In response to:

The Unknown Freud: An Exchange from the February 3, 1994 issue

To the Editors:

In answering the objectors to his “The Unknown Freud,”. Frederick Crews wrote [Letters, NYR, February 3]: “To my knowledge, no modern analyst has renounced the cardinal Freudian investigative tool of ‘free association,’ which is inherently incapable of yielding knowledge about the determinants of dreams and symptoms” (p. 41, col. 2). Not to my knowledge, either, and for good reason: because the alleged inherent incapability has not been demonstrated. Instead of attempting to do that, Crews simply refers us to Macmillan’s Freud Evaluated, Chapter 15, which he calls “a devastating critique.”

The burden of that chapter is easily summarized. Macmillan reaches the radical conclusion that there “can not be any guidelines to how these data [those of the free association method] should be interpreted” (p. 549) because “the absence of a second script prevents any rules from ever being formulated” (p. 564). That is, taking as inherent to the psychoanalytic method Freud’s notion that for every manifest dream, there exists a latent dream the text (script) of which can be discovered, Macmillan makes the quite plausible argument that we have no way of ever directly accessing latent dreams. If that were the only possible way of devising guidelines for analyzing psychoanalytic data, then the conclusion—the inherent incapability of which Crews speaks—would indeed follow. But it is not, and hence Macmillan’s case collapses.

About fifty years ago, when I was learning the art of diagnostic psychological testing,1 I was confronted with the task of reaching valid conclusions about the persons studied in personality research and about psychiatric patients on the basis of their free verbal responses to such projective techniques as the Rorschach ink blots and the pictures of the Thematic Apperception Test (TAT). The problem with the resulting data was that they were much more like free associations than like the results of administering intelligence tests or personality questionnaires. My mentors, Henry A. Murray and David Rapaport, were at that time fully aware of the pitfalls of psychoanalytic interpretation, which make analysts disagree with one another to an alarming extent, and had discovered a way out. I have since tried to detail it and to show its basic similarity to scientific method.2

The starting point is the realization that neither free association, the TAT or the Rorschach is a test—the sort of attempt to measure a specific aspect of personality or ability that we mean by that term. Hence, there is no single criterion, no independent measure of the same construct against which we can check the instrument we wish to validate. Instead, free association (and the projective techniques inspired by it) are sources of data, like interviewing or direct observation of behavior. Neither Crews nor Macmillan seems to have grasped this basic fact. But how should you work with such data? You can do so casually, irresponsibly, or in earnest ignorance of ways to do it with any control over manifold sources of error, as most psychoanalysts do. Or you can record your data objectively and analyze them in disciplined ways.

For that, you have to be clear on the difference between getting insights and checking them.3 Alas, not only the practitioners of psychoanalysis but many of their critics seem unaware that every science has two phases, which make quite different cognitive demands: first you have to get hunches, insights, or hypotheses; and then you have to test them. In the former, creative phase, there are and can be no complete rules; but that is true in all sciences. Kekule got the hypothesis of the ring structure at the heart of organic molecules in a dream-like reverie, but he realized that it was only a bright idea until it was checked against independent data. Not so easy, for at the time there did not exist the chemical equivalent of a “second script”—a direct method of seeing the carbon ring. So he had to draw inferences about what would happen in various chemical experiments if his surmise was correct or if other possible structures existed, and then check them out.

To work scientifically with psychoanalytic data, you have to make inferences from them (interpret them) and then attempt to refute those inferences by confronting them with additional, independent information. That is a great deal easier for clinical psychologists, who have at their disposal data from a variety of techniques of gathering relevant information, than it is for practicing psychoanalysts. It has been difficult for the latter to take seriously their plight: treating patients exposes them to extraordinarily rich data, but the nature of the situation restricts them to forming hypotheses for others to verify or refute. Moreover, scientific work on free associations requires that they be objectively recorded. Hence, Hartvig Dahl has for decades been attempting to persuade other psychoanalysts to join in assembling a library of tape-recorded psychoanalyses.4 With the complete transcript in hand, one can get consensual judgments on the degree to which any particular segment of data are contaminated by suggestion. All that Grünbaum has done—pace Crews—is to make a strong a priori case that psychoanalytic data are thus contaminated; he has not proved that any particular set of free associations are useless as the basis of a given type of inference.5 All scientific data are subject to contamination; that doesn’t make science impossible, only difficult. Scientific methodology is the study of ways to reach useful and asymptotically valid conclusions on the basis of fallible data gathered by fallible human beings.


