The Case-Study Method in Prediction

Leonard S. Cottrell, Jr.
Cornell University

I do not conceive the purpose of this paper to carry on the discussion of the relative merits of case-study versus statistical methods for analysis and prediction in the field of social behavior. Nor do I see the utility of attempting at this time to demonstrate further qualitative differences inherent in the two methods, for I am in substantial agreement with Stouffer's excellent logical synthesis of what most of us have understood to be two different operations. Rather it is my purpose to point to certain shifts in orientation which now appear to me to be necessary for users of both methods. This shift, it should be added, is rapidly taking place.[1] It should be further added that the shift is taking place and should take place in case-study methods first. Such changes will sooner or later probably require modification in measurement and statistical procedures. If the presentation in this paper appears dogmatic it is due primarily to efforts at condensation and not to any finality of conviction on the questions involved. It is hoped that the discussion of the ideas presented will yield more valid conceptions and a more tenable perspective on the problem.


In my opinion, users of what are termed case-study methods should make the situation a unit of description and analysis if these methods are to make maximum contributions to our ability to predict human behavior. The term situation

( 359) is widely as well as loosely used. The fact that it is ambiguously used and never explicitly defined may be taken as evidence that our habitual focus of attention has been on individual persons and their alleged traits rather than on interact units of observation and description. I mean here by situation a reciprocally related system of social selves. A social situation is always constituted in such a way that the behavior of any part of it can be understood only by taking into account that part's relation to the rest of the system. This is a very crude definition and is only approximate. A social situation is as hard to define as is an historical event. Theoretically and logically the whole world might constitute a social situation and in some ways it does. But many of the interrelationships among the parts become so attenuated that for most practical purposes we can ignore them. It should be noted that a situation or a situational context is a perspective bound conception. That is to say a situational field can only be perceived and described from one position or rôle perspective at a time. Each member of the situation is responding to it not as seen by some master mind above and outside the situation but as he perceives it. It should also be noted that situational fields vary in size, duration, and frequency of repetition. Situations as the term is used here may be interpersonal, intrapersonal, inter- and intragroup. Terms at present in the service of description, analysis and interpretation such as wishes, drives, traits, etc., which individuals are supposed to have indicate the habit of explaining behavior by' ascribing it to attributes possessed by the individual rather than by regarding it as a function of position in an interact system. Words like wish, drive, etc., do not stand for independent entities but acts involving counter acts, incompleted or resisted movements. There is no wish without blockage or drive without something to drive toward or against. Try to have the experience of pushing without having something which resists the push or of wishing in the midst of gratification. Or try to manifest a trait of aggressiveness without an imaginary projected or actually present counteractor to your aggression. Such attempts should lead to a realization that the behavioral components we tend to speak of as attributes more or less independently possessed by individuals are actually relational functions.

Now, lest I be accused of boxing with straw men, I should state my awareness of the fact that this is old and much traversed territory. Moreover I know that attribute theorists will say, "Of course we know that traits named are only names; they stand for dynamic interact processes." That is fine; I am

( 360) glad they know it, but I can not escape the disconcerting feeling that having dismissed my questioning with such an answer, they then proceed to construct instruments, develop theories, analyze and interpret behavior in attribute-terms and to measure characteristics "belonging" to individuals quite thoroughly abstracted from any specified interact contexts. To assist in maintaining such oblivion, there is the tendency to give a certain lip service to "the situation." We talk about "the general situation," the "family situation," the "community situation" quite glibly without anything like sufficiently attentive analysis and specifications of the interact pattern we refer to. Our correlations of traits with "environmental factors" is usually naive and without any conception of the significance of what we are trying to do in terms of situational analysis. We set up devices for measuring attitudes which reveal our basic assumptions about having attitudes in a vacuum. Of course we all insist we can have attitudes only toward something. But then we proceed to make that something a static reified something without reference to its rôle in a specified context.

