Reduced autobiographical memory specificity predicts depression and posttraumatic stress disorder after recent trauma.

In this prospective longitudinal study, the authors examined the relationship between reduced specificity in autobiographical memory retrieval and the development of depression, posttraumatic stress disorder (PTSD), and specific phobia after injury in an assault. Assault survivors (N = 203) completed the Autobiographical Memory Test (J. M. G. Williams & K. Broadbent, 1986) at 2 weeks after the trauma as well as structured clinical interviews at 2 weeks and 6 months. Participants with acute stress disorder or major depression at 2 weeks, but not those with phobia, retrieved fewer specific autobiographical memories than those without the respective disorder. Reduced memory specificity at 2 weeks also predicted subsequent PTSD and major depression at 6 months over and above what could be predicted from initial diagnoses and symptom severity. Moderator analyses showed that low memory specificity predicted later depression in participants with prior episodes of major depression but not in those without prior depression. Mediation analyses suggested that rumination partly mediated and perceived permanent change fully mediated the effects of low memory specificity on posttrauma psychopathology at follow-up. PsycINFO Database Record (c) 2008 APA, all rights reserved.

Overgeneral Memory and Depression

People with major depression often show difficulty in retrieving specific autobiographical memories (Williams & Broadbent, 1986; for a review see Williams et al., 2007). When asked to remember a specific event from their lives in response to a cue word (e.g., “happy”), individuals with depression show an overgeneral memory bias (OGM), that is, they tend to reply with descriptions that summarize several different events (e.g., “I am always happy when I visit friends”) rather than a specific instance (“I was happy when my school friend rang me last week”). The origins and consequences of OGM in depression have attracted much attention (see Williams et al., 2007, for a review). Depression is often triggered by stressful life events (e.g., Kendler, Karkowski, & Prescott, 1999). This raises the issue of whether stressful or traumatic life events play a role in the development of OGM retrieval. Williams (1996) suggested that an overgeneral retrieval style may develop as a means of regulating affect (affect-regulation hypothesis) after stressful events. He proposed that survivors of trauma or severe stress may learn to avoid the painful emotions that would be provoked if they recalled their negative experiences by halting autobiographical memory retrieval prematurely, that is, before a specific memory is retrieved. Over time, the truncated search may become generalized, leading to an overgeneral level of autobiographical remembering. Such an overgeneral retrieval style is thought to be an index of vulnerability to subsequent depression. OGM has indeed been shown to predict depression and anxiety after a failed in vitro fertilization (van Minnen, Wessel, Verhaak, & Smeenk, 2005) and to interact with the occurrence of stressful life events in predicting subsequent depression in college students (Gibbs & Rude, 2004).

Overgeneral Memory and Trauma

The affect-regulation hypothesis suggests that exposure to stressful events is a precursor of OGM. Empirical research only offers partial support for a relationship between OGM and trauma (Moore & Zoellner, 2007) in that some, but not all, studies found an association. Among depressed samples, a large proportion of individuals with OGM also reported a history of trauma and trauma-related intrusive and avoidance symptoms (e.g., Dalgleish et al., 2003; Hermans et al., 2004; Kuyken & Brewin, 1995). Henderson, Hargreaves, Gregory, and Williams (2002) found in a nonclinical sample that women who reported a history of childhood abuse had difficulties retrieving specific memories in response to cue words. However, some studies failed to find an association between childhood trauma and OGM (e.g., Johnson, Greenhoot, Glisky, & McCloskey, 2005; Wessel, Meeren, Peeters, Arntz, & Merckelbach, 2001).

OGM and Posttraumatic Stress Disorder (PTSD)

The inconsistent findings on the relationship between trauma and OGM raise the possibility that OGM may not be a necessary consequence of trauma (Moore & Zoellner, 2007). However, OGM may influence how well people adapt to such extremely stressful experiences. Problems in autobiographical memory retrieval may prevent trauma survivors from coming to terms with the trauma and from reclaiming their former lives. Researchers and clinicians have observed that people with PTSD appear “frozen” at the time of the traumatic event (e.g., Herman, 1992) and that they have difficulty remembering what they used to be like before the traumatic event (e.g., Ehlers, Maercker, & Boos, 2000). OGM may contribute to such changes in self-perception.

Several studies have found that trauma survivors with PTSD or acute stress disorder (ASD) show less specific autobiographical memory retrieval than those without PTSD or ASD (Harvey, Bryant, & Dang, 1998; Kangas, Henry, & Bryant, 2005; McNally, Lasko, Macklin, & Pitman, 1995; Schönfeld & Ehlers, 2006; Schönfeld, Ehlers, Böllinghaus, & Rief, 2007).

OGM and Other Anxiety Disorders

Interestingly, a series of studies have found that OGM does not appear to be associated with most other anxiety disorders (Burke & Mathews, 1992; Wenzel & Jordan, 2005; Wenzel, Werner, Cochran, & Holt, 2004). Thus, the literature suggests that OGM may be a specific feature of depression and ASD/PTSD. However, there is a lack of studies directly comparing the relationships between OGM and different psychological outcomes after stress.

