Research on Intergroup Relations
Muzafer Sherif and Carolyn W. Sherif 
The topic assigned for this paper—Research on Intergroup Relations —entails staggering tasks far beyond the scope and competence of one person, or even one discipline. Problems of intergroup relations are undoubtedly the most crucial, the most fateful of all problems in human relations today, especially with the ominous shadow of war hanging over the human race. The increasing research activity by social psychologists, political scientists, sociologists and other social scientists reflects the impact of the problems in various spheres of life. The following lines of work which can be classified under intergroup relations are illustrative of this thriving activity:
Studies on attitudes, stereotypes and social distance, and various methods of changing them (for example, through exchange of students and professors, person-to-person contacts, conference methods, communication, and so on).
Studies related to psychological theorizing, both by psychologists and nonpsychologists, on human impulses toward aggression, cooperation, dominance, submission; and on possible consequences of frustration, displaced aggression, and certain child-rearing practices for intergroup hostility and the development of aggressive attitudes.
Studies of social, legal, political and economic aspects of dis-
( 154) -crimination, with particular reference to issues raised by segregation practices in various countries.
Studies of intergroup relations in industry, especially with reference to labor-management problems—some of which are subsumed under the "human relations" approach in the United States.
Studies of political-party rivalries, party tactics, and pressure groups.
Studies related to problems of emerging nationalism vis-à-vis a declining colonialism.
Studies of "factionalism" in cultures undergoing an accelerated pace of acculturation under the impact of modern technology and military arrangements.
Studies of diplomacy, of war and peace, using such methods as content analysis and "rational" models of decision making.In discussing research on intergroup relations in social psychology, one obvious alternative, and probably the expected one, is to summarize these lines of research and make some evaluative remarks on the underlying theory, the research methodology, results obtained through various approaches and possible practical implications in dealing with intergroup problems. This alternative would require a great deal of space; and in the end, one has another cataloging of theory and research, evaluated in a serial way, with suggestive hints of some incompatible differences in the basic assumptions underlying the various approaches. This kind of survey is already available as represented by Robin Williams' chapter in Review o f Sociology (1957), by the Harding et al., survey in the Lindzey-edited Handbook of Social Psychology (1954), by T. H. Pear's chapter in the UNESCO volume on The Nature of Conflict (1957), and by Sherif and Sherif in Groups in Harmony and Tension (1953).
It is evident that research on intergroup relations is shooting in all directions, that assumptions about the nature of the problem of intergroup relations vary and even conflict, and that methods and techniques for their study are diverse. With current research in this state, it may be more to the point to choose a second alternative for the present discussion—namely, to start with a clear and elemental definition of the problem of intergroup relations, then
(155) derive pointers from this demarcation of the problem area for effective research strategies and practical implications of generalizations based on them.
Delineation of the problem will make it clear that research on intergroup relations entails more than study of the intergroup behavior of individuals—which is the proper level of analysis for social psychology. It also entails problems of institutionalized power relations and complex organizational systems, both formal and informal, which require analysis at their own level by political scientists, sociologists, anthropologists and others.
The discussion will be pursued, therefore, in terms of what appear to us as appropriate tasks of experimental social psychology. The proper tasks of social psychologists in this enormous problem area include, at least, the following:
1. Formulating a clear definition of what constitutes the properties of intergroup behavior and associated positive or negative attitudes in a way that is not essentially at variance with characterizations of intergroup relations at the more complex level of analysis of the various social sciences.
2. Formulating research strategies based on the above delineation of intergroup behavior, translating these into appropriate research designs, and drawing tentative generalizations based on data-collecting operations.
3. Cross-evaluating the generalizations, as well as the research strategies and designs, in relation to generalizations presented by social scientists working on a more comprehensive level of analysis.
In the last part of the paper, certain research directions of theoretical and practical import will be mentioned, which are suggested by the preceding social-psychological analysis.
DEMARCATION OF THE PROBLEM AREA OF INTERGROUP BEHAVIOR
Much confusion in the study of intergroup behavior stems from failure to delineate the nature of intergroup behavior, and failure to conceive of it in the specific context of intergroup relations of which it is a part.
Obviously, intergroup relations refer to states of friendship or hostility, cooperation or competition, alliance or enmity, peace or
( 156) war between two or more groups and their respective members. But, not every friendly or unfriendly act towards other persons is a case of intergroup behavior.
Psychologically, intergroup behavior involves positive or negative attitudes, experiences of frustration or gratification, experiences of scoring a success or suffering failure, feelings of superiority or inferiority, feelings of love or hate. But every experience of frustration or gratification, success or failure, superiority or inferiority, love or hate does not necessarily fall within the domain of intergroup behavior. Only those behaviors and associated attitudes that stem from membership in or aspired membership in a human group are properly eases of intergroup relations.
The term group in the phrase "intergroup behavior" should not be slurred over. If we claim to study relations between groups and their members, the properties of groups and the consequences of membership for individuals have to be brought into the picture. Otherwise, whatever we may be studying, we are not studying intergroup relations.
A group is a delineated social unit with properties which can be measured and which have consequences for the behavior of its members. These include, at least (1) structure or organization—that is, a power dimension as measured by effective initiative of members, and (2) a set of norms regulating behavior of the members in pursuing goals, in relationships with one another and with outgroups and their members—that is, evaluative dimensions which can be assessed in terms of what is upheld in the group and what is treated as deviate. Of course, groups may be part of larger organizations, and they usually are.
