From time to time, various scientific articles appear which discuss issues of childhood sexual abuse, memory, and responses to trauma. Since such studies are often widely cited in the scientific and popular press, it is critical to recognize their methodological limits. It is particularly important to understand what conclusions can and cannot legitimately be drawn from these studies on the basis of the data presented.
In 1997, Harrison Pope, Jr., M.D. wrote a series of columns for the FMSF Newsletter that became the nucleus for the book "Psychology Astray: Fallacies in Studies of ‘Repressed Memory’ of Childhood Trauma." Those columns are collected in this section. Readers who are interested in the issues presented in these columns may wish to purchase the book that includes 10 important chapters that were not a part of the newsletter column. A review of the book in the prestigious "Nature" (7/17/97) says that it is a "model of clear thinking and clear exposition." Harrison Pope is Professor of Psychiatry at Harvard University Medical School and Chief of the Biological Psychiatry Laboratory at McLean Hospital in Belmont, Massachusetts. He is a member of the FMSF Scientific Advisory Board.
Psychology Astray: Fallacies in Studies of "Repressed Memory" and Childhood Trauma. Boca Raton, Fl: Upton Books (1997)
by Harrison Pope, Jr.
The next five short articles address research focused on the question of whether
there is scientific evidence that proves that individuals are able to repress
memories of traumatic events.
The Children of Wish-Ton-Wish
To some people, it seems perfectly natural that memories can be "repressed." If one experiences a tragedy too terrible to contemplate, is it not only reasonable that the mind would try to expel the memory from consciousness?
Actually, from a Darwinian point of view, repression is anything but reasonable. If, for example, one did not vividly remember being attacked by a lion, but instead "repressed" the memory, then one would be liable to wander in front of other lions in the future -- with inauspicious consequences both for one’s own survival and one’s chances of perpetuating the species. Surely it would seem more logical that Mother Nature would have designed us to remember traumatic events vividly, so that we could avoid a repetition of them in the future. And for most of us, this has been our personal experience: horrible things that have happened to us are still ingrained in our minds years after they occurred.
In a recent study, for example, members of our research group interviewed 53 victims of a freak tornado which struck the town of Great Barrington, Massachusetts, in the Spring of 1995. One woman was trapped in her car when the storm hit; a tree fell across the road immediately in front of her, and live power lines collapsed onto the pavement behind. The car shook; the walls of a neighboring garage blew away like playing cards. In the back seat, her children were screaming.
"Did you have any loss of memory for that experience?" we asked. She looked at us in disbelief and said, "are you kidding?"
As this woman and many others can attest, terrifying experiences leave indelible memories. Therefore, where and when did the idea arise that the opposite could happen -- that a traumatic memory could be completely banished from consciousness?
One way to examine this question is to look at world literature. As we look at stories, poems, and dramas written throughout the ages in different places and different cultures, where do we find characters who "repressed" and then perhaps later "recovered" memories of traumatic events?
We have put this question to a number of experts in literature. Such a survey, admittedly, is hardly a formal scientific study, but it is nevertheless revealing. Throughout most of history, it appears, no one in any story in the world’s literature appears to have developed amnesia for a seemingly unforgettable traumatic event and later recovered the memory into consciousness. No one in the Bible, for example, seems to have repressed and then recovered a memory. Nor in Shakespeare -- a veritable catalog of the possible permutations of the human psyche -- do we find a clear instance of repression. No one has been able to show us a clear case of repression in classical Greek or Roman literature, in Islamic literature, or anywhere else in Western literature until well into the l9th century. Then, and only then, does repression begin to crop up (1).
As best as we can tell, one of the first cases of repression and recovery of memory appears in James Fenimore Cooper’s 1829 novel, The Wept of Wish-Ton-Wish (2). In this tale, set in the mid-seventeenth century, Indians attack the little settlement of Wish-Ton-Wish in Connecticut and abduct two children. One is a teenager named Whittal Ring, and the other is a little girl named Ruth Heathcote. Years later, Rueben Ring comes upon his lost brother Whittal in the woods. Whittal is now dressed as an Indian; he is wearing war paint and calls himself Nipset. He has complete amnesia for his past as a White man. His sister, Faith, recognizes her brother, but is unable to persuade him of his former identity, even when he looks at his own white skin.
Later, Ruth is also found. She, too, has become an Indian and goes by the name of Narra-mattah. Her memories of childhood are also completely repressed, but she has recurring images of her mother in dreams:
"Narra-mattah has forgotten all ... But she sees one that the wives of the Narragansetts do not see. She sees a woman with white skin; her eyes look softly on her child ..."
Ruth’s mother tries to help her child recover her lost memories, but in vain. Then, at the very end of the novel, the child falls ill and lies dying. And there, in the lush romantic prose of Cooper, we witness what just might be literature’s first case of a repressed memory. The mother of the dying child speaks to her:
"Look on thy friends, long-mourned and much suffering daughter! ‘Tis she who sorrowed over thy infant afflictions, who rejoiced in thy childish happiness, and who hath so bitterly wept thy loss, that craveth the boon. In this awful moment, recall the lessons of youth. Surely, surely, the God that bestowed thee in mercy, though he hath led thee on a wonderful and inscrutable path, will not desert thee at the end! Think of thy early instruction, child of my love; feeble of spirit as thou art, the seed may yet quicken, though it hath been cast where the glory of the promise hath so long been hid."
"Mother!" said a low struggling voice in reply. The word reached every ear, and it caused a general and breathless attention. The sound was soft and low, perhaps infantile, but it was uttered without accent, and clearly. "Mother -- why are we in the forest?" continued the speaker. "Have any robbed us of our home, that we dwell beneath the trees?" Ruth raised a hand imploringly, for none to interrupt the illusion.
"Nature hath revived the recollections of her youth," she whispered. "Let the spirit depart, if such be his holy will, in the blessedness of infant innocence!"
Another possible case of repression arises in 1859, in Charles Dicken’s novel, Tale of Two Cities. Dr. Manette, after 18-years imprisonment in the Bastille, has developed amnesia for long intervals of his past, including the period surrounding his release (3). He describes his amnesia in courtroom testimony:
"Has it been your misfortune to undergo a long imprisonment, without trial, or even accusation, in your native country, Doctor Manette?"
"He answered in a tone that went to every heart, "A long imprisonment."
"Were you newly released on the occasion in question?"
"They tell me so."
"Have you no remembrance of the occasion?"
"None. My mind is a blank, for some time -- I cannot even say what time -- when I employed myself, in my captivity, in making shoes, to the time when I found myself living in London with my dear daughter here. She had become familiar to me, when a gracious God restored my faculties; but, I am unable even to say how she had become familiar. I have no remembrance of the process."
And a few years later, in approximately 1862, Emily Dickinson in a poem implies more specifically that an event could breed amnesia simply because it is too traumatic to contemplate (4).
There is a pain - so utter
It swallows substance up
Then covers the Abyss with Trance
So Memory can step
Around - across - upon it
As one within a Swoon
Goes safely - where an open eye
Would drop Him - Bone by Bone.
By the end of the century, we find that "repression" and "recovery" of memory have entered romantic fiction in full-blown form. A typical case appears in the 1896 children’s novel, Captains Courageous, by a Nobel prize winner, Rudyard Kipling (5). One of the characters in the novel is a former preacher, Penn, who had long ago lost his entire family before his eyes in a tragic flood. After the flood, Penn has completely repressed the memory of the entire trauma, and has even forgotten that he ever was a preacher or had a family. We find him instead working for Captain Disko as a fisherman on a Grand Banks schooner, oblivious to his past. One day, a passing ocean liner carves a neighboring fishing schooner in two, killing its hands, including the captain’s son. The surviving captain is rescued by Disko’s crew and brought aboard. At this moment, Penn abruptly undergoes a transformation. He suddenly recovers the memory of the loss of his own family, and his voice transforms from his usual "pitiful little titter" to the authoritative tones of a preacher. He consoles the grieving captain, prays for him, and shares with him the memory of the tragic loss of his own loved ones years ago. And then, within hours, Penn "re-represses" the memory. He again forgets his past, reverts to a simple fisherman, and asks for his customary game of checkers.
With the coming of modern times, repression has found a new and even more fertile soil in that uniquely 20th century art form, film. From the thrillers of Alfred Hitchcock to the childhood trauma of Batman, characters in the movies regularly experience amnesia for traumatic events, and then, at some dramatic moment, recover the memory. Indeed, repression is the perfect device for Hollywood. Many a celluloid hero is seen having a "flashback" -- a fleeting, freeze-frame image, perhaps slightly out of focus -- of a long-forgotten event. What is the dark secret from the past? Perhaps, if the hero could make sense of this recurring image, recover the repressed memory, all would be explained. By the end of the movie, this is usually just what has happened.
In short, for all of us who have grown up in the 20th century, repression seems like a natural phenomenon; we have read of it in novels and seen it in the movies all our lives. Perhaps this is why so many people accept the concept without bothering to question it. But we must stop to remember that repression actually appears to be a parochial notion, seemingly restricted only to recent times and only to Western European culture. And we must also remember that repression was not a scientific hypothesis first proposed by Sigmund Freud or Pierre Janet. Rather, it seems to have arisen as a romantic notion in the Victorian era, somewhere in the middle of the 19th century. It had entered poetry and prose well before Freud and Janet were even born. It has continued to flourish in literature and cinema throughout the 20th century. It is a powerful dramatic device that makes for good fiction.
But does it make for good science?
1. The notion of repression also began to evolve in the writings of 19th century philosophers such as Schopenhauer and Nietzsche. For a detailed discussion of these beginnings, see Ellenberger, H. The Discovery of the Unconscious. New York: Basic Books, 1970.
2. Cooper JF. (1829) The Wept of Wish-Ton-Wish.
3. Dickens C. (1859) A Tale of Two Cities. New York: Dodd, Mead & Co., 1942. See Book the Second, Chapter 3.
4. Johnson TH (ed.) The Complete Poems of Emily Dickinson. Boston: Little, Brown & Co., Boston, 1960, page 294, No. 599. I am indebted to Dr. Gail S. Goodman and her colleagues for having discovered this poem. See Goodman GS, Quas JA, Batterman Faunce JM, Riddlesberger MM, Kuhn J. Predictors of accurate and inaccurate memories of traumatic events experienced in childhood. Consciousness and Cognition 4:269-274, 1994.
5. Kipling R. (1896) Captains Courageous: A Story of the Grand Banks. New York: Doubleday, Page & Co., New York, 1925, See chapters 3 and 7.
The Emperor’s Tailoring
Modern technology has greatly expanded our ability to study the structure and functions of the brain. Neuropsychological testing techniques, measurements of neurotransmitters (chemical messengers), and other advances in the understanding of brain chemistry have allowed us to probe ever deeper into the central nervous system. Perhaps the most dazzling technological strides have occurred in neuroimaging: with magnetic resonance imaging (MRI) and positron emission tomography (PET), we can now see images of living brain structures. With the appropriate computer software, we can detect subtle differences in anatomy or blood flow between different brain regions, or between patient populations and normal comparison subjects. Can we apply this technological arsenal to answer the question of whether it is possible to "repress" the memory of traumatic events?
At first, the answer to this question would seem to be yes. The literature has recently been filled with studies using various forms of high technology to study trauma and its consequences (1). For example, neuropsychological testing techniques have been used to quantify memory function in trauma victims as compared to non- traumatized comparison subjects. These studies have generally found that individuals diagnosed with post-traumatic stress disorder (PTSD) have greater difficulty remembering test items than do normal comparison subjects. Biochemical studies, similarly, have shown that numerous chemicals critical to the function of the nervous system, such as neuropeptides and neurotransmitters, may be affected in various ways by stress. Among the chemicals studied in this manner are epinephrine; norepinephrine; corticosteroids; pituitary and hypothalamic hormones that stimulate the release of corticosteroids; opioid peptides; gamma aminobutyric acid; vasopressin; and many others. Not surprisingly, data suggest that these substances are disrupted in various ways during the experience of trauma. They may even remain disrupted long afterwards in trauma survivors with PTSD. Furthermore, other studies have shown that many of these same neuropeptides and neurotransmitters have various effects on memory function, either enhancing or impairing memory under particular conditions.
Even more striking, however, are the latest studies of brain anatomy and metabolism in trauma victims. For example, studies of monkeys exposed to prolonged and fatal stress have shown damage in an area of the brain called the hippocampus (2). And damage to the hippocampus, in turn, has been shown to be associated with impairments in memory function. MRI studies of the hippocampus in humans diagnosed with PTSD have now also shown abnormalities in comparisons with healthy control subjects. Even more colorful and impressive are the findings of PET scans in trauma victims. For example, PET technology has been used to measure cerebral glucose metabolism (an index of brain activity) in combat veterans as compared to normal controls. Differences between the two groups were found in a number of different cortical areas (3). In another study, 8 patients with post-traumatic stress disorder, two of whom were victims of childhood sexual abuse, were exposed to "traumatic" vs. "neutral" scripts while undergoing positron emission tomography. In the traumatic scripts, audiotapes describing a personal traumatic experience, such as sexual abuse or a car accident, were played to subjects. The neutral scripts were audiotapes describing mundane experiences, such as emptying the dishwasher. Statistical mapping techniques were then applied to the PET scan results to identify which areas of the brain’s cortex displayed significant activation under these conditions (4). The 8 subjects with PTSD were specifically selected for the study because they had already been shown to demonstrate a physiologic response (in other words, measurable physical changes) in response to traumatic scripts. Therefore, it is probably not surprising that, when stimulated to remember their trauma, the subjects’ brains showed changes; there were significant differences in blood flow in various parts of their brains when they were thinking about watching their loved ones die in a car accident as opposed to, say, thinking about brushing their teeth. Upon comparing the color-enhanced computer printouts of the PET scans, even a layman can see obvious differences in the pictures taken in the traumatic condition vs. the neutral condition.