OGM and the Maintenance of Depression and PTSD

It has been suggested that OGM plays a role in the maintenance of depression (e.g., Gibbs & Rude, 2004; Peeters, Wessel, Merckelbach, & Boon-Vermeeren, 2002; Williams et al., 2006). OGM has been shown to predict poorer long-term clinical outcome in people with depression (e.g., Brittlebank, Scott, Williams, & Ferrier, 1993; Raes, Hermans, Williams, Beyers, et al., 2006; Williams, 1996). It has also been shown to be associated with deficits in social problem solving (Goddard, Dritschel, & Burton, 1996) and low ability to imagine future events (Williams et al., 1996), both of which may contribute to the maintenance of negative affect and hopelessness.

It remains as yet unclear whether OGM is also involved in the maintenance of PTSD. Theories of PTSD suggest that recovery from trauma requires the integration of the traumatic experience with the person's schemas about the self and the world (e.g., Resick & Schnicke, 1993) and with other autobiographical memories that determine the person's sense of self (e.g., Conway & Pleydell-Pearce, 2000; Ehlers & Clark, 2000). An impaired ability to retrieve specific memories from one's past (in particular positive memories) may impede such integration and thus delay recovery (see also Ehlers & Clark, 2000; Harvey et al., 1998; van Minnen et al., 2005; Williams et al., 2006). Prospective longitudinal studies are sparse and have generated mixed evidence as to whether OGM predicts greater long-term posttraumatic stress symptom severity. Harvey et al. (1998) assessed OGM shortly following accidents and found that low specificity in retrieving memories from the period following the trauma to positive cues predicted subsequent PTSD. However, Kangas et al. (2005) did not find that OGM predicted subsequent PTSD in recently diagnosed cancer patients, although OGM was related to concurrent ASD. In the latter study, memory retrieval was not restricted to the time following the trauma. In the present study, we investigated prospectively whether OGM predicts PTSD and depression in a sample of assault survivors.

Possible Mechanisms of Maintenance

What are the possible mechanisms by which OGM may put people at risk of chronic depression and PTSD after trauma? One possible way in which OGM may impede recovery is through a reciprocal relationship with rumination (Williams, 1996). According to the capture and rumination hypothesis (Williams et al., 2006), the cues used in the standard measure of OGM, the Autobiographical Memory Test (AMT; Williams & Broadbent, 1986), may activate rumination and abstract negative self-schemas so that respondents get “captured” in abstract cognitions and have problems progressing to the retrieval of specific memories. The failure to retrieve specific memories may in turn make it more difficult to correct the negative self-schemas and, thus, maintain negative thinking and related negative affect. People with depression or PTSD may be vulnerable to this maintaining process, as these disorders are characterized by negative self-related schemas and rumination (e.g., Foa, Ehlers, Clark, Tolin, & Orsillo, 1999; Ingram, Miranda, & Segal, 1998; Murray, Ehlers, & Mayou, 2002; Nolen-Hoeksema & Morrow, 1991). In contrast, people with other anxiety disorders—such as specific phobias—may be less susceptible, as their anxiety is mainly focused on external situations, and their self-schema may be less affected. In PTSD, failure to retrieve specific memories may be particularly relevant for appraisals of permanent change, that is, the appraisal that one and one's life have irreversibly changed for the worse by the trauma. Schönfeld and Ehlers (2006) found that OGM was associated with rumination and with perceived permanent change in trauma survivors. Perceived permanent change may interfere with people reclaiming their former lives after trauma and delay recovery. It has been shown to predict chronic PTSD (Dunmore, Clark, & Ehlers, 2001; Ehlers et al., 2000).

The above considerations raise the possibility that rumination and negative self-related schemas mediate the relationship between OGM and psychopathology. Preliminary data in support of this hypothesis were presented by Raes, Hermans, Williams, Beyers, et al. (2006), who showed that rumination partly mediated the relationship between OGM and subsequent depression. Data on PTSD and self-related schemas are lacking. In the present study, therefore, we considered rumination and perceived permanent change as potential mediators between OGM and depression and PTSD after trauma.

Aims of the Study

The purpose of the present study was to examine autobiographical memory retrieval in a sample exposed to a recent severe stressor, injury in a violent assault. We studied whether memory specificity is related to major depression, ASD, and specific assault-related phobia at 2 weeks postassault. To our knowledge, this is the first study to investigate the role of specificity in predicting these three disorders in the same sample of trauma survivors. On the basis of previous findings (Gibbs & Rude, 2004; Harvey et al., 1998; Wenzel, Jackson, & Holt, 2002), we hypothesized that assault survivors with major depression and those with ASD, but not those with specific phobia, would retrieve fewer specific memories than those without these disorders. Furthermore, we expected that memory specificity at 2 weeks would predict major depression and PTSD at 6 months following assault, and we tested whether it predicted over and above initial symptoms. In this study, we also tested whether rumination and perceived permanent change would mediate the influence of OGM on depression and PTSD. Finally, we considered history of childhood abuse, history of major depression, sex, and ethnicity as potential moderators of the relationship between memory specificity and posttrauma psychopathology (Williams et al., 2007).

Method

Participants

Participants were recruited from assault survivors who were treated for their injuries at the emergency department of a large urban teaching hospital during the period between July 2003 and December 2004. To be eligible for the study, participants had to understand and speak English fluently enough to be able to answer interview questions and fill in questionnaires. Participants with current psychosis and substance dependence, as well as those who could not remember the event (e.g., because of a head injury), were excluded.

