مجله رفتار درمانی و تجربی روانپزشکی

Abstract Background Intrusive reexperiencing in posttraumatic stress disorder (PTSD) is commonly triggered by stimuli with perceptual similarity to those present during the trauma. Information processing theories suggest that perceptual processing during the trauma and enhanced perceptual priming contribute to the easy triggering of intrusive memories by these cues. Methods Healthy volunteers (N = 51) watched neutral and trauma picture stories on a computer screen. Neutral objects that were unrelated to the content of the stories briefly appeared in the interval between the pictures. Dissociation and data-driven processing (as indicators of perceptual processing) and state anxiety during the stories were assessed with self-report questionnaires. After filler tasks, participants completed a blurred object identification task to assess priming and a recognition memory task. Intrusive memories were assessed with telephone interviews 2 weeks and 3 months later. Results Neutral objects were more strongly primed if they occurred in the context of trauma stories than if they occurred during neutral stories, although the effect size was only moderate (ηp2=.08)(ηp2=.08) and only significant when trauma stories were presented first. Regardless of story order, enhanced perceptual priming predicted intrusive memories at 2-week follow-up (N = 51), but not at 3 months (n = 40). Data-driven processing, dissociation and anxiety increases during the trauma stories also predicted intrusive memories. Enhanced perceptual priming and data-driven processing were associated with lower verbal intelligence. Limitations It is unclear to what extent these findings generalize to real-life traumatic events and whether they are specific to negative emotional events. Conclusions The results provide some support for the role of perceptual processing and perceptual priming in reexperiencing symptoms.

Introduction Intrusive memories are considered the hallmark symptom of posttraumatic stress disorder (PTSD). They commonly consist of relatively brief sensory impressions from the trauma (Hackmann, Ehlers, Speckens, & Clark, 2004; Reynolds & Brewin, 1999; Van der Kolk & Fisler, 1995), that are experienced as happening in the “here and now” rather than being a memory from the past (Hackmann et al., 2004; Michael, Ehlers, Halligan, & Clark, 2005), and are easily triggered by a wide range of internal and external cues (Southwick et al., 1993; Van der Kolk & Fisler, 1995). Intrusive memories and their triggers can include those that bear no meaningful relationship with the traumatic event but are only temporally associated (Ehlers & Clark, 2000; Ehlers et al., 2002; Speckens, Ehlers, Hackmann, Ruths, & Clark, 2007). The present experiment investigated two related processes that may be involved in the easy triggering of intrusive memories, perceptual processing during the trauma and perceptual priming. 1.1. Perceptual priming Priming is a type of implicit memory that is characterized by enhanced identification of previously seen stimuli (Schacter, 1992). Ehlers and Clark (2000) proposed that the easy triggering of intrusive memories by perceptually similar stimuli is, in part, a function of strong perceptual priming. They suggested that trauma survivors who acquire strong priming for stimuli that they encountered during the traumatic event have a reduced perceptual threshold for these stimuli. This makes it more likely that they detect potential triggers of intrusive memories in their environment. There is preliminary evidence for a role of perceptual priming in PTSD. Several studies found that after encoding trauma-related and neutral material, people with PTSD show greater perceptual priming for trauma-related material compared to controls in word-stem completion or perceptual identification tasks (e.g., Amir, Leiner, & Bomyea, 2010; Ehring & Ehlers, 2011; Michael, Ehlers, & Halligan, 2005). Priming predicted PTSD severity 6 months later (Ehring & Ehlers, 2011; Michael, Ehlers, & Halligan, 2005). Kleim, Ehring, and Ehlers (2012) further found that trauma survivors with PTSD show perceptual processing advantages in identifying blurred trauma-related pictures compared to those without PTSD. Experimental analogue studies provided initial support for the role of perceptual priming in the development of intrusive trauma memories (Arntz, de Groot, & Kindt, 2005; Ehlers, Mauchnik, & Handley, in press; Ehlers, Michael, Chen, Payne, & Shan, 2006; Michael & Ehlers, 2007). For example, a paradigm developed by Ehlers et al. (2006) investigates priming for neutral objects that are perceived just before something “traumatic” happens. Participants watch a series of “traumatic” and neutral picture stories, each comprising three pictures. The content of the first picture is neutral, introduces the main character of the story, and contains neutral objects for which priming was later measured. The plot then unfolds in the second picture either in a traumatic or in a neutral way. The last picture depicts the outcome of the story. With this paradigm, neutral stimuli that preceded a “traumatic” event showed enhanced perceptual priming and predicted intrusive memories (Ehlers et al., 2006; Michael & Ehlers, 2007). 1.2. Perceptual processing during trauma Information processing models of PTSD suggest that strong encoding of perceptual information and relatively weak encoding of contextual information during trauma predicts subsequent reexperiencing symptoms (Brewin, Dalgleish, & Joseph, 1996; Brewin, Gregory, Lipton, & Burgess, 2010; Ehlers & Clark, 2000). Two related ways of processing that may facilitate strong encoding of perceptual information have been investigated in PTSD; (i) data-driven processing (Roediger, 1990) refers to encoding that focuses on the surface level features of a situation (i.e., sensory details such as colours and shapes) rather than its meaning (Ehlers & Clark, 2000), and (ii) dissociation is a complex concept involving a lack of integration of subjective experiences including depersonalization, derealization, altered time perception and emotional numbing (e.g., Van der Kolk & Fisler, 1995). It has been suggested that dissociation during trauma may in part predict PTSD because, like data-driven processing, it leads to preferential encoding of perceptual information (Ehlers & Clark, 2000). It has further been suggested that dissociation during a traumatic event decreases focal attention, thereby interfering with meaningful processing of the traumatic event and promoting a nonverbal, perceptual processing style (Brewin et al., 1996; Siegel, 1995). Clinical and analogue studies found that data-driven processing and dissociation predict later reexperiencing and PTSD (e.g., Ehring, Ehlers, & Glucksman, 2008; Halligan, Michael, Clark, & Ehlers, 2003; Holmes, Brewin, & Hennessy, 2004). It is not yet clear by which mechanisms these forms of processing during the trauma contribute to PTSD. One possible pathway is perceptual priming (Ehlers & Clark, 2000) which is thought to rely heavily on perceptual operations and unconscious processes and should therefore benefit from encoding styles that favour perceptual information (Roediger, 1990). In line with this hypothesis, data-driven processing and dissociation correlated with priming in clinical and analogue studies (Ehlers et al., 2006; Lyttle, Dorahy, Hanna, & Huntjens, 2010; Michael & Ehlers, 2007). 1.3. Aims of the study A new version of Ehlers et al.'s (2006) paradigm to investigate perceptual priming for objects from traumatic contexts and its relationship with processing styles and intrusions was developed to address some limitations of the earlier paradigm. First, the number of primed neutral objects was increased. Second, objects were no longer presented within the slides of the picture stories, but interspersed in the intervals between story pictures so that they could be counterbalanced across trauma and neutral stories. The study investigated the following hypotheses: Hypothesis 1. Neutral stimuli are more strongly primed if they occur in a traumatic context than if they occur in a neutral context. Hypothesis 2. Enhanced perceptual priming for objects previously seen in a traumatic context predicts subsequent intrusive memories. Hypothesis 3. Data-driven processing and state dissociation during traumatic stories predict subsequent intrusive memories. Hypothesis 4. Data-driven processing and state dissociation predict enhanced perceptual priming for trauma-related stimuli. In addition, we assessed recognition memory for objects previously seen in traumatic and neutral contexts to check the possible influence of explicit memory on priming effects. On the basis of previous results with a similar paradigm (Ehlers et al., 2006; Michael & Ehlers, 2007), we did not expect an influence of story context on recognition memory. We also assessed state anxiety and verbal intelligence to explore their relationship with priming and intrusive memories.