The term intergroup relations refers to relations between two or more groups and their respective members. Whenever individuals belonging to one group interact, collectively or individually, with another group or its members in terms of their group identification, we have an instance of intergroup behavior. In other words, intergroup behavior occurs when an individual behaves toward another group in terms of his reference group ties. Therefore, the appropriate frame of reference for studying intergroup behavior is the functional relations between two or more groups and the products of these relations, which may be positive or negative.
If we try to deal with the role of individual factors—such as individual motivations, frustrations or aspirations—without including the consequences of group relatedness in the picture, we find ourselves in severe dilemmas of contradictory evidence. There is good reason to believe that some persons who have been severely frustrated in their life histories may become more intense in their intergroup prejudices and hostilities than some others more fortunate. But it is quite another thing to try to explain intergroup stereotypes, hostility, and conflict between groups on the basis of individual frustrations. As Otto Klineberg noted in his survey of Tensions Affecting International Understanding (1950), acts of violence against other groups, such as lynchings, do not necessarily occur more frequently in societies where the probability of individual frustration is greater.
From the UNESCO studies in India, Gardner Murphy (1953) concluded that being a "good" Hindu or a "good" Muslim at that time implied believing the nasty qualities and practices attributed by one's group to the adversary. The problems of intergroup behavior which are so crucial in human affairs are not primarily the problems of the deviate personality. They are problems of participation by the majority of the membership within the scale of social distances from various out-groups maintained by their group in more stable times, and participation in developing trends between groups in periods of flux and change, such as that more characteristic of the present time.
Nor is intergroup behavior simply and directly a problem of the modal practices applied within the bounds of the in-group to other members. Especially after World War II, we had a rash of explanations of the warlike tendencies of Germany and Japan on the basis of their traditionally authoritarian family structures and educational practices. This line of explanation does not square with certain facts, for example, with the statistics compiled by L. F. Richardson (in Pear, 1950) on the frequency of wars engaged in by world powers from 1850-1941. Britain heads the list with twenty —more than the Japanese (with nine), the Germans (with eight) or the United States (with seven). If there is a direct relationship between authoritarianism of procedures within a country and warlike behavior, these findings would put Britain ahead in both respects. It seems at least reasonable to ask a question which his-
(158) -torians and social scientists can help us answer: "Doesn't having an empire with far-flung interests to be protected and expanded have anything to do with the frequency of wars?"
Levels of Interaction in Intergroup Behavior
The foregoing considerations lead to conclusions concerning the effects of motivational components (for example, aspirations, frustrations, aggressive impulses) at different levels of interaction—individual, group and intergroup interaction (compare Katz, 1960).
We cannot legitimately extrapolate from the individual's motivational urges and frustrations to his experience and behavior in group situations as if interaction processes and reciprocities within a group were a play of shadows. It is equally erroneous to extrapolate from the properties of relations within a group to explain relations between groups, as though the area of interaction between groups consisted of a vacuum. Due consideration must he given to the relationship of the participants to their own groups and the constellation of relations with other groups, some of which may not be manifestly involved in the interaction at the time. The character of relations and norms that prevail within groups does influence their relations with other groups, but the process of interaction between the groups has properties which themselves shape the course of subsequent intergroup encounters. The give and take between groups may be full of conflict or in a state of flow. And this area of conflict, in the case of negative relations, or of flow, in the cases of cooperation or alliance, may reverberate within the groups themselves, producing changes in their internal structure and altering the focus of their chief concerns (Sherif & Sherif, 1953.).
What determines the positive or negative nature of interaction between groups? In large part, it is determined by the reciprocal interests and the goals of the groups involved. The issues at stake are usually interests of considerable consequence to the groups, if they are to play a part in intergroup relations. They may be a real or imagined threat to the safety of the group as a whole, economic interests, a political advantage, a military consideration, prestige, or a number of others. Once a particular issue comes to the foreground as the dominant influence in intergroup relations,
( 159) it may become the limiting factor, the main anchorage in interaction processes between them, coloring their evaluations of all other issues which arise.
GENERALIZATIONS FROM EXPERIMENTS ON INTERGROUP RELATIONS
Now we turn to a summary of research in which strategies were guided by the characterization of the problem as presented here. The research was initiated with the conviction that experimentation with small groups in social psychology should address itself to testing hypotheses based on significant and recurrent phenomena in the larger scene.
The objective of the experiments was to study under controlled conditions the rise and stabilization of unfavorable intergroup stereotypes and social distances between groups in order to specify the conditions in which they could be changed. Three large-scale experiments were carried out in 1949, 1953, and 1954 (Sherif & Sherif, 1953; Sherif, White & Harvey, 1955; Sherif, Harvey, White, Hood, & Sherif, 1961). In the last—the Robbers Cave experiment—the difficult task of reducing intergroup conflict, negative stereotypes, and social distance between groups was effectively undertaken.
