Journal of Consulting and Clinical Psychology (Dec, l994)
Elizabeth F.Loftus, Maryanne Garry, Julie Feldman University of Washington
Science and therapy
Dale Akiki, a former Sunday school teacher, was accused in 1991 of sexually abusing, terrorizing, and kidnapping children in his care at Faith Chapel, a church near San Diego. Although the case lacked corroboration, the prosecution claimed that many victims had repressed their memories of these horrible incidents and recovered them in therapy. Akiki was acquitted of all charges in November, 1993, after spending several years in jail (Sauer, 1993). While the trial lasted more than a half year, there was one short minute of testimony that captures the essence of a fundamental conflict between science and at least one testifying therapist in the case: Is therapy grounded in science? One of Akiki's defense attorneys asked the therapist if a diagnosis in therapy needed to be based on research-verified science: A: "That's a different type of science...I think there's....there are different types of science. There's research science and ...." Q: What type of science do you practice? A: Artful science Q: O.k. So you're saying it's an art, it's not a science? A: A little of both...I think there's a lot of counseling theory that has not necessarily been proven in scientific research.
The theory behind counseling, or all of psychotherapy for that matter, is best served if grounded in science. Williams' prospective study is an important empirical contribution to the much needed area of memory for childhood trauma. Ultimately it is studies like this one that will assist mental health professionals in doing as good a job with clients as they possibly can. It is important not only that the study be well done -- which it is, but also that it not be misinterpreted -- which, unfortunately, it already has been. This commentary focuses primarily on proper and improper ways to think about these novel findings.
Williams' Preliminary Report
Several years ago, Williams (1992) published a preliminary report of the study that is now reported more fully here. In the preliminary report, Williams told readers that she followed up 200 females who reported sexual abuse in the early 1970's. Her data were based on interviews with 100 of those women, primarily African-Americans. All had, some 17 years earlier, been girls aged infant to 12 years old when they were brought to a city hospital emergency department for treatment and collection of forensic evidence after a sexual assault (even when there was no physical trauma present). The sexual abuse ranged from sexual intercourse in about a third of the cases, to fondling (also about a third of the cases). The results showed that 38% did not recall the abuse or chose not to report it. For several years before formal publication, the preliminary report found its way into the hands and minds of mental health professionals and journalists. Thus we have a unique opportunity to observe how the study was being cited, and miscited, before it even formally appeared.
Citings of the Williams research.
Williams (1992) herself was fairly careful in her discussion of her own research. She summed it up this way: "...a large proportion of women do not recall childhood sexual victimization experiences." (p. 21). Some characterizations of the findings have been reasonably careful as well. For example, Henderson (1993, p. 43) put it simply: ".... among childhood sexual abuse survivors whose abuse had been reported when it happened, 38% claimed no memory of it when questioned l7 years later." Other writers have misstated the findings in small or even large ways, or they have drawn speculative or even unwarranted inferences. For example, one writer claimed that Williams found that "Thirty-eight percent of this sample reported that they had not been abused, in spite of hospital records to the contrary" (Wylie, 1993, p 42-43). In fact 38% did not say that had not been abused at all; rather, the majority of these women told about other abuse that they had experienced. Another writer claimed that Williams' study "supports Putnam's theory that traumatic events can be stored in the brain in a way that doesn't conform to the standard model of memory" (Wartik, 1993, p 65). In fact, the study says nothing about how traumatic events are stored in the brain differently from standard models of memory. Another scholar claimed that Williams, in her interviews 17 years after the abuse event, had found that "36% had no memory for that event including some of the people who were...15, 16, 17 years of age ." (Pezdek, 1993). In fact, Williams' followup study only involved women who were ages 10 months to 12 years at the time of their abuse. Other writers have claimed that the study supports "repression". One article called the study "one of the most systematic efforts to track memory repression" (Kandel & Kandel, 1994, p 34). Another used it as support for the claim that many "sexual-abuse victims repress memories for a period of time" (Horn, 1993, p 56). A third went even further: "This was repression or some other extreme "forgetting" defense in action" (Terr, 1994, p 53). In fact, while the study is often used to support the idea that repression of sex abuse memories is common, we argue later that in reality the findings do not constitute cogent support for anything like a repression mechanism of the sort originally speculated about by Freud.
