stage 1)
When individuals interact under conditions stated in hypothesis 1, concomitant with the formation of group structure, norms will be standardized regulating their behavior in relations with one another and in practices and activities commonly engaged in.

This hypothesis is also based on empirical findings by sociologists and on studies of adolescent cliques, and will be experimentally verified in this study.

The group norms which are standardized will be expressed as attitudes and conforming behavior of individual members. The production of a set of standards or norms can be verified by observing the reaction of group members to deviations from it. When there is a norm regulating the interpersonal relations of in-group members in terms of their established statuses and roles or regulating behavior in some practice or activity, it can be predicted that behavior by a group member deviating from the norm will arouse corrective reactions from other group members. (This applies also to norms regulating behavior toward out-groups which will become prominent in Stage 2.) The corrective measures or sanctions may range from actual punishment meted out to the deviate through "silent treatment", scorn, ridicule, criticism, expressions of disapproval, to amusement, varying according to the importance of the norm violated, the degree of deviation, and the status of the individual. Facts relating to reactions to deviation are reported by sociologists and also in the experiment by Schachter and others.

Stages of Intergroup Relations (2 and 3)

As stated earlier in our definition, intergroup relations refer to interaction between two or more groups collectively or between their respective members. In our study, intergroup relations refer to interaction between the two experimentally produced groups (as formed in Stage 1) and their respective members.

Stages 2 and 3 constitute the main stages of this experiment. All of the previous work in Stage 1 (in-group formation) leads up [p. 45] to them. Stage 2 is the tension or friction phase of intergroup relations. Stage 3 is the integration phase of intergroup relations.

Stage 2: (4-6 days) Intergroup RelationsFriction Phase

Relations between the experimentally produced groups start with a friction phase because the major problem of intergroup relations today is the reduction of existing frictions between various groups. For this reason, the phase of friction is preceding the attempt to reduce tension and to integrate groups into cooperative activities with common goals.

Friction between the two groups will be brought about through the introduction of two sets of conditions:

(a) During this stage the two groups will be brought into contact in a series of competitive activities in the form of a tournament of events which will yield cumulative scores with a reward for each member of the winning team. However, these individual rewards can be obtained only by being a member of the winning group and cannot be won individually. In other words, in order to win the award individually the members of each group are to contribute their individual bits to the winning of the team.

(b) Introduction of situations which will be perceived by one group as frustrating and which will be perceived as caused by the other group, and not by the camp administration. This was tried with positive results in 1949. The situations will embody goals which can be attained by one group and not by the other, in such a way that both groups will perceive the other as an obstacle in its way to attaining the goal.

In line with the methodological point that the subjects should not perceive this as an experiment on intergroup relations, conditions set up in Stage 2 and 3 conducive to group frustration and friction, or to integration as the case may be, must be designed in such a way that the subjects cannot assign the source of these conditions to the staff. They must be planned in a way such that it is not possible for group members to ascertain by checking verbally with the members of the other group that someone (the staff) has been manipulating conditions.

[p. 46] Our general hypothesis is that subjects who did not have appreciable contact with members of the opposite group during Stage 1 will develop negative attitudes verging on enmity towards the out-group which is perceived to be in their way for the attainment of goals shared in common within their group. Negative intergroup attitudes, such as prejudice, develop whenever any out-group is perceived as frustrating or as an obstacle. (In short, norms regulating behavior toward out-groups, like social distance norms, are standardized group products.) Negative attitudes toward out-groups will be generated situationally under these conditions and will tend to persist even though the individual members in question have not undergone any special degree of frustration in their life histories. Applying this general statement to the particular case of intergroup relations in this study, our specific hypotheses will be:

Hypothesis 1 (Stage 2)
In the course of competition and frustrating relations between two groups, unfavorable stereotypes will come into use in relation to the out-group and its members and will be standardized in time, placing the out-group at a certain social distance (proportional to the degree of negative relations between groups).

Evidence for the rise of stereotypes will be obtained by recording derogatory adjectives and phrases that are used to refer to the out-group. The specific competitive and frustrating situations and the activities and verbal utterances relating to out-groups will be noted. If possible, the frequency of references made to out-groups (positive or negative) and of activities undertaken relating to out-groups, both in intra- and intergroup situations, should be recorded. Such conditions, verbal utterances and activities in relation to the out-group constitute the steps on the basis of which stereotypes are built. In time all members of the out-group will be perceived in terms of the generalizations encompassed in the standardized stereotypes. This aspect of our study constitutes a contribution to the formation of norms of social distance (prejudice) which prevail in social groups. The tendency toward stereotype formation was noted in our 1949 study and verified in a more systematic way in R. Avigdor's doctoral thesis (Note 7).