The design for the three experiments was a longitudinal one in successive stages. As the first step, autonomous in-groups were produced among unacquainted individuals with homogeneous background characteristics. To have intergroup behavior, with associated attitudes towards the out-group(s), there have to be two or more groups with their distinctive organization and set of values or norms. Groups with anything resembling the properties of actual groups are historical affairs. There is no getting around it. Those investigators who try to do so, by ignoring the organizational and evaluative properties of actual groups and defining any collection of individuals as a "group," may achieve interesting results. These results may bear on some significant problem of social interaction; but they are not studying group interaction merely by calling the collection a group.
Once groups were formed whose natural histories had been
( 160) traced step by step, they were brought into functional contact in conditions which were reciprocally competitive and frustrating. specifically, the success of one group inevitably meant defeat for the other. Finally, conditions conducive to the reduction of conflict between the groups were introduced.
Now, as briefly as possible, the main generalizations drawn from the experiments will be summarized, starting after distinct and autonomous groups had been formed.
1. When groups engaged in reciprocally competitive and frustrating activities, such that the gain of desired goals by one results in loss for the other, unfavorable stereotypes of the out-group and its members come into use. In time, these unfavorable attitudes are standardized in a group, and the out-group is placed at a prejudicial distance, even to the point that group members want nothing whatever to do with members of the other group. This occurred even though members of the two groups did not differ in background and were matched as closely as possible in physical and personal characteristics. From a theoretical point of view, it is perhaps more important that they were socially well adjusted and had not experienced undue frustration in their life histories.
2. With the rise of prejudicial attitudes toward the other group, self-glorifying or self-justifying attitudes toward one's own group are strengthened. The performance of the out-group is deprecated and viewed with suspicion.
3. As implied by the last generalization, the state of intergroup relations has an impact on relations within groups. Here we are reminded of Kenneth Clark's observations in Little Rock (1958) that citizens whose constitutional rights were being threatened stood together with greater solidarity than previously.
But further, both in intergroup conflict and in the trend toward cooperation between groups in the 1954 experiment, the very patterns of relationships within the groups were altered. In 1949, the leader of one group became demoralized as his group faced defeat after defeat and another member took over. In 1954, one leader vacillated in engagements with the adversary and was replaced after a few such episodes. While practices within the groups became more genuinely reciprocal and group decisions involved more of the membership, treatment of the out-group was at its
( 161) worst. These findings suggest that in-group democracy need not necessarily imply democratic attitudes toward out-groups. In-group cooperativeness need not be extended to out-groups; and in fact, depending on the nature of intergroup relations, it may be associated with heightened exclusion of out-groups and accentuated animosity toward them.
Once the groups were in conflict and manifested hostile attitudes, unfavorable stereotypes and the rudiments of social distance, the problem of reducing these could be tackled. The findings may be summarized as follows:
1. Contact between the groups involving close proximity in activities enjoyed by members of each group does not necessarily produce a decrease in the existing intergroup hostility. Behaviorally, such occasions were utilized for exchanges of invectives, and for blaming the out-group for the existing state of affairs. The heavy hand of the past history of conflict between the groups guided their views of each other, and the products of that conflict were further institutionalized. These findings may necessitate some revision of the assumption that mere contacts between groups—without regard to the conditions of contact—through exchange of persons and social get-togethers will by themselves reduce unfavorable intergroup attitudes.
2. Conflict between groups and the products of that conflict were reduced through the introduction of superordinate goals. We define superordinate goals as goals which are compelling for members of two or more groups and cannot be ignored, but which cannot be achieved by the efforts and resources of one group alone. They require the coordinated efforts and resources of all of the groups involved. They are superordinate, rather than merely common goals, in the sense that they must override some of the goals of both groups which are incompatible with them. When groups in conflict come into contact under conditions embodying such goals, they tend to cooperate toward them, as we found in the research.
But conflict and its products were not reduced in one episode. Here too the time dimension has to be brought in as a major variable.
3. Various superordinate goals over a period of time were
( 162) necessary to sustain cooperation between groups, to permit procedures acceptable to both to he established and then transferred from one situation to the next. In the process, friction between groups and unfavorable stereotypes were reduced. The resulting changes in intergroup behavior were dramatic.
In concluding these generalizations, we venture to state some things that we have learned about the reduction of intergroup conflict. It is true that lines of communication between groups must be opened and contacts established before prevailing conflicts can be reduced. But contact and communication without superordinate goals may serve as occasions for recrimination, for accusations and endless reference to the problem of "Who's to blame?"
When contact and communication involve cooperative efforts toward superordinate goals, they are utilized in the direction of reducing conflict in order to attain the goals. Then information about the out-group, which was ignored or rejected when the groups were in conflict, is seen in a new light and the probability of its effectiveness is enormously increased.
Leaders then are in a position to take bolder steps toward bringing about understanding and harmonious relations. We observed one instance of a high status member who tried his hand at reducing conflict by going alone on a peace mission: he was expelled by the other group and considered somewhat a "traitor" by his own group. Such moves are simply not within the latitude o f acceptance for a group in conflict. But when compelling superordinate goals are articulated, the leader can make moves to further cooperation and he will receive support from other members.
In short, various measures which have been suggested for the reduction of intergroup conflict acquire new significance and new effectiveness when they become integral parts of interaction processes between groups oriented toward superordinate goals which have real and compelling value for all groups concerned.