A total of 1,063 assault survivors who attended the emergency department during the recruitment period were contacted, of whom 389 did not fulfill inclusion criteria, and 197 declined to take part. Another 255 were initially interested but failed to schedule or attend the research session within the designated time period. A total of 222 participants consented to participate in a research session at 2 weeks after the assault, and 203 (91%) of these completed the AMT. Reasons for noncompletion were too little time to complete the test and inability to concentrate for an extended period of time. Table 1 presents the demographic and clinical characteristics of the study sample. Participants had mainly been physically assaulted (n = 200, 99%; n = 3 sexual assaults). At 6-month postassault, 190 of the participants who had completed the AMT (83%) completed a further diagnostic interview.

Table 1

Sample and Assault Characteristics (N = 203)

To test whether the sample was representative of assault survivors presenting to this emergency department, we compared their demographic characteristics with those of a random sample drawn from all 2,785 assault survivors treated in the department during a period of 1 year. The study group and the random sample did not differ in age, sex, or severity of the injuries (all ps > .296). However, comparison with hospital data showed that the study sample may1 have comprised significantly more Caucasian participants than the random sample, χ2(1, N = 416) = 17.12, p < .001, namely 57.2% compared with a recorded 36.9% in the hospital sample. This difference may be due to the fact that some non-Caucasians had to be excluded because of language difficulties. We therefore tested whether ethnicity moderated the effects of memory specificity on psychopathology.

This figure presents an underestimate of the number of Caucasian hospital attendees. The figures are based on emergency department codes that require the staff to code ethnicity in 80 categories on the basis of country of origin. In practice, the category of “other” was often for non-British participants, regardless of their ethnic background.

AMT

The AMT (Williams & Broadbent, 1986) was administered individually by Birgit Kleim following the procedures outlined in Williams's (2003) manual. Participants saw 12 cue words (6 positive words: cheer, pleased, relieved, lively, glorious, peaceful; 6 negative words: worse, guilty, hopeless, awful, grave, ugly), which occurred in random order on a computer screen. Positive and negative words were matched for word frequency derived from the London–Lund corpus of English conversations (see Brown, 1984). The participants' task was to retrieve, and briefly describe, a specific personal memory in response to each cue word. The experimenter explained that a specific memory is a past experience that happened on a particular day. Participants were told that the event could be important or trivial. The time period from which events could be recalled was not specified. The experimenter gave examples of appropriate specific memories and inappropriate general memories, and gave participants the opportunity to practice the task with practice words. The AMT was not started until the participants had retrieved at least one specific memory to one of the practice words. Participants were given a maximum of 30 s to retrieve a memory for each word. If they did not provide a response in that time, this was scored as an omission. Responses were tape-recorded and later transcribed and scored for the percentage of first memories that were specific memories and for the assault-relatedness of specific memories. A memory was rated as assault-related when it involved the assault or its consequences (e.g., “when I looked into the mirror and saw the scar from the assault this morning”). A trained psychology graduate student who was blind to participants' diagnostic status rated all memories. A second independent rater who was blind to participants' diagnostic status scored a random sample of 50 oral AMT responses; good interrater agreement was obtained for the categorization of specific versus nonspecific responses (κ = .87) and for the categorization of assault-relatedness (κ = .78).

Assessment of Diagnoses and Symptom Severities

At 2 weeks, diagnoses of current and past (i.e., prior to the assault) major depression, ASD, and assault-related specific phobia were established with standard structured clinical interviews—the Structured Clinical Interview for DSM–IV (SCID; First, Spitzer, Gibbon, & Williams, 1996) and the Acute Stress Disorder Scale (ASDS, Interview Version; Bryant & Harvey, 2000), respectively. A trained psychologist (Birgit Kleim) conducted all interviews under the supervision of Anke Ehlers. Note that Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM–IV; American Psychiatric Association, 1994) proposes a hierarchical relationship in the diagnosis of some of the psychiatric disorders, stipulating that phobia may not be diagnosed if it is better accounted for by PTSD. To allow the investigation of phobic responses independent of the presence of PTSD symptoms, we omitted this rule in the present study, and participants could thus meet diagnoses of both ASD (or PTSD) and phobia if they had phobic symptoms severe enough to cause clinically significant distress.

The ASDS rates the presence of 19 ASD symptoms, each on a scale ranging from 0 (not at all) to 4 (very much). It determines a diagnosis of ASD as well as a symptom severity score (the sum of the scores for each symptom). In parallel, a depression symptom severity score was calculated as the sum of DSM–IV depression symptoms in the SCID (coded as 1 = absent, 2 = subthreshold, and 3 = present). Severity of assault-related phobia was assessed with the Assault Phobia Questionnaire. This 10-item scale follows DSM–IV criteria for specific phobias. It is similar to a scale for accident-related phobia developed by Ehring, Ehlers, and Glucksman (2006) that distinguishes well between people with and without travel phobias. Participants are asked to identify the situation, person, or object that they feared most since the assault. They indicate on a scale ranging from 0 (not at all) to 4 (very strongly) how afraid they were of this situation, person, or object (e.g., “I am very afraid of…”) and how much they avoided it (e.g., “I always avoid…”). The scale showed excellent internal consistency (α = .94).

At 6 months, the SCID modules for PTSD, major depression, and assault-related phobia were administered over the telephone by the same interviewer. The severity of PTSD symptoms was assessed with the PTSD Symptom Scale—Interview Version (PSS–I; Foa, Riggs, Dancu, & Rothbaum, 1993). The interviewer rated each of the PTSD symptoms on a scale ranging from 0 (not at all) to 3 (5 or more times per week/very much). The total PSS–I score is the sum of the ratings for the 17 items.