Many critics of psychoanalysis (Crews and Macmillan are not the only ones) make the understandable error of believing that it is intrinsically impossible to do disciplined, responsible scientific work with free verbal data on aspects of psychoanalytic theory. Having devoted most of my career to learning how to do it and, with the aid of my colleagues at New York University’s Research Center for Mental Health, carrying out a good deal of such research, I can testify that it is not easy, quick, or inexpensive; it is no longer fashionable or easy to fund; and the undertaking cannot be recommended to a young scientist eager to get ahead in an academic career—the payoff in publishable findings is too slow. Hence, the future of psychoanalysis as a science is hardly rosy; but neither is it merely a dream of self-deluded people.

Robert R. Holt

Professor of Psychology Emeritus, NYU

Truro, Massachusetts

To the Editors:

There is one issue that, I believe, is the most substantively central one in evaluating psychoanalytic theory or any other theory that is not adequately addressed in Crews’ original essay, in the letters responding to that essay, or in Crews’ reply to these letters. That issue is the heuristic value of psychoanalytic theory.

In presenting his indictment of psychoanalysis as a pseudoscience, Crews appears to assume that psychoanalytic theory belongs to the psychoanalytic establishment and that its parochial practices are decisive for determining the status of the theory. But psychoanalytic theory belongs to the intellectual and scientific community and, as I will show, has had a significant heuristic impact on that community. He also assumes—and is able to do so because broadsides substitute for careful argumentation—that psychoanalysis is a monolithic entity. It is not. It consists of a somewhat heterogeneous body of propositions, formulations, assumptions and hypotheses, some of which are foundational assumptions, some of which are indeed relatively immune to refutation, and some of which are eminently testable. For example, Freud’s claim that the major part of mental life goes on outside of conscious awareness could be seen as a foundational assumption which, by the way, is shared by most contemporary cognitive scientists. Freud’s ideas about life and death instincts can serve as a good example of a proposition relatively immune to empirical test. And as a final example, some of Freud’s propositions regarding repression and the wish-fulfillment theory of dreams seem eminently testable. (It seems clear that sufficient evidence has accumulated to lead one to conclude that the claim that unconscious wishes represent an invariant component of dreams has been falsified). But in order to make these distinctions, Crews would have to be more knowledgeable about and more interested in the details of psychoanalytic propositions than in other matters with which he is preoccupied.

A critical consideration in assessing the status of any theory is its heuristic value in generating research and further theory-building. This issue, far more important than any of the other matters taken up in his essay, Crews totally ignores. Let me provide merely one example: During the last number of years, an exciting and important body of research on “repressive style” and its correlates has appeared in scientific psychology journals and books.6 This work indicates that people who are characterized as employing a repressive style are more likely to show, among other things, higher physiological arousal (including higher blood pressure and cardiac rates) during stress, are more susceptible to certain physical illnesses (e.g., hypertension), and show poorer immune responses. Most important in the present context, this work is clearly and explicitly generated by Freud’s concept of repression. Furthermore, the research thus generated is likely to feed back and further operationalize and modify both the concept of repression and the hypotheses surrounding it. Crews does not appear to be aware of this kind of work. It certainly does not enter into his evaluation of psychoanalytic theory.

There may be and undoubtedly are other aspects of Freudian theory that have not been heuristic and, indeed, may be misleading and harmful. The task, then, for the intellectual and scientific community is to identify and either modify or reject these features of the theory. But such efforts are characterized by careful, discriminating, and detailed appraisals rather than wholesale condemnation or wholesale loyalties.


I would submit that in any serious evaluation of the status of psychoanalytic theory, the kind of heuristic impact I have briefly described is of far greater import and significance than the personal and admittedly juicier tidbits emphasized by Crews.

Morris Eagle, Ph.D.

The Austen Riggs Center

Stockbridge, Massachusetts

To the Editors:

I would like to correct some misconceptions which occur in J. Schimek’s response to Frederick Crews’s article “The Unknown Freud.” Schimek’s findings contained in his 1987 paper “Fact and fantasy in the Seduction Theory” were not “taken out of context and misused” in my book Seductive Mirage; nor did I “conclude from [his] arguments” that Freud made up all the sexual material he got from his patients. I wrote the first draft of my account of the infantile seduction theory episode in 1984; as I state explicitly in my book, the chapter in question was written completely independently of Schimek’s research. Hence I neither misused nor drew any conclusions from his paper, the existence of which I was unaware of until December 1988; the writings from which I drew conclusions were Freud’s own. Schimek is also mistaken in supposing that it was on the basis of the seduction theory episode that I drew the conclusion that Freud made up all the sexual material he got from his patients. (I would prefer to say that, though he certainly resorted to invention on occasion, he inferred most of that material on grossly inadequate grounds and misleadingly presented it as his “findings of analytic research.”) As my book demonstrates, the evidence for the dubious nature of almost all of Freud’s supposed sexual findings can be found throughout his work.