Over a period of time, the interaction of groups toward superordinate goals which have genuine and compelling value for all groups begins to assume an institutional form, in the sense that there are mutually accepted procedures for intergroup activities and mutually understood practices in intergroup behavior.
PROBLEMS OF VALIDITY
The generalizations just presented are based on experimental data summarizing the behavior of young boys in small groups. The essentials of the design and procedures have been replicated by Blake and Mouton (1961 ) with adults in industrial settings, and these major conclusions confirmed.
However, it is necessary to raise the much neglected problem of validity, if generalizations from small-group experiments are to have relevance for actual groups, small or large. The more restricted questions of procedure and methodology will be raised first, and then problems of broader import related to cross-disciplinary validation of findings. The issues will be illustrated through reference to the intergroup experiments discussed above.
First, consider the problem of experimenter bias and selectivity of observation, a problem in any research, experimental or nonexperimental. As Boring pointed out in his History o f Experimental Psychology, investigators at Cornell found sensations but no images, and investigators at Würzburg found images but not sensations. Rosenthal (1961) has recently documented experimenter bias in experiments using both animal and human subjects. The problem is not one of duplicity, but of the all-too-human tendency to search and to see facts which support a cherished hypothesis.
The problem was recognized in our intergroup research by using a combination o f methods for data collection on every hypothesis, yielding a variety of dependent measures which had to check if an hypothesis was considered confirmed. For example, the existence of unfavorable stereotypes was first determined by observations of name-calling and other negative reactions to the outgroup, then through the group members' ratings of the personal characteristics of members of the other group, and through their quantitative judgments of performance in a situation allowing a margin of error in any direction. Our conclusions about unfavorable stereotypes, therefore, were supported by observations, by the individuals' own evaluations of the other group, and by errors in judging performance which occurred even though the task was objective and which systematically overestimated performance by fellow group members while depreciating the achievement of the
( 164) out-group. Through the use of a combination of methods, the probability of drawing conclusions on selectively chosen data can be reduced.
Second, in any experimentation or measurement involving human subjects, there is the even more serious problem of the "demand character" of the situation, as Martin Orne (1961) has demonstrated experimentally. This problem stems from the individual's awareness that he is a "subject" and that the investigator is a psychologist or social scientist, with all that those terms can mean to the unskeptical. In our intergroup experiments, this problem was considered so serious that subjects did not know at the time that they were taking part in an experiment. They believed that the camp where the experiment took place was a genuine summer camp. The research personnel appeared as custodial and administrative staff in the camp, thus lacking the magic halo accorded representatives of science. Observers were instructed not to record observations in the sight of the subjects. Thus through the cooperation of research staff and parents, the situation seemed lifelike and natural to the subjects.
However, the foregoing problems relating to validity are trifling compared with the doubts raised by social scientists dealing with hard "facts of life," for example by David Truman in discussing research frontiers in polities (1955) following similar questions by V. O. Key; by Conrad Arensberg (1951) in discussing research on small groups and large organizations; and by the sociologist Arnold Rose (1954) in his comments on intergroup experiments. In different ways, these authors question the relevance of findings in experimental situations for problems of real life and, in particular, the pertinence of generalizations based on small-group research for behavior in social organizations facing important and vital problems.
On the other hand, confirmed experimentalists contend that generalizations based on experimental evidence can be applicable to a variety of real-life problems. Cartwright (1960, p. 25) has maintained that the experimenter can use a stimulus situation which may not represent a content area likely to be encountered in everyday life and actually achieve generality by doing so. This is plausible. But there should be at the same time strong emphasis on an additional and necessary qualification.
If experiments in social psychology are to have generality, they must be preceded by extensive surveys of the phenomena in question and the conditions in which they occur. The experimental model must be empirically based on close familiarity with the phenomena, rather than selected as a convenient or fashionable model from those available in the physical sciences or mathematics. The latter practice of science by analogy tears theory from its empirical basis and, indeed, produces experiments with little relevance (Oppenheimer, 1956).
If experimental models are empirically based, the problem of validity of experimental findings converges with the problem of validity of other research methods. The generality of conclusions based on research findings is bounded by the special conditions under which the study is conducted, and this limitation applies to field research as well. For example, an investigation of large-scale political or economic organization is necessarily limited by the period in which data are obtained. It is entirely possible that within a given historical period, some major variables maintain a fairly constant level and are, therefore, not even taken into account. Generalizations based on the data may be valid for that period, but not during times when the "constant conditions" begin to change.
However, there are customs and practices peculiar to psychological experimentation which encourage the accumulation of artifacts. For example, experimenters frequently specify only those variables which they actually manipulate and fail to specify other conditions which are not intentionally manipulated, but which, nonetheless, affect the outcome. Especially when he lacks sufficient grounding in the actualities of a problem area, an experimenter following this practice may very well report results applicable only to a particular and unrepresentative set of conditions; and it may not be possible to determine from his report what these conditions are.
For example, in recent years, more and more experimenters have designed and conducted experiments on small groups with very little concern about the properties of human groups and the consequences of membership in real life. At times, it is not even thought necessary to specify what relationships the individual subjects have outside the laboratory. Frequently the experimenter pro-
( 166) -vides no basis for determining whether he is studying groups or not.
