An Evaluation of Empirical Research Linked to Psychoanalytic Theory
[Paper presented on June 11, 1999, at the Annual Meeting of the Rapaport-Klein Study Group]
In the only instance where Freudís voice was recorded and made available to the public, he recounted very briefly the history of the psychoanalytic movement, adding that "resistance was unrelenting", concluding that the struggle for recognition was "not yet over". I doubt that even Freud could have anticipated that well over 50 years after that comment, the hostility to psychoanalytic ideas, some of it venomous, some of it badly uninformed, would still be in force.
Until fairly recently, for better or worse, psychoanalysis and psychoanalytic thought dominated American intellectual life. William Buckley reported that a survey of citations in the Social Sciences Citation Index for the period 1969-1977 found that Freud was the most frequently cited author with almost twice as many citations, 12,319, as the next most cited writer, Karl Marx, with 6,807 citations. Obviously something went terribly wrong because that is no longer the case. Citations to psychoanalytic literature are declining (Friman, Ien, Kerwin, and Larzelere, 1993). A report in the February, 1999 issue of the American Psychologist (Robins, Gosling, and Craik, 1999) deserves special comment. The authors reviewed the extent to which four content areas in psychology -- psychoanalysis, behaviorism, cognitive science, and neuroscience -- are cited by what the authors call "Psychology's flagship publications". Their grim conclusion was that "work appearing in contemporary psychoanalytic journals does not seem to be assimilated directly into the mainstream of scientific psychology. Rather, psychoanalysis seems to be a relatively self-contained camp..." (pp. 123-124).
Only half as many clinical psychologists, 18%, embrace psychodynamic theory today as they did in 1960, and there has been a wholesale abandonment of psychoanalysis in academic departments (Norcross, Karg, and Prochaska, 1997). Aside from the metropolitan New York area, probably fewer than 10 Ph.D. programs in clinical psychology, if that many, offer even the most meager training in psychoanalytic thought. The number of psychoanalytic patients apparently has been declining by 10% each year since 1979 (Jeffrey, 1998). Psychoanalysis is becoming marginalized in both academic life and clinical work, and the rate is accelerating.
The irony is that there is a great deal more science generated by psychoanalytic ideas and much more empirical evidence consistent with psychoanalytic ideas than is generally acknowledged. Kandel (Biology and the future of psychoanalysis: a new intellectual framework for psychiatry revisited. American Journal of Psychiatry, 1999, 156: 505≠524), who says he is a friend of psychoanalysis and who espouses a liaison between psychoanalysis and biology, in an article published only two months ago, wrote that "Although psychoanalysis has historically been scientific in its aim, it has rarely been scientific in its methods; it has failed over the years to submit its assumptions to testable experimentation" (p. 506). That bit of news, that research has failed to test psychoanalytic ideas, may come as a surprise to everyone in this room.
Even when it would be in the best interests of organized psychoanalysis to cite favorable empirical evidence, it has failed to do so. I do not recall one instance when a critic like Crews (1994) has been answered by referring to empirical data. The Freud exhibit in the Library of Congress had many interesting displays, but not one showing the extent to which there is scientific support for psychoanalytic ideas. Cartoons and film clips yes, data no. A meeting in Washington to honor the opening of the Freud exhibit featured a number of distinguished scholars, but not one to discuss the vast literature documenting the heuristic quality of psychoanalytic theories. Although simple organisms can defend themselves when attacked, the gate keepers of psychoanalysis continue to act as though there were no adequate counter arguments against external threats. This passive response to a very real danger is both dysfunctional and disheartening and is a mystery in search of a psychodynamic explanation.
