Issues in Research in the Psychoanalytic Process, Part 4
In keeping with this, the psychoanalytic discourse on prediction and on the use of prediction to cope with the problem of circularity has not been one-sidedly pessimistic. One of us has elsewhere (Wallerstein 1964) and with collaborators (Sargent et al. 1968) reviewed the psychoanalytic literature on the place of both the principle and the tool of prediction in psychoanalytic investigation, developing a rationale and structure for the formal use of predictions to more systematically and precisely link the data of psychoanalysis to the theory of psychoanalysis, and thereby creating a wedge by which to test and extend the theory. A manual has been elaborated for such usage, together with a fully-written-out case illustration, including the explicit formalized predictions made in that case and the process by which these predictions were empirically tested. As stated elsewhere (Sargent et at. 1968), despite the difficulty in making them, clinical diagnostic and therapeutic work rests on predictions that are inherent in the clinical undertaking. To quote:
Every responsible action in diagnosis and treatment involves one or more predictions derived from clinical experience or from theoretical hypotheses. When the clinical team in a case conference assigns a patient to treatment, a prediction is implied; the course recommended is expected to benefit the patient in certain ways, some of which are more, some less, spelled out in case discussion. In such deliberations, many other predictions are made, implicitly or explicitly, relating to possible contraindications, vicissitudes of the treatment course, and/or specific outcomes to be hoped for. Furthermore, clinical predictions, followed by observations of the course and outcome of therapy, provide myriad potential experiments which could test the hypotheses by which the predictions and treatment recommendations were guided. (p. 3)
Thus prediction is actually a pervasive, even universal, clinical phenomenon, usually implicit, and, as such, unremarked. The research task is to make explicit, so as to set up the conditions for formal testing, a range of predictions in a sample of patients entering psychoanalytic therapy, predictions relating, in the project discussed, to ‘the anticipated course and outcome of the recommended therapy, the nature of the problems to arise in the therapy in terms of expected transference paradigms, major resistances, and foreseeable external events that might (favorably or unfavorably) be expected to bear on the treatment course, and prognostic estimates in regard to expected or hoped-for changes in symptoms, in impulse-defense configurations, in manifest behavior patterns, and in level and nature of achieved insights’ (Wallerstein 1964, p. 684). It is at this point that critical control of circularity enters by setting down in advance the entire predictive complex of conditions, predictions proper and assumptive clauses, together with the predetermined evidence, in fact or in judgment, that will subsequently be necessary in order to sustain or refute the predicted outcome. It is by thus forcing the whole sequence of statements and supporting reasons in advance that observation is controlled and post hoc reconstruction, according to which almost any outcome can be plausibly rationalized in terms of a retrospective weighing of contending forces, avoided.
In this way, despite the many conceptual and practical difficulties in implementation, the systematic use of the predictive method can overcome the problems of circularity and permit us to conduct ‘experiments in nature’ (that is, where we do not control the antecedent conditions but designate their presence and hypothesize about their consequences) within the clinical research context.
This is not to say that prediction is the only way to avoid circularity or the only way to guard against confounding and error. For example, much of child analytic research is based on the direct observational method applied to the study of child development. Data from such observational and longitudinal study not only supplement but also check the retrospective and reconstructive data derived from the therapeutic process, whether of adults or of children. The congruence of the formulations derived from the data of the two independent, observational sources can be assessed. Additionally, and from within the psychoanalytic method proper, the fate of Freud’s original traumatic (seduction) theory of the psychoneuroses is a demonstration not only of the major fallibility of the retrospective method but also of the capacity of the superior mind to discern the increasing deviations from reality that such a false formulation progressively imposes and, within the method, turn it to a successful reformulation more loyal to reality; that what was once considered fact, an experiential vicissitude, is now to be considered fantasy, a maturational unfolding of an internal drive representation. This whole process of error and of rectification took place purely within the classical psychoanalytic method, without benefit either of ‘outside confirmation’ or of predictive safeguard.