Empirical findings by students of human groups and institutions indicate that groups do have certain properties and consequences for member attitudes and behavior. In our intergroup experiments, we used these findings as the basis for criteria to determine when groups had been formed, namely, some degree of organization, or patterning of status positions, and at least some standards or norms shared by the members. Surely these are minimal criteria for determining whether one is studying processes in a human group or simply a transitory interaction situation; and they have the advantage that they reflect properties of every human group functioning in real life. Operationally, the criteria can be determined through concordance among various independent measures, for instance, in the case of status positions, among ratings of different observers at different times and sociometric ratings.
In a sense, more than half the battle is won or lost in experimentation when the problem is formulated. Thus, if leadership is defined in the problem as individual prominence in a collection of casually related individuals performing motivationally neutral tasks (which may be called "problem-solving situations"), leadership is likely to be specific to particular situations and tasks. But if the problem of leadership is defined in terms of effective initiative and control in a group structure, and the individuals perform tasks that are genuine problem situations for members of that group, leadership is found to be more general from task to task.
Such contradictory results do not mean that experimentation per se is invalid and fruitless. If the conditions of the various experiments are specified, the results fall into a meaningful pattern. And if the experimenter has adequate grounding in the problem area, if his experimental model embodies the major variables involved in the phenomenon, experimentation can become a "crowning touch" of analysis.
In research on intergroup behavior, therefore, the remedy is not concentration on experimental work alone, or on surveys alone, or on field work alone, while deprecating or ignoring the others. No matter what the data-gathering operations, the best insurance against trivial, irrelevant, or invalid results is constant evaluation of the way problems are formulated and the conditions in which
( 167) they are studied in the light of persistent and recurrent empirical findings in the various social-science disciplines—findings by social scientists studying intergroup problems at their own level of analysis in labor-management relations, ethnic relations, factions and national movements, parties and pressure groups, and the sometimes grim problems between nations.
BROADENING THE METHODOLOGICAL STRATEGY IN THE STUDY OF INTERGROUP BEHAVIOR
To evaluate the conditions bounding particular intergroup phenomena, it is necessary to view them through the perspective of the system of relationships in which they are part processes. This oft-repeated dictum will continue to remain lip service unless we develop research methods for actually doing so. Otherwise, recognition of the multivariate determination of intergroup behavior may lead, as it has, to assigning equal importance to all variables, with a plea of ignorance concerning their relative contributions.
The problem can be illustrated with reference to the behavior of leaders representing their groups in communication, negotiation, or bargaining with other groups. Clearly, leaders are distinctive both as unique personalities and in terms of their status in their own groups. Yet, their behavior is a part process within their groups and in negotiations with other groups. Lewis Killian (1962) has recently analyzed the changing Negro leadership in the South and shown that the selective processes bringing certain personalities to the foreground inevitably involve the relationships among Negro groups vis-à-vis the white community, which are, in turn, affected by variables extrinsic to both of them (for example, legal decisions from the Supreme Court, the position of a federal government aware of international criticism of domestic segregation). If we want to understand the behavior of leaders in specific intergroup contacts, their personalities and the immediate interaction processes must be studied within this context.
Another example comes from Robert Dubin's discussion of labor-management relations in the United States as an intergroup system (1962). In the reputable sociological tradition, Dubin emphasizes "constraints" on representatives from their own groups, but he also notes constraints stemming from the intergroup bases
( 168) of negotiation accepted by both sides. The latter are revealed through the alternatives which the respective representatives never contemplate. For example, the American union leader does not consider requesting management to turn the plant over to the union, and management does not consider asking the union to give up all claims concerning worker welfare.
Now the question is: How are we to do justice, as social psychologists, to the study of individual behavior while recognizing it as part of group relations and systems of intergroup relations? How can we assess the relative contributions of the variables composing the context of intergroup behavior and those associated with a particular interaction situation? The traditional answers in psychology and sociology turn upon cross-sectional, multivariate research designs and analysis of factor weights.
An alternative approach which is feasible when conditions can be varied is through the longitudinal design in successive stages, introducing new variables at more complex levels of interaction at appropriate points in time. In the intergroup experiments referred to earlier, we started with the study of interpersonal attractions, a point at which many "group" experiments stop. Then, successively, we changed the ecological arrangements so that two collectives were separated, and then introduced conditions requiring interdependent activities within each. At this early point, the ratings of effective initiative of various individuals in different situations, made by independent observers, showed very little agreement. In fact, different members took the initiative in different tasks. Over a period of time as the group faced varied tasks, however, their relative ranks as to effective initiative displayed in group activities and decisions began to stabilize. An observer who had never rated the group before could rank the members in much the same way as one who had been with the group all along. Task variation no longer had the same importance for this behavioral dimension.
In the data obtained on over fifteen experimental and natural groups in our research, one of the facts pertaining to differentiation of status positions may be basic in the area of social judgment. Invariably, top and bottom positions are differentiated more readily and prior to others in time by the observers. Positions near the ends and in the middle are differentiated last, and judgments of
( 169) them by the same or by different observers are less reliable. Underlying this finding may .be the well-established phenomenon of the salience of end positions serving as anchors for judgment of others (Sherif & Hovland, 1961). Our data suggest that the phenomenon may be operative both in the discriminations of observers rating status positions and also in the judgments of members of groups in a state of formation. In the latter instance, the process of status stabilization seems to involve a recognition by members of the polarization of high and low positions before other positions are settled even for a time.