In regard to the seduction theory episode, it is a pity that James Hopkins, like Schimek, did not take the trouble to actually read my chapter on the subject before making assumptions about its contents. Had he done so he would have found (p. 20) that it contains an implicit refutation of his claim that when Freud later wrote of his female patients reporting paternal seductions in the context of his “error” he was referring not to his self-proclaimed “source of the Nile” discovery put forward publicly in the seduction theory papers of 1896, but to an unpublished transitional notion he adopted privately a little later. As Crews writes, this latter notion was no more than a speculative theory for which Freud determinedly sought analytic corroboration. Hopkins expresses scepticism about the “alleged” contradictions and anomalies to which several scholars have now directed attention, implying that for the most part they result from misinterpretations of Freud’s words. The anomalies are too numerous to cite here, but one example (the one Hopkins attempts to deal with) is the discrepancy between the original claim of the uncovering by analytic inference, in every one of his current cases (mostly women), of infantile sexual molestation by a variety of carefully listed and grouped assailants (but no fathers), and the belated (post-1924) claim that “almost all” of his female patients around that same period reported paternal childhood seductions. (The notion that the discrepancy can be accounted for by discretion on Freud’s part is totally incompatible with the facts and is disposed of in my book. (p. 20–21) It can also be seen that Hopkins’ purported explanation would have been inadequate even had it not had an erroneous basis.) Then there is the curious fact that it was only for the period during which he held to the seduction theory that Freud claimed to have received such numerous reports from his female patients. Again, one wonders why Hopkins apparently is untroubled by Freud’s claiming one hundred percent success (in uncovering repressed infantile sexual traumas) in 1896, an achievement unheard of in the annals of psychiatry in regard to the confirmation of any theory—and made even more remarkable by the fact that it was a theory he was to abandon within a short time!

It can scarcely be a coincidence that it was at precisely the time (1924–5) that Freud turned his mind to the detailed development of his theories of female sexuality that the phony story of reports from his early female patients of paternal seductions was first published, thereby underwriting the Oedipal theory. A corresponding only apparent coincidence occurs in relation to the belated (1925) psychoanalytic discovery that the first attachment of infant girls is to the mother. [SE, 19:251] In his second paper on female sexuality (1931) Freud suddenly announced the “very common” occurrence of [analytically reconstructed] phantasies of maternal seductions in regard to his female patients. [SE, 21:232] Within a short time the story had become that his female patients “regularly accuse” their mothers of seducing them. [SE, 21:238] The parallels with the seduction theory discrepancy discussed above are too obvious to labour the point, but they indicate that, far from that discrepancy being innocuous, as Schimek would have us believe, it is an early example of the fact (repeatedly demonstrated in my book) that Freud’s reporting of his clinical experiences is not to be trusted.

May I conclude with a suggestion (proposed by Arthur Koestler in a rather different context) for those who, while not being zealots, still remain under the spell of Freud. They should be sentenced to a year’s hard reading. Crews has supplied the initial booklist; in particular, anyone who has not read Macmillan’s comprehensive critique Freud Evaluated has scarcely begun to take the measure of current scholarly criticism of the Freudian enterprise. And there is the promise of more volumes to come from two other scholars, Han Israëls in Holland and, in Sweden, Max Scharnberg, who has already published three volumes of his projected opus The Non-Authentic Nature of Freud’s Observations (Acta Universitatis Upsaliensis, Uppsala Studies in Education, Uppsala, Sweden).

Allen Esterson

London, England

Frederick Crews replies:

In their long-standing concern to discriminate between the wheat and chaff in psychoanalytic theory, Robert R. Holt and Morris Eagle represent the highest standard of empirical responsibility to which the Freudian tradition can lay claim. It is telling, then, that neither of them appears capable of even addressing, much less confuting, the specific charges against psychoanalysis levied in my essay and in my subsequent response to critics.

The issue dodged in Holt’s letter is whether, as I claimed, free association is “inherently incapable of yielding knowledge about the determinants of dreams and symptoms” (italics added). Is there, even in principle, any trustworthy path of inference from a patient’s verbal associations to the causes of a given dream or symptom? My contention, cited but then ignored by Holt, is that you can’t get there from here. Let us review three well-established reasons why this is so.