The Real Williams Study
The analyses are based upon interviews with 129 women who, as girls, reported sexual assault and were examined at the emergency room at (a large urban) General Hospital in the early l970's (McCahill, Meyer & Fischman, 1979). Some l7 years later, when these women were interviewed, 38% failed to report the "index incident" that took them to the emergency room. What The Data Mean Williams claims that it is "quite dramatic" that 38% of the women did not tell the interviewer about the child sexual abuse. But to fully appreciate the drama of forgetting, we need to keep in mind that people can forget all kinds of things that might, at first thought, seem surprising. For example, people (over a quarter of those interviewed) have failed to recall automobile accidents 9-12 months after their occurrence, even though someone else in the car had been injured (Cash & Moss, 1972; described in Loftus, l982). People (over 20%) who had a family member die when they were four years old have failed to recall a single detail about the death (Usher & Neisser, 1993). People (over 15%) have failed to recall a hospitalization approximately nine months after discharge (National Center, 1965; described in Loftus, 1982). Patients have failed to recall visits to a health maintenance organization (HMO) that they made within the previous year for something that was "serious" or even "very serious" (Means & Loftus, 1991). In light of figures such as these, it does not seem quite so dramatic that people might also fail to recall a sexual trauma that occurred to them 17 years earlier.
Do the data mean that the women have completely repressed their
memories of abuse?
When women do not recall an abusive incident from nearly two decades earlier, does this mean that they repressed their memories? Not necessarily. In order to establish repression, it is necessary to first show that a memory existed in the first place . Perhaps some of the younger children who were examined, interviewed, or treated for sexual assault did not understand the meaning of their experience. It would be reasonable to assume that a parent might try to keep this knowledge from a young child in order to facilitate treatment or minimize damage to the child. If so, then some of these children never "knew" they were traumatized and thus had nothing to repress. This is not simply speculation, as there exist known cases of abused children who are interviewed by a psychotherapist but have no idea why they are being interviewed or are otherwise completely in the dark about the nature of their problem (Wagenaar, 1994). We should keep in mind that Williams herself reports that many of the girls did not allege assault themselves but were reported by others to have been assaulted. Even if the event were understood and stored in memory, there are many reasons why it might not be recalled later on. Normal forgetting of all sorts of events is a fact of life, but is not thought to involve some special repression mechanism. Forgetting of all sorts of events happens, even significant events, even traumatic ones like hospitalizations, accidents, and deaths, but is not thought to involve a repression mechanism. Others have opted to describe these data without using the term "repression." Perhaps they do so because the meaning of repression depends on who is using it. Freud himself evolved through a number of meanings, none of which were very well defined. What are some alternatives to the term "repression"? Herman and Harvey (1993) use the term "amnesia." They describe the Williams' (l992) study in the context of amnesia, referring to it immediately after this sentence: "Partial or even complete amnesia for childhood trauma is well documented." (p. 5) One could, of course, say that the women whom Williams studied constitute cases of complete amnesia. Williams herself told a Science News reporter (Bower, l993): "I don't know exactly what psychological process causes the amnesia." (p. l85). But this would mean that when we forget anything, it is an example of complete amnesia for that thing. It dilutes the meaning of the term amnesia, which has often been reserved for discussing a pathological sort of forgetting. It is rather like using the word "assassination" to describe the squashing of a bug. Given the new broad-ranging definition of amnesia, what would we say happens to the person who goes into the supermarket specifically to get aspirin, and leaves ten minutes later with a Snicker's bar, a magazine, a box of Pop-Tarts, and no aspirin? Is this amnesia? Or is it simply a case of forgetting? Amid all the arguing and rhetoric about real and suggested memories of abuse, what sometimes gets lost is this: forgetting is an ordinary phenomenon. Remembering the past in detail can be considered the exception. Forgetting something is so utterly common that trying to puff it up with a scientific name to make it appear exotic is unnecessary, and may in fact be an example of psychological spin-doctoring, the merging of science and politics. Perhaps the best way to think about the 38% in Williams' sample who don't remember is to say it simply: they don't remember.
What do the data say about remembering and repressing secretive incestuous abuse?