After we ascertained by explicit criteria that a status organization and norms had developed in each group, the resulting groups became the necessary condition for the study of intergroup behavior in the next stage of the study. Given the two groups, mutually incompatible goals could be introduced in order to study the rise of unfavorable attitudes and social distance between the groups. But then, as intergroup hostility was generated, the organization and certain norms within the groups—in other words the conditions necessary for studying intergroup behavior—began to change. The rankings of members on effective initiative, for example, shifted, with some members moving upward and some down.
Intergroup hostility and social distance, in turn, were the necessary condition for studying the effectiveness of superordinate goals in reducing unfavorable attitudes and stereotypes of the out-group. Thus the research design progressed from interpersonal encounters, through group interaction, then into two stages of intergroup interaction—one negative and one positive.
What can one say from this design about determinants of, say, friendliness or hostility at different levels of interaction? Since there were no groups and no intergroup relations at the start, the initially low level of hostile behavior had no group reference. After intergroup conflict, the introduction of superordinate goals produced a sharp decrease in expressions of hostility and an increase in friendly contacts with the out-group members. If an investigator had performed a multivariate analysis on a cross-section of behavior at three points in time, without information about the intervening changes in experimental conditions, there is little doubt that he
( 170) would have found different levels in hostility expressed, marked changes in the rank of various individuals in this respect, and possibly different factors. But he would probably be at a loss to account for the differences. Here lies the advantage of the longitudinal design introducing variables at different levels of interaction and at appropriate times.
Thus, it would seem possible to develop research strategies which take their point of departure from the view of intergroup behavior as part of systems of group and intergroup relations at increasingly more complex levels of analysis. Currently, at the Institute of Group Relations of the University of Oklahoma, an approach using the same logic is being used in field research on individual behavior and group processes within different sociocultural and ecological settings. The settings or areas of study are first defined in sociological fashion through a variety of socioeconomic and cultural measures. Summary indices which permit the settings to be ranked (high, middle, low) become a baseline for study of evaluative reference scales prevailing among the youth of each area (compare C. W. Sherif, 1961).
In other words, the values or norms—the latitude of acceptable behavior and the goals in so many respects—are determined for each area. Once these reference scales are determined, they become the baseline for assessing the norms of particular groups in the area. Particular small groups are being studied with reference to the areas in which they function and the prevailing values there. The organization and norms of a group, in turn, as well as its solidarity, its exclusiveness and other properties, form the setting for studying the behavior of particular individuals interacting as members.
In short, what is called behavioral and what is called institutional or organizational in the social sciences can cease to be antithetical in concept or in research practice through research designs for the study of part-processes in terms of their appropriate levels of interaction. Thus conceived, both approaches and both kinds of data are necessary to complete the picture. It amounts to conception of the "behavioral" in an institutional setting, and to a more complete and operational picture of the "institutional" in terms of actual behaviors of specific individuals.
TWO AREAS OF NEEDED RESEARCH
From the numerous problems of intergroup relations needing research, two areas will he discussed briefly. The first deals with ingroup and out-group images or stereotypes. The second concerns the reciprocal impact of in-group and intergroup values and trends.
Prevailing Intergroup Images and Aspects of Changing Them
The literature on intergroup relations contains any number of studies on how various groups or nations view one another. We have learned a great deal from these studies, and they will continue to be useful. Yet there seems little further profit in more and more inventories of trait attribution or stereotypes. One thing they have shown us is that one can predict fairly well what the pattern of images of other people will be from the history of relations among groups. In general, we have nice words for our own groups and our friends, and nasty ones for those seen as enemies, even when referring to the same actions. For example, when we kill the enemy, we arebrave or heroic. When he kills us, he is inhuman and fanatical.
From a practical point of view, inventories of how nations and groups view one another published under the prestigeful aura of science may be positively harmful. Consider the effects of feedback to each group of "how others view us." Even within the United States, the experience of seeing the rank of the country or ethnic group of one's origin can be a spine-chilling experience.
Investigations of the psychological bases for views of ourselves and others might more fruitfully center on the process of categorical thinking. From past traditions, somehow we psychologists still incline to the view that thinking, being a universal human phenomenon, can be studied devoid of content, personal or group reference.
The implications of Eugene Hartley's pioneering work on attitudes have never been fully exploited in this respect. In essence, Hartley (1936, 1944) showed that the individual's attitudes towards other groups stem primarily from contacts within his reference group, rather than from personal experiences with the outgroup in question. The powerful role of group reference was documented again in Pars Ram's UNESCO study (1955) showing
( 172) that intensity of personal suffering in intergroup conflict is not directly related to the strength of hostility toward the group seen as the cause of the suffering. Recently, the strength of the dictum—the deduction from verbal categorization, as compared with induction from a series of specific direct experiences—was demonstrated experimentally by W. R. Hood (1961) .
Categorizations are not ordinarily arrived at inductively through "rational" evaluation of direct experience in social life. Everything we know about human culture, and language in particular, informs us that our reference groups have ready-made categories for us which color even our direct experiences and trial-and-error encounters with the social world around us. The needed research must link thought process and categorical mentality with problems of group reference and values. It is hardly begun. We are still plagued by the arbitrary historical dichotomy between "rational" and "irrational" thinking, and too frequently the "irrational" is a label for thinking which leads to conclusions we do not like.