Our language reflects our essential thought processes, and it is interesting to note the difficulties we encounter and the clumsiness we show in trying to describe an interact situation. Statistical language and theory in its social psychological applications is pointed primarily to an individual attribute rather than an interact context description and analysis. I do not intend to suggest by this that statistics is inherently limited to the attribute frame of reference; on the contrary, it is no more limited than any other language form. But in its present state it reflects our basic orientation to the problem of explaining social behavior.

With respect to language difficulties, it is my opinion that attribute descriptions of social behavior can rather readily be translated into interact context terms with a considerable gain in clarity and precision for such description. But for such translations to be more than a verbal game a genuine shift in frames of reference must be achieved.

Now what has all this to do with prediction and particularly prediction by case-study methods? In my opinion it has a great deal to do with prediction. The usual way prediction problems have been posed in social psychology is as follows. Some criterion is set up which is the standard by which behavior is judged to be "successful" or not. Individual attributes, skills, and performances are then weighted in accordance with their correlations with the criteria. Composite scoring of the predictive items then serves as an index placing

( 361) the individual in a class with some determined probability of success.

In the first place I think we need to go further than we have in distinguishing between predictions of specific behavior on the one hand and judgments or evaluations or degrees of success according to some standard or other. It is one thing to predict how a couple will perform in a situation developing around the use of money and another to judge the contribution these behaviors make to a "successful adjustment."

The foregoing suggestion now opens a way to the added suggestion that in my opinion our predictive items must have reference to a specified type of interact context and our predictions must be made with respect to a specified situation. When I say specified I wish to make the requirement as rigorous as possible. I do not mean that we specify the situations by referring vaguely to "the marriage situation" or "the office situation" or the "factory situation." I mean in each case the particular rôle composition that furnishes the interact context in which the person whose behavior is predicted is to be placed. Needless to say, this specificity has not thus far been approximated in social psychological prediction procedures.

This suggests a need for (1) developing adequate and convenient ways of describing situations; and, (2) a system of classification of the main types of situational contexts for various areas of social life in which predictions are to be made. I have elsewhere made specific suggestions for research projects in this connection.[2] Some of these suggestions are added here to make more concrete what I have in mind.


"There is a need for exploring the problems suggested by the following questions:

1. Is it possible to classify the concrete interpersonal relationships In which social activity takes place into a limited number of types?

2. What will be the utility of this kind of classification for prediction?

Elsewhere in this work we have pointed to the

( 362)

need for a comprehensive attempt to identify the basic human traits and abilities in terms of which personalities could be unambiguously described. However, social behavior does not occur in a vacuum but always in some interact context. Hence in order to utilize knowledge about individual tendencies and abilities an understanding of the context in which they are manifested is indispensable. Indeed it is impossible to discuss individual characteristics except as components of an interrelated system of activity.

With reference to problems of prediction, it can be safely asserted that predictions are always made with reference to some type of context, though many such references are ambiguous and implicit.It is our contention that the more explicitly the prediction is referred to a specified context the more accurate the prediction is likely to be.

For these reasons we propose that an attempt be made to identify the major types of social situations. Little attention has been given to this problem by research workers, and except for some suggestive leads from theorists like Georg Simmel, von Wiese, and others, there is little previous systematic research work to use as a point of departure.

A number of lines for exploration suggest themselves. Four that offer some promise are mentioned here.

1. The materials assembled by the Cross Cultural Survey at the Institute of Human Relations, Yale University, offer an excellent opportunity to make a survey of the different patterns of Interpersonal relations which have been seized upon and formalized by a wide variety of cultures. It is quite possible that a multiple factor analysis of certain of these materials may furnish leads for establishing a basis of classification.

2. Sociologists, psychologists, psychiatrists and others have accumulated a large number of personal documents in which there is a great deal of descriptive material about the kinds of social situations in which the individual has lived. For the most part these documents have been studied with a focus of attention on the characteristics of the individual. A good deal of this type of material should be analyzed with the focus of attention on the situational patterns explicitly and implicitly defined in the document.