Finally, as a last major issue to be discussed, we turn to the question: To what extent can one generalize from an N of 1 or of very few? Probably no one would cavil with Strupp’s (1960) admiring remark, from the perspective of an empirical psychotherapy researcher, that it is a tribute to Freud’s genius ‘that he succeeded in making valid generalizations on the basis of exceedingly small samples’ (p. 63). Yetit is a fundament of modern empirical science that generalization across cases requires a sampling of many. As Janis (1958) put it:
An obvious weakness of a single case study … is that it can provide no indication as to whether the relationship applies to all other, many other, a few other, or no other human beings. Thus, even when a causal sequence is repeatedly found in a given person, the investigator cannot be sure that his findings can be generalized to any broad class of persons because the relationship may occur only in an unspecifiable, restricted class of persons sharing a unique constellation of complex predispositional attributes. [Or] to put the matter in more technical language … a major limitation of the findings is that there are zero degrees of freedom with respect to individual differences, even though each finding may be based on hundreds of degrees of freedom with respect to the samples of the subject’s behavior that enter into the correlation between the independent and dependent variables. (pp. 23-24)
In the face of such seemingly elementary considerations, how can one account for the success of so many N = 1, or N = very few, studies in psychological science? Support can be found both among statisticians and clinicians for the position that for many clinical research purposes more can, perhaps, be learned from a smaller than from a larger number of cases. From the statistical point of view, Edwards and Cronbach (1952) have stated: ‘Information gained from an experiment mounts more or less in proportion to factorial n where n is the number of uncorrelated response variables. By this estimate five tests can report 120 times as much knowledge as a single test in the same investigation! … Effort to refine measurement has the same beneficial effect on the power of an investigation as adding to the number of cases .. .’ (pp. 55-56). From the clinical point of view, Gill (in Brenman 1947) has stated:
The clinical researcher must compare situations in which a number of variables are varying at once, thus differing from the experimentalist who can attempt to hold all the variables but one constant. The clinical worker must find patterns and principles of relationships which must be true to account for the observed variations. The more simultaneously varying variables he must deal with, the more uniquely determined is the hypothesis he must deduce to fit the observations … Instead of saying that many variables force a multiplication of cases,we would say that they make necessary only a relatively few cases. (pp. 220-226)
The literature of experimental psychology has, in fact, also reviewed the variety of circumstances under which the even more limiting condition of N = 1 can still mark an appropriate and useful, sometimes the only possible, research strategy. Dukes (1965), in a paper entitled N = 1, discusses the conditions that warrant employing an N of 1 under four headings:
(1) If uniqueness is involved, a sample of one exhausts the population. At the other extreme, an N of 1 is also appropriate if complete population generality exists (or can reasonably be assumed to exist). That is, when between-individual variability for the function under scrutiny is known to be negligible … [(2) The dissonant nature of the findings]: In contrast to its limited usefulness in establishing generalizations from ‘positive’ evidence, an N of 1 when the evidence is ‘negative’ is as useful as an N of 1,000 in rejecting an asserted or assumed universal relationship … [(3)When there is a limited opportunity to observe: When] individuals in the population under study may be so sparsely distributed spatially or temporally that the psychologist can observe only one case, a report of which may be useful as a part of a cumulative record [and] (situational complexity as well as subject sparsity may limit the opportunity to observe) … (4) Problem-centered research on only one subject may, by clarifying questions, defining variables, and indicating approaches, make substantial contributions to the study of behavior. (pp. 77-78)
In this connection, Ebbinghaus’ classic and still fundamental work on memory, done in 1885, on only one subject, himself, was quoted.
Granting then the critical importance and even the established value of the intensive study of one or a few cases in order to discover relationships or under special circumstances, is there some point at which psychoanalytic research on the therapeutic process must become large scale in order to ‘prove’ the hypotheses developed on the few cases? If certain mechanisms can be demonstrated as components of change in the course of a completed psychoanalysis under study, just what is really involved in being able to prove that this is generally true in successful psychoanalyses? In part, the answer to this is linked to the distinction delineated by Bakan (1955) between general-type propositions and aggregate-type propositions. General-type propositions assert something which is presumably true of each and every member of a designable class. They are given increased support with each successive positive instance, though never ‘proved’ in a formal sense. With the first negative instance they are either overthrown in toto or, more likely, the class boundaries must be further circumscribed to effect a new, more limiting definition that excludes the negative event and maintains the new, more narrowed, and hence more precise, proposition. In this sense, the ‘truth’ is progressively approached via a succession of single cases. The situation is, of course, very different with aggregate-type propositions which assert things presumably true only of the class considered as an aggregate and where increasing exactitude and significance accrue with the increasing size and representativeness of the sample.