Cases were selected randomly for interrater reliability ratings, which were high (κ = .97 for ASD, κ = 1.00 for depression, κ = .85 for phobia, κ = .82 for PTSD; on the basis of 56 interviews, two raters who were each uninformed as to the other rater's diagnoses).

The interviewer asked participants when their symptoms started to ensure that they were responses to the assault. If participants reported that they had already been depressed or had other symptoms prior to the assault, the interviewer enquired whether the symptoms worsened as a consequence of the assault.

Assessment of Potential Mediators

Response to Intrusions Questionnaire

The Response to Intrusions Questionnaire (Clohessy & Ehlers, 1999; Murray et al., 2002) assesses dysfunctional cognitive strategies in response to intrusive memories. It was developed over a series of studies and shows good internal consistency and validity in predicting chronic PTSD (Clohessy & Ehlers, 1999; Murray et al., 2002). The Rumination subscale (α = .84 in the present study) comprises seven items asking participants to rate what they do when memories of the assault pop into their mind (e.g., “I dwell on how the event could have been prevented,” or “I dwell on how I used to be before the event”), each on a scale ranging from 0 (not at all) to 3 (always).

Perceived permanent change

This was assessed with the four relevant items from the Posttraumatic Cognitions Inventory (Foa et al., 1999)—for example, “I have permanently changed for the worse” (α = .77 in the present sample). The Posttraumatic Cognitions Inventory measures trauma-related thoughts and beliefs that have been shown to discriminate well between trauma survivors with and without PTSD. It has been shown to have good internal consistency and retest reliability.

Verbal Intelligence

Participants completed the National Adult Reading Test (NART; Nelson, 1991), a measure of verbal intelligence that is a widely accepted method of estimating premorbid intelligence levels. The NART requires participants to read out loud a list of 50 irregularly spelled words in order of increasing difficulty. Responses are individually scored as correct or incorrect, according to their pronunciation. The number of words read correctly comprises the final score. The NART has excellent reliability and construct validity (Crawford, Parker, Allan, Jack, & Morrison, 1991). It correlates highly with other measures of intelligence and allows the prediction of full-scale IQ scores (Nelson, 1991). Correlations with the Wechsler Adult Intelligence Scale (Wechsler, 1997) in healthy participants were around r = .75 (Bright, Jaldow, & Kopelman, 2002). Performance is characteristically preserved relative to other measures of cognitive function (e.g., Crawford, Steward, Garthwaite, Parker, & Besson, 1988). Recent studies have shown that current NART performance is superior in estimating a premorbid intelligence to scores based on demographic variables, such as education (e.g., Bright et al., 2002).

History of Childhood Abuse and Adult Trauma

Trauma history

Participants were asked whether, during their childhood, they had ever been physically abused or whether they had had any unwanted sexual experience as part of an interview about previous traumas (adapted from Kubany et al., 2000). They also stated the nature of the abuse, their age, and distress at the time. Participants further reported whether they had experienced any from a list of other traumatic events, such as an earlier assault, a sexual assault, a motor vehicle accident, warfare or combat, a robbery, threat of death or serious body harm, or any other life threatening or highly disturbing event, such as a natural disaster. The number of traumatic events experienced was used as an index of adult trauma history.

Objective assault severity

Objective severity of the recent assault was scored as a composite score from the following information: number of assailants (0 = one assailant, 1 = two or more assailants), assault duration (0 = 5 min or less, 1 = more than 5 min), injury severity (0 = mild injuries, 1 = moderate injuries, 2 = severe injuries), and weapon usage (0 = no weapon used, 1 = weapon used; see Dunmore et al., 2001).

Procedure

The study was approved by the local ethics committees and undertaken between July 2003 and December 2004. A few days after their admission at the emergency department, assault survivors received information about the study by mail and were invited to participate in a research session at 2 weeks after the assault. The invitation letter was followed by a telephone call in which further information about the study was given. Upon arrival at the session, participants had the opportunity to ask further questions and gave written informed consent. In the session, participants provided some details about their assaults, gave a short trauma narrative, did a picture identification task (results are reported elsewhere), and filled in questionnaires. They then completed the AMT. The diagnostic interviews followed. At 6 months, participants completed the relevant SCID modules over the telephone and filled in questionnaires. Participants were reimbursed £50 ($97) for their time and travel expenses.

Data Analysis

The Statistical Package for the Social Sciences (SPSS 15.0) was used for all analyses. Following common practice in investigations of OGM (see Williams et al., 2006, 2007), the main dependent variable was the percentage of the first memories that participants retrieved in response to the AMT cue words that were specific (we refer to this variable as memory specificity). We tested differences between participants with and without the disorders under investigation at 2 weeks (e.g., ASD vs. no ASD; major depression vs. no depression) with analyses of variance using the general linear modeling procedure, with the between-factors diagnostic group and the within-subject factor AMT cue valence (positive vs. negative). Logistic regression analyses tested whether memory specificity contributed to the prediction of PTSD, depression, and phobia at 6 months over and above what could be predicted from initial symptoms. Multiple regression analyses were conducted for moderator and mediator analyses (Baron & Kenny, 1986). PSS–I scores were square root transformed to normalize distributions. Mediation analyses were conducted in accord with McKinnon and Dwyer (1993; see also McKinnon, Fairchild, & Fritz, 2007). The Sobel test tested the significance of mediation effects.

