اثر بار شناختی و انگیختگی بیش از حد در خاطرات منفی مزاحم

Abstract Clinical theories of post-traumatic stress suggest that encoding processes at the time of a trauma are critical in determining whether intrusive memories will develop. Potential mechanisms that might influence the development of intrusive memories were studied, as was objective memory performance. In an analogue design, 65 participants were randomised to three conditions (cognitive load, hyperventilation, and control), and then watched a film of traumatic content. Intrusive memories were recorded during the experimental phase and at 1-week follow-up. Support was found for the prediction that verbal cognitive load and hyperventilation would facilitate intrusion development immediately following exposure to the trauma film; however, this was not maintained at follow-up. Consistent with cognitive models of post-traumatic stress, thought suppression and the distress associated with intrusive experiences mediated the relationship between distress caused by the film and intrusions at 1-week follow-up. Objective memory testing indicated that the three experimental groups showed similar recall and recognition performance for the content of the film; however, relative to the control group, individuals in the cognitive load condition were significantly less able to place film scenes in the correct order.

Introduction Following a traumatic event, it is common for individuals to re-experience the event in the form of intrusive thoughts and memories, and persistent intrusions are a hallmark feature of post-traumatic stress disorder (PTSD), one of the most common disorders to develop following traumatic events (American Psychiatric Association, 1994). In recent years, there has been an interest in understanding the mechanisms that result in unpleasant memories being experienced in an intrusive fashion. Much of this research has been conducted adopting a theoretical framework from the post-traumatic stress literature given that PTSD is characterised by unwanted memories of the traumatic event. In this context, cognitive and emotional processes during a traumatic event have been argued to be critical in determining how the event is experienced and whether intrusive memories are likely to develop and persist (Brewin, Dalgleish, & Joseph, 1996; Ehlers & Clark, 2000). Accordingly, theoretical models of PTSD hold that individuals who are unable to make sense of their experience have difficulty encoding the experience in a logical manner, or experience confusion or dissociation at the time of trauma are at risk of developing a memory of the event that is not fully ‘processed’. This refers to a memory that is incomplete or disjointed, which is associated with extremely strong affect and which is not placed into autobiographical context. That is, it is not remembered as a past event that is no longer dangerous, but as an event still associated with a sense of current threat. It has been hypothesised that a traumatic experience is processed through two memory systems labelled ‘verbally accessible memory’ (VAM) and ‘situationally accessible memory’ (SAM) (Brewin et al., 1996). Dual-representation theory of PTSD argues that information at the time of the trauma that receives a high level of cognitive processing is likely to be encoded in memory and later retrieved from the VAM system. That is, as an autobiographical representation of the event that can be voluntarily recalled. Owing to attentional capacity limitations, not all information encoded at the time of the trauma can be consciously processed in this fashion. This information is more likely to be stored in the SAM system which is primarily responsible for storing sensory-type information, including imagery, physical sensations, etc. The SAM system is more likely to be accessed automatically, in particular in response to external cues (e.g., seeing someone of similar appearance to one's assailant), as well as internal cues (e.g., heightened anxiety). It is the SAM system that is thought to be responsible for involuntary intrusive experiences such as unwanted memories and flashbacks (reliving of the traumatic experience). There are great individual differences in the experience of intrusive memories, and clinical and experimental research is beginning to identify the critical processes involved. The majority of this work has used trauma films as an analogue stressor, given the obvious problems of experimental control for real trauma events. Researchers have found that placing attentional demands on the VAM and SAM systems during an analogue stressor (i.e., while watching a film of traumatic content) can increase and decrease intrusion frequency, respectively (Brewin & Saunders, 2001; Holmes, Brewin, & Hennessy, 2004). Consistent with dual-representation theory, tasks that place a cognitive load on the VAM system (e.g., counting backwards in 3 s) are hypothesised to prevent verbal information from being encoded comprehensively, hindering the development of a coherent, autobiographic