Nonetheless, we can and do justify our efforts on single or few cases and the literature cited indicates that there is cogent scientific justification for this position depending on purpose and on circumstance of the study. Intensive case-by-case study may ultimately not be required for the testing of psychoanalytic propositions and, at such time, appropriate canons must be devised for the transition from hypothesis-formulating studies of single or few cases, intensively scrutinized, to hypothesis-testing studies on the appropriately larger samples. But in clinical research we do start with an approximation of N = 1 for at least two good reasons. The first is that we want to make reasonably sure that we are not oversimplifying the world in abstracting certain processes but are rather appropriately simplifying the world by identifying salient, invariant relationships. We prefer to seek these relationships within individual, intensively studied cases because this provides a needed anchorage in the complexity of clinical reality and, therefore, some protection against naive conceptions of how things hang together. Second, psychoanalytic researchers are historically just now in the process of developing research methods which formalize, systematize and render explicit the dimensions of clinical practice and clinical inference that up to now have remained informal, implicit and intuitive (the ‘art’ of psychotherapy). In order to make sure that these methods do not distort the processes they are intended to investigate, the methods have to be devised and studied in relation to well-understood individual cases. As we become more secure about our research technology in this area, we may imagine being able to deal with more extensive designs for cross-case studies.
In conclusion we would like to summarize the purposes of a discursive journey through the dilemmas posed by the many issues in research into the psychoanalytic process here discussed. We have attempted to confront side-by-side, with reference to both theory and practice, two questions relevant to our central thesis: Is it necessary to conduct more formalized and systematized studies of the therapeutic process in psychoanalysis? and Is such an endeavor possible? We maintain, on grounds that we hope are cogent and persuasive, that the answer to both questions today is an emphatic yes! And yet we also hope that we have not sought, however unwittingly, to minimize the manifold conceptual and technical difficulties encountered by the investigator who seeks to combine clinical relevance with scientific rigor.
Our central conviction is that the informal clinical case study, in spite of its compelling power, has certain real and obvious—indeed formidable—scientific limitations. The major task for research in the clinical field and the clinical process is the formalization of this highly artistic method into a disciplined research instrument which transcends our clinically satisfactory operating criteria of inner coherence and clinical conviction bred of experience, while approaching the scientific criteria of systematic replicability. Psychoanalysis has historically underrated these complex problems of hypothesis testing and verification. In part, this has been because it has not wished a sterile scientism to obstruct genuine exploratory and investigative zeal. But, in part, this has grown out of an historical tradition—and a particular constellation of scientific problems which were conducive to that tradition—which has placed exclusive reliance on a single method of naturalistic observation by trained participant-observers. It is our belief that it is appropriate, feasible and very necessary to supplement that tradition now in order to make further progress towards the solutions of the problems which we have here so urgently raised.
Bakan, D. (1955) ‘The general and the aggregate: A methodological distinction’. Perceptual and Motor Skills, 5,211-212.
Bellak, L. and Smith, M.B. (1956) ‘An experimental exploration of the psychoanalytic process: exemplification of a method’. Psychoanal. Quart., 25, 385-414.
Bergman, P. (1966) ‘An experiment in filmed psychotherapy’. In L.A. Gottschalk and A.H. Auerbach (eds) Methods of Research in Psychotherapy. New York: Appleton-Century-Crofts.
Brenman, M. (1947) ‘Problems in clinical research (Round table, 1946)’. Amer. J. Orthopsychiat., 17, 196-230.
Bronner, A.F. (1949) ‘The objective evaluation of psychotherapy (Round table, 1948)’. Amer J. Orthopsychiat., 19, 463-491.
Carmichael, H.T. (1956) ‘Sound film recording of psychoanalytic therapy: A therapist’s experience and reactions’. Iowa State Med. Soc., 46, 590-595.
Chassan, J.B. (1957) ‘On the unreliability of reliability and some other consequences of the assumption of probilistic patient-states’. Psychiat., 20, 163-171.
Dukes, W.F. (1965) ‘N = 1′. psychol. Bull., 64, 74-79.
Edwards, A.L. and Cronbach, L.J. (1952) ‘Experimental design for research in psychotherapy’.] Clin. Psychol., 8, 51-59.
Erikson, E. (1958) ‘The nature of clinical evidence’. Daedalus, 87, 65-87.
Ezriel, H. (1951) ‘The scientific testing of psychoanalytic findings and theory’. Brit.] Med. Psychol., 24, 30-34.
Ezriel, H. (1952) ‘Notes on psycho-analytic group therapy: interpretation and research’. Psychiat., 15, 119-126.
Freud, S. (1912) Recommendations to Physiciam Practising Psychoanalysis. Standard Edition. London: Hogarth Press.
Freud, S. (1920) The Psychogenesis of a Cctse of Homosexuality in a Woman. Standard Edition. London: Hogarth Press.