The significance level was set at p = .05, two-tailed. On the basis of the proportion of participants with the respective disorders, the study had 86%, 82%, and 91% power in detecting group differences in the order of d = 0.55 (estimated from Schönfeld & Ehlers, 2006; Schönfeld et al., 2007) at 2 weeks for depression, ASD, and specific phobia, respectively, for α = .05, two-tailed. At 6 months, the sample size of N = 190 had 78% power to detect an odds ratio of 1.7 in the prediction of PTSD or depression diagnosis with OGM, after adjusting for initial symptom severity (estimated from Bryant, Sutherland, & Guthrie, 2007; Schönfeld & Ehlers, 2006).

Results

Diagnoses

As shown in Table 1, 18.7% had major depression at 2 weeks, 16.7% participants had ASD, and 23.6% had an assault-related phobia. Of the 38 participants with major depression, 14 (36.8%) had become depressed after the assault, and 24 (63.2%) had already felt somewhat depressed before the assault. All of the latter participants reported that the depression had significantly worsened as a result of the assault; furthermore, none of these participants had the same degree of depression present before the assault. At 6 months, 16.3% of the participants had major depression, 24.2% had PTSD, and 20.0% had an assault-related specific phobia. There was considerable comorbidity between the three disorders under investigation. At 2 weeks, 41% of the participants who met diagnostic criteria for ASD also had major depression, and 37% of the participants who met diagnostic criteria for major depression also had ASD. At 6 months, 43% of participants with PTSD had comorbid major depression, and 64% of those with major depression also had PTSD.

Group Differences in Memory Specificity at 2 Weeks

On average, participants retrieved M = 7.63 (SD = 2.51) specific memories across the 12 trials (64%); for positive cues, M = 3.81 (SD = 1.47); for negative cues, M = 3.82 (SD = 1.46). Omissions were rare (see Table 2). Table 2 presents the AMT performance at 2 weeks for the ASD, major depression, and phobia groups. For comparison, results for assault survivors who did not have any of the diagnoses under consideration are also reported (see Table 2). As a conservative test of the association between low autobiographical memory specificity and ASD, major depression, and assault-related phobia, we compared participants meeting diagnostic criteria for each of the disorders with those who did not meet the criteria for that disorder2 (which could include participants meeting criteria for another disorder). These results are illustrated in Figure 1 (a table with means and standard deviations is available online as supplemental material). Significant main effects of diagnostic group indicated that participants with major depression, F(1, 201) = 6.08, p = .014, ηp2 = .029, and participants with ASD, F(1, 201) = 3.98, p = .047, ηp2 = .019, retrieved fewer specific autobiographical memories than participants without these disorders. In contrast, participants diagnosed with assault-related phobias did not show less specific memory retrieval than those without phobia, F(1, 201) = 0.17, p = .681, ηp2 = .001. There were no main effects of cue valence (all ps > .52; all ηp2 < .003) or interactions between valence and diagnostic group (all ps > .42; all ηp2 < .004).

Table 2

Autobiographical Memory Test Performance for Assault Survivors With Acute Stress Disorder (ASD), Major Depression (MD), and Specific Phobia (SP), and for Assault Survivors Without Any of These Disorders at 2 Weeks

Figure 1

Mean percentages of specific autobiographical memories in assault survivors with and without acute stress disorder (ASD), depression, and assault-related phobia, respectively. Black bars = positive cues; gray bars = negative cues; ns. = nonsignificant. ** p < .01.

The results were the same when the disorder groups were compared with participants without depression, ASD, or phobia.

There were no significant group differences between any of the diagnostic groups and the respective control groups without that disorder in the assault-relatedness of specific memories (all ps > .19; all ηp2 < .009). There was a significant main effect of valence: Across diagnostic groups, participants produced more assault-related memories to negative words than to positive words (all ps < .001; all ηp2 > .133). There was also a significant interaction between valence and ASD, F(1, 195) = 4.09, p = .045, ηp2 = .021, in that the ASD group retrieved more assault-related specific memories to negative cues than participants without ASD (MASD = 61.27, SD = 38.02 vs. MNoASD = 42.13, SD = 42.78, respectively), F(1, 202) = 5.86, p = .016. Interactions between valence and depression and phobia diagnostic groups were nonsignificant (both ps > .570; both ηp2 < .002). For omissions, there were no significant group differences (all ps > .32; all ηp2 < .006) or valence effects (all ps > .42; all ηp2 < .004). There was a marginally significant interaction between depression and valence, F(1, 201) = 3.42, p = .066, ηp2 = .017. Depressed participants tended to produce more omissions in response to positive cues than nondepressed participants.