These are all interesting and valuable studies, carefully performed under rigorous scientific conditions. What’s more, some have produced stunning findings. One can hardly examine the striking PET scan images without being impressed. Surely, then, this wealth of data provides mounting evidence that trauma does influence memory, and suggests neurological and biochemical mechanisms that might explain how trauma victims could develop amnesia for the event. One might believe, therefore, that we are finally accumulating scientific proof that memories of trauma can be repressed.
But this last conclusion is a fallacy, and it is important to understand why. The logical flaw here is the assumption that one can take a series of scientific findings, link them together, and safely extrapolate to conclusions about some other phenomenon which one has not studied directly. An example will make this clear. Studies have established that there is a highly significant positive correlation, in the animal kingdom, between brain size and intelligence. In other words, the larger the brain, the larger the IQ: worms are not as intelligent as seagulls, and seagulls are not as intelligent as dogs. It has also been established, through years of neuroanatomical studies, that men have bigger brains, on average, than women. But even though these two findings are both correct, it would be erroneous to infer that men have higher IQs than women. If we are interested in knowing whether men are smarter than women, we should stop drawing dangerous inferences from neuroanatomy, and instead go out and test actual samples of men and women for intelligence. By analogy, if one wishes to test whether trauma victims can repress their memories, one should go out and simply ask a group of victims if they can remember their trauma. If instead we merely infer that repression might occur on the basis of people’s cerebral glucose metabolism, or images of their hippocampi, we might prove seriously mistaken.
This fallacy may seem obvious, but it has bedeviled even the most brilliant thinkers. For example, the great Austrian mathematician Godel once performed a series of mathematical calculations based on Einstein’s equations from relativity theory (5). Godel’s calculations suggested that certain solutions to Einstein’s equations produced closed time lines, which would theoretically allow for the existence of time travel. Unfortunately, Godel found that the amount of energy necessary for a human being to travel back in time would be excessively large. He apparently believed, however, that his calculations did offer possible scientific evidence for the existence of ghosts.
Even the best of scientific findings, in other words, can be misused to reach dubious conclusions. And the average observer, blinded by the technological sophistication of such findings, may lose track of the sleight-of-hand inferences that someone is making from them. Such fallacious reasoning has often infiltrated the courtroom, where juries may be presented with impressive scientific results, strung together to seemingly imply that, say, working at video display terminals can cause miscarriages, that the drug Bendectin can cause birth defects, that silicone breast implants can cause arthritic disease, or that low-intensity magnetic fields can cause cancer (6,7). The point, once again, is that if one wants to test these hypotheses, one should not dwell on inferences from laboratory studies; one should simply go out in Nature and test whether or not the hypotheses are true. When this was done with the above hypotheses, they all failed the test of actual epidemiologic study.
One last example, taken directly from modern psychology, is the technique of "eye movement desensitization and reprocessing," usually abbreviated as "EMDR." EMDR is a novel psychotherapeutic technique in which the patient is asked to follow a moving object with his or her eyes while thinking about a traumatic experience (8). On the basis of a series of inferences from various scientific findings, proponents of EMDR claim that it is effective for the treatment of a range of psychiatric disorders. The reasoning goes something like this: it is well established that during one stage of sleep, called rapid-eye- movement sleep, or REM sleep, the eyes move rapidly back and forth. It is also known that REM sleep is associated with dreaming. Dreaming, in turn, is often associated with intense emotions. And trauma victims, indisputably, may have dreams about traumatic events that they have experienced. Therefore, if a patient voluntarily engages in eye movements in the therapist’s office, while recalling a traumatic experience, perhaps he or she could more effectively reexperience and work through the traumatic experience, with consequent progress in therapy.
An interesting chain of inferences, perhaps, and one which is indeed based on legitimate scientific observations. But the fact remains that no one has shown, in a properly designed study, that EMDR actually works. That is, when we put inferences aside and go out to actually test the efficacy of EMDR in nature, the hypothesis is not supported (9).
What is the lesson of all this? It is that a phenomenon is not proven just because inferences from various studies suggest a mechanism for how it might theoretically occur. In other words, no matter how impressive the findings of neurotransmitter assays or how colorful the pictures from PET scans, and no matter how intriguing the brain mechanisms that these studies might suggest, we cannot logically conclude from these studies that people can actually repress memories of traumatic events.
In other words, one should not speculate about the details of the emperor’s tailoring until one has first assessed whether he has any clothes.
1. Bremner JD, Krystal JH, Charney DS, Southwick SM. Neural mechanisms in dissociative amnesia for childhood abuse: relevance to the current controversy surrounding the "False MemorySyndrome." Am J Psychiatry 1996 153:71-82.
2. Uno H, Tarara R, Else JG, Suleman MA, Sapolsky RM. Hippocampal damage associated with prolonged and fatal stress in primates. J Neurosci 1989; 9:1705-1711.
3. Bremner JD, Ng CK, Staib L, Markey J, Duncan KJ, Zubal G, Krystal JH, Massa S, Rich D, Southwick SM Capelli S, Seibyl JP, Dey H, Soufer R, Charney DS, Innis RB. PET measurement of cerebral metabolism following a noradrenergic challenge in patients with posttraumatic stress disorder and in healthy subjects (abstract). J Nucl Med 1993; 34: 205P-206P.
4. Rauch SL, van der Kolk BA, Fisler RE, Alpert NM, Orr SP, Savage CR, Fischman AJ, Jenike MA, Pitman RK. A symptom provocation study of posttraumatic stress disorder using positron emission tomography and script-driven imagery. Arch Gen Psychiatry 1996; 53:380-387.
5. Stillwell J., Mathematics and its History. Springer-Verlag, New York, 1989. pp 330 331.
6. Foster KR, Bernstein DE, Huber PW (eds). Phantom Risk. MIT Press, Cambridge, Massachusetts, 1993.
7. Angell M. Shattuck Lecure - Evaluating the health risks of breast implants: the interplay of medical science, the law, and public opinion. New Engl J Med. 1996; 334: 1513-1518.
8. Hudson JI, Chase EA, Pope HG Jr. Eye movement densitization and reprocessing in eating disorders: caution against premature acceptance. Int J Eating Disorders. In press.
Admittedly, there are some studies which claim to show a benefit for EMDR. These include, for example, Silver SM, Brooks A, Obenchain J. Treatment of Vietnam war veterans with PTSD: A comparison of eye movement desensitization and reprocessing, biofeedback, and relaxation training. J Traumatic Stress 1996; 8:337-343; and Montgomery R, Ayllon T. Eye movement desensitization across subjects: Subjective and physiological measures of treatment efficacy. J Behavior Therapy Exp Psychiatry 1994; 25; 25:217-230. However, even these studies are subject to such serious methodological flaws that their findings are highly questionable. For a discussion of these flaws, see the review in footnote 8, above, and also Steketee G, Goldstein A. Reflections on Shapiro’s reflections: Testing EMDR within a theoretical context. The Beh Therapist 1994; 17:156-157, or any of several other critical reviews cited by Hudson and colleagues in their review.
Can People Repress Memories? Evidence of Prospective Studies, Part I
In previous columns in this Newsletter, we have examined retrospective studies that purport to show scientific evidence of "repressed memory." As we have noted, however, such studies are fraught with methodological flaws, in that they rely on individuals’ unconfirmed recollections of whether or not they forgot an event for some period of time earlier in their lives. Such designs, which we have previously termed "do-you-remember-whether-you-forgot" studies, do not pass scientific muster as satisfactory tests of the "repression" hypothesis.
Instead, a much better way to test the repression hypothesis would be to design a prospective study, in which one is not dependent upon somebody’s unconfirmed recollections of any sort of information. Such a design would not be difficult. First, one would obtain the names for a large group of people who had undergone a known, documented trauma. For example, one could go to the records of a hospital emergency room to find 50 children who were seen for trauma -- severe injuries, physical abuse, or sexual abuse -- and where there were specific medical findings in the records to show that the trauma actually occurred. Alternatively, one could identify 50 children who underwent a traumatic medical procedure, such as a painful rectal or gynecologic examination. One could get 50 assault or rape victims from police records, 50 victims of a tornado, or any other group where the facts of their personal traumatic experience were known and documented. Then, one would locate all of those trauma victims several years later, interview them, and simply ask them if they remembered the traumatic event. If a certain percentage of the subjects reported that they had completely forgotten the event, then we would have persuasive evidence that some people can repress the memory of trauma. On the other hand, if none of the subjects in any of the studies reported forgetting the trauma, then we would suspect that repression does not really happen -- except, of course, in the movies.
We would have to be careful about several confounding effects in such a study. The first is the normal amnesia of early childhood. If someone has no memory of having been brought to the emergency ward at age 1 or 2, such a case clearly provides no evidence of repression. We all have amnesia for most events before the age of 3, and even most events before age 6. Second, we would have to exclude neurological or medical causes of amnesia. If an individual was knocked unconscious in an accident, or if she received anesthesia for a medical procedure, we would expect her to have amnesia without any need to postulate repression. Similarly, combat veterans would represent a poor choice for our study, because head injuries, severe sleep deprivation, and other neurological insults to the brain are so common in wartime. Third, we would not want to study people with only mild trauma, because then we could not rule out the possibility that the subject was just experiencing ordinary forgetfulness for an event that was not particularly memorable. In other words, to test whether one can truly repress a memory, one would have to study a group of subjects who experienced a trauma that no ordinary person would be expected to forget. Fourth, when we interviewed our subjects to ask them about their memories, we would have to take care to make sure that they were disclosing all that they remembered. We will discuss this issue in detail in the next two columns, but an example will suffice here. Suppose that a girl undergoes a painful and embarrassing gynecologic procedure at age 10. When she reaches age 15, a researcher sees her for an interview and asks her if she has undergone any unusual medical procedures. Even if the interviewer is careful and sympathetic, the girl may still answer, "no," even though she actually remembers the event. To minimize such non-disclosure, the interviewer may need to ask the subject about the specific event in a more direct manner: "I know from your medical records that when you were 10, you were seen at the hospital for a special medical examination. Do you remember that?"
In summary, then, a satisfactory scientific test of repression would have to follow only a few simple rules: 1) locate a group of people who were victims of a documented trauma, and 2) interview them some years later to see if any of them report amnesia for the trauma. We would exclude cases where the failure to report might be due to a) early childhood amnesia, b) neurological or medical causes, c) ordinary forgetfulness, or d) deliberate non-disclosure. If after these exclusions, we were still left with a fair number of patients who described amnesia for the event, we would have evidence that repression really does occur. Those are the ground rules. What is the verdict? To our knowledge every study in the world literature which has come even remotely close to the above standards has failed to show any evidence that people can repress memories.
Here are some examples. In the 1960s, Leopold and Dillon (1) studied 34 men who had survived a terrible explosion when two ships collided. In interviews conducted about four years after the explosion, many of the men reported serious post-traumatic psychopathology, but none displayed amnesia. The authors wrote, "repression does not appear possible." In another study, Terr (2,3) interviewed 25 children who had been kidnapped and buried alive in a school bus four years earlier. She found that "each child could give a fully detailed account of the experience." Malt (4) interviewed 107 individuals who had been seen at an emergency ward for traumatic injuries 16 to 51 months previously. The only amnesia found in these individuals was that due to neurological injuries; no one was described as having repressed the memory. Wagenaar and Groeneweg (5) studied 78 subjects who were seen in relation to a Nazi war crimes trial in the 1980s. These subjects were asked about their memories of having been in a concentration camp 40 years earlier. Although many of the subjects were quite elderly by the 1980s, most remembered the camp "in great detail." Although the subjects had forgotten various specific items from their experience, they had forgotten non-traumatic items just as much as traumatic items; there was no evidence that they had selectively repressed traumatic memories. Interestingly, there were six men who had testified to various specific traumatic experiences when they were originally liberated from the camp in the 1940s, but who did not describe these memories when they were re-interviewed in the 1980s. However, when they were reminded of their earlier testimony, all but one of them promptly recalled the particular events. This is a remarkable record when it is considered that these former inmates were 65 to 82 years old by the 1980s -- and hence vulnerable to biological amnesia. Peterson and Bell (6) interviewed 90 children who had been seen at a hospital in Newfoundland for traumatic injuries six months earlier. It appears that every child, including even those only two years old at the time, remembered the event. Among the children who were 9 to 13 years old at the time of their injuries, so few made errors in their recall that the investigators could not even include them in a statistical analysis of the causes of errors of memory.