3. J. L. Moreno has developed methods for research and therapy which involve verbal, graphic and dramatic portrayal of social situations. The use of his methods should prove highly valuable in the type of research

( 363)

suggested here. It is quite possible that a study of Dr. Moreno's materials accumulated from past work could be used in conjunction with the attempt to classify types of interpersonal situations from the documentary material suggested in 2.

4. A factor analysis of words used to describe concrete social situations might yield some valuable suggestions for bases of classification, particularly if this analysis is paralleled by less formal studies of situational description. The classifications derived from the less formal procedures could then be compared with the clusters produced by the factor procedures.

Some valuable results of these and similar exploratory research efforts might be:

(I) Preliminary answers to the question of whether or not types of situational patterns emerge from such efforts.

(2) A development of clearer and more standardized ways of describing interpersonal situations. Most descriptions now are ambiguous and haphazard. This condition necessarily handicaps efforts to achieve specificity and clarity in describing personality, as well as situations, to say nothing of making specific predictions.

(3) Such research should serve to demonstrate the feasibility and utility of situational description.

(4) If the foregoing expectations are even partially realized, they may stimulate work in the direction of predicting situational changes."

This discussion of the importance of a situational or interact unit rather than an individual unit has relevance for the consideration of the case-study in prediction on at least two counts.

1. It Is through a knowledge of how the person consciously and unconsciously perceives (structures or defines) the major types of situations In his life activity that we gain a maximum predictive power. This knowledge is gained through an intensive analysis of the person's important incorporated self-other patterns, their genesis and intrapersonal organization. Since these patterns and their interrelations tend to be unique in many aspects, we are confronted with the condition suggested by Stouffer in his note on the statistical theory of the unique case. Since present statistical methods are not yet – flexible enough nor our information on a large number of personality systems adequate enough for the application of statistical methods, we must necessarily rely on case-study procedure.

( 364)

2. The burden of developing a solid basis of experience in describing and analyzing interact behavior and of developing fruitful hypotheses as well as developing a much needed descriptive terminology will obviously rest on flexible procedures that allow highly unique as well as complex processes to be considered without requiring a large number of instances on which to operate. This again points to the case-study method as a necessary part of our research efforts in social psychology at this juncture.


All of this talk about the situation as a unit does not enable us to avoid studying the individual actor nor to minimize the importance of such study. It does, however, require that we have a theory of the individual personality consistent with the findings of situational analyses.

The insights and formulations of Baldwin, Cooley, and Mead, about the nature of the human personality have never been fully exploited for implicit as well as explicit contributions to our understanding of the processes of personality development and organization, the processes of interaction and communication, of how we "know" another person, the nature of collective unity and disintegration, and other important social psychological phenomena. In a sense, modern social psychology is just beginning to catch up with the ideas developed by these men.

This is not the place to attempt an elaboration of an interactional theory of personality. It will be sufficient for our present discussion simply to say that the conception of the incorporation of the acts of the other as well as of the self—the importation of the social act—is of central importance. This process is perhaps basic to all social integration.

We are interested here in pointing out that the human organism going through this process gets equipped with a system of self-other patterns. The acts of the others are just as truly present (represented as expectational attitudes or incipient act tendencies) as the response mobilization identified as the self-behavior. The two parts are inseparable segments of one process. This means that if we can determine the self-other or response-expectation pattern of a person, we know how he will tend to perceive, or structure, or define a situation involving others in different areas of activity. We are then in

( 365) position to be precise and specific in prediction about what he will tend to see and the position he will tend to occupy in the situation as he perceives it. Another important possible increase in predictive power is involved in the fact that the importation process makes It possible for the self-"location to shift from the position of the present self to that of the other or others in the social act. This happens under certain conditions. When we know more about the determinants of "center of gravity" of self-other systems we shall be in position to make predictive statements about conditions under which profound shifts in self-behavior may be expected and what those shifts are likely to be.


