injusticebusters logo

Richard McNally

Richard McNally

Harvard researcher looks at memory
The sewer gas theory of remembering trauma

From Adriaan Mak

In the world of science theories (hypotheses) abound, but are accepted, and even then only tentatively so, if solid research supports the former. Only when subsequent studies are done and have the same outcomes does the theory gain acceptance. Yet even then, it is always possible that more refined research may alter our perception of what may have been accepted for a while as truth.

Jennifer Freyd, the estranged research psychologist daughter of Pamela and Peter Freyd, about a decade ago developed a theory called "betrayal trauma theory".

This theory in essence claims that, when children or adolescents are sexually abused by a primary caregiver, the memory of the abuse, each time after it happens, is possibly stored in some special compartment of the brain. This holding pen does not allow access to bring the abuse into conscious memory, while the child is growing up, it is argued. Betrayal trauma theory, Jennifer believes, allows the abused child or adolescent, which is still dependent on the abuser for food, clothing and shelter, to continue functioning in spite of the abuse. The teenager may have been raped by Daddy or Mommy the night before, but next morning has absolutely no consiousness of that evil, instead kisses Daddy and Mommy goodbye before happily going off to school.

Since then, to support her theory, Jennifer has published research, using people assumed to have recovered hidden memories of abuse. She and a collaborator did so in a 2004 study, where people, such as our accusing adult daughters or sons, were deficient in recalling words related to abuse. In science, such a study needs to be replicated by other scientists and yield similar results for the outcome of the study to be accepted as valid.

Richard McNally, whom some of you met at our last Toronto meeting, did just that and here is what he found: "Contrary to prediction, individuals reporting recovered memories of CSA did not exhibit memory deficits for trauma words they were told to remember when their attention was divided."

N.B. The introduction of "divided attention" in both the original study and its replication was needed to simulate the theoretical (hypothethetical) mechanism of "dissociation". This allegedly causes the abuse and its trauma to be diverted from being consciously remembered, and put into this again theoretical (hypothetical) cesspool, where traumatic memories are fermenting only to come bubbling to the surface in the form of sewer gas when the cesspool is stirred. The agent doing the stirring may be your friendly neighbourhood therapist with an agenda, or a lovely textbook such as "The Courage to Heal", although I would argue that the stink really was produced by the therapis or the book's authors.

My friend Claudette Grieb puts it: "Bullshit baffles brains". Memory research offers the simple solution that traumatic events are stored no differently than any other event. Piling hypothesis upon hypothesis upon hypothesis is the mark of poor science.

Adriaan Mak

PSYCHOLOGICAL SCIENCE Volume 16 Issue 4 Page 336 - April 2005



Forgetting of Trauma Cues in Adults Reporting Continuous or Recovered Memories of Childhood Sexual Abuse

Richard J. McNally [1], Carel S. Ristuccia [1], and Carol A. Perlman [1]

AFFILIATIONS: [1] Harvard University

CORRESPONDENCE: Address correspondence to Richard J. McNally, Department of Psychology, Harvard University, 33 Kirkland St., Cambridge, MA 02138; e-mail:


According to betrayal trauma theory, adult survivors of childhood sexual abuse (CSA) who were molested by their caretakers (e.g., a father) are especially likely to dissociate ("repress") their memories of abuse. Testing college students, some reporting CSA, DePrince and Freyd (2004) found that those scoring high on a dissociation questionnaire exhibited memory deficits for trauma words when they viewed these words under divided-attention conditions. Replicating DePrince and Freyd's procedure, we tested for memory deficits for trauma words relative to neutral words in adults reporting either continuous or recovered memories of CSA versus adults denying a history of CSA. A memory deficit for trauma words under divided attention was expected in the recovered-memory group. Results were inconsistent with this prediction, as all three groups exhibited better recall of trauma words than neutral words, irrespective of encoding conditions.