Though Psychology 101 texts say otherwise, psychoanalytic thought has proven to be surprisingly heuristic and has probably generated more research than any other theory of personality, and perhaps more than all other theories combined. The eight volumes of the series I edit, "Empirical Studies of Psychoanalytic Theories" (Hillsdale, NJ: Analytic Press), contain 327 pages of references, with an estimated 6300 citations and Fisher's (1986) two volume work on the body image has about 3600 references. Even granting that some of these citations are repeated and some refer to clinical literature rather than experimental, it is obvious that a sizeable number of studies based on psychoanalytic ideas have been published. The psychoanalytic establishment deals with this fact by ignoring it. Critics like Crews (1996) deal with this fact by denying it. But anyone prepared to acknowledge this research must conclude that psychoanalytic ideas have fared better than one could have expected of theories now 100 years old.
For the remainder of my talk I will present a highly subjective review of a small segment of this empirical literature. I will first examine the extent to which psychoanalytic theories satisfy the criteria of a good scientific theory. A good theory must (l) be heuristic; (2) explain phenomena that cannot be accounted for by any other theory; (3) be confirmed by experimental test; and (4) be revised and then ultimately replaced by more accurate explanations.
I have already stated that psychoanalytic thought has been vigorously heuristic. As is appropriate for a good scientific theory, a number of its postulates have been tried and found wanting. Given the evidence available, I believe a number of Freud's hypotheses have outlived their original usefulness and should either be revised or abandoned. For example, his ideas about women have frequently been wrong. Despite his expectations, women are not more dissatisfied with their bodies than men; in fact, the opposite is true. By most standards, women show more superego strength than men, not less. Evidently women experience only one kind of orgasm, the clitoral, not two as Freud hypothesized (Masters & Johnson, 1966).
In fact, Freud often erred when he discussed women. He was not simply posing as an innocent when he asked Marie Bonaparte, "What does a woman want?". Evidently he did not know. In my own research, I have more frequently obtained significant results with men than with women. Staffan Sohlberg in Sweden recently told me this has been his experience, too. Perhaps the theoretical framework I use is essentially based on a male psychology. This fall I will have a graduate student review the research literature to determine the frequency with which significant results are obtained from male subjects compared to studies correctly predicting female behavior. Psychoanalytic theory might be more a psychology of males than many of us are comfortable with. Although sex differences are ubiquitous in studies of personality, many experiments either do not report the sex of the subjects or pool male and female data without first analyzing separately for sex.
Early psychoanalytic theory was also incorrect about the origins of many physical disorders. The analytic penchant for psychic determinism can nowhere be better seen than in the descriptions of such ailments as ulcers, asthma, allergies, and even astigmatism. The conjectures that "asthma is the strangled cry for help" and that astigmatism results from sexual looking (Fenichel, 1942) contain more poetry than truth and psychoanalytic case histories of sick patients frequently confuse a correlation for causation. The specificity concept, that each ailment results from unique psychological characteristics, has received little empirical support and should be discarded. Both stress and dependency are nonspecific contributors to a variety of physical disorders.
Childhood autism, once a favorite topic for psychoanalytic explanation and treatment, can now be more accurately seen as an organic problem (Volkmar, Klin, Marans, & McDougle, 1996). Blaming a cold, rejecting mother for autism is yet another example of confusing a correlation with causation. Patient, extremely minute behavioral interventions not only are more effective with the autistic than psychoanalytic treatment but also can reach more autistic children than psychoanalytic treatment.
Some of Freud's views about child development need to be revised (Beebe, 1986; D. Silverman, 1986). Ample evidence documents a functioning ego in infants, even in neonates, contrary to Freud's speculation in his 1914 paper On Narcissism. His statement "It is impossible to suppose that a unity comparable to the ego can exist in the individual from the very start", (pp. 76-77) is quite wrong. The neonate shows surprising ability to track moving targets, to differentiate a familiar from an unfamiliar stimulus, and to react meaningfully with the care giver. Further, children show signs of superego behavior earlier than Freud's suggestion that it does not arise until after the Oedipal Complex has been resolved.