In the first place, “free” associations aren’t free at all, since they have been amply shown to lack the imperviousness to suggestibility that Freud rashly ascribed to them. Freudian therapists and patients come to share a causal outlook predetermining the kinds of factors that both parties will consider significant, and both the patient’s verbal productions and the therapist’s thematically pointed selection among them cannot escape being influenced by that bias. As Morris Eagle himself once put it, “suggestion and compliance, however subtle and complex, are critical factors in generating these supposedly confirming data,” and thus “each therapist, whatever his or her theoretical persuasion, can reach the verdict that they constitute confirmations.”7

Second, a patient’s free associations point at best to current preoccupations that may or may not have begun before the dream or symptom that is to be explained. Even an enduring preoccupation is by no means necessarily a pathogen. Only by invoking Freud’s purported “logic of the unconscious” could one maintain the contrary, but that logic is precisely what stands in question here. Unfortunately, Freud “discovered” it by fiat and defended it, as his followers still do, merely by applying it interpretatively and treating the resultant speculations as corroborative findings. To my knowledge, the vast theoretical literature of psychoanalysis contains not a single cogent effort to show how a mere theme in a patient’s contemporary mind can be certified as an early determinant of that patient’s neurotic disposition.

And third, even if associations did point to potential causal factors, psychoanalysts would still face the obstacle of trying to decide which ones actually governed a particular dream or symptom. Freud himself acknowledged that some factors must be “suppressed by others because they are too weak, and they therefore do not affect the final result” (SE, 18:168). Moreover, Freud conceded, at the time of analysis there is no way of knowing which factor is dominant: “We only say at the end that those which succeeded must have been the stronger” (SE, 18:168). Thus psychoanalysts guess blindly at causes but later seek reassurance from the fact that a long concatenation of such hunches has yielded a self-consistent view of the patient’s case. Yet if the method of drawing causal inferences from free associations is wild in each individual occurrence, it must also be wild in the aggregate.

These circumstances help to explain why psychoanalysts, as Holt reports, “disagree with one another to an alarming extent.” Quite simply, they are winging it from start to finish. This is not just my judgment but, if one reads carefully, Holt’s as well. Freudian practitioners, he tells us, habitually neglect the crucial difference between forming hypotheses and adequately testing them, and the “hardly rosy” future of psychoanalysis as a science should be consigned to hands other than theirs.

Holt’s proposed rescue operation, such as it is, entails the extraction of taped free associations as data that can be assessed (presumably by investigators of a Freudian bent, since no one else would be interested) for the degree to which those data are contaminated by suggestion. Two questions about this scheme immediately spring to mind. First, how could psychoanalytically trained researchers be expected to mark the limits of suggestion in any given case when, as critics have decisively shown, the whole process of analyst-patient interaction is steeped in indoctrinating effects? And second, how much impact could such studies have on what Holt calls the “earnest ignorance” of analysts in their daily therapeutic work?

I am afraid that Holt’s invoking of methodological commonplaces from Hans Reichenbach and Karl Popper serves no purpose beyond a cosmetic “scientizing” one. More to the point is Popper’s judgment of psychoanalysis as it is actually conducted: “those ‘clinical observations’ which analysts naively believe confirm their theory cannot do this any more than the daily confirmations which astrologers find in their practice.”8 Though Holt does not say so explicitly, it is apparent that he agrees with Popper in this regard.

Morris Eagle’s letter can be answered more briefly, for it rests on a manifestly untenable premise. This is that “heuristic value,” or stimulation to further insight, ought to be weighed more heavily than intrinsic plausibility in the scrutinizing of propositions about the mind. When, I wonder, did Eagle arrive at this quixotic means of safeguarding his favorite notions? I find no trace of it in any of his previous writings—including the most recent—that evaluate points of psychological doctrine.9 Quite typical, for example, is a paper of 1983 where he denounces certain post-Freudian claims that “are either incoherent or without any evidential support.”10 According to his revised outlook, Eagle should instead have scolded people who fail to look beyond such narrow considerations as coherence and supporting evidence.

Psychoanalysis has indeed borne many heuristic consequences, but most of them have proved, to borrow Eagle’s words, “misleading and harmful.” Moreover, repression—Eagle’s one example of a heuristically fruitful psychoanalytic idea—has been involved in every pernicious instance that could be named, from the supposed causes of homosexuality and female irrationality through false memory syndrome. One may doubt whether the alleged conceptual breakthrough of “repressive style” is adequate recompense for all the inconvenience, anguish, and confusion wrought by Freud’s theory of repression. And even if it were, repression would still have to stand or fall on its own empirical merits, not on its having contributed an adjective to the name of a cluster of behaviors. As Eagle well knows, but now chooses to forget, the identifying of an authentic psychological “style” does not thereby validate any given explanation of how that style gets developmentally formed.

Finally, Eagle is mistaken in saying that I overlook the heterogeneity of psychoanalytic propositions. They are so heterogeneous, I have argued, as to constitute a self-condemning jumble of dubious and incompatible claims. I am concerned to explore the laxity of method that makes such chaos inevitable. So, too, is Allen Esterson, whose Seductive Mirage can be recommended to all readers who have sensed that Freud’s “scientific greatness” is now ripe for thoroughgoing reappraisal.11

This Issue

April 21, 1994