Some clinicians believe that children can blunt the emotional impact of sexual abuse by dissociating their attention during episodes of molestation (e.g., Terr, 1991). They also believe that such an avoidant, dissociative coping style may render it difficult for survivors to recall their abuse years later. Indeed, some clinicians believe that the more frequently a child has been abused, the more likely he or she will have amnesia for the abuse. As one distinguished psychiatrist claimed, "the nature of traumatic dissociative amnesia is such that it is not subject to the same rules of ordinary forgetting: it is more, rather than less, common after repeated episodes; involves strong affect; and is resistant to retrieval through salient cues" (Spiegel, 1997, p. 6).

The aforementioned theories imply that survivors of childhood sexual abuse (CSA) develop superior ability to disengage attention from trauma-related cues, so that their later recall for such cues is impaired. Researchers have used two directed-forgetting methods to test this hypothesis in the laboratory (Johnson, 1994). In the item method, the participant views a series of words on a computer screen. Shortly after the appearance of each word, instructions appear, telling the participant either to remember or to forget the previous word. The participant is later asked to recall all words, regardless of initial instruction. Typically, participants recall more words they were told to remember than words they were told to forget-an effect usually attributed to attentional disengagement from words followed by forget instructions.

In the list method, participants view a series of words and halfway through the list are told to forget all words they have seen so far. After viewing the remainder of the list, participants are asked to recall all the words, regardless of whether they were from the first or second half of the list. Typically, participants exhibit a recall deficit for words from the block they were told to forget-an effect that often disappears on a recognition test. This initial memory impairment is often attributed to retrieval inhibition of encoded words that have been rendered temporarily inaccessible.

These cognitive methods have been used to study memory in people reporting histories of CSA (for a review, see McNally, 2003, pp. 260-274). Item-method directed-forgetting studies have failed to support the hypothesis that adult survivors of CSA possess superior ability to disengage attention from threat words (e.g., incest), and consequently have impaired subsequent recollection of those words (McNally, Clancy, & Schacter, 2001; McNally, Metzger, Lasko, Clancy, & Pitman, 1998). Likewise, a list-method directed-forgetting experiment failed to support the hypothesis that adult survivors of abuse possess superior ability to inhibit retrieval of trauma words they have encoded (McNally, Clancy, Barrett, & Parker, 2004). Moreover, two of these studies involved participants who either said they believed they harbored repressed (dissociated) memories of CSA or had reported recovering long-forgotten memories of CSA (McNally et al., 2001, 2004). That is, heightened ability to forget trauma cues has not been detected even among those persons who would presumably be most likely to possess this skill.

DePrince and Freyd (2004), however, claimed that such experiments provide inadequate tests of the hypothesis of superior forgetting in dissociative CSA survivors. According to these authors, CSA survivors coped with abuse episodes by dissociating or disengaging their attention. Therefore, an adequate test of the hypothesis would require participants to perform the directed-forgetting task under conditions of divided (rather than selective) attention. Testing college students (some with reported trauma histories), DePrince and Freyd found that those scoring high on a questionnaire of dissociation proneness (the Dissociative Experiences Scale, DES; Bernstein & Putnam, 1986) exhibited impaired recall of trauma words under divided- but not selective-attention conditions, relative to participants scoring low on the questionnaire. DePrince and Freyd concluded that memory impairment for trauma material should be evident under divided-attention conditions because these conditions mimic those prevailing during CSA episodes.

DePrince and Freyd's research was inspired by Freyd's (1996) betrayal trauma theory, which asserts that individuals who were sexually abused during childhood by a caretaker are especially likely to experience amnesia for their abuse. According to the theory, these children are motivated to make sense out of a senseless situation: Their assailants are the very same people on whom they must rely for food, shelter, and clothing. To resolve this conflict, the children block out their memories of abuse, and thereby can maintain their bond to the (abusive) caretaker. Freyd's theory is designed to explain, for example, why a woman who had been sexually assaulted by her father throughout her adolescence might be entirely unaware of ever having been abused until encountering salient reminders of the attacks many years later.