The above studies span a range of traumas, from single events like the marine explosion to longstanding events like the concentration camp experience. Some of the subjects in some of the studies had spoken at length about their experiences to other people, or undergone prior interviews, and hence might be expected to have particularly clear memories. On the other hand, some of the subjects were being studied for the first time, and had had no opportunity to "rehearse" their memories previously. The one feature shared by the subjects in every study was that they remembered their trauma.
Some critics might still object to our evidence here. They would argue that explosions, kidnappings, concentration camps and hospital visits are very different from "secret" traumas such as childhood sexual abuse. Even allowing that repression does not occur for ordinary traumas, perhaps it might still occur in certain special situtaions, like that of a child who is forced to undergo repeated sexual assaults from someone whom she is supposed to love. Therefore, rather than be too quick to dismiss the possibility of repression, we owe it to ourselves to examine prospective studies that look specifically at the memories of victims of childhood sexual abuse. However, as will be seen in the next two columns, these studies also fail to provide any methodologically sound evidence that repression can occur.
1. Leopold, R.L. and Dillon, H: Psycho-anatomy of a disaster: a long term study of post-traumatic neuroses in survivors of a marine explosion. Am J Psychiatry 1963: 119-921.
2. Terr, L.C.: Children of Chowchilla: a study of psychic trauma. Psychoanal Study Child 34: 552-623, 1979.
3. Terr, L.C.: Chowchilla revisited: the effects of psychic trauma four years after a school-bus kidnapping. Am J Psychol 140: 1543-1550, 1983.
4. Malt, U: The long-term psychiatric consequences of accidental injury: a longitudinal study of 107 adults. Br J. Psychiatry 153: 810-818, 1988.
5. Wagenaar, W.A. & Groeneweg, J: The memory of concentration camp survivors. Appl Cog Psychol 4: 77-87, 1990.
6. Peterson, C. & Bell, M.: Children’s memory for trauma injury. Child Develop 67:3045-3070, 1996.
Can People Repress Memories? Evidence of Prospective Studies, Part II: ’Cuz I Wanted to Forget
In a previous column, we have seen that prospective studies of trauma victims consistently fail to show evidence of repression. But there remain four other prospective studies, to our knowledge, which have looked specifically at memories of childhood sexual abuse. If repression can be found anywhere, perhaps these studies are the place to look.
The first such study was published in 1990 by Donna Della Femina and her colleagues . These investigators interviewed 69 young adults in Connecticut as part of a follow-up study of formerly incarcerated youths. The investigators possessed detailed information, obtained in evaluations may years earlier, about physical and sexual abuse which these subjects had experienced as children. Upon interviewing these subjects as young adults, the investigators found that 26, or 38 percent of them, reported childhood histories which were inconsistent with the information gathered on their earlier evaluations. Specifically, 18 of these 26 subjects were known to have been severely abused in childhood, yet they denied or minimized any history of abuse when they were interviewed. Had those 18 individuals repressed the memory of their trauma? And had the other 8 subjects, who suddenly revealed new information about childhood abuse, recovered old memories which had been repressed at the time of their initial evaluations?
Fortunately, Femina and her colleagues did not jump to these conclusions. Instead, they decided to re-contact the 26 subjects who had given interview information discrepant with their known histories. They managed to track down 11 of the 26 for a second interview, which they called the "clarification" interview. In the clarification interview, they took care to establish rapport, and then confronted the subjects with the known discrepancies in their responses.
On the clarification interview, it appears, all 11 of the subjects admitted that they had always remembered their abuse, but had simply chosen not to disclose the information to the interviewer, either during the first interview (for the 8 subjects who denied their previously recorded history of abuse) or on their initial evaluation (for the 3 subjects who revealed new information about abuse not recorded on their initial evaluation). When asked to explain the reasons that they had chosen to withhold the information, the subjects gave a variety of responses. For example, one girl, who was known to have been sexually abused by her father, and whose mother had attempted to drown her as a child, had minimized any abuse at all on the initial interview. When asked on the clarification interview why she did not previously disclose the information, she burst into tears and said, "I didn’t say it ’cuz I wanted to forget. I wanted it to be private. I only cry when I think about it." Similarly, a boy who had been repeatedly beaten by his psychotic father denied abuse on his initial interview but then admitted to it on the clarification interview. When asked why, he said, "My father is doing well now. If I told now, I think he would kill himself." Another subject, who had also initially failed to disclose a known history of physical abuse, explained on the clarification interview that he had simply not liked the original interviewer.
Finally, looking at the 3 subjects who described histories of abuse on interviews that were not recorded in their initial evaluations, the investigators again found that deliberate nondisclosure, rather than repression, was responsible for the discrepancies. Two of the subjects revealed that at the time of their earlier evaluations, they were too embarrassed to reveal what they had suffered. The third had refused to disclose his abuse because at the time he did not trust anybody.
In summary, then, the investigators documented many reasons why subjects might not reveal a history of childhood sexual or physical abuse on follow-up interviews. These include "embarrassment, a wish to protect parents, a sense of having deserved the abuse, a conscious wish to forget the past, and a lack of rapport with the interviewer." In no instance, however, was any subject found to have displayed repression.
What is the lesson of this? It is that people will sometimes choose not to disclose information, especially sensitive and embarrassing information like a history of childhood sexual abuse, on interviews. If this happens, it would be naive for us to jump to the conclusion that they have repressed their memories. Instead, as illustrated so well by Femina and her colleagues, it is critical to ask any non- disclosing subjects directly about their known trauma history, to see whether they will then acknowledge that they remember it.
The same considerations apply to a second recent study conducted in London, where investigators interviewed 20 women who had been removed from their homes as children by social service agencies . These women were known to have been sexually abused, both from their own reports as children and by the report of an adult who had been familiar with the household at the time. The follow-up interviews were conducted when the subjects were 18-24 years old. On these interviews, three (15 percent) of the 20 women failed to give a "yes" answer to the question, "were you sexually abused as a child?" However, in this case, by the author’s admission, none of the subjects was directly asked about her known sexual abuse history, and no "clarification interviews" were conducted later. Therefore, when it is considered that the rate of negative responses in this study was well below the 38 percent rate in the Femina study described above, it again seems that these findings can be easily explained on the basis of non-disclosure alone, without any need to postulate repression.
One other recent small study followed up 22 children who had reportedly been abused in one of three day-care centers . Again, three of the children were said to display complete amnesia for their abuse when interviewed 5 to 10 years later. However, in this case there is a different problem: the children were reported to be a median of 2 1/2 years old at the time of the original abuse. Among the examples provided in the study, one child was in day care between the ages of 6 weeks and 12 months of age; another was 21 months old. As we have mentioned in previous columns, there is a large literature showing that children remember almost nothing from before the age of three, much less at 12 or 21 months. Thus the findings in this study appear readily explainable on the basis of childhood amnesia alone, again without any need to postulate repression. In addition, there remains the possibility of deliberate non-disclosure in this study as well, but the issue is not mentioned in the paper.
In summary, these three follow-up studies of victims of childhood sexual and physical abuse all fail to produce any evidence of repression. As Femina and her colleagues so clearly demonstrated in their "clarifications interviews," cases of seeming repression on initial interview regularly turn out to represent deliberate non-disclosure.
This leaves us with only one remaining prospective study of victims of childhood sexual abuse - our last chance, in effect to find an acceptable demonstration of repression in the scientific literature. But this study, performed by Linda Meyer Williams, is perhaps the most widely quoted study of them all. If there is any study which provides legitimate evidence of repression, many experts would say, the Williams study would be the one. Therefore, the Williams study deserves a particularly careful analysis in a column all to itself.
1. Femina, D.D., Yeager, C.A., Lewis, D.O.: Child abuse: adolescent records vs. adult recall. Child Abuse Negl 145: 227-231, 1990.
2. Bagley, C. Child sexual abuse and mental health in adolescents and adults: British and Canadian perspectives. Aldershot, UK: Avebury, 1995.
3. Burgess, A.W., Hartman, C.R., Baker, T. Memory presentations of childhood sexual abuse. J Psychosocial Nursing 33:9-15, 1995.
Garbage In, Garbage Out
Little noticed in the annals of social science research, but good reading for any beginning student of psychology, is the Tucson Garbage Project (1). In this study, a group of archaeologists decided to study the garbage discarded by randomly selected households in Tucson, Arizona during 1973 and 1974. More than 70 student volunteers, dressed in lab coats, surgical masks, and gloves, sorted through the garbage of 624 Tucson households and divided the refuse into more than 200 categories. Meanwhile, a group of trained personnel went out and interviewed individuals in a random sample of 1 percent of the households in the city. The interviewers asked, among other questions, how many cans or bottles of beer were consumed in the household in an average week. Then the data from each of Tucson’s census tracts were analyzed. The average reported weekly beer consumption of all households in a given census tract (standardized as the number of 12-ounce bottles or cans) was compared with the actual number of bottles and cans found in the garbage.
The reader has probably already guessed what happened. The number of beer cans and bottles in the garbage vastly exceeded the number that people had admitted to in their interviews. Looking, for example, at Tucson’s census tract number 10, more than 86 percent of the household reported to interviewers that they did not consume any beer at all in an average week, and not a single household (out of 60 interviewed) claimed a weekly consumption of more than 8 cans. But the garbage from tract 10 told another story. Only 23 percent of the households had no beer cans in their garbage, whereas 54 percent of households had more than 8 cans. In fact, the average number of cans in the garbage from that 54 percent of households was 15 per week -- in other words, 2 1/2 six-packs. And even these findings may underestimate the true discrepancy between interview data and garbage data, because, in 1973, most beer cans in Tucson were recyclable.
What does this have to do with studies of repression? Those who have read our previous columns (see FMSF Newsletters Nov/Dec 1996 and Nov 1997) will quickly recognize the point: people regularly fail to disclose sensitive information to interviewers. Like the subjects in the Femina study, who remembered but chose not to reveal their histories of childhood physical and sexual abuse, the people of Tucson were unwilling to tell an interviewer their true histories of beer consumption. They had not repressed the memory of all those beer cans; they just did not want to tell a stranger about it.
As with other concepts in epidemiology discussed elsewhere in past columns, this phenomenon has a name: response bias. Response bias has been studied extensively, in hundreds of investigations, for at least 50 years, and we now know a great deal about it. But before continuing with this discussion, we must take some time out to introduce the best known prospective study which has been claimed to show that people repress memories of childhood sexual abuse -- the study of Linda Meyer Williams (2).
Many readers will already have heard of the Williams study. It is regularly cited as the single most powerful piece of evidence that it is actually possible to repress memories. Frequently, in the popular media, in scientific articles, and even in courtrooms, the study is cited as though its findings were established, without even a passing mention of its methodological flaws (3). But these flaws are so critical that they deserve a careful review, and hence we describe the methods of the study in some detail.
Williams examined 129 women who had been evaluated at a city hospital in Philadelphia in the early 1970s for possible sexual abuse. At the time of that evaluation, which might be called the "index episode," these subjects were young girls between 10 months and 12 years of age. Williams possessed the hospital records from this "index episode." Then, approximately 17 years after the time of the index episode, Williams arranged for two interviewers to locate these women and ask them about their histories. The women were not informed that the investigators were specifically looking at their histories of childhood sexual abuse; they were simply told that they were being asked to participate in an important follow-up study of people who had been seen years earlier at the city hospital. During the course of the interview, each woman was asked about various types of traumatic experiences which she might have experienced during childhood, including sexual abuse. The interviewers also asked the women to describe any episodes which they themselves had not considered to be sexual abuse, but which other people had considered as such. However -- and this is the important part -- the two investigators interviewing the women were blind to all information about the women’s sexual abuse history; in other words, they had no knowledge of the specifics of the "index episode" when they interviewed their subjects, and they asked the subjects only in general terms about sexual abuse. The subjects WERE NEVER SPECIFICALLY ASKED ABOUT THE INDEX EPISODE ITSELF.
Forty-nine, or 38 percent of the 129 women did not describe the index episode of alleged sexual abuse in the course of the interview. Williams suggests in her paper that these women "did not recall" the episode. She supports this interpretation by noting that many of the women reported other traumatic events, or sensitive details of their histories -- such as substance abuse, sexually transmitted diseases, and even other instances of physical or sexual abuse -- while still not reporting the index episode. Therefore, Williams argues, it seems likely that the women would have reported the index episode if they had remembered it. But can we conclude that any of these 49 women had actually repressed the memory of the index episode? Several methodological problems immediately become apparent. First, only 37, or 28 percent of the 129 women had been found to display genital trauma when they were examined by the doctors at the time of the index evaluation. By contrast, as discussed in our column of January 1996, studies by gynecologists have shown that as many as 96 percent of girls subjected to genito-genital contact will display genital tract findings even on an unaided medical examination (4). Clearly, something is wrong here. It appears that a majority of Williams’ subjects, if they were sexually abused, were not victims of genito-genital penetration.
Williams admits to this. In another paper, in fact, she notes that approximately one-third of the cases involved only "touching and fondling." (5) And in an earlier description of this same sample of subjects, written back in 1979, Williams and her colleagues imply that for many of the girls, the alleged instance of sexual abuse was not particularly traumatic and therefore not particularly memorable:
"Whereas the event [the index episode] is disturbing to the victim, it is perhaps no more disturbing than many other aspects of a child’s life. In the first year following the rape [in the broad, statutory definition of the term], the victim’s family may deliberately maintain an "everything-is-normal" posture. These efforts, combined with the child’s natural tendencies to forget and to replace bad feelings with good feelings, usually result in the appearance of few adjustment problems..." (6; bracketed inserts ours).