Prediction of PTSD, Major Depression, and Specific Phobia at 6 Months

Logistic regression analyses tested whether memory specificity at 2 weeks contributed to the prediction of PTSD, depression, and assault-related specific phobia at 6 months over and above initial symptoms. For PTSD at 6 months, ASD symptom severity at 2 weeks (ASDS) was forced into the regression in the first step and explained 43% of the variance, χ2(1, N = 190) = 64.33, p < .001, Nagelkerke R2 = .43. Low memory specificity significantly improved the prediction, χ2(1, N = 190) = 7.04, p = .008, total Nagelkerke R2 = .47, ΔR2 = .04. We repeated the analysis to test whether memory specificity contributed to the prediction of PTSD without comorbid depression or phobia. Low memory specificity again improved the prediction over and above initial symptom severity at trend level, χ2(1, N = 139) = 3.18, p = .075, total Nagelkerke R2 = .369, ΔR2 = .04. For major depression at 6 months, depression severity at 2 weeks (number of symptoms on SCID) was forced into the regression in the first step and explained 13% of the variance, χ2(1, N = 187) = 14.59, p < .001, Nagelkerke R2 = .13. Low memory specificity significantly improved the prediction, χ2(1, N = 187) = 7.45, p = .006, total Nagelkerke R2 = .19, ΔR2 = .06. For specific phobia at 6 months, phobia severity at 2 weeks (Assault Phobia Questionnaire) was forced into the regression in the first step and explained 27% of the variance, χ2(1, N = 187) = 32.59, p < .001, Nagelkerke R2 = .27. Low memory specificity did not add significantly to the prediction, χ2(1, N = 187) = 1.45, p = .230, total Nagelkerke R2 = .28, ΔR2 = .01.

We also tested whether memory specificity predicts PTSD and depression diagnoses at 6 months over and above initial depression and ASD diagnoses. This was the case; memory specificity predicted both PTSD and depression over and above initial ASD and depression diagnoses, χ2(1, N = 190) = 7.29, p = .008, total Nagelkerke R2 = .292, ΔR2 = .047 (PTSD), and χ2(1, N = 190) = 6.23, p = .014, total Nagelkerke R2 = .229, ΔR2 = .049 (depression).

Hierarchical linear regression analyses using the continuous PSS–I and depression severity scores at 6 months as outcome measures obtained the same results. In the first step, symptom severity at 2 weeks was forced into the regression function, and OGM in the second step. For both PTSD and depression, OGM predicted over and above initial symptom severity: PTSD prediction, Step 1: F(1, 188) = 152.90, p < .001, R2 = .45; Step 2: ΔF(1, 187) = 6.20, p = .014, ΔR2 = .02, OGM β = −.14; depression prediction, Step 1: F(1, 186) = 27.00, p < .001, R2 = .13; Step 2: ΔF(1, 185) = 4.18, p = .042, ΔR2 = .02, OGM β = −.14.

OGM in PTSD and Depression: Correlates, Moderation Analyses, and Mediation Analyses

Table 3 shows the correlations between OGM, symptoms of PTSD and depression, trauma variables, intelligence, and the variables considered for moderation and mediation analyses. OGM correlated with the severity of the present assault and low intelligence but not with childhood abuse and past traumas. To test whether memory specificity predicts over and above intelligence and assault severity, we entered NART scores and assault severity in the first step of a hierarchical logistic regression analysis, and we entered memory specificity in the second step. OGM at 2 weeks predicted both PTSD and major depression at 6 months over and above what could be predicted from intelligence and assault severity: for PTSD, χ2(1, N = 181) = 3.68, p = .055, ΔNagelkerke R2 = .029; for depression, χ2(1, N = 181) = 5.76, p = .016, ΔNagelkerke R2 = .052.

Table 3

Pearson Correlations Between Memory Specificity, Acute/Posttraumatic Stress, and Depression Severity at 2 Weeks and at 6 Months With Trauma and Cognitive Variables

Moderation effects

To test for moderation, we entered memory specificity in the first step of a hierarchical multiple regression analysis, followed by the moderator variable under investigation, followed by the interaction of memory specificity and the moderator (Baron & Kenny, 1986). Sex, childhood abuse, and ethnicity did not moderate the effect of specificity on PTSD or depression; all interaction terms remained nonsignificant in the hierarchical regression equations (all ps > .11).

History of major depression prior to the assault moderated the effect of memory specificity on depression severity, as indicated by a significant interaction (β = −.19, p = .042), which contributed significantly to the explained variance in depression symptom severity at 6 months, ΔR2 = .02, F(3, 186) = 6.86, p < .001. Further analyses showed that for participants with a history of major depression (n = 79), memory specificity significantly predicted depression severity at 6 months (β = −.32, p = .004). Among participants who had never had a major depressive episode before the assault (n = 106), however, specificity was not a significant predictor of depression symptom severity at follow-up (β = −.12, p = .232). The interaction between history of major depression and memory specificity was not significant for PTSD symptom severity (β = −.11, p = .231).

Mediation effects

We used mediation analyses to test whether rumination and permanent change mediated the effects of memory specificity on PTSD and depression, using the three-step regression analyses approach suggested by Baron and Kenny (1986). Step 1 requires that the hypothesized mediator correlates with the independent variable (memory specificity). As shown in Table 3, both rumination and permanent change correlated significantly with memory specificity. Step 2 involves predicting the dependent variable with the independent variable alone, and Step 3 involves predicting the dependent variable with both the independent variable and the mediator. Table 4 summarizes the results for Steps 2 and 3. Mediation is demonstrated when the amount of variance explained by the independent variable is smaller when the mediator is included in the regression function compared with when the independent variable is considered alone. This was the case for both hypothesized mediators. For both PTSD and depression severity, the amount of variance explained by memory specificity was reduced when rumination or perceived permanent change was included in the regression, as indicated by significant Sobel tests (all ps < .042).