In the experiment reported here, we replicated DePrince and Freyd's (2004) procedure with three groups of adults recruited from the community. Participants in the continuous-memory group reported always having remembered their CSA, those in the recovered-memory group reported recalling their CSA after many years of not having thought about it, and those in the control group reported never having experienced CSA. Although DePrince and Freyd's procedure was inspired by betrayal trauma theory, we could not optimally test Freyd's theory because hardly any of our participants reported sexual abuse by a caretaker. Contrary to the claim that victims are more likely to forget having been sexually assaulted by their fathers than having been sexually abused by uncles, cousins, teachers, and other individuals, we have found that most people who report remembering episodes of CSA that they had not thought about in many years report having been abused by noncaretakers. For example, in a recent study of 27 individuals with recovered memories of CSA, only 8 victims mentioned having been sexually abused by either a father, a stepfather, or a mother (Clancy & McNally, 2004).

In fact, DePrince and Freyd's (2004) study did not provide an optimal test of betrayal trauma theory either. That is, they provided no evidence that college students who reported betrayal trauma had ever forgotten their abuse, and having forgotten one's abuse is precisely what betrayal trauma theory is designed to explain.

In any event, if DePrince and Freyd's (2004) paradigm taps mechanisms relevant to forgetting one's sexual abuse, then participants reporting having forgotten and then recovered their memories of CSA should exhibit memory impairment for trauma words relative to neutral words under divided-attention conditions. This effect should be less evident (or not evident) in participants with continuous memory of CSA and in control groups.



The participants in this study had previously responded to newspaper advertisements soliciting adults who either had or had not experienced CSA for a research project on trauma and memory. During prior visits to our laboratory over the past several years, these volunteers had received a memory interview that enabled their assignment to the recovered-memory, continuous-memory, or control group; had completed a battery of questionnaires; had undergone structured psychiatric diagnostic interviews; and had participated in other studies. Reported abuse ranged from fondling to rape.

Recruitment and testing were in accordance with the American Psychological Association's ethical guidelines regarding the use of human participants. The protocol and informed consent form were approved by the Harvard University Committee on the Use of Human Subjects.

The continuous-memory group consisted of 21 adults (16 women) who reported always having remembered their CSA, and whose mean age was 43.6 years (SD = 10.4). We were able to obtain external corroboration of the abuse for 2 of these participants. Only 4 participants, all women, reported abuse by a caretaker. The recovered-memory group consisted of 11 adults (5 women) who reported recalling memories of CSA after many years of not having thought about their abuse, and whose mean age was 46.3 years (SD = 13.9). None of these individuals was either able or willing to provide a corroborating informant. Only 2 participants, both women, reported abuse by a caretaker. The control group consisted of 16 adults (10 women) who reported never having been sexually abused as children, and whose mean age was 44.9 years (SD = 15.0).


Stimuli were drawn from previous research (DePrince & Freyd, 2004; McNally et al., 1998). The trauma category comprised 24 words related to sexual assault (e.g., rape, incest), whereas the neutral category comprised 24 words related to household objects (e.g., lamp, curtain). Another set of 12 trauma words and 11 neutral words appeared only as distractors on the recognition test.


Each stimulus word appeared in lowercase letters on a Macintosh G4 Powerbook laptop computer and remained at center screen for 6 s before being replaced by the next word. The experimental words appeared in four blocks of 12 words each. Each word was assigned to one of the blocks, with the proviso that each block had to include 6 trauma words and 6 neutral words. The words within a block were assigned to a single random sequence. Twenty-four different word sequences were then created by combining the blocks in all possible orders. Participants in the recovered-memory group saw Sequences 1 through 11, those in the continuous- memory group saw Sequences 1 through 21, and those in the control group saw Sequences 1 through 16.

For each participant, two blocks appeared under selective-attention conditions, and two blocks appeared under divided-attention conditions. Under selective- attention conditions, each word appeared in black letters against a white background. Under divided-attention conditions, words randomly changed colors from red to blue (and vice versa) and appeared against a white background. So, for example, the word molested might appear in red letters for 2 s, switch to blue letters for 1 s, and then switch back to red letters for the final 3 s of the 6-s presentation time. Participants were told to press the space bar on the computer whenever a word changed its color. Hence, under divided-attention conditions, participants had two tasks: encode the word and track color changes.