In other words, looking both at the lack of medical evidence and at Williams’ own words, it seems that many of these girls may have experienced episodes which were not particularly severe. An episode of only touching and fondling, without any medical evidence of penetration, might not be perceived as particularly traumatic or particularly memorable to a young child, even though an adult might recognize it as clear sexual abuse. When we consider that Williams herself found these episodes "no more disturbing than many other aspects of a child’s life," and subject to "the child’s natural tendencies to forget," it becomes clear that many of the women, interviewed 17 years later, might simply have forgotten the event. They had not repressed the memory of the index episode; it had simply seemed too minor to be worth remembering.
Of course, we can debate back and forth the question of how many of the women might fall into this category. But at the least, it seems clear that the most scientifically reasonable approach is to restrict our analysis in the Williams study to the 37 women who did show evidence of genital trauma at the time of the index evaluation. These represent the cases where there can be no dispute that serious sexual abuse really occurred, and where the victim would not be expected simply to forget. Among these 37 cases, we are left with 18 who failed to report the episode in the follow-up interview.
But this number may need to be reduced even further when we allow for the effects of early childhood amnesia. Recall that the subjects were as little as 10 months old at the time of the index episode. As we have mentioned in past columns, failure to recall an event from one’s infancy clearly does not represent evidence of repression. Looking at Williams’ data, we find that about one quarter of the total sample of 49 non-reporting women were aged 4 years or younger at the time of the index episode. Applying this ratio to the subgroup of 18 cases described above, we would estimate that there were only about 14 women who 1) had medically documented genital trauma; 2) were old enough at the time to remember the experience; and 3) did not report the experience on the follow-up interview 17 years later. In short, we are left with only about 14 subjects in the only remaining study which we have left to analyze. The case for repression of memories of childhood sexual abuse, therefore, now hangs on only 14 people. But we have not yet considered the problem raised at the beginning of this column, response bias.
When we factor in response bias, what is left of the Williams study collapses completely. Remember that none of the subjects in the study was ever asked directly whether or not she remembered the known index episode; none of the non-reporting subjects was ever given a "clarification interview" in the manner of the Femina study described in the column of November, 1997. Recall also that 38 percent of the subjects in the Femina study chose not to disclose their history of abuse during an initial interview -- but when given clarification interviews, 100 percent revealed that they actually remembered. When we consider the roughly 14 still-unexplained cases out of the 129 subjects in the Williams study, we see that this number falls well within the range to be expected from non-disclosure alone-indeed, it is surprisingly small -- without any need to postulate the existence of "repression."
Response bias due to non-disclosure is a well recognized problem in social science research, documented in hundreds of studies throughout the last 50 years. In 1956, for example, the United States Congress authorized a continuing program of health surveys by the Public Health Service to provide reliable statistical information about health status in the United States population. This mandate produced a long series of studies over the next 20 years, in which scientists examined the accuracy of survey methodology. They found that people, even when carefully interviewed by trained personnel, consistently underreported life events which were known to have occurred. In one study, for example, 28 percent of subjects failed to report a one-day hospitalization which they were known to have undergone within the past year (7). In another, approximately 30 percent of subjects did not disclose a known car accident (without head injury or loss of consciousness) which was documented to have occurred 9 to 12 months previously (8). In yet another, 35 percent of subjects did not report a doctor’s visit which they were known to have made just within the last two weeks (9). Clearly, these subjects had not repressed the memory of having just gone to the doctor; the interviewers were simply witnessing response bias.
The scientists in these studies performed numerous analyses to determine what caused underreporting of life events (10). They found, for example, that people were more likely to withhold information about undesirable, threatening, or sensitive material as opposed to neutral material. They also discovered that non-disclosure of information was generally more common among non-White subjects than among White subjects, and more common among subjects of lower socioeconomic class than among subjects of higher socioeconomic class. It is worth noting, in this connection, that Williams’ subjects were mostly African-American women of lower socioeconomic class. And it need hardly be added that childhood sexual abuse would certainly rank among the most sensitive categories of information.
Another typical study of response bias was the National Crime Survey (11). Several studies in this survey used a "reverse record" system to validate reports of victimization. This technique involved sampling victims of crime from a record system, such as police files, and then locating the victims and interviewing them using a survey questionnaire. Information from interviews was then compared to actual records to establish the accuracy of the survey instrument. The studies consistently found that victims often failed to disclose crimes which they had recently experienced. In one study in Baltimore, for example, victims underreported burglaries by 14 percent, robberies by 24 percent, and assault by 64 percent. In another study in San Jose, assault was underreported by 52 percent and rape by 33 percent. In several of the studies, the interviewers probed in detail about the victims’ histories, while still not directly confronting the subjects regarding the known crime. But even with probing, high rates of underreporting persisted. Again, there is nothing to suggest that these people repressed the memory of the crimes; a certain percentage of them simply withheld the information on interview.
The list of studies of non-disclosure goes on and on (12). In every study, people have been found to underreport sensitive or embarrassing information of all types, such as alcohol consumption (13), drug use (14), having declared bankruptcy (15), drunk driving charges (15), arrest records (16), HIV infection (17), other medical conditions (18), psychiatric history (19), and, of course, childhood sexual abuse (20-22). Indeed, in one of these latter studies (21), no less than 72 percent of 116 self-acknowledged victims of childhood sexual abuse said that they had denied their history of abuse when initially interviewed -- a figure even more striking than the 38 percent non-disclosure rate in the Femina study. The recurring theme from all of this literature is obvious: when interviewees fail to report sensitive information from their histories, the investigators should immediately suspect response bias. Until they have addressed this problem (for example, by means of clarification interviews), they absolutely, positively, must not slip into the assumption that their subjects have forgotten (much less repressed) the information.
We return, now, to the Williams study. Remarkably, Williams does not mention any of the literature on non-disclosure which we have briefly reviewed above. Even the Femina study is not cited. Of course, Williams admits that none of her subjects was directly asked about the known index episode. She also admits to the existence of response bias. But she does not seem to recognize that many of the women in her own study might have chosen to withhold information about their index episode of childhood sexual abuse. If 35 percent of interviewees in a government study fail to disclose a simple doctor’s visit occurring within the last two weeks, and 64 percent of recent assault victims fail to reveal the incident even when interviewed in detail, how many victims of childhood sexual abuse, interviewed by an unfamiliar person, of higher socioeconomic class, 17 years later, might choose to withhold information which they actually remembered?
And if this is not enough, it is worth noting that Williams herself is an author of a large review article which seems to contradict the conclusions of her own study (23). In collaboration with two other authors, she reviewed the aftereffects of childhood sexual abuse in 45 studies examining 3,369 victims. As far as can be seen from the review, none of the victims in any of these studies was described as showing repression.
In a word, then, despite its wide publicity and frequent uncritical acceptance, the Williams study suffers from methodological problems which collectively render its results completely inadequate as a demonstration of repression. Indeed, when we add together the factors of lack of documentation, ordinary forgetfulness, childhood amnesia, and deliberate non-disclosure, it seems remarkable that only 38 percent of the women failed to report the index episode. In other words, the observation that a full 62 percent of the women described an event that had occurred 17 years earlier -- in the face of all of these opposing factors, and even when they were not asked specifically about it -- would seem to offer a persuasive demonstration that repression does not occur.
In conclusion, we do not mean to be unduly harsh on Williams. Her study methodology is vastly superior to most of the previous studies of repression discussed in previous columns. But the study is still subject to certain methodological limitations. In short, when assessing any prospective study of this type, the reader would be wise to remember the Tucson Garbage Project.
1. Rathje, W.L. & Hughes, W.W. The garbage project as a nonreactive approach: Garbage in...garbage out? in Sinaiko, H.W. and Broedlins, L.A. (Eds.) Perspectives on attitude assessment: Surveys and their alternatives. Washington, DC: Smithsonian Institution. 1975.
2. Williams, L.M. Recall of childhood trauma. A prospective study of women’s memories of child sexual abuse. J. Consult Clin Psychology 62: 1167-1176, 1994.?
3. For a detailed discussion of the misuse of the Williams study in courtrooms, see Hagen, M.A. Whores of the court: The fraud of psychiatric testimony and the rape of American justice. New York: Regan Books, 1997.
4. Muriam, D. Child sexual abuse -- genital tract findings in prepubertal girls I. The unaided medical examination. Am J Obstet Gynecol 160: 328-333, 1989.
5. Williams, L.M. Adult memories of child sexual abuse: Preliminary findings from a longitudinal study. American Society for Prevention of Child Abuse Advisor 5: 19-20, 1992.
6. McCahill, T.W., Meyer, L.C., & Fischman, A.M. The aftermath of rape. Lexington, MA: Lexington Books, 1979.
7. National Center for Health Statistics: Reporting of hospitalization in the Health Interview Survey: A methodological study of several factors affecting the reporting of hospital episodes. Washington, U.S. Dept. of Public Health, Education, and Welfare, Publication No. 584-D4, May 1961.
8. National Center for Health Statistics: Optimum recall period for reporting persons injured in motor vehicle accidents. Vital and Health Statistics. Series 20No. 50. Washington, U.S. Dept. of Public Health, Education, and Welfare, Publication No. 72-1050, April 1972.
9. National Center for Health Statistics: Health interview responses compared with medical records. Vital and Health Statistics. Washington, Public Health Service, PHS Pub. No. 1000-Series 20 No.7, July 1965.
10. United States Department of Health, Education, and Welfare, Health Resources Administration. A summary of studies of interviewing methodology. Washington, DC: DHEW Publication No. (HRA) 77-1343, 1977.
11. Lehnen, R.G. & Skogan, W.G (Eds.). The National Crime Survey: Working Papers. Vol I: Current and historical perspectives. U.S. Department of Justice Bureau of Justice Statistics, NCJ-75374, December 1981.
12. Numerous reviews and entire books have been written about non-disclosure and other pitfalls of interviewing techniques. See for example Taner, J.M. (Ed.) Questions about Questions. Russell Sage Foundation, New York, 1992; Mangione, T.W., Hingson, R. & Garrett, J. Collecting sensitive data. Soc Methods & Res 10: 337-346, 1982; Belson, W.A. Validity in survey research. Grower Publishing Co., Aldershot, England, 1986; Fowler, F.J. Improving survey questions design and evaluation. Sage Publications, Thousand Oaks, CA, 1995; Fowler,F.J., Survey research methods, 2nd edition. Newbury Park, Sage Publications, 1993, pp 69-93; and Zdep, S.M., Rhodes, I.N, Schwartz, R.M. & Kilkenny, M.J. The validity of the randomized response technique. In Singer E. & Presser S. (Eds.) Survey Research Methods: A Reader. University of Chicago Press, Chicago, IL, 1989.
13. See among many studies in this area, Polich, J.M., Armor, D. & Braiker, H.B. The course of alcoholism: Four years after treatment. Santa Monica: The Rand Corporation, 1979; Cooke, D.J. & Allan, C.A. Self-reported alcohol consumption and dissimulation in a Scottish urban sample. J. Stud Alcohol 4: 617-629, 1983; and many other such studies reviewed by Midanic, L.T. Validity of self-reported alcohol use: A literature review and assessment. Br. J. Addiction 83: 1019-1029, 1988.
14. Among the numerous studies on this topic are: Swerdlow, N.R., Geyer, M.A., Perry, W., Cadenhead, K., & Braff, D.L. Drug screening in "normal" controls. Biol Psychiatry 38: 123-124, 1995; and Blynn, S.M., Gruder, C.L. & Jegerski, J.A. Effects of biochemical validation of self-reported cigarette smoking on treatment success and on misreporting abstinence. Health Psychol 5: 125-136, 1986.
15. Locander, W., Sudman, S. & Bradburn, N. An investigation of interview method, threat and response distortion. J. Am Stat Assoc 1971; 71:269-275.
16. Tracy, P.E. & Fox, J.A. The validity of randomized response for sensitive measurements. Am Sociological Rev 1981; 46:187-200.
17. A large literature has arisen in this area. Among the many studies, see Marks G., Bundek. N.I., Richardson, J.L., Ruiz, M.S., Maldonado, N. & Mason, H.R. Self-disclosure of HIV infection: Preliminary results from a sample of Hispanic men. Health Psychology 11: 300-306, 1992; and McCarthy, G.M, Haji, F.S. & Mackie, I.D. HIV-infected patients and dental care: nondisclosure of HIV status and rejection for treatment. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 80: 655-659, 1995.
18. See, for example, National Center for Health statistics. Interview data on chronic conditions compared with information derived from medical records. Vital and health statistics data: Evaluation and methods research. Washington, DC, U.S. Department of Health, Education and Welfare, Publication No 1000-Series 2- No. 23, May, 1967; and Salinsky, M.C., Wegener, K. & Sinnema, F. Epilepsy, driving laws, and patient disclosure to physicians. Epilepsia 33: 469-472, 1992.
19. Examples include Sacks, M.H., Gunn, J.H. & Frosch, W.A. Withholding of information by psychiatric inpatients. Hosp Comm Psychiatry 32: 424-425, 1981; and Bennett, M. & Rutledge, J. Self-disclosure in a clinical context by Asian and British psychiatric outpatients. Br J Clin Psychol 28: 155-163, 1989.