Modern work on brain activity during sleep strongly suggests that dreams do not always represent efforts at wish fulfillment. Winson discusses this issue in the Barron, Eagle, and Wolitzky book (Interface of Psychoanalysis and Psychology. Washington, D.C.: American Psychological Association, 1992). Freud's notion, so clearly delineated in "Civilization and its Discontents" (1929) that energy comes in fixed amounts and that investment in one activity thereby reduces available interest in others, has no physiological basis. His instinct theory has not stood up well to experimental test. In their chapter in the Barron, Eagle, and Wolitzky 1992 book, Lachman and Beebe write that "Various assumptions about early development contained in the drive-discharge-conflict theory... are disputed by empirical studies" (p. 134). Modern data do not support Freud's contentions that homosexuality is a pathology and paranoid ideation is an effort to deal with homosexual impulses.
This is an incomplete list of instances where Freud's intuitions are inconsistent with empirical data. Unfortunately, there seems to be no official recognition that a basic overhaul of psychoanalytic theory is needed, that new ideas should be added, and invalid hypotheses dropped from the canon. Overall, however, it is remarkable that the list is not even longer given that these ideas were formulated close to 100 years ago by people who had no other source of data than their unsystematic observations of their patients and their own children.
For the rest of my talk I will discuss two areas of analytic thought -- unconscious processes and personality types -- that have withstood the vicissitudes of time and experimentation. I chose those topics for no better reason than I have found them interesting, susceptible to operational definition, and have worked with each with some success.
Perhaps the most crucial concept in all of psychoanalysis, one that had been controversial and subject to intense criticism until fairly recently, is that human behavior is marked by an active set of unconscious mechanisms that receive, process, integrate, and interpret stimuli below the level of conscious awareness. For years, academic psychologists dismissed the notion of unconscious mechanisms out of hand. William James (1890) in a felicitous statement declared that the unconscious "is the sovereign means for believing what one likes in psychology, and of turning what might be a science into a tumbling ground for whimsies" (pp. 163-164). Failure to have found experimental support for the concept of unconscious processes would have been extremely embarrassing to psychoanalysis. Happily, empirical data demonstrate that the unconscious need not be a tumbling ground for whimsies as James feared but is a fit subject for experimental inquiry.
Freud had two views of unconscious processes. His first description, as found in such publications as "The Interpretation of Dreams" (1899), "Papers on Technique" (1911-12), "Papers on Metapsychology" (1915), and the case histories, held that unconscious mental life was disorganized, chaotic, and had no purpose other than immediate gratification of impulses.
The second theory evolved gradually and piece meal and was published in a number of places, "Beyond the Pleasure Principle" (1920), "Inhibitions, Symptoms and Anxieties" (1925), "The Ego and the Id" (1922) and "Outline of Psychoanalysis" (1938). In these publications Freud suggested that the unconscious might be organized, purposeful, and directed to goals more complex than simple impulse gratification. Instead of being controlled by the pleasure principle, unconscious processes were now seen as regulated by higher mental functions.
The research evidence is unequivocal: The second theory predicts better than the first. A former member of this group, Lloyd Silverman (1983) led the way in demonstrating the influence of unconscious processing in his work with subliminal stimulation, showing repeatedly and with a considerable variety of psychiatric and nonpsychiatric patients that a message presented to subjects below the level of conscious awareness can affect meaningful units of behavior.
Apparently independently of Lloyd, social psychologists and cognitive psychologists also found that subliminally presented stimuli influence behavior. In social psychology, the mere exposure experiments of Zajonc (1968) demonstrate how subjects' behavior can be altered by simply and repeatedly exposing stimuli below the level of awareness. Cognitive psychologists (Jacoby, Woloshyn, & Kelley, 1989) have shown that the repeated subliminal exposure of a name will cause that name to appear familiar and famous. In perhaps the neatest demonstration of all, the cognitive psychologist, Fowler (1981), demonstrated that subjects could correctly select a word that meant the same as a subliminally exposed work, even though they could not identify a word that looked like the target word or a word that sounded like it. Fowler's experiment raises this tantalizing question: How can people know what a word means if they do not know what it looks like or sounds like?