Results 3.1. Validity of the picture stories As expected and in line with pilot study 3, participants perceived trauma stories as significantly more unpleasant (M = −6.4, SD = 2.7 vs. M = 5.4, SD = 3.2; F(1, 50) = 408.0, p < .001, ηp2=.89ηp2=.89) and activating (M = 3.6, SD = 3.7 vs. M = −4.2, SD = 3.7; F(1, 50) = 175.8, p < .001, ηp2=.78ηp2=.78) than neutral stories. In addition, participants reported significantly greater state anxiety during traumatic picture stories, M = 35.9, SD = 9.2, compared to neutral stories, M = 29.3, SD = 7.8; F(1, 50) = 17.1, p < .001, ηp2=.31ηp2=.31, and baseline, M = 30.4, SD = 6.4; F(1, 50) = 22.56, p < .001, whereas state anxiety during neutral stories did not differ from baseline, F(1, 46) = 1.53, p = .22. Participants engaged significantly more in data-driven processing (M = 0.9, SD = 0.6 vs. M = 0.6, SD = 0.4; F(1,50) = 9.4, p = .004, ηp2=.16ηp2=.16) and state dissociation (M = 0.3, SD = 0.4 vs. M = 0.1, SD = 0.2; F(1,50) = 10.4, p = .002, ηp2=.18ηp2=.18) during trauma stories than during neutral stories. 3.2. Perceptual priming task 3.2.1. General priming effect To check for a general priming effect, a repeated measures ANOVA compared identification rates for all objects that participants had previously seen interspersed into the picture stories with those for new objects. Primed objects were more readily identified, M = .60, SD = .14, than unprimed objects, M = .40, SD = .15, F(1, 50) = 101.0, p < .001, ηp2=.70ηp2=.70. 3.2.2. Effects of trauma context on priming (Hypothesis 1) Table 2 shows the effects of story context on object identification in the blurred picture task. In line with Hypothesis 1, objects previously seen in the context of trauma stories were identified with greater probability than objects previously seen in the context of neutral stories. Table 2. Results of priming and recognition tests. Memory task Picture stories Statistic Trauma Neutral F(1,50) p ηp2ηp2 Priming task Identification score .63 (.16) .58 (.15) 4.40 .042 .080 Recognition task Sensitivity (d′) 1.40 (.96) 1.50 (1.01) .74 .395 .014 Response bias (c) .03 (.51) .05 (.47) .11 .740 .002 Table options 3.2.2.1. Further analyses Twenty-six participants (51%) reported having tried to actively recall the objects during the priming task; identification rates active recall: Mtrauma = .66, SDtrauma = .15; Mneutral = .65, SDneutral = .13; no active recall: Mtrauma = .60, SDtrauma = .18; Mneutral = .52, SDneutral = .14. A 2 × 2 ANOVA with story context (trauma vs. neutral) as the within factor and active recall (yes vs. no) as the between factor, showed a main effect for active recall, F(1, 49) = 6.41, p = .015, ηp2=.12ηp2=.12, indicating that participants who used the active recall strategy had significantly higher identification rates than those who did not. Further, there was a trend for an interaction Story Context × Active Recall, F(1, 49) = 2.61, p = .113, ηp2=.05ηp2=.05. Bonferroni-controlled post-hoc analyses revealed that participants who did not actively recall the objects identified objects from trauma stories significantly more often than objects from neutral stories, F(1, 49) = 6.9, p = .011, ηp2=.12ηp2=.12, whereas no significant difference was found in the active-recall group, F(1, 49) = .15, p = .703, ηp2=.00ηp2=.00. Therefore, active recall was controlled for in the following analyses where appropriate. To test whether order of story block influenced identification rates, a mixed 2 × 2 ANOVA with order (trauma stories vs. neutral stories presented first) as the between factor and story context (trauma vs. neutral) as the within factor and was computed. There was no main effect for order, F(1, 49) = .005, p = .946, ηp2=.00ηp2=.00; but the main effect of story context, F(1, 49) = 5.46, p = .047, ηp2=.078ηp2=.078, was qualified by a significant Order × Story Context interaction, F(1, 49) = 5.46, p = .024, ηp2=.10ηp2=.10. Post hoc analyses revealed that the EPP effect was only significant for participants who saw the trauma stories first, F(1,49) = 10.16, p = .001, ηp2=.17ηp2=.17; and could not be demonstrated for those who saw the neutral stories first, F(1,49) = .042, p = . 838, ηp2=.00ηp2=.00. Therefore, order of story block was controlled in the following analyses where appropriate. 