narrative of the trauma experience, and such tasks have been found to result in more intrusive memories of the film over a 1-week period (Holmes et al., 2004). Similarly, visuo-spatial-type tasks that interfered with information being encoded in the SAM system resulted in participants experiencing fewer intrusions (Brewin & Saunders, 2001; Holmes et al., 2004; Stuart, Holmes, & Brewin, 2006). A separate line of research suggests that intense hyperarousal and anxiety play an important role in the development of intrusive experiences following trauma. Information processing models of trauma response posit that fear networks are developed in which representations of memories, affective and somatic responses, and attributions of the trauma are readily activated in the individual because of sensitivity to trauma-related stimuli (Foa & Kozak, 1986; Harvey & Bryant, 2002). Thus, intense affect and its association with cues at the time of the trauma increases the likelihood that these cues will become prompts for later intrusions. There is also indirect evidence of a relationship between hyperarousal and reexperiencing symptoms in clinical samples. For example, PTSD participants have been observed to demonstrate greater autonomic responses during the writing of flashback accounts than of ordinary memories (Hellawell & Brewin, 2002), peritraumatic arousal predicts subsequent intrusive symptoms (Resnick, 1997), and chemically induced arousal elicits flashback and intrusive phenomena (e.g., Bremner et al., 1997; Kellner, Levengood, Yehuda, & Wiedemann, 1998). More recently, inducing hyperarousal via a hyperventilation provocation test has been shown to result in an increase in intrusive memories among individuals with acute stress disorder (ASD) relative to non-ASD trauma survivors (Hopwood & Bryant, 2006; Nixon & Bryant, 2005). As discussed earlier, maladaptive processing of the traumatic event is argued to lead to memories that are incomplete, fragmented, and disorganised, and clinical studies of individuals with ASD and PTSD seem to support this (e.g., Foa, Molnar, & Cashman, 1995; Halligan, Michael, Clark, & Ehlers, 2003; Harvey & Bryant, 1999). A limitation of this research is that the exact, objective details of trauma are not known to the researchers; thus, it is difficult to be confident about the accuracy of the content of the memory. Experimental research with analogue trauma paradigms provides somewhat mixed findings in regard to the effect of negative experiences (typically trauma films) upon memory content and organisation. For example, across three studies Holmes et al. (2004) generally found few differences between encoding conditions in explicit memory (cued recall and recognition) of trauma film content. Other studies have found that various levels of cognitive processing can influence the degree to which film material is recalled in order and the coherence of the memory account (e.g., Halligan, Clark, & Ehlers, 2002, Study 1; Wegner, Quillian, & Houston, 1996); however, this is not an universal finding (e.g., Halligan et al., 2002, Study 2; Kindt & van den Hout, 2003; Rassin, Merckelbach, & Muris, 1997), with there being some suggestion that reports of disorganisation are more likely to be meta-memory judgements rather than an actual occurrence (e.g., Halligan et al., 2002; Kindt & van den Hout, 2003). Finally, three other processes are likely to influence the development of intrusive memories: thought suppression, negative interpretation of intrusions, and working memory capacity. For individuals with PTSD, thought suppression is a common technique used aimed at controlling these thoughts and memories. However, the very act of trying to suppress thoughts has been shown to be counterproductive, as it leads to an increase in the very thought one is trying to avoid (for reviews, see Rassin, Merckelbach, & Muris, 2000; Wenzlaff & Wegner, 2000). Individuals with PTSD are motivated to suppress memories of the trauma not only because they are distressing but also because they also hold distorted appraisals about the meaning of intrusions, for example, that there is something seriously wrong with them (Ehlers, Hackmann, & Michael, 2004). Given that the act of thought suppression places a load on cognitive capacity, the role of working memory becomes relevant. Working memory is sometimes defined as the capacity for sustained and controlled attention, when faced with distracting or interfering tasks (Brewin & Beaton, 2002). Brewin and his colleague observed that in university students, ability to suppress a nontraumatic experimental image (an image of a white bear) was significantly and positively correlated to the individual's working memory ability. It has been proposed that poorer working memory is (in part) responsible for the inability to inhibit information when it needs to be suppressed or controlled in some fashion, especially in the face of competing demands (Brewin & Beaton, 2002; Rosen & Engle, 1998). This relationship has been replicated by Brewin