20. Farrell, L.R. Factors that affect a victim’s self-disclosure in father-daughter incest. Child Welfare League of America 67: 462-468, 1988.
21. Sorensen, T. & Show, B. How children tell: The process of disclosure in child sexual abuse. Child Welfare League of America 70: 3-15, 1991.
22. Faulkner, N. Sexual abuse recognition and non-disclosure inventory of young adolescents. Doctoral dissertation. Available through University of Michigan Library, 300 North Zeeb Road, P.O. Box 1346, Ann Arbor, MI 48106-1346.
23. Kendall-Tackett, K.A., Williams, L.M. & Finkelhor, D. Impact of sexual abuse on children: A review and synthesis of recent empirical studies. Psychol Bull 113: 164-180, 1993.
by Harrison Pope, Jr.
The next four articles examine the question of whether childhood sexual abuse causes psychiatric disorders in adulthood. The series is not intended to "forgive" or exonerate the morally repugnant phenomenon of child sexual abuse in any way.
Does Smoking Cause Arthritis? Hundreds, if not thousands, of simply designed studies have now appeared in the literature, examining the prevalence of childhood sexual abuse in various populations of patients with psychiatric disorders. The typical hypothetical study goes something like this (1):
Drs. Harrison and James interviewed 50 women at a clinic who were being treated for eating disorders. Some of the women had bulimia nervosa, a disorder characterized by compulsive eating binges, followed by self-induced vomiting. Other women had anorexia nervosa; they had dieted until they weighed much less than they ought to weigh, but they still perceived themselves to be too fat. Many of the women had experienced both disorders at various times over the years. For comparison, Drs. Harrison and James also interviewed 50 women of the same age who were recruited from the community at large. The community women were included only if they showed no evidence of a major psychiatric disorder. The investigators found that 25 (50 percent) of the 50 women with eating disorders reported a history of childhood sexual abuse, as compared to only 5 (10 percent) of the comparison women from the community. This difference proved to be highly "statistically significant." Specifically, using a statistical test called Fisher’s exact test, the investigators calculated that the odds of such a difference occurring by chance alone were less than one in ten thousand. Thus, in the text of the paper, the authors added the phrase "p<0.0001 by Fisher’s exact test, two tailed." On the basis of this highly significant finding, they concluded that childhood sexual abuse played an important causal role in the development of eating disorders.
Are the conclusions of this hypothetical study justified? The answer is no, for a long series of reasons. First, our hypothetical investigators have failed to consider possible methodological errors in their design that might produce an apparent association between childhood sexual abuse and adult psychiatric disorders, even though a true association might not exist. Second, even if we allow that there is a true association between childhood sexual abuse and eating disorders, the investigators still have failed to demonstrate that the association is a causal association. In this article and the next, we consider flaws in our hypothetical study which might have caused the finding of a false association. In the following article, we move on to the issue of causality.
The first possible cause of a false association in our hypothetical study -- and in countless actual studies in the literature -- is the problem of selection bias. Selection bias refers to the possibility that the investigators have selected subjects who are not representative of the overall population of such people in the world at large. This bias could appear in two places. First, the women coming to the clinic for treatment of eating disorders may have a higher or lower prevalence of childhood sexual abuse than women with eating disorders in the general population. Second, the comparison women recruited from the community may have a higher or lower incidence of sexual abuse than community women as a whole. Let us look at each of these possibilities.
First, are Harrison and James’ subjects with eating disorders representative, in other words typical of people with eating disorders as a whole? One can think of many reasons why they may not be. For example, women with eating disorders who also happen to have a history of childhood sexual abuse may be more likely to seek psychological treatment than women with eating disorders who have no history of childhood trauma at all. Another possibility is that Drs. Harrison and James may be well known for their interest in childhood trauma. If so, then women with eating disorders who also happen to have a sexual abuse history may be somewhat more likely to seek treatment at Harrison and James’ clinic, whereas women without sexual abuse are somewhat more likely to visit a different clinic across town.
For reasons such as these, the sample of women investigated by Harrison and James will probably show a higher prevalence of sexual abuse than women with eating disorders as a whole. But there is likely an equally serious reverse bias in the investigators’ comparison group. Suppose that we find that Drs. Harrison and James chose their comparison subjects by posting an advertisement around their local medical area seeking "woman for a study involving interviews regarding their psychological symptoms" and offering them $30 to participate. Clearly, the women who are willing to respond to such an advertisement are not a random sample. In particular, women with a history of serious childhood sexual abuse may be embarrassed to sign up for a psychiatric interview. It is not worth $30 to them to contemplate a stranger asking them about their childhood experiences. Thus, the women who actually show up in Harrison and James’ offices, ready to be interviewed about their psychiatric histories, will likely exhibit a much lower rate of childhood sexual abuse than the true rate in the population.
Then, Harrison and James compound the problem even further with their requirement that these comparison subjects be free of psychiatric disorder. This criterion introduces a further selection bias into the comparison group, in that it creates a sample of "supernormals" who now have a much lower prevalence of psychiatric disorder as a whole than the natural rate in the community.
What is so bad about using "supernormals?" Consider an analogy from medicine. Suppose that we wish to test the hypothesis that cigarette smoking causes people to develop rheumatoid arthritis. (This hypothesis, as the reader probably knows, is completely false.) We examine 50 patients with confirmed rheumatoid arthritis and obtain detailed histories of their lifetime cigarette consumption. We find that 50 percent report some history of cigarette use. We then get a comparison group of individuals from the community at large, choosing them so that their average age and male/female ratio match closely with the rheumatoid arthritis group, and exclude from this group any individuals who show evidence of any significant medical disease. In this group of "healthy controls," we find, not surprisingly, that the lifetime prevalence of cigarette smoking is markedly lower than in the patients with rheumatoid arthritis. Can we conclude therefore that smoking causes rheumatoid arthritis? Of course not. We have simply selected against cigarette smokers in the comparison group by our insistence that they be "supernormals" with no serious medical illness of any type.
How are Drs. Harrison and James to deal with these problems? Fortunately, these methodologic difficulties are well understood in epidemiological research, and established methods exist to address them. For example, Drs. Harrison and James could obtain data from 1,000 women in the community at large. Upon examining the histories of these women, let us say that they find that 50 of the 1,000 display eating disorders. These women are unselected, in that they represent every case of eating disorders in the sample, regardless of whether or not they were seeking treatment. Then, Drs. Harrison and James select from the remaining 950 women an age-matched group of 40 comparison subjects without regard to the presence or absence of psychiatric disorders (except, of course, an eating disorder). Assuming that all of the 50 subjects in each of the two groups agree to cooperate with the investigation (thus minimizing any bias from self-selection), Drs. Harrison and James will have two groups unlikely to be seriously affected by selection bias.
Although these methods are admittedly more tedious and expensive than the "quick and dirty" hypothetical study described earlier, it is easy to see that they would produce much more reliable results. It is remarkable, then, to find that the great majority of published studies of childhood sexual abuse and adult psychiatric disorder fail to control for selection bias, and thus may produce findings just as suspect as our bogus conclusion that smoking causes arthritis. In short, by insisting on studies which have adequately addressed the issue of selection bias, we have already greatly narrowed the field of studies which meet our methodological standards for testing the relationship between childhood sexual abuse and adult psychiatric disorder.
This hypothetical study, and most of the material in the next three sections of this series are taken from a journal article which we have previously published: Pope, H.G. Jr., Hudson, J. I. "Does childhood sexual abuse cause adult psychiatric disorders? Essentials of methodology." J Psychiatry Law Fall, 1995: 363-381. We refer the reader desiring a full scientific presentation of these arguments to the original article.
The Pregnant Women and the Power Lines
Let us suppose that Drs. Harrison and James, the apocryphal investigators in our previous column, have now received a large research grant. Using this money, they design a new and far superior study to avoid the problem of selection bias. They obtain 50 women with eating disorders and 50 comparison women from a large community sample in the manner previously described. In this new improved study, the difference in the prevalence of sexual abuse between the eating- disordered patients and the comparison group patients has narrowed. Now, they find that only 20 (40 percent) of the 50 women with eating disorders report a history of childhood sexual abuse as compared to 10 (20 percent) of the controls. The difference between groups in the prevalence of history of childhood sexual abuse is not nearly as robust as in the previous, more seriously biased design, but it is still statistically significant (p<.05 by Fisher’s exact test, two-tailed). Can our investigators now conclude that sexual abuse has an etiologic role, albeit a more modest one, in eating disorders?
Unfortunately, they cannot, because they still have failed to deal with the equally serious potential problem of information bias. This form of bias refers to the error caused when the investigators obtain inaccurate information (for whatever reason) from subjects in one or both study samples, leading them to overestimate or underestimate the true prevalence of childhood sexual abuse in the groups.
How might such bias occur? To begin with, it can occur if the investigators are not blinded, meaning that they know whether they are interviewing a subject with an eating disorder or a comparison subject. When interviewing a subject with eating disorders, an unblinded investigator may perhaps probe slightly more carefully, ask slightly more detailed questions about a history of childhood sexual abuse, than when interviewing a comparison subject. Such a bias might be very subtle, and the investigator might introduce it quite unconsciously, yet it could slightly skew the responses of the eating-disordered women in favor of reporting a history of childhood sexual abuse and the responses of the comparison women against such reporting.
Of course, the investigators could deal with this problem by obtaining a sexual abuse history while blinding themselves to the group status of the subject. In other words, one investigator could obtain the eating disorder history on the subject, and then present the subject to a second investigator who would inquire about a sexual abuse history without any knowledge of whether the interviewee was an eating-disordered subject or a comparison subject. But this strategy does not completely resolve the problem of information bias, because even if the investigators are blinded, the subject herself is not. And she may be powerfully biased by a phenomenon known in psychology as "effort after meaning."
Effort after meaning refers to the natural human tendency to seek an explanation for our suffering(1). For example, if one were to become severely depressed at this moment, one could easily construct a very plausible explanation of why the depression started now as opposed to six months ago or six months in the future. One tends to do this automatically, because it is difficult to accept that sheer bad fortune, or random acts of nature, can account for one’s psychiatric problems. By analogy, it is likely that women with eating disorders (or patients with any sort of psychiatric disorder), in their effort after meaning, have carefully reflected about events in their past lives. They are likely to have thought about any traumatic or unusual situations which they endured, perhaps wondering if these traumas may have contributed to their current symptoms. Moreover, if they have seen recent popular presentations in the media about the issue of childhood sexual abuse and psychiatric disorders, they might be particularly likely to have reflected carefully upon possible abuse experiences, even relatively minor ones, in their history. By contrast, subjects with no psychiatric disorder may have devoted little thought to their childhood experiences, because they had no motivation to engage in an effort after meaning.
Effort after meaning produces a type of information bias known as recall bias, and this bias is frequently encountered throughout medicine. For example, suppose that we decide to study 50 mothers who have just given birth to an infant with congenital malformations, and then interview a comparison group of 50 mothers whose infants were entirely normal. We ask both groups of mothers if they can remember having been in the vicinity of high tension power lines at any time during their pregnancy. It would not be surprising if the mothers of the malformed infants recalled a higher frequency of such exposure - not because their true exposure was any higher, but because these mothers had spent long and tortuous hours reflecting upon every possible adverse experience during their pregnancy that might possibly have caused the infant becoming malformed. The mothers of the normal infants, on the other hand, would have devoted little thought to their experiences with power lines, even though their average level of exposure probably was about the same as that of the first group of women. Indeed, one recent study of mothers of malformed infants came out with just this sort of finding. These mothers correctly recalled various types of exposure that had actually been documented in their medical records, such a urinary tract or yeast infections, antibiotic drug use, and use of birth controls after conception, much more often than did a comparison group of mothers of normal infants.(2)
As can be seen, recall bias poses a serious problem for Drs. Harrison and James in their hypothetical study. Unlike the study of mothers just described, where one can check medical records to confirm various exposure factors, our investigators have little opportunity to confirm reports of childhood sexual abuse in their two groups of women. Indeed, the inability to confirm reports about the past is always and forever a problem in retrospective studies, as we have discussed earlier in our articles regarding retrospective studies of the repression hypothesis. The only definitive way to resolve the problem, as we have also discussed earlier, would be to do a prospective study. Such a design is proposed in the final column of this series.
1. This phenomenon is discussed in Barlett F. C.: Remembering: A Study in Experimental and Social Psychology. Cambridge University Press: Cambridge, 1932. See also Tennant C: Life events and psychological morbidity: the evidence from prospective studies. Psychological Medicine 13:483-486, 1983.
2. Werler, M.M., Pober, B.R., Nelson, K, Holmes, L.B. Reporting accuracy among mothers of malformed and nonmalformed infants. Am J Epidemiology 129:415-421, 1989.
