Arresting visuospatial stimulation is insufficient to disrupt analogue traumatic intrusions

Thomas Meyer; Chris R. Brewin; John A. King; Desiree Nijmeijer; Marcella L. Woud; Eni S. Becker

doi:10.1371/journal.pone.0228416

Abstract

Intrusive memories are a core symptom of Post-Traumatic Stress Disorder (PTSD). A growing body of analogue studies using trauma films suggest that carrying out specific demanding tasks (e.g., playing the video game Tetris, pattern tapping) after the analogue trauma can reduce intrusive memories. To examine the mechanism behind this effect, we tested whether mere engagement with attention-grabbing and interesting visual stimuli disrupts intrusive memories, and whether this depends on working memory resources and/or the concurrent activation of trauma film memories. In a total sample of 234 healthy participants, we compared no-task control conditions to a perceptual rating task with visually arresting video clips (i.e., non-emotional, complex, moving displays), to a less arresting task with non-moving, blurred pictures (Study 1), and to more demanding imagery tasks with and without repetitive reminders of the trauma film (Study 2). Generally, we found moderate to strong evidence that none of the conditions lead to differences in intrusion frequency. Moreover, our data suggest that intrusive memories were neither related to individual differences in working memory capacity (i.e., operation span performance; Study 1), nor to the degree of engagement with a visuospatial task (i.e., one-week recognition performance; Study 2). Taken together, our findings suggest that the boundary conditions for successful interference with traumatic intrusions may be more complex and subtle than assumed. Future studies may want to test the role of prediction errors during (re-)consolidation, deliberate efforts to suppress thoughts, or the compatibility of the task demands with the individual’s skills.

Citation: Meyer T, Brewin CR, King JA, Nijmeijer D, Woud ML, Becker ES (2020) Arresting visuospatial stimulation is insufficient to disrupt analogue traumatic intrusions. PLoS ONE 15(2): e0228416. https://doi.org/10.1371/journal.pone.0228416

Editor: Ilan Harpaz-Rotem, Yale University, UNITED STATES

Received: July 29, 2019; Accepted: January 14, 2020; Published: February 3, 2020

Copyright: © 2020 Meyer et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The analysed dataset can be obtained via the Open Science Framework using the following link: https://osf.io/f38cd/.

Funding: The authors received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Introduction

A large number of individuals are exposed to potentially traumatic events at some point in their life, such as war, life-threatening accidents, or being held captive [1]. Many trauma victims continue to suffer from recurrent memories that intrude into consciousness in the form of flashbacks, or intrusions. Intrusive memories are a core symptom of post-traumatic stress disorder (PTSD), a disorder with devastating consequences for functioning and well-being [2, 3]. This makes intrusive memories a key target in efforts to develop early interventions for trauma victims.

Various analogue studies with trauma films have shown that carrying out specific demanding tasks (e.g., playing the video game Tetris, pattern tapping, N-back tasks) during or after film viewing can reduce involuntary film memories in the following weeks (for reviews, see [4, 5, 6]). For instance, studies have found reductions in intrusive memories in participants who played Tetris immediately after a viewing a trauma film [7], after 30 min (e.g., [8, 9]), or after 4 h [10] The fact that tasks administered after film viewing can be effective suggests that traumatic memories are sensitive to disruption while they undergo the processes of consolidation or reconsolidation/re-encoding (i.e., during and shortly after formation or activation of the memory trace), in line with the view that newly formed memories are prone to retroactive interference for up to several hours [11, 12]. In parallel to these analogue studies, accumulating evidence suggests that already consolidated distressing autobiographical memories can lose their vividness and emotionality when the individual repeatedly executes guided eye movements whilst remembering [13, 14]. Based on these findings, many clinicians and researchers are hopeful that a disruptive effect on intrusive trauma memories can be exploited in effective treatment and preventive interventions for trauma victims (e.g., [15, 16, 17]).

