A Discussion of the Issues: (II)

Muzafer Sherif, Alan Ross Anderson, Charles Ackerman and Gordon J. DiRenzo.


I would like to stick to the basic orientation of this Symposium, namely, definitions and theoretical orientations, rather than going to procedural matters, which are relatively easy. We are interested in sociology and we are interested in psychology. We are interested both in personality and in group and in culture. But on so many topics, our interests are dichotomized and elaborated as distinct and almost insulated "approaches." So we have personality-oriented approaches, group-oriented approaches, or culture-oriented approaches.

It seems to me that there is a place for both a group-oriented and a personality-oriented approach, and both are needed. But this is not an endorsement of eclecticism. I am following Durkheim who stated, in The Rules of the Sociological Method and in The Elementary Forms of Religious Life for example, that social life is not a continuation of natural life, but an emergent level. This is where he ties in the criteria of exteriority and constraint. With this orientation, we can attack the problem of integrating the group or cultural level and the individual level, including individual personality.

Take social and cultural phenomena as things out there and analyze them at a sociological level of analysis, with proper units for that level, such as group organization and institutional structure, customary procedures, routines, and social norms. When we deal with individual behavior, on the other hand, we are working at a different level of analysis, a psychological level; we have to deal in terms of psychological units of analysis. However, the two levels of analysis—the sociological and the psychological should not be contradictory, if they are valid. On the contrary,

they must supplement one another.

Let me illustrate with the case of color blindness. If the

( 221) physiologist is studying color blindness, his units of analysis arc the receptors, the optic nerve, cerebral functioning, and so on. But, if we are psychologists, we have to use units of analysis appropriate to the psychological level. Here we deal with the individual's perception and behavior in response to selected stimuli, such as the color plates or skeins of yarn in a color vision test. We ask what the person sees on a figure; for example, does he or does he not see the numeral seven? This is a psychological question. There is still the possibility that on casual inspection a person may not respond as expected because the colors in question are not discriminated in the lexicon of colors in his culture. Now, the physiological level of analysis, the psychological level of analysis, and the cultural or sociological level of analysis should supplement one another in elucidating why an individual responds in a particular fashion. It is absurd to struggle with the question of whose prerogative it is to do what. If their findings are correct, the analysis of physiologists, psychologists, and social scientists supplement one another, as indicated in the case of color blindness. The issue is the same in the question of studying personality and group or culture.

The point can be illustrated through our research, begun in 1958, in which we have been studying small groups of boys for a period of from six months to a year. These groups are composed of boys who come together of their own choice and who do not know that they are being studied, the observer appearing to them in the guise of a somewhat older friend. Without exception, the interaction in such clusters produces some kind of pattern. In the most important dimension of power, we have found in every case a hierarchical arrangement (that is, a status hierarchy). Some pattern is an invariant product if a number of individuals interact with one another over a given time span. Now we are involved in determining cutting points to differentiate the various divisions of the status hierarchy in different organizations of this kind. These are not new notions, but we are taking steps toward operationalizing them in terms of specific indicators of power.

Now these patterns of relationships in small groups are not ahistorical phenomena. As Rome was not built in a day, so groups

( 223) are not built in one or a few days. That is why the concept and the analysis of a group differ completely from the ahistoricalism that characterized the earlier period of the movement called Group Dynamics, which was based upon physical analogies. Sociologically, in the study of groups we have no choice but to proceed longitudinally or historically.

The data in this regard warrant a generalization on a sociological level. When a number of individuals with a common interest interact with one another repeatedly over a period of time, they form some kind of pattern or structure. This pattern or organization is obviously a phenomenon at the sociological level of analysis. But, let us turn now to personality. Who occupies what position in the structure and why? Of course, who occupies what position is a function of the unique characteristics of particular individuals constituting the pattern in relation to the activities they engage in most frequently. Only when we can put the individuals in their proper context—namely, into the pattern of relationships in the group—can we come to see the personality characteristics of the individual members in a meaningful way. And even then we will find that the achievement of particular status and roles in a group is not explained by personality characteristics in the abstract. Personality characteristics become relevant in particular situations. Of course this is not a new finding. Thrasher's study The Gang includes reports of the variations in the personal characteristics that are the most prized by different groups. In one gang, where toughness was important, the leader would be a tough guy. In another, the leader was a hunchback, who could not actually carry out plans himself, but who became the leader because shrewdness in the ability to steal was at a premium.

In short, psychology and sociology need not wrangle over prerogatives. They should supplement each other, and if each will use the proper level of analysis with its appropriate units, they can validate each other's findings.