Don’t Buy that Lawrence Welk Recording! The Problem of Confounding
In the previous two columns, we have shown how selection bias and information bias seriously compromise virtually all retrospective studies. But let us suppose that our intrepid investigators of our previous two columns, Drs. Harrison and James, have now received an even larger research grant to do an even more refined study. They obtain a huge community sample, and select individuals with eating disorders and matched control subjects with careful attention to minimize selection bias. Then they interview subjects in both groups under blinded conditions to avoid any information bias introduced by the investigator. Instances of sexual abuse in both the eating disordered and control group are scored only if they are unequivocal and meet rigorous diagnostic criteria of demonstrated reliability. Further, let us suppose that the investigators are able to obtain confirmatory evidence in some manner to show that the sexual abuse actually did occur in the cases in which it is reported (this last item is probably a somewhat unrealistic expectation, but let us grant it for the purposes of argument). Let us suppose that, even with all of these rigorous methods to control for bias, Drs. Harrison and James still are able to show a statistically significant difference when they compare the prevalence of sexual abuse in the subjects with eating disorders and the control subjects. Can they now, at last, conclude that childhood sexual abuse contributes to the etiology of eating disorders?
Unfortunately, they still cannot. We will now grant that they have shown as association between childhood sexual abuse and eating disorders. However, as we have stated earlier, the fact that there is an association between A and B does not necessarily mean that A caused B. In fact, logically, there are three alternative explanations for the association as shown in the following figure:
1) A -> B
2) B -> A
3) A <- C -> B
In examining this figure, let us assume that childhood sexual abuse is "A" and adult psychiatric disorder is "B." The first possibility, as the figure shows, is that A causes B. In this case, that would be the possibility that childhood sexual abuse causes adult psychiatric disorders. This of course in the hypothesis that we wish to test. But to establish this possibility, we must first rule out two other possibilities. First, we must consider the possibility that B causes A (i.e., that psychiatric disorder somehow predisposes to sexual abuse), and second, we must allow for the possibility that B and A do not cause one another, but both are caused by a third factor, C (which is often called a "confounding variable").
Let us look at these alternative possibilities. First, consider the possibility that B causes A. There are many examples of this type of association in ordinary life and in clinical medicine. Suppose, for example, that we interview 100 overweight subjects and ask them if they have a history of having used artificial sweeteners in their coffee at some time in the last year. We then pose the same question to 100 thin subjects. We find a highly significant difference showing a clear association between use of artificial sweeteners and being overweight. Do we conclude, therefore, that artificial sweeteners cause obesity? Clearly not. The true direction of causality is that B causes A, namely that being overweight leads individuals to use artificial sweeteners more frequently.
More difficult and less trivial examples come from clinical medicine. Thirty years ago, for example, a study found that agricultural workers who were more physically active were less likely to develop heart disease than sedentary agricultural workers . Would it be correct to conclude, therefore, that being sedentary contributes to the evolution of heart disease? No. We must allow for the alternative possibility that workers who already had early symptoms of incipient heart disease (e.g. chest pain on exertion) would be more likely to choose sedentary agricultural jobs than their counterparts who had no symptoms of evolving cardiac disease. In other words, B may have led to A, rather than A to B.
Another obvious example exists in the area of sexual abuse: individuals with mental retardation are more likely to have experienced sexual abuse than individuals of normal intelligence . But clearly it would be illogical to conclude that childhood sexual abuse causes mental retardation. Rather, mentally retarded individuals are more at risk for victimization because they are less able to defend themselves against abuse.
But do these arguments extend to adult psychiatric disorders? Is it reasonable to argue that bulimia nervosa or depressive illness, or anxiety disorders, appearing in an individual at the age of 20, could possibly have predisposed him or her to have been sexually abused at the age of eight? This possibility is not as far fetched as it might seem. Specifically, studies have shown that individuals with adult psychiatric disorders have often experienced prodromal symptoms (in other words, premonitory symptoms) of their disorders extending far back into childhood. For example, adults who display panic disorder or bulimia nervosa are more likely to have experienced fear of going to school ("school phobia"), fear of being separated from their mothers ("separation anxiety") or bedwetting ("primary enuresis") in childhood . Similarly, individuals with eating disorders as adults often have histories of depressive or anxiety disorders long prior to the onset of the eating disorder4. Therefore, it is possible that some individuals with eating disorders in Harrison and James’ study may have displayed a degree of depression or other psychological distress, even years ago in childhood, that rendered them more vulnerable to being preyed upon by potential abusers.
Admittedly, this particular direction of causality might account for only a small portion of the possible association between childhood sexual abuse and adult psychiatric disorders. However, we still have to rule out the last of the possibilities shown in the figure above, namely that A and B are both caused by a confounding variable, C. The issue of confounding is again a constant problem, both in ordinary life and in clinical medicine. To begin with a simple example, suppose that we were to study 100 residents of a nursing home and ask them if they had ever purchased a Lawrence Welk recording. Lawrence Welk was a famous performer many years ago, and his recording were very popular in the 1930’s and 40’s. Therefore, we would likely find that a high percentage of the nursing home residents reported that they had bought at least one such a recording at some time. If we then obtained a comparison group from the community at large, we would undoubtedly find that a far smaller percentage of our comparison subjects had made such a purchase. In fact, some of them would probably report that they had not even have heard of Lawrence Welk. Does it follow from our findings that buying a Lawrence Welk recording will cause you to end up in a nursing home? Should we put warning stickers on all Lawrence Welk recordings in record stores, alerting potential purchasers of the risk? Clearly not. In fact, the association between ownership of Lawrence Welk recordings and nursing home residents is simply attributable to the confounding variable of advanced age. In other words, age is the "C" in the figure above.
Turning to medicine, the literature is filled with the corpses of theories that failed to take into account the possibility of confounding variables. Even elegant and expensive studies, involving big teams of investigators and hundreds of thousands of dollars in costs, have sometimes proved dead wrong when it was later discovered that a confounding variable had created a mere illusion of causality. Of many examples that could be cited, one was the finding of an association between the use of inhaled nitrites (so-called "poppers") and the development of AIDS . In the early 1980’s before the human immunodeficiency virus (HIV) had been isolated, various epidemiologic studies were conducted to assess what factors might cause people to develop AIDS. It was found that homosexual men who used "poppers" to get a "rush" during sexual activity were markedly more likely to develop AIDS than homosexual men who had not used these drugs. Some studies even conducted elaborate statistical tests, called regression analysis (to be discussed in more detail next month), in an attempt to rule out possible confounding variables. Nevertheless, inhaled nitrites still emerged as a statistically significant factor, and it was concluded that they might cause, or at least contribute to, the development of AIDS.
Now, of course, we know that nitrites do not cause AIDS, and that the disease is instead caused simply by infection with a specific virus, HIV. It turns out, in retrospect, that certain sexual practices that predispose to HIV transmission (especially receptive anal intercourse) are closely associated with use of "poppers." In other words, the association between nitrite use and AIDS was a real one, but it was not a causal association at all. Instead, the association was caused by the presence of a confounding variable, namely specific sexual practices.
Returning, then, to the new hypothetical study by Drs. Harrison and James, we see that an association between childhood sexual abuse and adult psychiatric disorder, however rigorously proved, might not be a causal association at all. it might simply be due to any of a number of confounding variables. Individuals who have been sexually abused in childhood are also likely to have been physically abused, neglected, or subjected to all manner of other difficulties while growing up. Even more importantly, there may have been a genetic loading in their families for disorders such as alcohol dependence or manic-depressive illness . Relatives with alcoholism or manic episodes (the "high" periods of manic-depressive illness), in turn, may be more likely to abuse a child in the family. But that abuse victim already carries the genetic predisposition to develop psychopathlogy, even if she were not sexually abused. In other words, childhood sexual abuse and psychopathology would be expected to "travel together" down the family tree as a result of the confounding variable of genetics alone, even if the sexual abuse did not itself cause psychiatric disorders.
Data that support this speculation come from one recent study that described 12 sexually abused women with bulimia nervosa . This study was one of the few in which the psychiatric diagnosis of the perpetrator, as well as that of the victim, was assessed. Of the eight women in this study found to have been abused by a biological relative, six (75 percent) were abused by a family member diagnosed with alcohol dependence, a major mood disorder (such as manic-depressive illness), or both. Now, there is substantial evidence that alcoholism and major mood disorders are more prevalent in the family trees of individuals with bulimia nervosa than in the population at large, raising the possibility that there is a genetic link among these various disorders . It is possible, therefore, that genetic factors alone might account for the association of sexual abuse and bulimia nervosa observed in this investigation, and that sexual abuse itself had no role in causing the adult eating disorder at all.
In summary, association does not prove causality. This is not a difficult concept. It represents one of the most basic teachings of "Psychology 101." And it is easy to illustrate, as shown by our examples of the association between use of artificial sweeteners and obesity, or purchase of a Lawrence Welk recording and nursing home residence. Yet this elementary principle is ignored, or only barely acknowledged, in many scientific studies of childhood sexual abuse. It is even more rarely noted in popular reports of these studies in the media. The lay reader, hearing the latest media report of a new "major study" like that of Drs. Harrison and James, must be wary. A history of childhood sexual abuse may well be associated with some adult psychiatric disorders, but it is premature to jump from this finding to an assumption of causality.
. McDonough, J.R., Hames, C.G., Stulb, S.C., & Garrison, G.E., Coronary heart disease among Negroes and whites in Evans County, Georgia, J. Chron Dis 18: 443-458, 1965.
. Tharinger, D., Horton, C. B., & Millea, S. Sexual abuse and exploitation of children and adults with mental retardation. Child Abuse Negl 14: 301-312, 1990, and Stromsness, M. M. Sexually abused women with mental retardation: Hidden victims, absent resources. Women and Therapy 14: 139-152, 1993.
. Robinson, P. H. & Holden, N. L. Bulimia nervosa in the male. Psychol Med. 16: 795-803, 1986, and Perugi, G., Deltito, J., Soriani, A., Musetti, L, et al. Relationships between panic disorder and separation anxiety with school phobia. Compr Psychiatry 29: 98-107, 1988.
. Hudson J. I., Pope, H. G., Jr., Yurgelun-Todd, D., Jonas, J. M., & Frankenburg, F. R. A controlled study of lifetime prevalence of affective and other psychiatric disorders in bulimic outpatients. Am J. Psychiatry 144: 1283-1287, 1987, and Brewerton, T. D., Lydiard, R. B., Herzog, D. B., Brotman, A.W., O’Neil, P. M., & Ballenger, J. D. Comorbidity of axis I psychiatric disorders. J. Clin Psychiatry 56: 77-80, 1995.
. For a discussion of how an association was erroneously assumed to be causal in this case, see Vandenbroucke, J. P., Pardoel, VPAM: An autopsy of epidemiologic methods: the case of "poppers" in the early epidemic of the acquired immunodeficiency syndrome. Am J. Epidemiol 129: 455-457, 1989.
. Cadoret, R. J. Genetics of alcoholism. In: Alcohol and the Family: Research and Clinical Perspectives. Edited by R. L. Collins, K. E. Leonard & J. S. Searles. New York, Guildford Press, (1990); Tsuang, M. T. & Faraone, S. V. The Genetics of Mood Disorders. Johns Hopkins University Press: Baltimore, (1990).
. Bulik, C. M., Sullivan, P. F. & Rorty, M. Childhood sexual abuse in women with bulimia. J. Clin Psychiatry 50: 460-464, 1989. . Hudson, J. I. & Pope, H. G., Jr. Affective spectrum disorder: Does antidepressant response identify a family of disorders with a common pathophysiology? Am J. Psychiatry 147: 552-564, 1990.
Alcohol Consumption, Lung Cancer and Higher Mathematics
As illustrated by the experiences of our apocryphal investigators in the previous three issues of this column, the problems of selection bias, information bias, and confounding seriously compromise the conclusions of virtually all published studies examining the relationship between childhood sexual abuse and adult psychiatric disorders. As these flaws are more widely acknowledged, newer and more sophisticated studies are attempting to address the question of causality.
Are these newer efforts helping us to reach a more definite answer?
Consider one recent study in the prestigious American Journal of Public Health (AJPH). The authors examined survey responses from 1099 randomly selected American women regarding their experiences of childhood sexual abuse and history of bulimic behaviors. The methodology was careful and sophisticated. The women were administered face-to-face structured interviews in which they were asked detailed questions. Specified operational diagnostic criteria were used both for the diagnosis of bulimic behaviors and for childhood sexual abuse. These strategies would be expected to minimize the problems of selection bias and information bias, discussed in our previous columns. When the investigators analyzed the results of the study, they found that, even with these biases controlled, a clear association between bulimic behaviors and childhood sexual abuse still remained. In fact, binge-eating was twice as common among abused respondents as compared to non-abused respondents, and a full bulimic syndrome (binge-eating, overconcern about body weight, and a history of vomiting, laxative abuse, or similar measures) was three times as common in abused as opposed to non-abused women.
So much for association. But is the association causal? Or are bulimic behaviors and sexual abuse mutually caused by confounding variables? To deal with this question, the authors used a sophisticated mathematical technique called regression analysis, in which they calculated the "population attributable risk," or ARc, where
ARc = (x-y)/x
and where y, in turn, is defined by the formula
y = Sigma n I *
j j j
Without stopping to explain all of these symbols (they are all defined in the paper), suffice it to say that even after mathematically controlling for the effects of age, ethnic group, and parents’ educational level, the authors found bulimic behavior was still associated with childhood sexual abuse. Specifically, the authors suggest that among the women with the full bulimic syndrome just described, as many as 34 percent of the cases would not have occurred in the absence of childhood sexual abuse.