The study of a case involves the use of the observer's personality as an instrument of observation of an ongoing series of interacts. This is particularly true where the case-study includes personal contact. When the study is limited to documentary materials, the essential processes involved in case analysis are obviously reduced.

The question then is what is this method by which one comes to "know" a person? I suppose it is fair to state that this knowing means an ability to state the way in which he will view his life situations (his point of view), how he came to have such perceptions (genesis) and what his attitudes and overt responses are most likely to be under specified circumstances (prediction). It may be pointed out that this is a kind of knowing found in friendship and family relations though in most cases overlaid with highly stereotyped self-other patterns. Now pointing to the similarities between knowledge of a case and of a friend does not necessarily add to our understanding of the processes involved. I indicate the partial [3] similarity merely to suggest the fact that an observer-subject relation is a social interaction not too unlike many other social relations, and to suggest that if we understand the processes of social interaction involved in what I have called incorporation of self-other

( 366) patterns, we may more deliberately and with greater precision use these processes for the study of cases and for predicting their behavior in specified Interact fields.

Obviously these statements point to the assertion that the distinctive method of the case-study involves the conscious and skillful use of the incorporative or role-taking processes which go on most fully in the more intimate Interpersonal relationships. This is sometimes referred to in terms of varying ambiguity as sympathetic introspection, empathetic introspection, identification-projection, analysis of transference, etc. Ambiguous though these terms be, they need not be given up as referring to permanently mysterious and unanalyzable intuitional processes. The task is to become more alert and aware of our own processes, more skillful in placing ourselves in the act positions of others in the case-study interrelations and more explicit in stating to ourselves and others what goes on. With such sharpening of our wits and sensitivities, we should not only become more capable as case analysts but should achieve more of a consensus on the difficult questions of the validity of the procedure.

When a charming young woman comes in with many problems and after the preliminary interview states how much better she feels; that she knows I will show her the way out of her difficulties; that she wishes she had come in sooner, etc., etc., I may experience an inner glow at being a hero and helping an attractive lady in distress. If my registry of the situation stops there I a m indeed naive and not using good case-study skill. Moreover, if I do help the lady it will be by a fortunate accident rather than by a self-consciously used skill. I should, from the very start, attempt to get her perspective on me. What is her "structuring of the situation" involving the two of us? What is her self-conception and conception of me? From what I know of her past, if I had been in her rôle and had said and acted as she did, how would I have conceived the interviewer and my relation to him? Her words might have signified a great relief at finding someone at last who would take responsibility for making important decisions for her and "tell her what she ought to do." They might have meant that she had now found someone who could be used as a cat's-paw to lay down the law to her husband, parents, etc. They may have meant that she wished to be loved and was trying to be pleasant and attractive. They might have meant strong hostile aggressions which were overlaid by a veneer of "sweetness and light." They might have meant an honest satisfaction in finding a relationship in which she would have help and an opportunity to discover the sources of her difficulties and learn ways

(367) of solving them. These and many other tentative formulations of the situation I arrive at by taking her possible rôles and indicating to myself the possible definitions of the situation in which her behavior in our relationship will have meaning.    I can not afford to take one definition and throw the other possibilities out until much more of her attitudes and behavior have been experienced and tested by this process. If I learn that she can't stand her father on account of his unreasonable and overbearing behavior; that she separated from her husband because of his foolish insistence on having his own way; and that she is now having a feud with her school principal because he is jealous of her superior knowledge about how to run a school —I may continue to consider as one of her possible structurings of the situation a simple positive dependent relation on a benevolent rescuing hero. But I will keep more than one eye on the possibility that I may be a target for a lot of destructive aggression against men. Whatever this structuring may turn out to be, it must be clearly seen by the observer and gradually made clear to the subject in order to give her an understanding of the situation and her position in it. Through such an analysis and understanding she can come to a clear realization of the meaning of her attitudes and actions in similarly structured situations.