3.3. Recognition task Table 2 shows the results of the object recognition task. As expected, there were no main effects or interactions of story context for sensitivity of discrimination (d′) or response bias (c). Equally there were no differences in hits (both 70%) or false alarm rates (25–27%) for the two story contexts. Additional analyses found no main effects or interactions for order of story block. 3.4. Predictors of intrusive memories At 2 weeks 10 participants (19.6% of 51) reported intrusive memories of images from the picture stories in the preceding week with mean distress ratings of M = 5.7, SD = 2.4, range = 2–9; mean intrusion frequency M = .43, SD = .99, range = 0–4. At 3 months, 5 participants (12.5% of 40) reported intrusions with mean distress ratings of M = 5.5, SD = 1.9, range = 4–8: mean intrusion frequency M = .5, SD = 1.45, range = 0–7. In addition, a few participants reported intrusions of the interspersed neutral objects, but these were not included in the correlational analyses as these may have been influenced by the two memory tests ( Krans, Näring, Holmes, & Becker, 2009). In accordance with Hypothesis 2, EPP predicted intrusive memories in the second week after the experiment, r = .29, p = .04. This relationship remained significant when controlling for order of story block, r = .29, p = .04 (trauma stories first, r = .32; neutral stories first, r = .30), and active recall, r = .29, p = .04. However, EPP did not predict intrusions at 3 months, r = .02, p = .91, n = 40. As expected, recognition sensitivity (d′) for objects from trauma stories did not correlate with intrusive memories reported at 2 weeks, r = −.05, p = .710. There was a trend for a negative relationship with intrusions at 3 months, r = −.28, p = .076. In line with Hypothesis 3, both data-driven processing, r = .33, p = .018, and state dissociation during trauma stories, r = .53, p < .001, predicted the frequency of intrusive memories in the second week after the experiment. Data-driven processing and dissociation no longer significantly predicted intrusive memories at 3 months (r's < .21, p's > .20). Increases in state anxiety with the trauma stories also predicted intrusive memories at 2 weeks, r = .36, p = .012, but not at 3 months, r = −.07, p = .68. There were no significant correlations with intelligence. A multiple regression analysis tested the relative contribution of EPP, recognition sensitivity, data-driven processing and dissociation, pleasantness and arousal ratings, anxiety response, and verbal intelligence (NART) to the prediction of intrusive memories at 2 weeks. The predictors were entered simultaneously and explained 40.6% of the variance, R = .637, F(8, 45) = 3.16, p = .008. EPP, β = .30, p = .041, explained unique variance; there was a trend for intelligence, β = .24, p = .098, but no unique effects of anxiety, β = .24, p = .124, arousal β = −.22, p = .157, pleasantness, β = −.113, p = .457, and recognition sensitivity, β = .01, p = .942. Data-driven processing, β = .247, p = .181, and dissociation, β = .249, p = .177, which are conceptually related measures, correlated substantially, r = .59, and did not explain unique variance when they were both entered together into the multiple regression. When they were entered individually, they each explained unique variance over and above the other predictors, data-driven processing, β = .374, p = .023, and dissociation, β = .374, p = .023. 3.5. Predictors of enhanced perceptual priming In contrast to Hypothesis 4, EPP did not correlate with self-reported data-driven processing, r = −.03, p = .836, nor with state dissociation, r = .19, p = .182. Controlling for order of story block and active recall did not affect these relationships. Other reactions during the trauma stories, including increases in state anxiety, perceived arousal and pleasantness did not correlate with EPP either (see Table 3). The only variable that predicted EPP was lower verbal intelligence as measured by the NART, r = −.30, p = .034. Lower intelligence also correlated with data-driven processing, r = −.31, p = .027. Table 3. Correlations between variables in multiple regression analysis. 2 3 4 5 6 7 8 9 1. Intrusions at 2 weeks .29* −.05 .33* .53*** .36* .05 .28 .03 2. Enhanced Priming .14 −.03 .19 .04 −.05 .06 −.30* 3. Recognition sensitivity −.20 −.13 −.15 −.05 .12 .02 4. Data-driven processing .59*** .30* .29* −.40** −.31* 5. Dissociation .44** .24 −.42** −.20 6. Anxiety .45** −.31* −.06 7. Arousal −.38** −.14 8. Pleasantness .16 9. Verbal Intelligence *p < .05, **p < .01, ***p < .001.