natural interaction. Cliques chosen for the final experiment were those in which there was high correspondence among the status ratings obtained.
The overall finding was that the higher the status of a member, the greater his tendency and that of other group members to overestimate his future performance. The lower the status of a group member, the less the tendency of other group members and of himself to overestimate his performance, even to the point of underestimating it. If these results are valid, it should prove possible to predict leaders and followers in informal groups through judgmental variations exhibited in over- and underestimations of performance.
The summer of 1953 marked our first attempt at a large-scale experiment starting with the experimental formation of ingroups themselves and embodying, as an integral part of the design, the assessment of psychological effects of various group products.8 This assessment involved laboratory-type tasks to be used in conjunction with observational and sociometric data. The overall plan of this experiment was essentially like that of the 1949 study summarized earlier. However, it required carrying through a stage of ingroup formation, to a stage of experimentally produced intergroup tension, and finally to integration of ingroups. The scope of this experiment, embodying laboratory-type procedures at crucial points in each stage, proved too great for a single attempt. During the period of intergroup relations, the study was terminated as an experiment owing to various difficulties and unfavorable conditions, including errors of judgment in the direction of the experiment.
The work completed covered the first two stages and will be summarized here very briefly. The plan and general hypotheses for these stages are similar, on the whole, to those of the 1949 study summarized earlier.
Prior to the experiment, subjects were interviewed and given selected tests administered by a clinical psychologist. The results of these assessments were to be related to ratings along several behavioral dimensions made by the experiment staff during the experiment proper when ingroup interaction had continued for some time.
At the end of the stage of group formation, two ingroups had formed as a consequence of the experimental conditions, although the rate of group formation and the degree of structure in the two groups were somewhat different.
Our hypothesis concerning experimental formation of ingroups, substantiated in the 1949 study, was supported. As a by-product of ingroup delineation, we again found shifts and reversals of friendship choices away from the spontaneous choices made prior to the division into groups and toward other members of the ingroup.
At the end of this phase of ingroup formation, just before the first scheduled event in a tournament between the two groups, psychological assessment of group members within each status structure was made through judgments obtained in a laboratory-type situation. In line with methodological concerns mentioned earlier in the chapter, a member of the staff introduced the experimental situation to each group with the proposal that they might like to get a little practice for the softball game scheduled later that day. When this proposal was accepted, the experimenter took each group, separately and at different times, to a large recreation hall where he suggested turning the practice into a game in which everyone took turns and made estimates of each other’s performance. This suggestion was accepted as a good idea. Thus each boy took a turn at throwing a ball at a target 25 times, and all members judged his performance after each trial.
It should be noted that in previous studies, judgments of future performance were used as an index. The important methodological departure here was using as the unit of measurement the difference between actual performance and judgment of that performance after it was executed. To do so, the stimulus situation had to be made as unstructured as possible so that the developing status relations would be the weighty factor in determining the direction of judgmental variations.
In line with our hypothesis in this experimental unit, the results indicated that variations in judgment of performance on the task were significantly related to status ranks in both groups (Sherif, White, and Harvey 1955). The performance of high-status members was overestimated by other group members; the performance of low-status members tended to be underestimated. The extent of over- or underestimation related positively to the status rankings. Variations in judgment of performance on this task did not significantly correlate with skill, or actual scores, of members. This finding should not be interpreted to mean that skill can be discarded as a factor, or that it would not be highly related to judgmental variation in a more structured task. Of the two groups, skill seemed to be of relatively greater importance in the group that achieved less stability and solidarity. This result is one of several indications that the relationship between judgmental variation and status rankings is closer in the group with greater solidarity and greater structural stability. This finding of a relationship between the degree of structural stability, on the one hand, and the psychological response of members as revealed in their judgments, on the other, points to the necessity for systematic concern with the degree of group structure and solidarity as a variable in small group studies. In particular, it should be brought systematically into the study of leadership and problems of conformity (Sherif 1954).
Following the experimental assessment of psychological effects of group structure in existing and in experimentally formed ingroups, the next step in our program of research was to extend the use of judgmental-variation techniques to the level of intergroup relations among already existing groups. Such an experimental unit was completed by O. J. Harvey in 1954. Harvey investigated relations between existing informally organized groups and their effects on ingroup functioning and on evaluations of the ingroup and outgroup. Organized cliques were chosen on the same basis as those in the study, already summarized, of status relations in existing informally organized groups. In the first experimental session, ingroup members judged each other’s performance on a task. In the second session, two cliques with either positive or negative relationships with each other were brought to the situation together. Here a similar procedure was followed, with ingroup members judging performance both of other ingroup members and of members of the functionally related outgroup. In addition, subjects rated ingroup and outgroup members on ten adjectival descriptions presented on a graphic scale. These ratings were included to yield data relevant to our hypothesis in the 1949 study concerning the nature of group stereotypes and to the hypotheses of Avigdor’s study (1952) on the rise of stereotypes among members of cooperating and rival groups.
Results obtained in this experiment bear out the hypotheses. Greater solidarity was evidenced in the ingroup when negatively related outgroups were present, as revealed by an increasing relationship between judgmental variation and status ranks and by greater overestimation of performance by ingroup members. Ingroup performance was judged significantly above that of outgroup members when the groups were antagonistic, which was not the case when the groups present were positively related to each other. Finally, results clearly show a much higher frequency of favorable attributes for ingroup members (e.g., “extremely considerate,” “extremely cooperative”) and a much higher frequency of unfavorable attributes for members of an antagonistic outgroup (e.g., “extremely inconsiderate,” “extremely uncooperative”). The difference between qualities attributed to ingroup members and members of friendly outgroups is much smaller and not so clear-cut, as would be expected.
Thus we demonstrated the feasibility of experimental study, through laboratory-type techniques, of norm formation, of status relations within groups, and of positive and negative attitudes between groups, on the one hand, and on the other, of experimental production of ingroups themselves, as evidenced in two previous studies. Our next step would be to carry through the large-scale experiment along the lines of our 1953 attempt, pulling together all of these various aspects into one design. Judgmental indices reflecting developing ingroup and intergroup relations would be obtained through laboratory-type techniques at choice points in a way that would not clutter the flow of the interaction process. These judgmental indices could be checked against data obtained through more familiar observational, rating, and sociometric methods. If indications of the findings through judgmental processes were in line with the trends obtained