It is problematic to apply the current data to a prototypic case of incestuous abuse in which the child is threatened to keep to herself the horrible secret. It has been argued that some children repress their abuse memories because they cannot talk about their abuse. Their whole world is telling them it didn't happen. The abuse experienced by the children in Williams' sample was generally not of this type. To enter the study in the first place, a female rape victim had to come to the (large urban) General Hospital Emergency Room alleging rape (McCahill et al, 1979, p. 10). If they first went to the police, the police were instructed to take them to the hospital. If they first went to the hospital, in virtually all cases, the police were notified ( p. 8l). The pre-treatment interviews were "extremely long" p. 87) -- sometimes more than three hours. The police directive in effect at the time required that all victims under l8, even though examined at (the urban) General Hospital, also be referred to the Youth Study Center for a medical exam (p. 93). Then the victims were paid at least one visit and sometimes four visits by a social worker. In many cases the victims were involved in preliminary hearings and in trials. Thus the typical victim talked with many individuals -- hospital employees, law enforcement, social workers, district attorneys and others. In some ways, this makes it even more surprising that the percentage who did not report the index abuse was as large as it was. Be that as it may, this study tells us more about traumatic events that were out of the closet than it does about ones that were hidden away.
What do the data say about remembering and repressing repeated episodes of abuse?
The abuse experienced by these children was generally not of this type. The index abuse, when it happened, was typically a one-time event. At the time the initial data were collected, the victimized children were classified according to whether they had been sexually assaulted before by the same or a different offender (McCahill et al, 1979, p. 8); for the overall sample of 1400, 80% were brought in because of a one-time assault. From the current analysis, it appears that 70% of the children in the sample were similarly brought in for a one time assault. The only example of a child case described in McCahill et al. is clearly of this type: a 7 year old girl who was raped by her babysitter's boyfriend a fortnight earlier -- the "only such attack " (p. l48). The remaining 30% of the sample, at the time of their appearance at the hospital in the 70's, had had a prior experience by the same perpetrator or by someone else. We are not told how many of the 15 who did not remember had had a prior assault by a different person. We are not told, since the data were not collected, how many of the 15, if any, had been assaulted by the same offender on numerous prior occasions.
The number of critical events is important for a complete understanding of how these events might be remembered. Research on memory, with children and adults, suggests that people are more likely to forget an isolated incidence of abuse than a series o f repeated events -- even though the repeated events may become blended into a typical script (Lindsay & Read, l994; Means & Loftus, 1991). Williams reports that 38% of women who had no prior experience of sexual abuse before the index event forgot the abuse, whereas 33% of women who had had a prior (possibly repeated) experience forgot. While this difference was not significant, it is in the direction opposite to that posited by Terr's (1991) Type I/Type II trauma hypothesis.
The Williams data are clear in showing that women often forget a single incident of sexual abuse. As Williams' recognizes, they do not show that women who deny any experience of sexual abuse whatsoever were actually molested numerous times in childhood.
Do they data tell us that people repress abuse and then reliably recover it later?
The women in Williams' sample forgot their abuse. Even if one takes the position that it has been repressed, it is still the case that it has not been de-repressed. Perhaps that could be done. For example, it would be interesting to take her sample (or one similar to it) and randomly divide the sample into two treatments -- one that gets "memory recovery" treatment and one that does not. Would the women in the memory recovery group later recover these memories? Before this type of study is undertaken, one would need to fully explore its ethics. Adequate research has not yet been done to show that digging out deeply repressed memories of molestation actually helps people get better. Potential investigators proposing such research would need to work through the consideration that such digging might make some of the women worse. Have the women forgotten their abuse or have they chosen not to report it? Because many of the women were willing to report other sexual victimizations although not the one in the hospital record, Williams believes that the bulk of the nonreports were due to women who actually did not remember. Of course, another possibility is that some women may have been trying to report the index event but their memory was sufficiently distorted that the researchers misclassified the woman as failing to remember the index event but remembering some other sexual abuse event. Misclassifications can arise because of the known difficulty of matching information from medical records to information obtained from people's memories (Loftus, Smith, Klinger, & Fiedler, 1992). To understand why women would report a different incident of abuse but not the index event, one should also examine when the various events happened relative to one another. If at the time of their appearance at the hospital in the l970's, 70% of women h ad never before been sexually assaulted, then the "different" incident they were reporting to the researcher who interviewed them in the l990's must have necessarily occurred more recently. Perhaps the more recent experiences were relatively salient in their minds, making the index events appear relatively less salient. A related issue concerns how often the women have visited hospital emergency rooms in general and the (large urban) hospital in particular. If some of them had been to the hospital on numerous occasions over their lifetime, one would expect that they may have some difficulty remembering the details of any one specific experience at the hospital.