[p. 47] In addition to observational data, the rise of stereotypes will be tapped through two experimental units introduced at this stage:

1. Experimental indices reflecting the reciprocal intergroup evaluations in terms of stereotype ratings (testing Hypothesis 1, Stage 2). This is essentially the technique used by Avigdor.

2. Experimental indices revealing overestimation of performance of in-group members and underestimation of performance of out-group members. In this unit a bean-toss contest between the 2 groups will be introduced. The contest consists of rapid gathering of as many beans as possible by all members of each group within a brief time period. After the contest, beans presumably picked up by each member will be projected on a screen, identifying with each projection the individual who presumably collected them. Actually the same number of items will be projected each time in the same confined area, the items being spread in somewhat different arrangements. Estimates of the number of beans will reflect overestimation of the performance of in-group members and underestimation of the performance of out-group members. This tendency can be stated in the form of specific hypotheses:

Hypothesis la (Stage 2)
In-group members will tend to overestimate the number of items purportedly obtained by in-group members and underestimate the number of items attributed to out-group members.
Hypothesis 1b (Stage 2)
The degree of this tendency manifested will vary according to the status (low or high) of in-group and out-group members in question.

The feasibility of the two experimental units, viz., assessment of differential judgments of performance of members of in-groups and out-groups and differential rating of qualities in so many relevant dimensions, has already been clearly established in an experimental study carried out in our project (Note 8).

[p. 48] These data from assessment techniques as well as sociometric choices will be obtained again at the end of Stage 3, and will serve as an index of decrease of unfavorable attitudes toward out-groups in that stage.

Hypothesis 2 (Stage 2)
The course of relations between two groups which are in a state of competition and frustration will tend to produce an increase in in-group solidarity.

Increased group solidarity will be revealed in the expressions of glorification of the in-group and of "feats" of members, especially those of high standing. Increased encouragement of efforts of in-group members in a way not manifested during the period when the in-group was not in contact with the out-group will be another indication. Additional behavioral data in support of this hypothesis will be derived from the experimental units described above.

Hypothesis 3 (Stage 2)
Functional relations between groups which are of consequence to the groups in question will tend to bring about changes in the pattern of relations within the in-groups involved.

This hypothesis should hold true for both positive and negative intergroup relations of consequence. (See also last paragraph of this chapter.) The changes in in-group relations can be measured in terms of popularity and status of in-group members in various respects. The degree of consequence of intergroup relations for the group in question can be measured (a) by the frequency of references to the out-group, and (b) by the amount of planning and activity engaged in within the in-groups in relation to the out-groups.

One way of testing this hypothesis is through special attention to ratings of status relations within the groups by participant observers. These ratings should be continued throughout the intergroup phases with the expectation that some important changes in the functional relations between groups will produce consequential changes in the in-group structure as stabilized at the end of [p. 49] Stage 1. The participant observers' ratings will be checked with independent ratings by other observers in contact with the groups, thus contributing to the reliability of the data.

The hypothesis is predicted for both parties (winning and losing groups in our study). In the case of the group suffering defeat the impact of intergroup relations may be to the extent of disorganization of the in-group pattern, which will be marked by shifts in status positions occupied by various members.

Related to the above hypothesis is a subsidiary one concerning the functioning of low status members of the two contending groups. This has theoretical implications in view of present-day controversies. It can be stated as follows:

Hypothesis 4 (Stage 2)
Low status members will tend to exert greater efforts which will be revealed in more intense forms of overt aggression and verbal expressions against the out-group as a means of improving their status within the in-group (Note 9).

An empirical test of this subsidiary hypothesis will be found in observation and comparison of the hostile and aggressive reactions of low status members toward the out-group (a) when reacting in the presence of in-group members high in status and (b) when reacting when high status members of their in-group are not in the immediate vicinity.

Stage 3 (6-7 days) Intergroup Relations: Integration Phase

This stage constitutes the crucial and novel aspect of this study. Deliberately the attempt to bring about cooperation between groups follows a stage of friction produced between them experimentally. This should be the attempt in studies aiming at reduction of group tensions. Production of harmony between groups which are not in a state of tension does not present much of a problem in terms of intergroup events today. There are various possibilities or alternatives for the study of reducing intergroup tensions. One alternative could be called [p. 50] the "common enemy" approach. Empirical evidence and a tryout of this measure as an expedient manner of reducing post-experimental hostility in 1949 indicates that this measure can be effectively used. But it implies conflict between larger group units.