Table 4

Results of Regression Analyses Testing Whether Rumination and Permanent Change Mediate the Relationship Between Memory Specificity at 2 Weeks and PTSD and Depression Severity at 6 Months

Discussion

In this study, we aimed to investigate the relationship between overgeneral autobiographical memory retrieval at 2-week posttrauma and acute and chronic psychological disorders after a traumatic experience. At 2 weeks, assault survivors with ASD generated fewer specific autobiographical memories in the AMT than participants without these disorders, replicating findings in other trauma populations (Harvey et al., 1998; Kangas et al., 2005). Furthermore, participants who had major depression after the trauma showed lower memory specificity compared with those who did not develop depression. These results are in line with a recent study by van Minnen et al. (2005), who found that low memory specificity predicted the exacerbation of depressive mood after a stressful event, and with previous research showing OGM in depressed people with a history of trauma (Hermans et al., 2004; Raes, Hermans, Williams, Beyers, et al., 2006). The lack of group differences in OGM between assault survivors with and without assault-related specific phobia is also consistent with the literature, as previous studies found no relationship between OGM and anxiety disorders other than PTSD (e.g., Wenzel et al., 2002, 2004). Thus, the findings support the specificity of OGM to depression and ASD/PTSD. To our knowledge, this is the first study to demonstrate specificity in the same population.

Low memory specificity at 2 weeks predicted both chronic PTSD and major depression at 6 months, and predicted over and above what could be predicted from initial symptom severity, regardless of whether diagnosis or symptom severity at 6 months was used as outcomes. The current result is in contrast to a recent study of cancer patients (Kangas et al., 2005), which found that deficits in the retrieval of specific memories were not predictive of subsequent PTSD. However, the smaller sample size and attrition during follow-up may have affected these results.

The present study did not find any effects of AMT cue valence. The literature shows inconsistent findings (van Vreeswjik & de Wilde, 2004), with some studies reporting patients to be less specific in response to positive cues (McNally, Litz, Prassas, Shin, & Weathers, 1994; Williams & Scott, 1988), whereas others report OGM effects in response to negative cues, rather than positive and neutral cues (Jones et al., 1999; Mackinger, Pachinger, Leibetseder, & Fartacek, 2000). In line with our findings, McNally et al. (1995), Goddard et al. (1996), Schönfeld and Ehlers (2006), and Schönfeld et al. (2007) reported OGM in patients with PTSD and depression in response to both positive and negative cues.

In the present study, we also explored correlates of OGM, and potential moderator and mediators of the relationship between OGM and posttrauma psychopathology. Sex, ethnicity, and history of childhood abuse did not interact with OGM in predicting PTSD or depression at 6 months in the present sample. There was no relationship between OGM and childhood abuse and number of adult traumas, but OGM correlated with the severity of the recent assault. Together, these results point to a relationship between OGM and trauma, but they also suggest that OGM does not necessarily develop after trauma. The results are in line with a recent review that concluded that trauma exposure per se is unlikely to be the primary mechanism leading to overgenerality (Moore & Zoellner, 2007). Instead, OGM may characterize those survivors of childhood or adult trauma who develop psychopathology. Some previous studies also failed to find a significant association between childhood trauma and OGM (Johnson et al., 2005; Wessel et al., 2001). However, our study only included a limited assessment of childhood abuse, and some participants may not have disclosed abuse or other previous traumas as no treatment was offered. A more detailed assessment of childhood abuse would have allowed us to investigate the relationship more comprehensively and conclusively.

Recent studies by Dalgleish et al. (2007) and Raes, Hermans, Williams, Demyttenaere, et al. (2006) have suggested that the AMT is sensitive to individual differences in executive capacity and depression-related reduction in working memory. In line with their findings, OGM correlated with a measure of verbal intelligence in the current study. However, OGM predicted both PTSD and depression at 6 months over and above what could be predicted from intelligence.

A history of major depression prior to the assault moderated the relationship between OGM and major depression, but not PTSD, at follow-up. These results are in accord with the view that prior depressive episodes constitute a vulnerability factor that may predispose individuals to more general memory retrieval (Gibbs & Rude, 2004). There is evidence that individuals remitted from depression show more overgeneral autobiographical memory retrieval than never-depressed controls (Kuyken & Brewin, 1995; Mackinger et al., 2000). The present results are of interest in the discussion about findings that risk of further depressive episodes increases with every consecutive episode and that successive episodes of major depression require less and less external provocation by stressful life events (e.g., Kendler, Thornton, & Gardner, 2000). It has been hypothesized that in people with repeated episodes of major depression, new episodes become increasingly mediated by internal processes, such as ruminative thinking cycles reactivated by dysphoric mood (e.g., Ma & Teasdale, 2004). OGM may be an indicator of this cognitive vulnerability to recurrence of depression.

As expected, low memory specificity was related to greater rumination about the trauma and greater perceived permanent change. These variables were also shown to fully or partly mediate the relationship between OGM and both PTSD and depression at follow-up. Previous research has shown that rumination increases the tendency to retrieve overgeneral memories (see Park, Goodyer, & Teasdale, 2004). This is consistent with Williams's (1996) hypothesis that self-focused rumination contributes to a block of specific autobiographical retrieval (“mnemonic interlock”). Previous studies have found associations between OGM and rumination in depression (Raes, Hermans, Williams, Beyers, et al., 2006) and rumination about the trauma in trauma survivors (Schönfeld & Ehlers, 2006). Like the present study, Raes, Hermans, Williams, Beyers, et al. (2006) also found that rumination partly mediated the relationship between OGM and depression 7 months later, although they used a different measure of rumination that was not trauma related. The result that perceived permanent change mediated the effects of OGM on PTSD severity is consistent with Schönfeld and Ehlers's (2006) suggestion that inability to retrieve specific memories from one's life may enhance the sense that one has changed for the worse as a person, which in turn contributes to the maintenance of PTSD (Dunmore et al., 2001; Ehlers et al., 2000).