Do the data mean that true prevalence of abuse has been underestimated? Based upon her findings, Williams concluded that self-reports of childhood experiences of sexual victimization are likely to result in an underestimation of the true prevalence of such abuse. As Dawes (1994) has aptly noted, we cannot support this claim by looking only at the genuine memory errors. Dawes' claim can best be appreciated by first realizing that when people are asked about any event in the past there are two types of errors that are made: some people experienced the event but don't report it (false negatives) and some people did not experience the event but do report it (false positives). The balance between these two errors determines whether the reporting rate is an underestimation or an overestimation of the true rate. If the event in question is one that occurs in less than half of the population, the balancing test is particularly crucial, for false positives can easily equal or even outnumber false negatives. Without an estimate for the proportion of non-abused people who erroneously remember abuse this issue cannot be resolved. Do these data justify therapeutic digging for lost memories of childhood sexual abuse? Although the data provide evidence of genuine memory failure, they do not eliminate concerns about the risks of using certain memory recovery techniques (Ceci, Loftus, Leichtman & Bruck, 1994; Lindsay & Read, 1994; Loftus, 1993). While it might be tempting to argue that since the rate of forgetting is so high (38%), it justifies the use of aggressive memory recovery techniques like hypnotic age regression, sodium amytal interviewing, sexualized dream interpretation, creative imagination, and a host of other questionable procedures. However, first it must be shown that recovering such memories is therapeutically beneficial to clients. Even if the benefits for actual victims can be shown, therapists still face the same dilemma experienced by the doctor who has invented a new drug for diabetes. Assume that her drug is one that will help the patient who already has diabetes, but will give the disease to the patient who did not initially have it. If the doctor does not know in advance the health status of her patient, should she be prescribing the new drug?
Conclusion
Back in the l970's, McCahill et al (1979) said something that is as true today as it was then: "Rape hurts everyone." (p. 244). This study makes an important contribution to our understanding of how children might later remember or forget a significant childhood trauma. But the study does not provide support for the notion that a long stream of childhood traumas is routinely banished from conscious awareness and then can later be reliably recovered. We must keep in mind that the 49 women (38%) have forgotten, but they have not later reliably remembered. Some of them were so young when the abuse happened, it is exceedingly unlikely that they will ever be able to later genuinely remember, although conceivably they could be told about their abuse and construct a "true" memory of the experience. Others were abused when they were older, but their failure to remember the abuse can be understood as normal forgetting that follows the same laws as forgetting of all sorts of other life events. Adapting an argument made by Pope and Hudson (1994), it is surprising we do not have more documented cases of total repression and reliable recovery later. If one conservatively estimates the number of Americans who have endured traumatic sexual abuse in the past, the figure could exceed 10 million. If repression occurred in only 10% of these cases, then at least one million adult Americans would presently harbor repressed memories of childhood sexual abuse. With these kind of numbers, Pope and Hudson wonder why there are no published studies of groups of patients exhibiting well-documented cases of total repression and reliable recovery later. Williams' groundbreaking study gives us an idea of the kind of study that might be done to provide evidence for this type of process. She has shown us how it is possible to go beyond the disputed cases that are based only on an alleged victim's de-repressed memory and an alleged perpetrator's denial -- tragic as these cases may be. She has shown us that we need not relax the requirement that there be proof that the trauma occurred before we can study the forgetting and remembering of that trauma. Not only does rape hurt everyone, but false memories of rape hurt everyone too. Hopefully, the important work of Williams will be used to help genuine victims heal from their past childhood traumas, and not be used to justify and rationalize the creation of new victims.
References Bower, B. (1993) Sudden recall. Science News, l44, l84-186. Cash, W. S. Moss, A. J. (1972, April) Optimum recall period for reporting persons injured in motor vehicle accidents. (National Center for Health Statistics. Vital and Health Statistics. Series 2-No 50. DHEW Publ. No (HSM) 72-l050. Health Service s and Mental Health Administration.) Washington: U.S. Govt Printing Office.
Ceci, S. J., Loftus, E. F., Leichtman, M.D., & Bruck, M. (1994) The role of source misattributions in the creation of false beliefs among preschoolers. International Journal of clinical and experimental hypnosis, in press.