Another alternative would be to arrange a series of events in which achievement of individuals can be made supreme. But this would simply achieve disruption of the in-groups. In terms of actual happenings in intergroup events, the use of this measure in an experimental study would be unrealistic and would have few if any realistic implications for the reduction of intergroup tensions. As noted earlier, actual intergroup tensions take place either collectively between group units or between individual members of the in-groups reacting in terms of their group identifications.

A third alternative would be through leadership techniques. With appropriate manipulation this measure can be made effective. But in actual groups, intrusion of an outside person as a leader is not a welcome one. In actual groups, leaders, too, are part of the group structure, and they have to function within certain bounds in whatever initiative they take. For this reason, manipulation of conditions through leaders who are not part and parcel of the groups in question has little implication for the state of intergroup relations that actually exist.

Such considerations led to the choice of the alternative to be used in this study. The main feature of the alternative chosen is the introduction of superordinate goals which are integral to the situation and which cannot be ignored by the groups in question. The main criteria in the choice of procedures to be introduced in this integration stage will be that goals of sufficient strength to the groups in question be superordinate, in the sense that the resources and energies of any single group will be inadequate for the attainment of the goal, thus creating a state of real and/or perceived interdependence. Situations will be planned and listed before the experiment in which such a state of interdependence inheres (a) keeping a sufficient level of motivation that members of groups are directed toward the superordinate goals, and (b) introducing a series of stimulus conditions which will make the facing of the superordinate goals and the modes of their attainment compelling.

[p. 51] The superordinate goals will not be introduced abruptly right after this stage starts. Initially some contact situations will be introduced. At these occasions the groups will have to be in close physical proximity under conditions in which expression of their hostility toward one another will not be very appropriate. Of course, mere get-togethers or contact will not materially help reduce the friction. The aim of this early period is to create the possibility of communication between members of the two respective groups. For example, the improvised birthday of an outsider (preferably a local personage not related to the subjects positively or negatively in an appreciable way) to which both parties are invited would be an example of such an occasion. The early phase of Stage 3 will thus consist of occasions that will give the two groups opportunity for contact or communication.

Hypothesis 1 (Stage 3)
It is predicted that the contact phase in itself will not produce marked decrease in the existing state of tension between groups.

The persistence of tension will be revealed in reactions showing resistance to cooperation with the out-group, in spite of contact, and persistence of negative stereotypes. If this prediction holds, it will eliminate the alternative hypothesis that contact in itself will bring about reduction of tensions.

After a series of contact situations, a series of superordinate goals will be introduced - - - goals which cannot help having appeal value to the members of both groups. The following are examples of superordinate goals inherent in a situation for members of both groups concerned, the attainment of which is dependent on collaboration on the part of both groups: (a) A project related to some improvement of the water tank on the hill and the pump near the reservoir, since the tank provides water for members of both groups. (b) Creating a situation of interdependence in a joint overnight camp in which members of both groups will need mutual aid for their meal and sleeping facilities. Probably the increased social suggestibility in new situations or situations of uncertainty may be utilized to enhance the effects of the conditions of interdependence. (c) Other examples already suggested by staff members are the possibilities of [p. 52] utilizing the swimming pool or the truck (which brings their provisions) e. g., having the truck in a rut deep enough to require the combined efforts of both groups to free it.

Hypothesis 2 (Stage 3)
When groups in a state of friction are brought into contact under conditions embodying superordinate goals, the attainment of which is compelling but which cannot be achieved by the efforts of one group alone, they will tend to cooperate toward the common goal.
Hypothesis 2a (Stage 3)
Cooperation between groups necessitated by a series of such situations embodying superordinate goals will have a cumulative effect in the direction of reduction of existing tensions between groups.

Even though the groups are brought into situations which permit communication between them and then situations requiring their collaboration toward a common goal, the effects of friction produced in Stage 2 will tend to persist, along with the by-products of this friction. One of the indices important in the study of the changes in this stage, in addition to observational data giving a gross account, will be the decrease in expressions of resistance to collaboration with the out-group, which will be strong at first.

Observational data will be collected in the mess hall and other situations involving choices (of seating arrangements, etc.) to check the extent of intermingling among members of the two groups.

Another way of gaining evidence of reduced tension will be a decrease in the actual use of derogatory terms and expressions toward the out-group. After the series of superordinate goal situations have exerted a cumulative effect, the rating of relevant stereotypes will be repeated. The "bean toss" experiment or a similar procedure will be applied here if it can be carried out without spoiling the flow of the interaction process.