At Buffalo, my students and I have successfully demonstrated a whole range of behaviors that can be affected by using an appropriate message, delivered subliminally. In Bornstein's dissertation (1987), male subjects formed an alliance with a person whose face had previously been exposed subliminally rather than with a person who face had not been shown. With Katkin, Bornstein, Poynton, and Reed I demonstrated (1991) that the message, "No One Loves Me", exposed subliminally, produced autonomic nervous system arousal as measured the subjects' electrodermal rate. The same phrase exposed supraliminally had no effect, nor did control phrases. Now we have proven that subliminal messages can reach into autonomic nervous system functioning, I hope some investigator will take the next logical step and try to determine if subliminal messages can affect other physiological mechanisms, such as immune functions or killer cell activity. Some years ago Lloyd Silverman suggested this as a possible area of experimentation.
Hansen's dissertation (1988) provided evidence that a fairly long sentence, "My Lover and I are One", can be processed and assimilated at the subliminal speed of 4 milliseconds, despite prior claims in cognitive psychology that a sentence as long as six words could not possibly be processed at a subliminal speed.
Patton's dissertation (1992) described the effects of exposing the subliminal message, "Mama is Leaving Me", to female bulimic and control subjects. After viewing either the experimental message or the control message, "Mama is Loaning it", subjects were asked to sample and rate crackers in a bogus tasting situation. Bulimic subjects who viewed the experimental message at a subliminal speed ate a mean of 19.40 crackers, whereas no other subjects in any of the other groups ate more than 9.6 crackers. In other words, by using an appropriate subliminal message we managed to induce the bulimics to eat more than twice as many crackers as any other subject. Ain't science grand?
Talbot's dissertation research (1991) showed male subjects either the experimental message, "Mommy is Leaving Me", or the control phrase, "Mona is Loaning it". The subjects also rated themselves on two different measures of their self-confidence in winning an attractive woman's affection. Talbot found that subjects exposed to the subliminal message, "Mommy is Leaving Me", expressed less self-confidence than all other subjects. Self-confidence was affected only by the experimental phrase exposed subliminally.
Another of my students, Jones (1994) tested one of Freud's more esoteric hypotheses, that a woman's wish for a penis can be replaced by a desire for a baby. Jones was convinced that an experimental finding by Greenberg & Fisher (1980) supporting the merit of Freud's penis = baby equation resulted from flaws and deficiencies in research design. Persuaded to conduct her own study of that hypothesis, she took Greenberg and Fisher's basic design and improved it. The original study did not include a supraliminal condition to ensure that an unconscious need was involved and she also added male subjects as an additional control. In addition, she used state-of-the art recording equipment to prepare and play her messages. The original script of Fisher and Greenberg unfortunately combined penetration imagery along with the pregnancy theme. Jones prepared and used separate penetration and pregnancy phrases.
Female subjects were randomly assigned to one of six experimental conditions: subliminal pregnancy message, subliminal original penetration message, subliminal revised penetration message, or supraliminal presentation of these three messages. Male subjects were placed in either the subliminal pregnancy or subliminal penetration conditions. Phallic imagery was scored from responses to either Form A or Form B of the Holtzman inkblot series. One form was administered before the auditory stimulation and one form after. We were all surprised by the results: At the .005 level, women who listened to the subliminal pregnancy message gave more phallic responses on post-test compared to pre-test than those exposed to the two penetration messages. The supraliminal presentation of the pregnancy message had no affect on phallic imagery. The Holtzman responses were also scored for sexual content: No message at any speed influenced sexual responses. The male subjects' Holtzman responses were not affected by any message at any exposure time. To this day, I am somewhat uneasy with these results but like a committed empiricist, I accept the data.
Perhaps the most important conclusion resulting from the many experiments using subliminal stimulation arose from Bornstein's meta-analysis of such investigations: Across studies, stimuli presented below the threshold of awareness have more potent effect than those shown supraliminally. I urge all of you to read Bornstein's lead article in the most recent issue of Psychoanalytic Psychology (1999) in which he carefully provides a compelling explanation for this phenomenon. Awareness can restrict behavior, as Spence and Holland (1962) demonstrated more than 30 years ago, whereas lack of awareness frees behavior from such constraints.