For each participant, one selective-attention block and one divided-attention block were followed by instructions to remember the words in the previous block, whereas the other selective-attention block and the other divided-attention block were followed by instructions to forget the words in the preceding block.

The four experimental blocks were preceded by a block of 12 country names (e.g., Mexico, Ireland) presented under divided-attention conditions and followed by forget instructions. Finally, the four experimental blocks were followed by a block of another 12 country names under selective-attention conditions and followed by forget instructions. The country names served as primacy and recency buffers, and these words, when recalled by participants, did not figure in any of the analyses.

Immediately after this encoding phase, participants were asked to recall and write down each of the words they had seen, regardless of whether the word had appeared in a remember block or a forget block. Participants received as much time as they needed, but most took only several minutes to recall as many words as possible. Participants were then handed a sheet of paper that listed all the trauma and neutral words that had been presented, plus the distractors. They were asked to circle any words they recognized having seen during the encoding phase.


Following DePrince and Freyd (2004), we counted the number of trauma and neutral words recalled under the two attention conditions (selective vs. divided) and under the two instructional conditions (remember vs. forget) for the continuous- memory, recovered-memory, and control groups (Table 1). Effects with p values greater than .10 are not reported.


Note. Standard deviations are in parentheses. Also following DePrince and Freyd (2004), we conducted a 3 (group) x 2 (word category) analysis of variance (ANOVA) for remember words under the divided- attention condition. In contrast to these investigators, we did not obtain a significant Group x Word Category interaction, F(2, 45) = 1.27, p = .29, eta = .28. Unlike DePrince and Freyd's high-dissociation group, our recovered- and continuous-memory groups did not exhibit recall deficits for trauma words relative to neutral words and relative to the control group. Indeed, these groups (and the control group) exhibited superior recall of trauma words relative to neutral words, as evinced by the significant effect of word category, F(1, 45) = 50.56, p = .0001, r = .73. [1]

Moreover, word-category effects indicated that all three groups recalled trauma words more often than neutral words under divided attention following forget instructions, F(1, 45) = 15.69, p = .0001, r = .51, and under selective attention following remember instructions, F(1, 45) = 3.02, p = .09, r = .25. However, all three groups recalled neutral words more often than trauma words under selective attention following forget instructions, F(1, 45) = 11.29, p = .002, r = .45.

Participants completed the DES (Bernstein & Putnam, 1986), and we followed DePrince and Freyd's (2004) guidelines for dividing participants into high- dissociative (n = 12) and low-dissociative (n = 22) groups. In contrast to what DePrince and Freyd found, the predicted Group (high vs. low dissociation) x Word Category interaction for remember words under divided attention was nonsignificant, F< 1, whereas both the high- and the low-dissociation groups recalled significantly more trauma words than neutral words under divided- attention conditions following remember instructions, F(1, 32) = 32.1, p = .0001, r = .71. The same pattern held for the recognition data, which yielded a nonsignificant interaction, F< 1, and showed that trauma words were recognized more often than neutral words, F(1, 32) = 23.4, p = .0001, r = .65.


Although DePrince and Freyd's (2004) experiment was inspired by betrayal trauma theory, neither their experiment nor ours provides an optimal test of this theory. For an optimal test, researchers would need to recruit adult survivors of CSA who had been abused by caretakers and who had forgotten their abuse for many years and then remembered it in adulthood. In DePrince and Freyd's study, high-DES college students did report more betrayal trauma than did low-DES students, but there was no evidence that those students who had experienced betrayal trauma had ever forgotten having been abused by their caretakers, and it is forgetting of abuse that Freyd's theory is designed to explain. In fact, previous experiments that inspired Freyd's theory (DePrince & Freyd, 1999, 2001) are even less relevant to betrayal trauma than their most recent one. In these previous studies on high- versus low-dissociative college students, DePrince and Freyd did not assess for CSA, let alone sexual abuse at the hands of caretakers that was forgotten for years and then remembered later.