Certainly this analysis represents an elegant mathematical treatment, far more sophisticated than the simplistic designs of Doctors Harrison and James in the previous columns. Do these findings now finally allow us to conclude that childhood sexual abuse causes bulimic behaviors, at least in some cases?
Absolutely not. Consider again an example from medicine. Suppose that we surveyed 10,000 patients for their history of alcohol consumption and their history of lung cancer, using the same survey and interviewing techniques described in the bulimia study. We would find an association between alcohol and lung cancer. But is this a causal association? As it turns out, it is not. Alcohol in itself has little or no role in causing lung cancer. But people who drink alcohol also smoke more cigarettes, and cigarettes really do cause lung cancer. The confounding variable, in other words, is tobacco consumption.
Now suppose that we do a logistic regression analysis, just as in the bulimia study described earlier, and calculate "adjusted odds ratios" to measure the contribution of alcohol to lung cancer. As with the above bulimia study, we control for numerous variables, such as age, ethnic group, and parents’ level of education. We even go further and control for additional variables: the respondents’ height, weight, and blood pressure, their religious affiliation, political persuasion, and even their favorite brand of breakfast cereal. But we still forget to control for their consumption of cigarettes. What happens? We have used a lot more computer time, and our mathematics looks even fancier, but our findings continue to give the erroneous impression that alcohol is the culprit. We have still missed the real cause of lung cancer.
Now it may still be correct, in a certain technical sense, to say that alcohol is a "risk factor" for lung cancer, because people who drink do show a higher rate of lung cancer. And we could blur the whole issue a bit by arguing that alcohol is part of a "multifactorial group of interacting etiological factors" in lung cancer or that it is part of an "integrated causal model" for lung cancer. All of this would sound very impressive if we did not know the simple truth. Alcohol doesn’t cause lung cancer; cigarettes do.
Returning then, to the bulimia study reviewed here, we must acknowledge that its methods represent a considerable advance over most of the studies previously published in this area. But despite its sophistication, it still ultimately founders on the issue of confounding variables. Perhaps childhood sexual abuse is one of several related factors in the cause of bulimia nervosa, but these results still do not permit us to conclude that it necessarily has any causal role at all.
1. Wonderlich, S.A., Wilsnack, R.W., Wilshack, S.C., Harris, T.R.: Childhood sexual abuse and bulimic behavior in a nationally representative sample. Am J. Public Health 86: 1082-1086, 1996.
Psychological Medicine, 1995, 25, 121-126.
CAN MEMORIES OF CHILDHOOD SEXUAL ABUSE BE REPRESSED?
Harrison G. Pope Jr. and James I. Hudson
From the Biological Psychiatry Laboratory, McLean Hospital, Belmont and Department of Psychiatry, Harvard Medical School, Boston, MA, USA.
We sought studies which have attempted to test whether memories of childhood sexual abuse can be repressed. Despite our broad search criteria, which excluded only unsystematic anecdotal reports, we found only four applicable studies. We then examined these studies to assess whether the investigators: (1) presented confirmatory evidence that abuse had actually occurred; and (2) demonstrated that their subjects had actually developed amnesia for the abuse. None of the four studies provided both clear confirmation of trauma and adequate documentation of amnesia in their subjects. Thus, present clinical evidence is insufficient to permit the conclusion that individuals can repress memories of childhood sexual abuse. This finding is surprising, since many writers have implied that hundreds of thousands, or even millions of persons harbor such repressed memories. In view of the widespread recent public and scientific interest in the areas of trauma and memory, it is important to investigate further whether memories of sexual abuse can be repressed.
Is it possible for victims of childhood sexual abuse to ‘repress’ their memories? Many recent writers, both popular and scientific, have suggested that repression is common (Bass & Davis, 1988; Blume, 1990; Fredrickson, 1992; Herman, 1992). On the other hand, experimental evidence for the existence of repression is less convincing. In an extensive review, Holmes (1990) argues that laboratory studies have failed to produce clear evidence for repression, despite more than 60 years of attempts to do so.
Of course, even if repression cannot be induced in the laboratory, it might still occur in real life, such as in victims of childhood sexual abuse. And even if repression occurred in only a fraction of such victims, one would still expect to see many cases clinically, since childhood sexual abuse is common (Nash & West, 1985; Wyatt, 1985; Russell, 1986; Bagley & Ramsay, 1986; Finkelhor et al. 1990; Anderson et al. 1993). Specifically, using the conservative estimate that 10 percent of women and 5 percent of men have endured serious childhood sexual abuse, then 14,000,000 adults in the United States alone are former victims. If repression occurred in only 10 percent of these cases, at least 1,400,000 Americans, and millions more worldwide, now harbor such repressed memories. Given this huge pool of predicted cases, one might expect to find in the literature various published studies of patients exhibiting well-documented cases of repression.
In this paper we review this literature. We first discuss the criteria necessary for a satisfactory confirmation of the hypothesis that repression can occur, and then we examine the relevant studies.
CRITERIA FOR AN ACCEPTABLE STUDY
Evidence that traumatic events occurred
To demonstrate repression, one must first confirm that the traumatic events actually occurred. In many studies of trauma, such documentation is straightforward, because the events are a matter of historical record. Examples include a 4-year follow-up of 26 children kidnapped on a school bus (Terr, 1979, 1983); a study of 16 children who witnessed a parent murdered (Malmquist, 1986); an investigation of 113 elementary school children involved in a sniper’s attack (Pynoos & Nader, 1989); a 4-year follow-up of 34 survivors of a marine explosion (Leopold & Dillon, 1963); an examination of 100 concentration camp survivors 12-20 years afterwards (Strom et al. 1962); and a follow-up of 23 victims of Nazi persecution, half of whom were age 19 or less at the onset of trauma (Chodoff, 1963). Interestingly, although many of the subjects in these studies exhibited severe post-traumatic psychopathology, it appears that no subject in any of these studies developed amnesia. Indeed, many recalled the events in extraordinary detail.
Of course, sexual abuse is more difficult to document than kidnappings, murders, or war crimes. But there are various types of documentation that most investigators would accept: contemporaneous medical evidence of the abuse; photographs; reports from reliable and unbiased witnesses; or confirmation by the perpetrator himself (or herself). Admittedly, many cases of alleged childhood sexual abuse can be neither definitely confirmed nor refuted, but given the large number of expected cases, there should still remain an adequate number sufficiently well-documented for study.
Evidence for ‘psychogenic’ amnesia
The second requirement for a satisfactory test of repression is to demonstrate that the victim actually developed ‘psychogenic’ amnesia for the trauma. To demonstrate amnesia, one must first exclude cases in which victims simply tried not to think about the events, pretended that the events never occurred, or appeared to derive secondary gain by merely claiming to have amnesia (i.e. to avoid embarrassment or to extend a legal statute of limitations).
Secondly, one must show that the failure of memory exceeds ordinary forgetfulness. Some experiences, though clearly meeting published research criteria for sexual abuse, may not be particularly memorable to a child (for examples, see Pope et al. 1994). Thus, a satisfactory test of the repression hypothesis must demonstrate amnesia for abuse sufficiently traumatic that no one would reasonably be expected to forget it.
Thirdly, to demonstrate ‘psychogenic’ amnesia, one must exclude cases in which amnesia developed for some biological reason, such as seizures, alcohol and drug intoxication, or head trauma. This last factor figures strongly in war neuroses: in one study of 200 cases, 50 percent of the soldiers had lost consciousness and another 22 percent were ‘dazed’ on the battlefield (Henderson & Moore, 1944). A further biological source of amnesia is the immaturity of the developing central nervous system in young children. Children have nearly complete amnesia for events before the age of three, and substantial amnesia for events before the age of six (Fivush & Hudson, 1990; Usher & Neisser, 1993). Of course, it might be argued that some sexual abuse would be so traumatic that it ought to be remembered even by a child aged five or younger. But again, given the high predicted prevalence of repression, one should expect to find ample numbers of cases without having to rely on under-age-six examples.
Provided that the above exclusion criteria are met, the postulated mechanism of the amnesia -- whether it be called ‘repression’, ‘dissociation’ or ‘traumatic amnesia’ -- is unimportant. It is sufficient that a study should simply exhibit individuals with complete amnesia for well-documented abuse that was too striking to be normally forgettable.
REVIEW OF THE EVIDENCE
In an attempt to find methodologically sound evidence for repression of memories of childhood sexual abuse, we searched the literature for studies that had examined this phenomenon in a series of patients. We included all reports that presented a group of patients analyzed in any quantitative manner; only unsystematic anecdotal reports were excluded. Despite this broad search, we found only four applicable studies in the literature (Herman & Schatzow, 1987; Briere & Conte, 1993; Loftus et al. 1994; Williams,1994). We review these studies below, using the criteria developed above.
The Herman & Schatzow study
In the first attempt to document repression of memories of childhood sexual abuse, Herman & Schatzow (1987) reported on 53 women who they treated in time-limited group therapy for ‘incest survivors’. The authors do not specify whether subjects were selected prospectively or retrospectively; it is also not clear whether they represented consecutive individuals or a chosen subsample.
Among the 53 patients, 14 (26 percent) were rated as having ‘severe’ amnesia for the presumed incest, and might, therefore, represent examples of true repression. However, since the mean age of onset (+ or - S.D.) of abuse in this group is reported as 4.9 + or - 2.4 years, abuse in some of these women apparently occurred during the period of normal childhood amnesia at age five or earlier. Thus, only a subset of the 14 women in the sample would meet the criterion of displaying full amnesia for events occurring at ages older than five.
But this subset shrinks further, or perhaps vanishes entirely, when we apply the second criterion; namely, the requirement that the abuse be confirmed. Only 21 (40 percent) of the 53 patients obtained ‘corroborating evidence’ of the incest, and it is not clear whether these 21 cases include any of the women with ‘severe’ amnesia described in the paragraph above. Admittedly, another 18 (34 percent) of the 53 patients were reported to have ‘discovered that another child, usually a sibling, had been abused by the same perpetrator’. But the evidence that supported these latter discoveries is not specified, neither does it follow that abuse of a sibling, even if true, confirms abuse in the index case.
In short, it is not certain that any of the 3 subjects met both the criteria of clear amnesia and clear confirmation of trauma. Indeed, of the four cases examples given, three do not meet the criteria (1 and 2 did not have amnesia, and 4 had virtually no confirmation). Case 3 apparently had at least partial amnesia, and good confirmatory evidence. But even this case, it appears, is not actually a real person, since the authors explain that ‘all examples cited are composites of several cases’.
Parenthetically, it is curious that Herman (1981) hardly mentions repression or amnesia in her study, Father-Daughter Incest, published only 6 years earlier. In this book, all of the 40 women in the case series displayed apparently clear and lasting memories of their abuse.
The Briere & Conte study
In the second study, Briere & Conte (1993) analyzed questionnaire responses of 450 patients ‘with self-reported histories of sexual abuse and who were currently in therapy’. The questionnaire contained a single question regarding amnesia for sexual abuse: ‘During the period of time between when the first forced sexual experience happened and your eighteenth birthday was there ever a time when you could not remember the forced sexual experience?’ A total of 267 (59 percent) of the 450 subjects answered ‘yes’.
This result is open to several methodological questions. Subjects were ‘recruited by their therapists’; further details of the inclusion or exclusion criteria are not provided. It is not clear whether the abuse events were confirmed in any of the cases. Neither is it clear what portion of the subjects had experienced abuse sufficiently traumatic that they would reasonably be expected to remember it always. And, most importantly, a ‘yes’ answer on this single question does not demonstrate clear repression of a traumatic memory. A subject answering ‘yes’ might mean only that he or she gave no thought to the event during some period, or attempted to deny or minimize the event. No follow-up questions were asked to assess these possibilities.
Finally, some subjects may have been influenced by suggestion. All were in treatment with therapists who were part of an ‘informal sexual abuse treatment referral network’, and who, therefore, may have communicated to their patients, explicitly or implicitly, that repression of traumatic experiences was to be expected. With this potential degree of expectation, and with therapists choosing which subjects would receive the questionnaire, it would not be surprising if many subjects answered ‘yes’ to a question that asked if there was ever a time when they could not remember an abuse experience.
The Loftus, Polonsky & Fullilove study
Similar limitations affect the recent study of Loftus et al. (1994). In a design comparable to that of the Briere & Conte study, these authors interviewed 52 women with a history of abuse and asked whether they had forgotten the abuse ‘for a period of time and only later [had] the memory return’. In contrast to the 59 percent rate found by Briere & Conte, only 10 (19 percent) of these 52 subjects reported a period of forgetfulness.
However, even the much lower 19 percent figure is subject to the same methodological questions as the previous study; none of the cases of abuse was independently confirmed, neither was the extent or nature of the ‘forgetfulness’ investigated. In other words, the evidence does not show that any of the 10 women displayed lasting amnesia for documented events that would normally be expected to be unforgettable.
The Williams study
In the most recent study, Williams (1994) has reported an investigation with more rigorous design. She presents data from interviews of 129 women who, as children, had been brought to a city hospital emergency department in 1973-1975 for treatment and collection of forensic evidence after reported sexual abuse, even if no physical trauma occurred. The abuse was thus documented in medical records and interviews with research staff conducted at the time. The author notes that not all cases were extreme: about a third involved only touching and fondling (Williams, 1992).