and other researchers (e.g., Brewin & Smart, 2005; Klein & Boals, 2001). The present research sought to examine a number of these factors in relation to intrusion development. Participants were randomised to three conditions (cognitive load, hyperventilation, and control) before watching a trauma film. Initial intrusions were assessed immediately after the film, and participants were followed up at 1 week. Thought suppression, working memory, intrusion distress, and premorbid post-traumatic stress was measured. The study had several aims. First, to replicate the finding by Holmes et al. (2004) that placing a verbal cognitive load during processing of a traumatic film will result in an increase in intrusive memories. Second, to test the proposition that hyperarousal during exposure to the film will strengthen encoding of material, leading to increased accessibility of the memories, and consequently more intrusions. Third, to examine the role of thought suppression, distress caused by intrusions, and working memory ability on intrusions. It was expected that regardless of condition, working memory ability, and thought suppression would be positively associated with intrusion development. We also predicted that thought suppression and distress caused by initial intrusions would mediate the relationship between distress caused by the film and later intrusions

Results One-tailed alpha values are reported for planned comparisons, and unless otherwise stated, all other values are 2-tailed. Effect sizes are reported as partial η2 for analysis of variance (ANOVA) or Hedges's g for comparisons (t tests). Hedges's g is interpreted with the same conventions as Cohen's d, but is adjusted for sample size, thus more suitable for comparisons with modest sample sizes. Eleven participants were lost to follow-up; however, these participants did not differ from completers on any baseline or experimental variables. Mean replacement was used for their missing data on the IES-R, and replicated the analyses conducted on the original data. Sample characteristics and manipulation checks ANOVAs for baseline variables showed the three groups were comparable on variables such as depression, anxiety, stress, post-traumatic symptoms to prior traumatic events, and working memory (see Table 1). Approximately, 83% of the sample had experienced a prior traumatic event and 30% fell above a clinical cut-off in terms of moderate symptom severity (⩾11, Foa, 1995). There were no significant differences between these groups on either of these variables, χ2(2, 64)=.72, p=.70, and χ2(2, 50)=1.77, p=.41, respectively. Table 1. Sample characteristics and outcome measures, including means, standard deviations, and inferential statistics Cognitive load (n=22) HVPT (n=22) Control (n=21) Statistic M (SD) M (SD) M (SD) DASS-Depression 2.77 (2.64) 4.05 (4.16) 3.57 (2.96) F(2, 62)=0.83 DASS-Anxiety 2.45 (2.56) 3.27 (3.55) 2.52 (2.09) F(2, 62)=0.57 DASS-Stress 6.23 (4.80) 7.30 (4.33) 6.33 (3.60) F(2, 62)=0.39 PDS 4.55 (5.38) 6.86 (10.09) 8.19 (7.76) F(2, 62)=1.15 OSPAN 10.58 (4.25) 8.68 (5.89) 9.05 (6.75) F(2, 62)=0.61 Film unpleasantness 9.14 (1.70) 8.55 (2.06) 9.43 (0.75) F(2, 62)=1.68 Distress from film 7.64 (2.54) 6.45 (2.81) 7.95 (1.91) F(2, 62)=2.25 Task confidence 8.86 (1.35) 7.50 (1.63) 8.12 (2.12) F(2, 61)=3.32* Immediate intrusions (frequency) 9.38 (4.80) 8.68 (5.08) 6.24 (2.19) F(2, 61)=3.18* Immediate intrusions (duration (s)) 83.57 (67.71) 43.46 (49.88) 66.47 (61.22) F(2, 61)=2.43† Suppression attempt 7.38 (2.58) 6.14 (2.68) 7.29 (2.05) F(2, 61)=1.72 Distress at intrusions 5.81 (2.36) 4.59 (2.68) 5.67 (2.27) F(2, 61)=1.61 One-week intrusions (IES-Intrusions) 1.03 (0.50) 0.71 (0.45) 1.01 (0.58) F(2, 62)=2.64† IES-Avoidance 1.06 (0.69) 0.66 (0.45) 0.97 (0.63) F(2, 62)=2.64† IES-Arousal 0.68 (0.55) 0.25 (0.23) 0.37 (0.40) F(2, 62)=6.23* IES-Total 0.91 (0.55) 0.57 (0.35) 0.76 (0.46) F(2, 62)=4.83* Recall sequence 2.05 (1.61) 2.57 (1.63) 3.10 (1.41) F(2, 58)=2.25 Recall performance 4.45 (1.84) 4.14 (1.78) 5.19 (1.75) F(2, 62)=1.95 Recognition performance (d’) 0.23 (0.65) 0.33 (0.78) 0.46 (0.56) F(2, 60)=0.62 Note: †p<.10, *p<.05. IES-I=Impact of Events Scale; PDS=Post-traumatic Stress Diagnostic Scale; DASS=Depression, Anxiety and Stress Scale; OSPAN=Operation-word span task. Table options The groups did not differ in terms of how unpleasant they found the film, F(2, 62)=2.06, p=.14, or how upset they felt after watching the film, F(2, 62)=2.25, p=.11. The groups did differ in their confidence that they adhered with the instructions given before the film, F(2, 62)=3.32, p=.04, η2=.10. Specifically, the HVPT group reported less confidence in keeping up with the hyperventilation tape than the cognitive load group did in remembering the 9-digit number string, t(41)=−2.97, p=.01, g=.89, although the mean value of 7.5 out of 10 for this rating was considered to reflect acceptable effort. There were no significant differences for adherence between the cognitive load and control group, t(40)=1.35, p=.19, nor between the