These studies on subliminal stimulation are consistent with observations from developmental psychology, dream research, and the work of Weiss, Sampson & the Mt. Zion Psychotherapy Research Group (The Psychoanalytic Process: Theory, Clinical Observation, and Empirical Research. New York: Guilford, 1986) all of which agree that contrary to Freud's first theory, unconscious processes are not chaotic, unorganized, and totally removed from reality considerations. Rather, the results from these diverse sources strongly support Freud's second theory about the unconscious, that there is a purpose, direction, and organization to complex mental activities that occur without conscious awareness. Sohlberg has made the interesting observation that cognitive psychologists study a dumb unconscious while psychoanalysts study a smart unconscious. Winson (1992), writing about dreams, summarized what research has demonstrated about unconscious functioning: "The unconscious is a cohesive, continually active mental structure that takes note of life's experiences. The scheme of interpretation is strongly influenced by early experiences as the result of the critical period of brain development... Dreams are not the guardians of sleep but are the products of a brain-based cognitive process" (pp. 354-355).
The other psychoanalytic concept my students and I have investigated is that human behavior can be characterized as showing stable, consistent patterns that are manifested over a number of different situations. Contrary to what one is likely to find in Psychology 101 textbooks, we believe that while situational influences cannot be ignored, behavior is the product of strongly internalized needs, that these needs direct behavior consistently, and that they can be assessed through appropriate techniques.
In studying this notion we relied on Freud's observation that behavior is organized and purposeful. In "Character and Anal Erotism" (1908) he noted that some people "were noteworthy for a regular combination of the following three characteristics. They are especially orderly, parsimonious, and obstinate" (p. 169). He also attributed these features to their refusal to become easily toilet trained because "they derive a subsidiary pleasure from defecating" (p. 170). The idea that the three traits of orderliness, parsimoniousness, and obstinacy should be found in the same people is easily tested by factor analysis and, in fact, has been examined at least six times by factor analysis and by several other correlational studies (Masling & Schwartz, 1979). All but one concluded that the three traits do form a constellation as Freud conjectured. Someone scoring high on one of these traits is also likely to score high on the other two. I know of no other theory to explain this result. The other part of Freud's hypothesis, that toilet training is somehow responsible for later behavior, has been frequently explored and generally been found wanting. Whatever causes anal behavior in adults, it is probably not the more formal characteristics of their toilet training.
The other important psychoanalytic character type -- the oral -- has been frequently investigated and just as often confirmed. I defined orality as the sum of all oral and dependent responses on the Rorschach test. Following the suggestions provided in Schafer's "Psychoanalytic Interpretation in Rorschach Testing" (New York: Grune & Stratton, 1954), I constructed a simple, lexical, highly reliable method for scoring oral and dependent Rorschach responses, though that surely could not have been Schafer's aim when he put together his suggested list. We used the group administration of the Rorschach in most studies, controlling for the number of responses by suggesting how many percepts we expected for each card. In using the group Rorschach we lost the richness of response of the individually administered test but gained in the ease of data collection. About 50 published articles have documented the ability of this simple Rorschach measure to predict behavior.
Freud, Abraham, Glover, and Fromm have all described what we can call the oral character. The central theme is that orals believe they cannot sustain themselves by their own efforts and find safety and comfort in being close to a nurturing other. The most obvious places to look for oral behavior are in eating and drinking. As expected, we found that those who report many oral and dependent images on the Rorschach tend to be obese and alcoholic, findings replicated in separate samples (Masling, 1986). Of course, we looked at the relationship between orality and smoking but have never found anything worth reporting. However, Ed Whitson (1983) demonstrated that smoking in females was positively correlated with an objective measure of hysteria and negatively correlated with obsessiveness. These two variables -- hysteria and obsessiveness -- separated smokers from non-smokers by almost a full standard deviation, an impressive difference in personality research. I emphasize that the results showed a correlation between smoking and hysteria because none of data we have ever collected permit any statement about causality.