Gill, M.M., Simon, J., Fink, G., Endicott, N.A. and Paul, 1. (1968) ‘Studies in audio-recorded psychoanalysis. I. general considerations’.] Amer. Psychoanal. Ass., 16,230-244.
Glover, E. (1952) ‘Research methods in psychoanalysis’. Internat. J. Psycho-Anal., 33, 403-409.
Haggard, L. A., Hiken, J.R. and Isaacs, K.S. (1965) ‘Some effects of recording and filming on the psychotherapeutic process’. Psychiat., 28, 169-191.
Hartmann, H. (1958) The Psychoanalytic Study of the Child. New York: International Universities Press.
Hartmann, H. (1959) ‘Psychoanalysis as a scientific theory’. In S. Hook (ed) Psychoanalysis Scientific Method and Philosophy. New York: New York University Press.
Janis, I.L. (1958) Psychological Stress. New York: John Wiley and Sons.
Knapp, P.H., Mushatt, C. and Nemetz, S.J. (1966) ‘Collection and utilization of data in a psychoanalytic psychosomatic study’. In L.A. Gottschalk and A.H. Auerbach (eds) Methods of Research in Psychotherapy. New York: Appleton-Century-Crofts.
Kris, E. (1947) ‘The nature of psychoanalytic propositions and their validation’. In S. Hook and M.R. Kowitz (eds) Freedom and Experience: Essays Presented to Horace] Kallen. New York: Cornell University Press.
Kubie, L.S. (1956) ‘The use of psychoanalysis as a research tool’. Psychiat. Res. Rep., 6, 112-136.
Loevinger, J. (1963) ‘Conflict of commitment in clinical research’. Amer. Psychologist, 18, 241-251.
Lustman, S.L. (1963) The Psychoanalytic Study of the Child. New York: International Universities Press.
Moore, B.E. and Fine, B.D. (1967) A Glossary of Psychoanalytic Terms and Concepts. New York: Amer. Psychoanal. Assn.
Rapaport, D. (1960) The Structure of Psychoanalytic Theory: A Systematizing Attempt J. Psychological Issues, Monogr. 6). New York: International Universities press.
Sargent, H.D. (1961) ‘Intrapsychic change: Methodological problems in psychotherapy research’. Psychiat. 24,93-108.
Sargent, H.D., Horwitz, L., Wallerstein, R.S. and Appelbaum, A. (1968) Prediction in Psychotherapy Research: a Methodfor the Transformation of Clinical Judgments into Testable Hypotheses [Psychological Issues, Monogr. 21). New York: International Universities Press.
Schafer, R. (1967) Projective Testing and Psychoanalysis. New York: International Universities Press.
Schlesinger, H.J. (unpubl. ms.) ‘Discussion of studies in audiorecorded psychoanalysis 1. General considerations by Merton M. Gill et al.‘. Meeting of Amer. Psychoanal. Assn. New York, Dee. 1967.
Schmidl, F. (1955) ‘The problem of scientific validation in psychoanalytic interpretation’. Int .J. Psycho-Anal., 36, 105-113.
Seitz, P.F.D. (1966) The Consensus Problem in Psychoanalytic Research. A.H. Auerbach. New York: Appleton -Century-Crofts.
Shakow, D. (1959) ‘Discussion’. In E.A. Rubinstein and M.B. Parloff (eds) Research in Psychotherapy. Washington, D.C.: Amer. Psychol. Assn ..
Shakow, D. (1960) ‘The recorded psychoanalytic interview as an objective approach to research in psychoanalysis’. Psychoanal. Quart., 29, 82-97.
Strupp, H.H. (1960) ‘Some comments on the future of research in psychotherapy’. Behavior. Sci., 5, 60-71.
Strupp, H.H., Chassan, J.B. and Ewing, J.A. (1966) ‘Toward the longitudinal study of the psychotherapeutic process’. In L.A. Gottschalk and A.H. Auerbach (eds) Methods of Research in Psychotherapy. New York: Appleton-Century-Crofts.
Waelder, R. (1962) ‘Psychoanalysis, scientific method and philosophy’. J. Amer. Psychoanal. Assn., 10, 617-637.
Waelder, R. (1963) ‘Psychic determinism and the possibility of predictions’. Psychoanal. Quart.,32, 15-42.
Wallerstein, R.S. (1964) ‘The role of prediction in theory building in psychoanalysis.J. Amer. Psychoanal. Assn., 12, 675-691.