One approach to the problem is through the study of categorical judgments. There is a considerable body of facts about categorical judgments, ranging from categorization of discriminable sensory dimensions to evaluative placement of social objects and opinions. Certain facts already stand out from this research with significance for problems of categorical mentality. Among these is the general finding that when an individual adopts a particular subset of values as his standard in judgment—whether the values he weights, inclinations, or social values—his judgments of the entire set are colored in predictable ways (Sherif and Hovland, 1961) .
Strong attachment to a particular stand toward social objects or issues promotes a tendency to deal with the domain with few categories—to distinguish "those who are for us" and "those against us," with few discriminative segments between. Furthermore, strong allegiance to a restricted range of values raises the threshold for acceptance of other views, with the result that judgments are distributed differently by the adherent and the casual evaluator. These two measures—number of categories and distribution of judgments within them—commend themselves as measures of categorical mentality which can be used with a variety of content
(173) as indicators in the study of developing adherence to a value system and ramifications of this adherence.
The relevance of categorical judgments to the patterned stereotypes with which peoples view one another is clear. The primary datum is the placement of themselves and others into categories of differential-affective loading. Behind this placement is the process of adherence to value systems, from which all of the stereotypes, conceptions, and traits follow logically. It is time, therefore, that research concentrate on the premises from which our views of others arise.
With intergroup behavior conceived as a part-process in ongoing systems of intergroup relations, research on categorization in social judgment may provide guidelines for the study of categorical thinking which will have generality and relevance for study of decisions by individuals at various levels of the social system.
STUDY OF THE RECIPROCAL IMPACT ON IN-GROUP AND INTERGROUP TRENDS
Phrasing the issues in broad terms first, problems of the reciprocal impact of trends within groups and between groups suggest (1) study of the compatibility and incompatibility of in-group and intergroup values and goals in significant dimensions and (2) study of the consequences of the compatibility or incompatibility.
The issues are forced to our consideration today because groups are no longer closed systems. The state of interdependence is enormously increased by modern modes of production, transportation, communication, and the associated social and political awakening of peoples toward larger social units, manifested in emerging nationalisms and trends toward internationalism. No matter how large or small, no social unit today is an island unto itself. Certain political, economic, and social issues have become issues of a common concern rather than being provincial and exclusive interests of particular social units.
The general phenomena are not foreign to this country. Interstate laws and regulations in commerce and transportation, for example, are taken for granted today. More recently, questions of segregation-desegregation within a given state have become more
( 174) and more the business of the whole country, and even of other countries, for that matter.
Even the question of whether or not a given social unit can stand on its feet is no longer the exclusive business of the members of that unit. Its organization or disorganization may have impact on others.
Therefore, there is a fundamental question to be clarified: What constitutes the exclusively internal affairs of social units, and in what conditions are they exclusively internal? When do issues become the vital business of other social units within a system of intergroup relations?
This fundamental question was raised recently with broad reference to the international scene by Otto Klineberg (1962): "There is a widespread view, incorporated in the Charter of the United Nations, that what happens inside the boundaries of a particular nation is the concern of that nation alone, and is therefore not a legitimate field of action or even inquiry by the international community. This view is undoubtedly justified for many internal phenomena, but at least in the special case of the relations between ethnic or `racial' groups, it cannot possibly be defended."
The question raises fundamental problems which are amenable to investigation, even though admittedly enormous research undertakings are required. Among these problems are the following:
(a) What values or trends within groups are incompatible with the prevailing or emerging intergroup trends?
(b) Why are these trends seen, at a particular time, as incompatible with progress in intergroup affairs?
(c) What are the islands of resistance working against intergroup trends, as conceived by this or that set of groups?
(d) In the past and present, what conditions have been effective in bringing incompatible in-group values and trends into accord with emerging intergroup trends?
Of course, answering these questions would represent huge research tasks, commensurate with the significance of the problems. They are multifaceted topics which require conceptually coordinated research programs in which political scientists, historians, sociologists, economists, and others should exchange data and accumulated experiences, and really try to integrate them, rather
( 175) than talking at each other across the boundaries of their own particular and cherished group lines.
If such questions can be answered, social psychology can contribute its bit. The essential features of the intergroup phenomena can be extracted from the work of social scientists working on more complex organizational levels. Then, and only then, the processes can be replicated in miniature to test conclusions and hunches derived from them.
On the basis of the generalizations derived from historical and other social-science inquiry, it should be possible to create operationally conditions in which, say, ten distinct and autonomous groups of manageable size develop independently of one another in separate locations. In various combinations, these groups can be brought together under conditions conducive to competition with one another or to cooperation with one another. It may not prove to be an unsurmountable operation, then, to introduce dimensions of significance to all groups which are compatible with the values and goals of some groups and incompatible for other groups.
If such a research program were carried out and replicated with sets of groups of similar size in different sociocultural settings, we should be able to learn in clearer focus when in-group values become the business of other groups and when in-group values are inimical to the broader trends in intergroup relations. We should be able to derive fruitful leads for generalization about the functioning of leadership, representation, and negotiation, in intergroup encounters. We may find more realistic criteria for assessing the relative weights of the so-called hard facts (for instance, resources, wealth, organization, and weapons for dealing with the adversary) and the soft facts (compare North, 1962) in intergroup relations. The term soft facts apparently refers to social and psychological factors, including relative stability of the leader in his own group, the latitude of acceptable behavior for the behavior in question, and the appraisals of the hard facts concerning in-group and out-group by the leaders.