Orals should be gullible, ready to rely on the judgement of others, ready to swallow anything. The Asch yielding experiments were a good place to investigate the extent to which oral subjects would ignore their own perceptions and would instead rely on the judgement of others. That is exactly what we found: More orals than non-orals yielded to the majority opinion in a standard yielding experiment (Masling, Weiss, and Rothschild, 1968).
They are also more compliant. O'Neill, Jayne, and I (1981) found that those students in the Psychology 101 subject pool who participated in the required experiments early in the semester had higher oral scores than those who waited until the last minute. Later Bornstein and I (1985) replicated these results with another student sample. When we divided the semester into four quarters, we found a straight line relationship between orality and date of participation in the experiments -- the earlier in the semester the greater number of oral and dependent images reported on the group Rorschach. It is evident that psychologists who draw from the subject pool early in the semester are studying different kinds of people than those who obtain their subjects later.
Weiss and I (Weiss, 1969) extended this research into the social psychology of the experiment by allowing subjects to infer what response we expected of them. In her dissertation, Weiss told one group that the study was part of a national effort to establish norms for visual perception and that college students like them tended to overestimate when asked to guess how many dots were briefly exposed on a screen. She informed another group that she expected them to underestimate the number of dots. Prior to the perceptual task she selected subjects based on their responses to the Blacky Test. Weiss found that female orals tried to be "good" subjects, their guesses following her suggestions. Some subjects, however, responded 180 degrees differently from what she had told them the norms were. Happily and to a statistically significant extent, these subjects tended to be anal. Weiss and I called the need of some subjects to oppose the hypothesis, "The Screw You" effect, though for some reason most text books refer to it as "deliberate subject noncompliance".
If orals believe they can survive only by obtaining external supplies, they would require accurate interpersonal skills to know how and when and from whom to get their needs met. My students and I conducted five separate investigations of this hypothesis, obtaining significant results each time. The basic design called for subjects to complete a personality or interest questionnaire as they thought the target person would. We found that oral subjects could predict the responses of housemates, classroom instructors, and casual acquaintances better than the low oral subjects. In general, male orals were better at this task than females, particularly if the target of perception were also male.
One experiment is particularly relevant to members of this group because with Shiffner and Shenfeld, I hypothesized (1980) that clients would observe their therapists as closely as therapists were observed them. When clients seen in our department clinic completed their third therapy session we asked them to predict how they thought their therapists would complete an interest questionnaire. In our small sample of 11 clients we found that the high orals predicted the responses of their therapists more accurately than the low orals; you will not be surprised to learn that the most accurate perceivers were high oral female clients of male therapists. Curiously, the ability to perceive the therapist accurately did not improve as therapy progressed; clients were no more accurate in predicting their therapists' responses at the end of therapy than they were after the third session, perhaps because the transference had developed sufficiently by then to warp perception.
If orals need the company of others to feel secure, they should experience distress when they are forced to remain alone. Price, Goldband, Katkin, and I (1981) tested this conjecture by placing male subjects in a small sound proof chamber. Half the subjects were alone in that chamber and half were placed with a male confederate who sat at right angles to him, both men instructed to remain silent. The only subjects who did not show increased autonomic nervous system responses arousal during the course of the experiment were the oral subjects seated with the confederate. All other subjects -- the orals seated alone and the low orals either alone or with the confederate -- showed signs of physiological arousal as assessed by the electrodermal response. For the high oral subjects the mere presence of another human who simply sat there and said nothing was evidently sufficiently comforting to dampen their physiological arousal.