HVPT and control group, t(41)=−1.08, p=.29. Accuracy of the 9-digit number for participants in the cognitive load group was reasonable, with 68% recalling the number string in the correct order. For the remaining 32%, their errors were relatively minor, and comprised either: (a) forgetting 1 digit, (b) recalling 2 digits in the wrong order, or (c) replacing a digit with one not in the original string. Immediate intrusions To test the proposal that the two experimental conditions would result in more intrusions than the control condition, a series of ANOVAs followed by planned comparisons was conducted. Descriptive statistics for these analyses are reported in Table 1. As predicted, the cognitive load and HVPT group experienced significantly more intrusions in the 5 min monitoring interval immediately post-film than the control group, t(40)=2.73, p<.01, g=.83, and t(41)=2.03, p=.02, g=.61, respectively, 1-tailed. In terms of duration of intrusions, differences between the two experimental conditions and the control condition were nonsignificant, t(40)=.86, p=.40, and t(41)=−1.35, p=.18. There were no significant differences between the groups in relation to their efforts to suppress intrusions and the distress caused by these intrusions during the monitoring period, F(2, 61)=1.72, p=.19, and F(2, 61)=1.61, p=.21, respectively. Intrusions at 1-week follow-up Of primary interest was the degree to which participants had intrusive experiences over the week following the experiment. Accordingly, planned comparisons were conducted using the Intrusion subscale of the IES-R as the dependent variable of interest. The analysis failed to support the prediction that the cognitive load group would have significantly more intrusion than the control group, t(40)=.10, p=.46 (1-tailed), and contrary to predictions, there was a nonsignificant trend for HVPT participants to have fewer intrusions than controls, t(41)=−1.89, p=.07 (2-tailed). Secondary analyses of the IES-R, that is, Avoidance and Arousal subscales, and overall IES-R score found the cognitive load group had significantly higher Arousal scores than the control group, t(40)=2.10, p=.02 (1-tailed), g=.63. There was a nonsignificant trend for the HVPT group to have lower scores on the Avoidance subscale, t(41)=−1.86, p=.07, and lower scores on the total IES-R score than the control group, t(41)=−1.83, p=.08. Memory performance Descriptive statistics are reported in Table 1. Planned comparisons demonstrated that the cognitive load group were significantly less likely to correctly order the selected sequence of events in the film than the control condition, t(38)=2.18, p=.02, (1-tailed), g=.68, although no difference was observed between the HVPT and control group on this measure, t(40)=1.11, p=.14 (1-tailed). There were no significant differences between groups in relation to cued recall performance, F(2, 62)=1.95, p=.15. Recognition data were analysed by computing d′ and c as set out by Stanislaw and Todorov (1999). The d′ statistic provides a measure of sensitivity, taking into account guessing and any resultant false-positives, and is independent of response bias (c). In the present analyses, d′ reflects the degree to which a participant was able to determine whether they had seen what was described in the statement or not (e.g., ‘The rapist wore a shirt with a collar’), with higher scores indicating better memory performance. d′ can range between +∞ and −∞, and a value of 0 indicates an inability to distinguish whether the statement described something previously seen from something new (unseen); ‘larger values indicate a correspondingly greater ability’ to distinguish between the two ( Stanislaw & Todorov, 1999, p. 139). In the present study, mean values of d′ were low, ranging between .23 and .46 across the groups, indicating low sensitivity, that is, participants found the recognition memory task difficult. A′ was also computed in the event the assumptions of normality and equivalent variance of d′ were not met ( Stanislaw & Todorov, 1999); however, these analyses replicated the findings obtained using d′, hence the latter are reported. The groups had comparable performance on the recognition task, F(2, 60)=.62, p=.54, and similarly did not differ in terms of biased responding, F(2, 60)=2.20, p=.12, with participants favouring ‘True’ responses. Working memory, thought suppression and intrusive thoughts distress Contrary to prediction, working memory ability (OSPAN) was not correlated with intrusions immediately after the film or at follow-up (IES-intrusion), p's ranging between .46 and .58. After dummy coding to control for experimental group membership, the hypothesis that thought suppression and distress caused by immediate intrusions would mediate the relationship between initial distress from the film and intrusive experiences over the week (IES-intrusion) was tested. The procedures for testing mediation as set out by Baron and Kenny (1986) were followed, and the significance of the mediated effect (reported as a z