( 224)


I would like to raise this question. In the case Professor Lazarsfeld discussed at length in his original talk, or in similar cases, what have these considerations to do with whether or not one conceives of correlations as explanations? Let me put the matter this way. The situation looks to me somewhat like Keplerian astronomy: we know as a matter of fact that the planets move in certain elliptical orbits, but it is not until we get a general theory of the Newtonian sort, from which one can deduce the fact that they ought to move in elliptical orbits, that one feels that one has an explanation of why they do it. I am not absolutely sure why it is that we have this feeling, but it certainly is true that one thinks that Newtonian mechanics explain the observations that Kepler made about the positions of the planets. I am just asking this as a question: in what sense do the correlations themselves count as explanations? It always seems to me that there are some additional facts that need to be explained.

In just the same way, for example, one might associate hair color with body type. (This is, of course, not a good illustration.) But, let us suppose that various body types always produce constant hair color. It would seem to me that this is a bit of information—something to be explained—but that the fact itself does not serve as an explanation of anything at all.


If you will permit me, I would like to address this problem by discussing the concept of "reference group." Suppose that I am a small businessman assessing how prosperous I am. If I take the "upper class" as my reference point, my judgment of my prosperity would shift downward. But if I take the salaries of teachers, my judgment of my prosperity would shift upward. Here is a bit of information which, as Professor Anderson notes, a correlation does not explain. From the reference group concept, we can predict such systematic variations because it, in turn, was based on the experimental psychology of judgment—that is, of the categorization of objects and events.


Judgments are never absolute. Judgment always involves a process of comparison. One has to have at least two things to compare in order to make a judgment. Now, judgment of particular things varies systematically depending upon the series or context of which they are parts and upon the standard or anchor used as the primary basis for comparison. Judgments vary systematically in direction according to the relationships among the particular objects judged and the stimulus value of the standard or anchor. The relationships governing shifts in judgment have been explored extensively in the psychological laboratory, and the principles discovered are readily generalized to the case in point.

The person uses the position of the group or set of people with whom he identifies as the major anchor or standard for assessing the positions of other groups on the social scale. With reference to upper-class income, the small businessman in our case exaggerates the difference between his own and theirs. This is a contrast effect, which occurs when the objective discrepancy between one's standard and the particular item judged is large. It follows that the businessman regards himself as less prosperous when comparing himself to the upper class than when comparing himself with teachers, whose incomes are also discrepant, but lower. On the other hand, if he were to assess the income of, say, academic administrators, the small difference between his own income and theirs would lead him to regard his prosperity as being on a par with university presidents. This is an assimilation effect, also well known in the psychology of judgment, which occurs when there are small differences between the standard for judgment and the item judged.

In other words, whether it is the question of judging physical objects or of judging social phenomena, you can predict systematic displacements in judgment provided that you know the anchor value and the values of the items being compared with it. The phenomena are not merely described, but are governed by lawful relationships among standards and comparison stimuli, as you may see in our books on the topic—Social Judgment and Attitude and Attitude Change.



Let us grant the fact—which would seem to be true—that every judgment we make is made relative to some standard or some, let us say, anchor point. Granting this, one might then be moved to ask "Why?" Why do we do it that way? Why do we not do it some other way?



Psychologically, the "why" is specified in terms of relationships between anchors (either the person's psychological anchor or an objective stimulus standard) and the objects of judgment. If you want a further answer to "why," you would have to work on a different level of analysis, namely the physiological or biochemical, in terms of units applicable to the central nervous system. Many principles of judgment are equally applicable to a variety of sensory and perceptual phenomena, indicating that they are not dependent solely upon the operation of the receptor organs. It turns out that they are also applicable to the ways an individual appraises himself relative to others, provided that you know his reference groups.



But, in Newtonian physics, one does not do that. What one does is to derive Kepler's laws from the more general principles.



Let me be clear that I am not stating a logical premise, but an empirical generalization. It is an explanation in the sense that you can derive other things from it. But we do not have our Newton yet in psychology or in social science. When we do, no doubt there will be a body of more general principles and their specific corollaries. If so, my guess is that they will include social, psychological, and physiological variables. As far as I am concerned I will be more than satisfied if we advance in social science from our primitive state in the twentieth century to the level that Kepler achieved in the first half of the seventeenth century.


I am quite in agreement with Professor Anderson that the correlation is a fact to be explained. What constitutes "explanation"? So far as I am concerned, the set of general statements from which one could generate the correlation constitutes the explanation. With reference to all the fours that have appeared in these discussions, I would submit (1) that we could generate these by the AGIL paradigm and that (2) therefore—until phenomena come along that we cannot subsume—action theory constitutes an explanation of them.


Thank you, gentlemen, for these comments. Your discussion has clarified some of the issues that we have been talking about today. Unfortunately, limited time inevitably results in having to leave behind many unanswered questions. We certainly did not intend to resolve the problems involved in the many issues that were discussed. Let us hope, nonetheless, that our deliberations may provide a significant contribution in that direction.



Valid HTML 4.01 Strict Valid CSS2