In the next test of this general hypothesis, O'Neill, Katkin, and I (1982) asked whether the presence of any other person, regardless of potential nurturing qualities, would help the orals dampen their arousal response. This time we varied the quality of the relationship between the subject and a confederate. The two men were first put in a small room together and instructed to try to get to know each other. Half the time the confederate was warm, interested, and friendly and half the time he was polite but distant and remote. Following this l0 minute interaction the two men were put in the sound proof chamber and electrodermal measurements taken. There was a highly significant difference at the .005 level in the high oral subjects as a result of the prior 10 minute interaction: Those seated with the "cold" confederate show dramatic EDR arousal; those seated with the warm subject demonstrated no such arousal and in fact, even manifested a slight decline in autonomic arousal from baseline to experimental condition. I wish now I had thought to interview the low oral subjects because I suspect they may not even have noticed the quality of the relationship with the confederate.
I will close by describing three other experiments, none published, that I conducted for no other reason than my curiosity about the groups I studied. If all politics are local perhaps all experiments have a political purpose. Intrigued by the motives of animal rights supporters who threaten the lives of scientists who use animals in their research, I hypothesized that their dislike of cruelty to animals was one way to deny their own hostility. I constructed a scale assessing attitudes toward using animals in research. You may be interested to know that 21% of our undergraduate sample would use drug addicts rather than animals in research, 6% would use the severely retarded rather than animals, and 11% would substitute alcoholics for animals in research. I selected the highest scoring third, and compared them with subjects who clustered around the mean. All subjects also completed a questionnaire in which they were asked to assign either jail or monetary sentences, or both, to those convicted of various crimes, from relatively minor matters to such serious crimes as child abuse and murder. In both the original study and its replication I found that the animal rights subjects meted out stiffer sentences, including the death penalty, than the control subjects. The extreme scorers, who had no problem accepting modern medical treatments based on animal research, had moral reservations about the use of animals in experiments at the same time they recommended the death penalty.
The last study also grew out of my hypothesizing reaction formation in an another extreme group, this time homophobics. My students and I constructed a scale assessing intolerance of homosexuality. To secure an adequate range of scores, along with some common sense items we deliberately wrote items we thought no rational person would ever endorse. To our dismay, we found that even the most extreme, wacko item was endorsed by some subjects. For example, 17% would castrate homosexuals, 15% agree that homosexuals lack the fundamental characteristics that make us human, 33% would not allow homosexuals to teach in elementary schools, 13% could not enjoy music written by a homosexual, 11% agree that homosexuals should be required to wear some kind of identifying mark. We then selected the extreme scorers from a subject pool sample, put them in front of a screen, and gave them a manual control that allowed them to view slides as long or as briefly as they liked. We told the subjects to examine the slides closely because they were to be tested for their memory of them at the completion of the experiment. The experimenter then left the room. All the slides were made from photographs clipped from popular magazines, some showing heterosexuals and heterosexual activity, others showing nude or partially nude males and homosexual activity. The slide projector was rigged so that we could record how long each slide was exposed. We hypothesized that the homophobic subjects would look longer at the homosexual stimuli than the heterosexual. The results were significant, as we like to say, at the oh shit! level. After the experiment I interviewed some of the homophobic subjects, expecting to find signs of dysfunctional or antisocial behavior in them, or evidence they grew up in difficult circumstances. I found no such patterns. Other than appearing to be quite rigid and conventional, they seemed no different from any other Buffalo student. I shelved the whole project as yet another good idea gone astray.
Some time after that fiasco I was delighted to discover that Adams, Wright, and Lohr conducted an experiment testing my hypothesis, but they used a more sensitive method. Homophobics and control subjects were exposed to explicitly erotic stimuli consisting of heterosexual, homosexual, and lesbian activities. The authors employed a far better measure of sexual interest than I did, however -- they used a penile plethysmograph, a condom like device that fits snugly around the penis. As the penis becomes engorged, the plethysmograph expands, moves a column of air, and makes traces on a moving graph. The results demonstrated what every clinician knows but can rarely prove: Overdoing can be a reaction formation against the threatening impulse. Although all subjects were sexually aroused by the heterosexual and female homosexual stimuli, only the homophobic subjects were sexually aroused by the male homosexual stimuli. Aggressive stimuli did not discriminate between groups. I note in passing that the authors managed the neat trick of discussing this manifestly psychoanalytic concept without once mentioning reaction formation or citing any psychoanalytic writing.