Dawes, R. M. (1994) House of Cards: Psychology and psychotherapy built on myth. NY: Free Press.
Henderson, R. (1993) The tangled net of memory. Common Boundary, 11, 38-45.
Herman, J. L. Harvey, M. R. (April, 1993) The false memory debate: social science or social backlash? Harvard Mental Health Letter, 4-6. Horn, M. (November 29, 1993) Memories lost and found. U.S. News and World Report, p 52-56, 60-63.
Kandel, M. Kandel, E. (1994) Flights of memory. Discover, 15, 32-38.
Lindsay, S. Read, D. (1994) Applied Cognitive Psychology, in press.
Loftus, E.F. (1982) Memory and its distortions. In Kraut, A.G. (Ed) The G. Stanley Hall Lecture Series. Washington DC: American Psychological Assn., p. 123-154.
Loftus, E. F. (1993) The reality of repressed memories. American Psychologist, 48, 5l8-537.
Loftus, E. F., Smith, K. D., Klinger, M. R. & Fiedler, J. (1992) Memory and mismemory for health events. In J. M. Tanur (Ed.) Questions about questions: Inquiries into the cognitive psychology of surveys. NY: Russell Sage. p. 102-137.
McCahill, T., Meyer, L. C. & Fischman, A. (1979) The aftermath of rape. Lexington, Ma: Lexington Books. Means, B. Loftus, E.F. (1991) When personal history repeats itself: decomposing memories for recurring events. Applied Cognitive Psychology, 5, 297-3l8.
National Center for Health Statistics (1965, July). Reporting of hospitalization in the Health Interview Survey. (Public Health Service, DHEW-HSM Publication No 73-1261). Washington, D.C.: U.S. Government Printing Office.
Pezdek, K. (1993, December 6) Transcript of radio interview with Michael McConnell of WLW Cincinnati.
Pope, H.G. Hudson, J.I. (1994) Can memories of childhood sexual abuse be repressed? Unpublished manuscript, McLean Hospital, Harvard Medical School.
Sauer, M. (November 21, 1993) Glad you're finally free. The San Diego Union-Tribune, p A-l, A-21.
Terr, L. C. (1991) Child traumas: An outline and overview. American Journal of Psychiatry, 148, 10-20.
Terr, L. (1994) Unchained memories. NY: Basic Books.
Usher, J. A. Neisser, U. (1993) Childhood amnesia and the beginnings of memory for four early life events. Journal of Experimental Psychology: General, 122, 155-165.
Wartik, N. (May, 1993) A question of abuse. American Health. p 62-67.
Wagenaar, W. A. (1994, March 28). Personal communication (via email from Leiden University.
Williams, L. M. (1992) Adult memories of childhood abuse: Preliminary findings from a longitudinal study. The Advisor, 5, 19-20.
Wylie, M. S. (September/October, 1993) Trauma and memory. Family Therapy Networker. Vol l7, No 5, p 42-43.
Please address all correspondence to Elizabeth Loftus, Dept. of Psychology, University of Washington, Seattle, Washington, 98195. The basic memory research conducted by EL has been supported by grants from NSF, NIMH, and NIH.
The fact that the preliminary report claimed 38% of l00 women could not remember and the Final report claimed that 38% of 129 women could not remember could be explained in several ways: l) The preliminary report on l00 women was actually a report on ro ughly l00 women. 2) An additional 29 women's data were included for the Final report and exactly 11 (38% of these) could not remember the abusive incident.
The fact that the preliminary report claimed 36% of cases involved sexual intercourse and the final report claims that 60% involved sexual penetration could be due to different classification schemes. In the preliminary report, Williams reported that 53% of women who failed to report the index event did report a different sexual abuse experience which occurred at some other time during their childhood.
In the final report, Williams claims that 68% of those who did not recall the index event told about some other sexual abuse. The reason for this discrepancy is not immediately apparent, but could be due to change in the way responses were classified.
To appreciate Dawes' logic, assume 24% of women are abused and a quarter of these forget. This means 6% of the total sample of women would be contributing to an underestimate. Now assume that 8% of the nonabused group erroneously reports abuse. This means that 6% of the total sample of women (8% x 76% = 6%) would be contributing to an overestimate. The two errors would balance one another and the prevalence estimates would then be accurate.
Elizabeth Loftus Psychology Department University of Washington email: eloftus@u.washington.eduSeattle, Washington 98195