This same rôle taking process takes place in trying to understand past or present non-interview situations. Jones tells of his domestic woes. If I should say as sincerely as he does what he says about his wife, I could not escape seeing her as a monster. Yet in my capacity as Jones I find I still live with this she-devil. In my Jones rôle supplied by his lengthy and vivid report, I lived through a crisis in which I mustered courage to leave her. After I left her I was lonely and "lost." I came back. I was very crestfallen but still angry and put on a front. I told her I was only coming back on account of my duty to the children. However, I remember dreaming that night that I was on a journey—a dark night, a strange road, and I was lost. I walked and walked and came to a house in a dark woods all lighted up and warm. There was good food. There was still the fear of the dark woods but inside the house I felt at home and comfortable. (One of his terms of endearment was "light of my life"). As I live through more and more of Mr. and Mrs. Jones's experiences I am able to get a firm grasp of the two or more self-other patterns in which their behavior is cast and in which it becomes intelligible. When I am able to communicate this perception of the actual ongoing situation and the rôles of the two people in them, their own behavior,

( 368) attitudes, and resentments are more intelligible to them, since they are seen now as functional parts of the rôles they are playing in the actual situation rather than in the stereotyped relationships in which they are seeking to function as husband and wife.

Note that I do not seek to determine what kind of man Jones is in the sense of what amounts of specified attributes he possesses. I seek to delineate the expectancy-response patterns with which he operates in the situations studied.  When these patterns and their development are clear, we should be in position to make reasonably precise predictions about Jones's tendencies in specified contexts. As I understand Moreno's psychodramatic procedures, he is doing the kind of analysis I am suggesting by using actual dramatic representations of situations, using actual persons as auxiliary egos. His efforts in this direction should be extremely valuable in testing the validity and utility of situational analyses. Unfortunately, so far as I know, there is yet no systematic check on the accuracy of prediction made by deliberate, clear-cut and skillful use of the procedures I am suggesting. My own experience, the work of J. L. Moreno, H. S. Sullivan, W. L. Warner, and others who represent the trend to situational as over against attribute analysis all lead me to be optimistic about the possibilities of this approach.

In addition to the difficulties in achieving precision and quantification, of "attribute" habits of thinking, and of the lack of a "situational language," there is one great difficulty inherent in this method. I refer to the opportunity for unconscious projection of the observer's own self-other (expectancy-response) patterns on to his cases. Two observations should be made here. One is that objectivity as we have ordinarily conceived of it undergoes considerable modification when one uses the method of rôle taking. One does not get his data by being detached. He is implicated in a social act and plays as real a part in it as the subject. It is only by reading the impact of the action on his own personality system that he is able to analyze what is happening. True detachment and "objectivity" in the old sense cannot yield the knowledge we need about human relationship. At the same time a case analyst must achieve a disciplined observer rôle that can take account of what is going on without being completely submerged in the part he is playing in the situation. This is a difficult accomplishment, and I suspect that Moreno's dramatic procedures may be necessary in difficult cases. Present training for use of case-study methods appears haphazard and for the most part naive as compared with the rigorous discipline and self-knowledge which the above requirement indicates.


A second observation is that projection actually does take place and is unavoidable. The more clearly this is recognized the safer and more valid the method is likely to become. What seems to happen to me in the soundings I make trying to get the subject's structuring of the situation is that I seek to place myself in the situation described by the subject and to construct what I would see if I behaved as he did or said he did. The self with its perspective of expectations which I find emerging I then, so to speak, "accuse him of being" or I "try on" him. In other words, I project on to him what I experience. I need to be exceedingly careful not to become so convinced of this self, so satisfied and secure with it because it is so intelligible—such a good hypothesis—that I am blinded to the poor fit when I try it on the victim. What happens is that I try to partially identify myself with the case in a specified situation; I take a reading of what happens to the identified me in conception of self and expectancies of others. This frame I then try on the case. If it seems to account for most of the attitudes and behavior and, most important, if it enables me to predict accurately for similar situational events in his experience, past or future, then I am fairly certain of the results of my identification-projection. If the fit is poor, I must identify and project again. From this it will be seen that prediction is an essential part of the method and not merely a desirable end result. It is an essential part of an operational statement of the identification-projection or empathetic introspective method.