The area of soft facts, in which the research task of social psychologists primarily lies, may not be such an insignificant one. We need not go too far to find illustrations in which appraisals of the capabilities and intent of one's own group and other groups
( 176) have tipped the decision-making process in favor of one alternative, rather than another, at times with disastrous consequences. If we have here emphasized the imperative need to study social-psychological processes within the framework of hard facts of the larger social scene, it was not to depreciate the importance of individual perceptions and judgments, but to direct attention toward ways of study through which their import may be properly assessed.
Arensberg C. H. Behavior and organization: industrial studies. In J. H. Rohrer and M. Sherif (eds.), Social psychology at the crossroads. New York: Harper, 1951.
Blake, R. R., and Mouton, Jane S. Competition, communication and conformity. In I. A. Berg and B. M. Bass (eds.), Conformity and deviation. New York: Harper, 1961.
Cartwright, D., and Zander, A. Group dynamics. Research and theory. (2nd ed.), Evanston, Ill.: Row Peterson, 1960.
Clark, K. B. Observations on Little Rock. New South. Atlanta, Ga.: Southern Regional Council, June, 1958. Pp. 3-8.
Dubin, R. Union-management relations as an intergroup system. In M. Sherif (ed.), Intergroup relations and leadership. New York: Wiley, 1962.
Harding, J., Kutner, B., Proshansky, II., and Chein, I. Prejudice and ethnic relations. In G. Lindzey (cd.), Handbook of social psychology. Vol. 2. Cambridge, Mass.: Addison-Wesley, 1954.
Hartley, E. L. (Horowitz). The development of attitudes toward the Negro. Arch. Psychol., 1936, No. 194.
Hartley, E. L. (Horowitz). Race attitudes. In O. Klineberg (ed.), Characteristics of the American Negro. New York: Harper, 1944.
Hood, W. R. The rigidity of concept utilization as a function of inductive and deductive derivation. Doctoral dissertation, on file in the Library, Univer. of Oklahoma, 1961.
Katz, D. Current and needed psychological research in international relations. Paper to annual meetings, American Psychological Assoc., 1960.
Killian, L. M. Leadership in the desegregation controversy: an institutional analysis. In M. Sherif (cd.), Intergroup relations and leadership. New York: Wiley, 1962.
Klineberg, O. Tensions affecting international understanding. New York, Soc. Sci. Res. Counc. Bull, 1950, No. 62.
Klineberg, O. Intergroup relations and international relations. In M. Sherif (ed.), Intergroup relations and leadership. New York: Wiley, 1962.
Murphy, G. In the minds of men. New York: Basic Books, 1953.
North, R. C. International conflict and integration: problems of research. In M. Sherif (ed.), Intergroup relations and leadership. New York: Wiley, 1962.
Oppenheimer, R. Analogy in science. Amer. Psycholog., 1956, 11, 127-135.
Orne, M. T. The demand characteristics of an experimental design and their implications. Paper to the annual meetings, American Psychological Assoc., 1961.
Pear, T. 11. The psychological study of tensions and conflict. In The nature of conflict. UNESCO, Tensions and technology series, 1957.
Ram, P. A UNESCO study of social tensions in Aligarh, 1950-51. Ahmedebad, India: New Order Book Co., 1955.
Richardson, L. F. Statistics of deadly quarrels. In T. H. Pear (ed.), Psychological factors of peace and war. New York: Philosophical Library, 1950.
Rose, A. M. Theory and method in the social sciences. Minneapolis: Univer. of Minnesota Press, 1954.
Rosenthal, R. On the social psychology of the psychological experiment: with particular reference to experimenter bias. Paper to annual meetings, American Psychological Assoc., 1961.
Sherif, Carolyn W. Self radius and goals of youth in different urban areas. Southwestern Soc. Sci. Quart., 1961, December, 259-270.
Sherif, M., Harvey, O. J., White, B. J., Hood, W. R, and Sherif, Carolyn W. Intergroup conflict and cooperation. Robbers Cave experiment. Norman, Okla.: University of Oklahoma Book Exchange, 1961.
Sherif, M., and Hovland, C. I. Social judgment: assimilation and contrast effects in communication and attitude change. New Haven: Yale Univer. Press, 1961.
Sherif, M., and Sherif, Carolyn W. Groups in harmony and tension. New York: Harper, 1953.
Sherif, M., White, B. J., and Harvey, O. J. Status in experimentally produced groups. Amer. J. Sociol., 1955, 60, 370-379.
Truman, D. B. The impact on political science of the revolution in the behavioral sciences. In Research frontiers in politics and government. Washington, D.C.: The Brookings Institution, 1955.
Williams, R. M., Jr. Racial and cultural relations. In J. B. Gittler (ed.), Review o f sociology. New York: Wiley, 1957.
- MUZAFER SHERIF Research Professor of Psychology and Director of Institute of Group Relations, and CAROLYN W. SHERIF, Research Associate, Institute of Group Relations, University of Oklahoma