The women were aged 10 months to 12 years at the time of the abuse. They were contacted and interviewed approximately 17 years later, at age 18 to 31. They were told that their names had been ‘selected from the records of people who went to the city hospital in 1973-1975’ for an ‘important follow-up study of the lives and health of women who during childhood received medical care at the city hospital’. The interviewers asked detailed questions about each woman’s history of sexual abuse, including questions about events that the patient herself perhaps did not define as abuse, but which others had. The women were also asked if anyone in their families ‘ever got in trouble for his/her sexual activities’. However, it appears that if the subjects still failed to report the known episode of abuse after these questions, they were not asked directly about their documented visit to the hospital to see whether they then acknowledged remembering it.
Forty-nine (38 percent) of the 129 women did not report the abuse event to the interviewer, and Williams speculates that most of these cases represent actual amnesia, rather than voluntary withholding of information. She supports this view by pointing out that many women gave detailed descriptions of other personal or embarrassing childhood events, including other experiences of sexual abuse, while still not acknowledging the index episode.
But it is hazardous to conclude that Williams’ 49 ‘non-reporters’ actually had amnesia. In considering this question, it is instructive to examine a similarly designed study by Femina and colleagues (1990). These investigators interviewed 69 young adults (mean age, 24 years) whose histories of abuse (primarily physical abuse) had been extensively documented years earlier during adolescence. On interview, 26 of the 69 subjects (also a proportion of 38 percent) gave responses discrepant with their previously documented histories. In particular, 18 of these 26 individuals were known to have been severely abused in childhood, yet they denied or minimized any experiences of abuse on interview as young adults.
Femina and colleagues then performed a second interview (which they called a ‘clarification interview’) with eight of these 18 ‘deniers’. When asked directly about their known abuse histories, it appears that all eight individuals admitted that they actually remembered, but had withheld the information during their first interview. For example, one woman, ‘whose mother had attempted to drown her in childhood and whose stepfather had sexually abused her, minimized any abuse at all on follow-up’. But in the clarification interview, when presented with this history, she admitted, ‘I didn’t say it cuz I wanted to forget. I wanted it to be private. I only cry when I think about it’. Similarly, one man, who as a boy was frequently beaten by his father, also minimized any history of abuse on the first interview. When presented with the history in the clarification interview, he acknowledged the beatings but said, ‘my father is doing well now. If I told now, I think he would kill himself’.
On the basis of these clarification interviews, Femina and colleagues list reasons for non-reporting of abuse as, ‘embarrassment, a wish to protect parents, a sense of having deserved the abuse, a conscious wish to forget the past, and a lack of rapport with the interviewer’. However, no case of non-reporting was ascribed to amnesia. Given this observation, it would be unwarranted to conclude that the 49 non-reporting subjects in Williams’ study actually had amnesia, since no clarification interviews were performed.
Indeed, the underreporting of life events on interview has been recognized and studied for several decades. For example, a study by the United States national Center for Health Statistics (1961) found that 28 percent of respondents, when interviewed in detail by trained workers, failed to report a one-day hospitalization that they were known to have undergone during the past year. Similar investigations have found that about 30 percent of individuals known to have been involved in an automobile accident (without recorded injury) did not report it on detailed interviews 9-12 months later (National Center for Health Statistics, 1972); 35 percent of another cohort did not report a visit that they had made to a doctor within the past 2 weeks (National Center for Health Statistics, 1965); and 54 percent did not report a hospital admission that they had undergone 10-11 months prior to the date of interview (National center for Health Statistics, 1965). In light of these figures, it does not seem necessary to posit repression to explain Williams’ 38 percent rate of non-reporting for events occurring 17 years earlier.
Williams’ results may also reflect normal childhood amnesia: 25 (51 percent) of her 49 non-reporting subjects had experienced their index episode of abuse at age six or earlier. Indeed, the only case presented in detail in the paper is of a woman whose abuse (of unspecified severity) occurred at age four. And even among individuals who were older at the time of abuse, one must allow for ordinary forgetfulness for events not perceived as strikingly memorable, especially among the one-third of subjects who experienced only touching and fondling. Thus, Williams’ 38 percent rate of non-reporting might be readily explained as a combination of cases of early childhood amnesia, cases or ordinary forgetfulness, and perhaps many cases of failure to report information actually remembered. Additional discussions of methodology of Williams’ study, and some of the other studies analyzed above, may be found in other recent works (Loftus, 1993; Ofshe & Singer, 1994).
DESIGN OF FUTURE STUDIES
Williams’ study provides a useful starting point for the design of a rigorous test of the repression hypothesis. First, one must seek a group of individuals unequivocally documented to have been traumatized, sexually or otherwise. For example, one might begin with a group of medical records, as Williams did, or one could trace all victims identified by a confessed perpetrator or a reliable witness. Opportunities of the latter type are uncommon, but may arise periodically in forensic settings. Secondly, one would select all individuals who were above the age of five at the time of abuse, and who were definitely known to have endured abuse too traumatic to be normally forgettable. Thirdly, one would locate and interview these individuals -- with suitable ethical and therapeutic precautions -- with regard to any past history of trauma.
Fourthly, subjects who still denied abuse on this general interview would then receive a ‘clarification interview’ in which they were asked directly about the known abuse event. If some subjects still reported amnesia even in response to the direct questions, this finding would suggest repression.
If one adhered to all aspects of this technique, a study with even a modest number of subjects might provide a useful test of the hypothesis that repression can occur. Indeed, even a series of several case reports, provided that they strictly met the criteria outlined above, could represent persuasive preliminary evidence for the existence, though not for the frequency, of repression.
Laboratory studies over the past 60 years have failed to demonstrate that individuals can ‘repress’ memories. Clinical studies, which extrapolate from the laboratory to the study of real-life traumas, must consequently start with the null hypothesis: that repression does not occur.
To reject the null hypothesis, and show that repression of childhood abuse memories can occur clinically, one must meet two requirements. First, one must confirm that traumatic abuse actually occurred. Secondly, one must demonstrate that individuals actually developed amnesia, of non-biological origin (and after the age of five), for this abuse. We performed a literature search for studies that have attempted to document repression of memories of childhood sexual abuse. Despite our broad search criteria, which excluded only unsystematic case reports, we located only four such studies, which we then examined on the basis of the above two criteria. None of the four studies presents data that satisfy both of the two requirements.
It must be emphasized that these four studies are the only applicable studies that we were able to locate. In other words, this brief review does not present merely a selection of the most important studies, but the entirety of all published studies, which to our knowledge have systematically tested whether repression of memories of childhood sexual abuse can occur.
It might be argued that this dearth of studies is due to the difficulty of documenting trauma and demonstrating amnesia. But if repression affects even a small fraction of abused individuals, one would expect hundreds of thousands, if not millions, of current cases in the United States alone, and even larger numbers worldwide. Thus, the difficulties of documenting repression should be more than counterbalanced by the large pool of cases.
In summary, present evidence is insufficient to permit the conclusion that individuals can ‘repress’ memories of childhood sexual abuse. This finding is surprising, since many writers have suggested that there is a high prevalence of repression in the population. Thus, this area of psychiatry begs further carefully designed studies to resolve one of its most critical questions.
Anderson, J., Martin, J., Mullen, P., Romans, S. & Herbison, P. (1993). Prevalence of childhood sexual experiences in a community sample of women. Journal of the American Academy of Child and Adolescent Psychiatry 32, 911-919.
Bagley, C. & Ramsay, R. (1986). Sexual abuse in childhood: psychosocial outcomes and implications for social work practice. Journal of Social Work and Human Sexuality 4, 33-47.
Bass, E. & Davis, L. (1988). The Courage to Heal, Harper & Row: New York.
Blume, E.S. (1990). Secret Survivors, John Wiley & Sons: New York.
Briere, J. & Conte, J. (1993). Self-reported amnesia for abuse in adults molested as children. Journal of Traumatic Stress 6, 21-31.
Chodoff, P. (1963). Late effects of the concentration camp syndrome. Archives of General Psychiatry 8, 323-333.
Finkelhor, D., Hotaling, G., Lewis, I.A. & Smith, C. (1990). Sexual abuse in a national survey of adult men and women: prevalence, characteristics, and risk factors. Child Abuse and Neglect 14, 19-28.
Femina, D.D., Yeager, C.A. & Lewis, D.O. (1990). Child abuse: adolescent records vs. adult recall. Child Abuse and Neglect 14, 227-231.
Fivush, R. & Hudson, J.A. (eds.) (1990). Knowing and Remembering in Young Children. Cambridge University Press: New York.
Fredrickson, R. (1992). Repressed Memories: A Journey to Recovery from Sexual Abuse. Simon & Shuster: New York.
Henderson, J.L. & Moore, M. M. (1944). The psychoneuroses of war. New England Journal of Medicine 230, 272-278.
Herman, J.L. (1981). Father-Daughter Incest. Harvard University Press: Cambridge, MA.
Herman, J.L. (1992). Trauma and Recovery. Basic Books: New York.
Herman, J.L. & Schatzow, E. (1987). Recovery and verification of memories of childhood sexual trauma. Psychoanalytic Psychology 4, 1-14.
Holmes, D. (1990). The evidence for repression: an examination of sixty years of research. In Repression and Dissociation: Implications for Personality, Theory, Psychopathology and Health (ed. J. Singer), pp. 85-102. University of Chicago Press: Chicago.
Leopold, R. L. & Dillon, H. (1963). Psycho-anatomy of a disaster: a long-term study of post-traumatic neuroses in survivors of a marine explosion. American Journal of Psychiatry 119, 913-921.
Loftus, E.F. (1993). The reality of repressed memory. American Psychologist 48, 518-537.
Loftus, E.F., Polonsky, S. & Fullilove, M.T. (1994). Memories of childhood sexual abuse; remembering and repressing. Psychology of Women Quarterly 18, 67-84.
Malmquist, C.P. (1986). Children who witness parental murder: post-traumatic aspects. Journal of the American Academy of Child Psychiatry 25, 320-325.
Nash, C.L. & West, D.J. (1985). Sexual molestation of young girls: a retrospective survey. In Sexual Victimization (ed. D.J. West), Part 1 pp. 1-92. Gower: Aldershot.
National Center for Health Statistics (May, 1961): Reporting of hospitalization in the Health Interview Survey: a methodological study of several factors affecting the reporting of hospital episodes. US Dept. of Public Health, Education, and Welfare. Publication No. 584-D4: Washington.
National Center for Health Statistics (July, 1965): Health interview responses compared with medical records. Vital and Health Statistics, PHS Pub. No. 1000-Series 2-No. 7. Public Health Service. Washington: US Government Printing Office: Washington.
National Center for Health Statistics (April, 1972): Optimum recall period for reporting persons injured in motor vehicle accidents. Vital and Health Statistics. Series 2-No. 50. DHEW Publ. No (HSM) 72-1050. Health Services and Mental Health Administration. US Government Printing Office: Washington.
Ofshe, R.J. & Singer, M.T. (1994) Recovered-memory therapy and robust repression: influence and pseudomemories. International Journal of Clinical and Experimental Hypnosis 42, 391-410.
Pope, H.G. Jr., Mangweth, B., Negrao, A.B., Hudson, J.I. & Cordas, T.A. (1994). Childhood sexual abuse and bulimia nervosa: a comparison of American, Austrian, and Brazilian women. American Journal of Psychiatry 151, 732-737.
Pynoos, R.S. & Nader, K. (1989). Children’s memory and proximity to violence. Journal of the American Academy of Child Psychiatry 28, 236-241.
Russell, D.E.H. (1986). The Secret Trauma: Incest in the Lives of Girls and Women. Basic Books: New York.
Strom, A., Rufsum, S.B., Eitinger, L., Gronvik, O., Lonnum, A., Engeset, A., Osvik, K. & Rogan, B. (1962). Examination of Norwegian ex-concentration camp prisoners. Journal of Neuropsychiatry 4, 43-62.
Terr, L.C. (1979). Children of Chowchilla. Psychoanalytic Study of the Child 34, 547-623.
Terr, L.C. (1983). Chowchilla revisited: the effects of psychic trauma four years after a school-bus kidnapping. American Journal of Psychiatry 140, 1543-1550.
Usher, J.A. & Neisser, V. (1993). Childhood amnesia and the beginnings of memory for four early life events. Journal of Experimental Psychology 122, 155-165.
Williams, L.M. (1992). Adult memories of child sexual abuse: preliminary findings from a longitudinal study. American Society for Prevention of Child Abuse Advisor 5, 19-20.
Williams, L.M. (1994). Recall of childhood trauma. A prospective study of women’s memories of child sexual abuse. Journal of Consulting and Clinical Psychology 62, 1167-1176.
Wyatt, G. (1985). The sexual abuse of Afro-American and white American women in childhood. Child Abuse and Neglect 9, 507-519.
Copyright 1995 Cambridge University Press.
Posted on the FMSF website
with permission of the author and publisher.
Last Updated: February 25, 2014
Except where noted, all material on this site is copyrighted © 2006-17 False Memory Syndrome Foundation.
PO Box 30044 • Philadelphia, PA 19103 • Telephone (215) 940-1040
Email: Send an e-mail to FMSF