score) was tested using the Aroian version of the Sobel test (see Baron & Kenny, 1986; Preacher & Hayes, 2004). As predicted, voluntary thought suppression mediated the relationship between distress from the film and later intrusions. Table 2 details two regressions summarising this effect. The first indicates the significant effect of the independent variable (film distress) on the dependent variable (later intrusions); the second indicates that controlling for the mediator (thought suppression) reduced the effect of film distress to nonsignificance, indicating full mediation, z=2.00, p=.04. As thought suppression was significantly correlated with the other IES subscales (i.e., avoidance and arousal reactions), it is important to test whether the mediating effect of thought suppression was specific to intrusion development over the week, or whether it was related to the other subscales of the IES. If the latter were observed, it would suggest the thought suppression was related to participants’ general maladjustment rather than intrusions per se. Accordingly, mediation analyses were conducted for the IES-Avoidance and IES-Arousal subscales. Interestingly, thought suppression did not act as a mediator in either case. When thought suppression was controlled, initial distress at the film was an independent predictor of avoidance behaviours at follow-up in the final model, F(1, 59)=4.53, p=.04, R2 change=.06 (film distress: β=.28, p=.04; thought suppression: β=.23, p=.08), and there was a trend for film distress to be a unique predictor of arousal reactions, F(1, 59)=3.94, p=.052, R2 change=.05 (film distress: β=.25, p=.052; thought suppression: β=.20, p=.11). Table 2. Summary of the mediating effect of thought suppression on the relationship between film distress and later intrusions (n=65) Variables β R2 Overall F Regression 1: film distress predicting later intrusions (IES-I) .31* .17 F(3, 61)=4.15* Regression 2: thought suppression and film distress predicting later intrusions (IES-I) Thought suppression .32* Film distress .18 .25 F(4, 59)=4.87** Note: *p<.05; **p<.01. IES-I=Impact of Events Scale-Intrusion subscale; group membership controlled (not reported), and only contributed a nonsignificant 8% of the variance in each regression). Table options A similar pattern of findings was observed when distress caused by immediate intrusions was tested for mediation. As indicated in Table 3, intrusion distress fully mediated the initial distress caused by the film and intrusions at follow-up, z=2.28, p=.02. Like thought suppression, this finding appeared to be specific to intrusions, with initial film distress, and not intrusion distress accounting for IES-avoidance scores, F(1, 59)=4.59, p=.04, R2 change=.06 (film distress: β=.28, p=.04; intrusion distress: β=.19, p=.15), and although the preconditions of mediation were met for predicting IES-Arousal scores, the mediated effect was not significant, z=1.89 p=.06, final regression model, F(1, 59)=2.80, p=.10, R2 change=.03 (film distress: β=.21, p=.10; intrusion distress: β=.28, p=.03). Table 3. Summary of the mediating effect of intrusion distress on the relationship between film distress and later intrusions (n=65) Variables β R2 Overall F Regression 1: film distress predicting later intrusions (IES-I) .31* .17 F(3, 61)=4.15* Regression 2: intrusion distress and film distress predicting later intrusions (IES-I) Intrusion distress .37** Film distress .15 .27 F(4, 59)=5.54*** Note: *p<.05; **p<.01; ***p<.001. IES-I=Impact of Events Scale-Intrusion subscale; Group membership controlled (not reported), and only contributed a nonsignificant 8% of the variance in each regression). Table options A final caution for these analyses is the problem of multicollinearity given that several variables were measured within relatively short periods of each other (i.e., film distress, thought suppression and intrusion distress), and it could be argued that rather than separate constructs being assessed, one measure (e.g., intrusion distress) simply served as a proxy for another (e.g., distress caused by the film). Although these variables were significantly correlated with one another (as required to meet the preconditions of mediation), these correlations were not so high to suggest that multicollinearity was a concern (r's were between .36 and .47, p's ranging from .001 to .004). Controlling for pre-film mood and post-traumatic stress symptoms There were no significant correlations between pre-film mood (as measured with the DASS-21) and intrusions in the 5 min monitoring phase or at 1-week follow-up (IES-intrusion), p's ranging from −.17 to .11. There were nonsignificant trends for prior post-traumatic stress symptom severity (PDS) to be positively correlated with intrusion duration in the monitoring phase r=.23 (p=.07), and with IES-intrusion scores at follow-up, r=.22 (p=.08). As stated earlier, the groups did not differ significantly on these variables, and controlling for PDS scores did not alter any findings pertaining to intrusions.