I have stated what I consider the most distinctive method in case-study analysis. I do not mean to imply that all case-studies are done with this method. (I would suspect that most of those engaged in research and therapeutic work requiring them to work with individual cases would question the accuracy of the foregoing statement.)

Much of the study of cases is aimed at isolating syndromes and typical personality patterns which experience has shown to be correlated with certain resulting behavior, problems in adjustment, success or failure in some activity, or what not. This process requires knowledge of symptomology and experience in case diagnosis, but ordinarily does not require the degree of empathetic skill which the procedures I discussed above require. It is essentially an informal statistical procedure and there is no reason why in time it may not be reduced to more explicit statistical operations. I should like to point out that this is a legitimate part of work with cases. The more intensive empathetic methods usually lead to 'a the establishment of knowledge on the basis of which the less

( 370) demanding diagnostic classification work may be done. it is at this stage of the knowledge about cases that the method merges into or lends itself to explicitly statistical tests of knowledge arrived at by case-study methods.

The insights and hypotheses come from the rôle taking process. The categories and functional relations so discovered are then used at the syndrome classification and diagnostic level. This is as it should be, but the difficulties of the first method, and the relative economy of the second make for a tendency to set up categories and types and to operate with those alone and never press explorations further. Unless insight and analytic skill are constantly kept fresh by frequent use of the rôle-taking process, we drift into static classification, the elements of which tend to become reified entities while our abilities to see actual dynamics of a case become correspondingly low. It is one of the tragedies of case research that the valuable insights of a skillful investigator are often taken over by followers who then proceed to apply them in a rule of thumb, symptom-tagging style—witness the history of psychiatry and psychoanalysis. [4] It is in this manner that knowledge frequently becomes sterile and actually blind. In my opinion, it is only by frequent recourse to the empathic processes of studying cases that we expand our hypotheses to cover human interaction and personality organization more completely. With these tentative formulations as guides we are then able to use fruitfully repeated observations and statistical manipulations for verification and for more efficient prediction.


  1. See L. S. Cottrell, Jr. and Ruth Gallagher, "Important Developments in American Social Psychology During the Past Decade," Sociometry, Vol. 4, Nos. 2 and 3 (May and August) 1941 for a statement of the main currents of change in contemporary social psychology.
  2. These suggestions were made to Paul Horst and Paul Wallin for their sections on suggestions for research in the forthcoming S.S.R.C. monograph on prediction of social adjustment. I do not know how much of the suggestion was used.
  3. The differences are of course also important. In friendship relations, the values and the maintenance of the relations are in the focus of awareness. The understanding gained is incidental and frequently not entirely conscious. Much of the time we neither know nor care how much we "know" about our friend. In the case analysis the focus of attention is on the very things that are incidental in the friendship.
  4. It is just this type of incrustation that H. S. Sullivan is breaking through in his efforts to formulate an interpersonal theory of psychotherapy. See his lectures, Conceptions of Modern Psychiatry: The First William Alanson White Memorial Lectures, Psychiatry, Vol. 3, No. I, February , 1940.
    It should be stated here that the point of view represented in this paper was developed independently of Sullivan's formulations. My first published effort at applying this point of view was an article entitled "Rôles and Marital Adjustment," Publication of the American Sociological Society, Vol. 27, 1933, pp. 107-115. In that article I indicated the chief sources of my own thinking. It is interesting to note the convergence toward this point of view of many independent workers. See Cottrell and Gallagher, op. cit.

Valid HTML 4.01 Strict Valid CSS2