Although our experiment tested for a possible mechanism enabling someone to forget CSA, only 2 participants reported having recovered memories of abuse by a caretaker. The scarcity of individuals who report having forgotten sexual abuse by a parental caretaker is fully consistent with Russell's (1986/1999) classic epidemiological study on incest. Not one of the incest survivors interviewed in that study ever mentioned having forgotten her sexual abuse (p. xxxvi).

In summary, we found that trauma words were recalled more often than neutral words, and this effect was just as evident in adults reporting either continuous memories of CSA or recovered memories of CSA as in the control group. Contrary to prediction, individuals reporting recovered memories of CSA did not exhibit memory deficits for trauma words they were told to remember when their attention was divided. Our findings suggest that DePrince and Freyd's (2004) results may not generalize to adults who report forgetting and then recovering memories of CSA. For these individuals, trauma words are strikingly memorable, regardless of whether they are encoded under divided- or selective-attention conditions. Preparation of this article was supported by National Institute of Mental Health Grant MH61268 awarded to the first author. We thank Anne DePrince for answering questions about her methods.


Bernstein, E.M., & Putnam, F.W. (1986). Development, reliability, and validity of a dissociation scale. Journal of Nervous and Mental Disease, 174, 727-735.

Clancy, S.A., & McNally, R.J. (2004). "Recovered" memories of childhood sexual abuse: Forgetting as a consequence of voluntary suppression. Manuscript submitted for publication.

DePrince, A.P., & Freyd, J.J. (1999). Dissociative tendencies, attention, and memory. Psychological Science, 10, 449-452.

DePrince, A.P., & Freyd, J.J. (2001). Memory and dissociative tendencies: The roles of attentional context and word meaning in a directed forgetting task. Journal of Trauma and Dissociation, 2, 67-82.

DePrince, A.P., & Freyd, J.J. (2004). Forgetting trauma stimuli. Psychological Science, 15, 488-492.

Freyd, J.J. (1996). Betrayal trauma: The logic of forgetting childhood abuse. Cambridge, MA: Harvard University Press.

Johnson, H.M. (1994). Processes of successful intentional forgetting. Psychological Bulletin, 116, 274-292.

McNally, R.J. (2003). Remembering trauma. Cambridge, MA: Belknap Press/Harvard University Press.

McNally, R.J., Clancy, S.A., Barrett, H.M., & Parker, H.A. (2004). Inhibiting retrieval of trauma cues in adults reporting histories of childhood sexual abuse. Cognition and Emotion, 18, 479-493.

McNally, R.J., Clancy, S.A., & Schacter, D.L. (2001). Directed forgetting of trauma cues in adults reporting repressed or recovered memories of childhood sexual abuse. Journal of Abnormal Psychology, 110, 151-156.

McNally, R.J., Metzger, L.J., Lasko, N.B., Clancy, S.A., & Pitman, R.K. (1998). Directed forgetting of trauma cues in adult survivors of childhood sexual abuse with and without posttraumatic stress disorder. Journal of Abnormal Psychology, 107, 596-601.

Russell, D.E.H. (1999). The secret trauma: Incest in the lives of girls and women (2nd ed.). New York: Basic Books. (Original work published 1986)

Spiegel, D. (1997). Foreword. In D. Spiegel (Ed.), Repressed memories (pp. 5-11). Washington, DC: American Psychiatric Press.

Terr, L.C. (1991). Childhood trauma: An outline and overview. American Journal of Psychiatry, 148, 10-20.


[1] The recognition data were strikingly similar to the recall data. For example, although DePrince and Freyd's (2004) Group x Word Category interaction was nearly significant for remember words in the divided-attention condition, ours was not (F< 1). All three of our groups recognized more trauma than neutral words, F(1, 45) = 40.7, p = .0001, r = .69.

(Received 3/22/04; Revision accepted 7/13/04)

To cite this article

McNally, Richard J., Ristuccia, Carel S. & Perlman, Carol A. (2005)

Forgetting of Trauma Cues in Adults Reporting Continuous or Recovered Memories of Childhood Sexual Abuse.

Psychological Science 16 (4), 336-340. doi: 10.1111/ j.0956-7976.2005.01536.x