Although the study has several strengths, including a large sample size, recruitment soon after a trauma, and a prospective longitudinal design, it also has several limitations. First, it remains unclear whether low autobiographical memory specificity is a predisposing factor or a consequence of the trauma. For practical reasons, it was not possible to assess OGM prior to the assault. Hence, future research will need to address whether OGM effects are present prior to the development of psychopathology (see Bryant et al., 2007). Second, there was high comorbidity among the disorders, which is common in survivors of violent assault (e.g., Kilpatrick et al., 2003). This raises the question of whether the outcomes studied represented different disorders. Prior studies suggest that PTSD and depression are best conceptualized as related, but distinct, responses to trauma (Blanchard, Buckley, Hickling, & Taylor, 1998; Breslau, Chase, & Anthony, 2002; O'Donnell, Creamer, Pattison, & Atkin, 2004). High comorbidity makes it difficult to demonstrate specificity. Nevertheless, our findings show that OGM was specific to depression and ASD, whereas participants with phobia did not differ from participants without phobia in autobiographical memory retrieval. It is noteworthy that the study was conservative in testing specificity, as it allowed comorbid ASD and specific phobia. If we had excluded comorbid cases with ASD from the phobia group, we would have biased the results in favor of finding specificity of OGM for ASD as opposed to phobia. However, the high comorbidity may have compromised the ability to detect differences between depression and PTSD. Nevertheless, given the high comorbidity between the disorders, as well as the symptom overlap in diagnostic criteria between PTSD and depression, and between PTSD and phobia, it might be useful for future researchers to adopt a dimensional rather than a categorical approach. This could involve using symptom clusters across disorders (e.g., phobic responses, numbing, and reexperiencing) rather than symptom dimensions that are based on different diagnostic entities as outcomes. Third, diagnoses at 6 months were based on a SCID interview conducted over the phone, rather than a face-to-face diagnostic interview. Such telephone interviews may make participants somewhat less forthcoming in providing information about their symptoms than face-to-face interviews. However, the interviewer had already established a relationship with the participants at the first assessment, which makes it unlikely that participants withheld information from her later. Fourth, the sample comprised nearly exclusively survivors of physical assaults. The results are therefore limited to survivors of physical assaults, and they need to be replicated with other trauma populations.

In the present study, we pointed to an “added value” of OGM in the prediction of chronic PTSD and depression over and above initial symptom severity. However, the observed effect sizes were in the small to medium range. OGM only predicted an additional 4%–6% of the variance of depression and PTSD at follow-up. Nevertheless, the results are of interest, as initial symptom severity is a strong predictor of PTSD (e.g., Rothbaum, Foa, Riggs, Murdock, & Walsh, 1992), and very few variables have been shown to predict PTSD over and above initial symptoms. These effect sizes observed in this study are in the range of the best replicated predictors of PTSD according to a recent meta-analysis by Ozer, Best, Lipsey, and Weiss (2003). Transformed into binominal effect sizes, the results suggest that a trauma survivor high in OGM has an approximately 20% greater chance of having PTSD or depression at 6 months than a survivor low in OGM, an effect size that exceeds that of many standard medical interventions (Rosnow & Rosenthal, 2003).

Interestingly, the two factors that were found to partly mediate the effect of OGM on future depression and PTSD, rumination about the trauma and perceived permanent change, are among the few variables that have been shown to have incremental predictive validity over and above initial symptoms (Dunmore et al., 2001; Ehlers, Mayou, & Bryant, 1998). Overall, given the small incremental predictive power and the results of the mediator analyses, practical applications of the AMT as a tool for identifying traumatized people at risk of chronic PTSD appear limited. Questionnaire measures, such as the rumination or perceived permanent change scales, are easier to administer and score than the AMT and may thus be more useful for screening programs.

Nevertheless, the results point to potential therapeutic strategies. It is possible that training in the accessibility of specific memories would help people break through the capture and rumination cycle (Williams et al., 2006) and help people put the trauma into perspective by integrating it into the context of other experiences. Our data would suggest that such training may be potentially relevant for trauma survivors with episodes of past depression. There is evidence that OGM is modifiable (McBride, Segal, Kennedy, & Gemar, 2007; Watkins, Teasdale, & Williams, 2000). Alternatively, one may argue that the most promising way to reduce OGM and poor posttrauma adaptation would be to target rumination. In a series of studies, Watkins and colleagues (e.g., Watkins & Teasdale, 2001; Watkins et al., 2000) found that reducing rumination leads to less overgeneral memory retrieval in depressed patients. Reducing rumination is a component of some effective treatments for PTSD (e.g., Ehlers, Clark, Hackmann, McManus, & Fennell, 2005). It would also be of interest to study whether treatment procedures used in effective treatments of PTSD to counteract perceived permanent change, such as “reclaiming your life” assignments (Ehlers & Clark, 2000), change OGM bias.