Thus far I have described experiments inspired rather directly from psychoanalytic thought. Considerable research, however, is only indirectly linked to analytic thought. In his chapter on attachment research Morris Eagle (1996) wrote "...an important impetus to the development and growth of psychoanalytic theories is research (and theorizing) in areas that are not expressly psychoanalytic but that are relevant to psychoanalytic theory. Ironically, such research is often more vital to psychoanalytic theory than are attempts at direct experimental testing of psychoanalytic hypotheses and concepts" (p. 105).
The next volume of the Empirical Studies is devoted to experiments based on second-order psychoanalytic thought, in this case studies of issues in health and illness. I will briefly summarize several chapters. Bornstein discusses how dependent behavior is related to a variety of illnesses. A retrospective design, in which personality is assessed only after a person has become ill, cannot determine whether the trait preceded or followed the disease. Two longitudinal studies of males subjects, however, did collect personality data in healthy people before the onset of illness. One investigated Harvard undergraduates (Vaillant, 1978) and the other male VA patients (Greenberg & Dattore, 1981). Both studies showed that dependent men were more likely than nondependent men to become ill many years later -- 30 years in one study, 10 years in the other. The illnesses included cancer, hypertension, ulcers, colitis, and neuromuscular problems. In a third study using a non-retrospective design, Bornstein (1993) tested college women when they were healthy. He found that those who scored high on dependency were more likely to report incidents of illness later in the semester than those who were not as dependent.
O'Neill's chapter on cardiac problems and personality resurrects the Type A personality syndrome, showing that self-reporting of Type A behavior is frequently misleading. Instead, when clinicians categorize patients as Type A, predictions of cardiac disorders become markedly more accurate. Paul Duberstein writes about the health of the surviving member of a couple after the partner dies, arguing that the responses to loss cannot be understood apart from longstanding temperament traits and psychodynamic themes of dependency and help-seeking, early loss, and ambivalence.
Arndt, Goldenberg, Greenberg, Pyszcznski, and Solomon describe their creative efforts to understand the consequences of being forced to confront the fact that one will ultimately die. Those lacking adequate defenses against death anxiety may become physically ill, depressed, and may experience sexual difficulties.
Finally and with some trepidation because the results seem so counter-intuitive to me, I will summarize the chapter of Smythe and Greenberg who describe the salutary effects of writing about emotional experiences. Subjects are asked to write essays expressing their feelings about the most traumatic or stressful event they can remember. Typically, they write for 20-30 minutes on 3-5 consecutive days. That is the only "therapeutic" intervention. The subsequent beneficial effects on health of this rather mechanical, non-interpersonal intervention are astonishing. A meta-analysis of 13 experiments studying over 800 healthy individuals demonstrated a 23% effect size favoring the experimental subjects. This effect size is much larger than the 9% effect size for psychological intervention for those convicted of driving while intoxicated and is only slightly lower than the effect size reported for psychotherapy. Outcome effects, assessed from 1 to 4 months post-writing, include improvements in depression, anxiety, visits to the doctor, immune system function, upper respiratory illness, and grade point average. I should emphasize that so far only healthy subjects show these benefits; those already seriously ill seem not to respond so well to the writing experience.
My paper describes a major paradox: An observable, measureable decline in the importance of psychoanalysis at the same time that psychodynamic ideas inspire a great deal of research, some of it directly found in analytic writing and others more distantly related. Able investigators in developmental, social, and cognitive psychology borrow concepts from psychoanalysis, frequently without attribution, at the same time that many academic departments are becoming narrowly behavioral and anti-dynamic. Parts of psychoanalysis probably should be consigned to Trotsky's "dust bin of history" but parts are as rich and insightful as they were 100 years ago. I leave it to each of you to resolve this paradox in your own way.
Web Editor: Paolo Migone. For suggestions or corrections, please e-mail to: email@example.com