Toward the end of Stage 3 sociometric choices will be [p. 53] obtained again. It is predicted that in comparison to those obtained at the end of Stage 2 there will be a marked increase in choices of out-group members.

As predicted in Hypothesis 3 (Stage 2), intergroup relations developing in interaction directed toward superordinate goals will also tend to bring about changes in in-group relations. As in the case of the friction phase (Stage 2), proportional to the demands for intergroup cooperation, there may be changes in in-group structure. A special note should be made here of those who are contributing more to intergroup cooperation, e.g., lieutenants who exhibit strivings toward still higher position in the in-group structure and those in marginal roles. Effective cooperation will be brought about when high status members or members on the move to higher status through activities in the area of intergroup relations take a hand in (a) initiating in-group moves toward cooperation and (b) in participating in intergroup communication related to superordinate goals.


[1] This chapter is an outline of the study prepared and distributed prior to the experiment in mimeographed form to staff members of the study and a number of colleagues interested in this problem area throughout the country. Since this paper gave the high points of the theoretical rationale and the blueprint to guide the actual experiment, it is presented here in substantially the same form, including the use of the future tense in referring to various procedures.


1. Leads derived from the field work of sociologists concerning relations of small groups are summarized in M. Sherif and H. Cantril, The Psychology of Ego-involvements (New York: Wiley, 1947), Chapter 10; M. Sherif, An Outline of Social Psychology (New York: Harper, 1948), Chapters 5-7; M. Sherif and C. W. Sherif, Groups in Harmony and Tension (New York: Harper, 1953), especially Chapter 8.

Psychological principles derived from the work of experimental psychologists and utilized in our previous work as well as the present undertaking are summarized in M. Sherif, The Psychology of Social Norms (New York: Harper, 1936), Chapter 3, "The Frame of Reference in Psychological Phenomena"; M. Sherif, An Outline of Social Psychology (1948), especially Chapters 4, 7, and 9; M. Sherif and C. W. Sherif, op. cit., especially Chapter 6.

2. The present approach is elaborated more fully in Groups in Harmony and Tension (op. cit.), especially pp. 13-15, 146-156, 182-190, 296-307.

3. A brief summary of our 1949 experiment was presented in Social Psychology at the Crossroads (J. Rohrer and M. Sherif, edits., New York: Harper, 1951), Chapter 17. A fuller of that experiment is given in M. Sherif and C. W. Sherif (1953), Chapters 9 and 10. A short report of the completed part of the 1953 experiment is given in M. Sherif, B. J. White, and O. J. Harvey in Status in experimentally produced groups. American Journal of Sociology, 1955, 60, 370-379.

4. It was thought that obtaining sociometric indices 3 times (once at the end of each 3 stages), asking the same or similar questions within a 3-week period might appear repetitious (if not suspicious) to the subjects. Therefore, in line with our main concern not to clutter the natural flow of the interaction process, it was decided prior to the actual start of the experiment to restrict sociometric choices to the intergroup stages (2 and 3) and forego them at the end of the in-group stage (1).

5. O. J. Harvey, An experimental approach to the study of status relations in informal groups. American Sociological Review, 1953, 18, 357-367.

6. M. Sherif, B. J. White, O. J. Harvey, op. cit.

7. R. Avigdor, The Development of Stereotypes as a Result of Group Interaction (on file in the New York University Library, 1951). For a brief summary, see M. Sherif and C. W. Sherif (1953), op. cit., 290-295.

8. O. J. Harvey, An Experimental Investigation of Negative and Positive Relationships between Small Informal Groups Through Judgmental Indices. Doctorate Dissertation, University of Oklahoma, 1954; also presented in M. Sherif, et. al., Theoretical and Experimental Studies in Interpersonal and Group Relations, Current Papers in Social Psychology, University of Oklahoma, 1954, Chapter 4, Sociometry, 1956, 19, 201-209.

9. This hypothesis does not imply that high status members will not initiate and actively participate in intergroup conflict. In line with one of the major tenets of Groups in Harmony and Tension (op. cit.), intergroup behavior in conflict or cooperation consists mainly in participation in the intergroup trends of one's group. A line of activity in positive or negative intergroup relations will be ineffective unless high status members either take a lead, join or assent to the developing intergroup trend.  If they stay in the way o fan unmistakable trend in intergroup relations or deviate from it markedly, the consequence will be sinking in the group hierarchy (See Hypothesis 3, Stage 2).