To date, however, the mechanisms by which dual tasks administered after trauma exposure can reduce intrusive trauma memories remain ill-understood. A widely held view is that interference with intrusive memories depends on a task’s ability to tap into limited cognitive resources that are required for trauma-related imagery and for the consolidation of the traumatic memories [4, 5]. Although there have been debates about the role of modality-specific working memory components that may be linked more specifically to imagery (e.g., the visuospatial sketchpad; [18]), many scholars tend to concur that imposing high levels of working memory load seems critical for the dual-task interference effects (e.g., [8, 14, 19, 20]). Indeed, several research lines highlight the role of working memory resources in the development of intrusive memories. For instance, a higher working memory capacity (WMC)–i.e. a larger amount of goal-relevant information that can be held active in working memory in the presence of distractors [21]–has been linked to fewer intrusive memories of negative personal experiences [22], or a better ability to suppress intrusive thoughts [23, 24]. Moreover, some studies suggest that WMC modulates the benefits of dual-task interference on the development of intrusive memories. For instance, Gunter and Bodner [25] found that three different visual or spatial tasks were able to reduce memory vividness and emotionality, and more strongly so in participants who had a low WMC, as measured by means of reading span.

However, some studies did not find an association between intrusive memories and WMC (e.g., as measured on an operation span task; [26], cf. [27]), whereas others could not find a dose-response relationship between the cognitive load of a dual task and benefits in the reduction of intrusions [28]. In addition, some studies have failed to replicate effects of visuospatial tasks on intrusions, or to link WM taxation to intrusion number and quality. For instance, Asselbergs et al. [29] found no experimental effects for two novel visuospatial gaming apps that required spatial planning and decision making. As a potential explanation, both of these apps may have required fewer visuospatial resources than tasks that have been shown to reduce intrusions such as Tetris. However, the authors matched one of their apps to Tetris in terms of its ability to interfere with a simple dual reaction time task, meaning that the null findings cannot be attributed to a general lack of WM taxation. In a similar vein, van Schie et al. [30] found that performing eye movements or counting whilst remembering the hotspot of a traumatic film reduces intrusive memories, but this effect could not be replicated in two out of three experiments. In addition, in all three experiments, they found no correlation between WM taxation (i.e., interference with dual reaction time task) and a reduction in intrusion vividness or emotionality. Together, these findings point out that there are disparities in the effectiveness of visuospatial tasks to reduce intrusive memories that cannot be entirely attributed to WM taxation and/or individual differences in WMC.

An alternative possibility to the working memory hypothesis is that the beneficial dual task effects are driven by attentional redirection. Accordingly, any task that attracts attention and diverts it away from the traumatic experience might be able to interfere with the consolidation of these distressing memories. Although some studies have addressed this idea (also coined "distraction" hypothesis; e.g., [31]), they compared demanding tasks differing in modality, rather than manipulating attention-grabbing characteristics within the same modality. Still, suggestive evidence comes from Kavanagh et al. [14], who found decreased vividness and emotionality for emotional memories when participants were stimulated with “visual noise” (i.e., a screen of flickering squares) during recall, as compared to no-task control. Although this reduction was not as effective as a more taxing eye-movement condition, it provides tentative evidence that a relatively passive viewing condition can reduce the emotional impact of traumatic memories. Notably, this would have important implications for theoretical models of intrusive memories, as well as practical implications for cheap and low-demanding interventions for trauma victims (e.g., in emergency care settings).

With these considerations in mind, we explored in two studies whether perceptually arresting tasks with high levels of visuospatial information have the potential to reduce intrusive memories of traumatic film fragments in the consolidation phase immediately following film viewing. Specifically, we investigated whether simple rating tasks with different degrees of visuospatial complexity (Study 1) or active engagement with complex visual stimuli via imagery (Study 2) would reduce the formation of intrusive memories, as compared with no-task control conditions.

Study 1

In Study 1, we used traumatic film fragments to induce intrusive memories in healthy individuals and compared the effects of three conditions, differing in the number and complexity of visual items involved in a task given directly after film viewing. In particular, we employed a highly visually arresting (H-VA) task that exposed participants to high levels of visuospatial information (i.e., arresting and attention-grabbing displays), but relatively low working memory load. As control conditions, we used a similar low visually arresting (L-VA) task with little visuospatial information, and a no-task resting condition. We hypothesized that the group performing the H-VA task would report fewer intrusive memories than both the L-VA group and the no-task condition. We also assessed individual differences in WMC using an operation span (OSPAN; [32]) task at baseline, in order to test whether a visuospatial stimulation would be particularly beneficial to individuals with lower WMC. Accordingly, we expected WMC to be a better predictor of fewer intrusive memories in the H-VA condition.

Materials and methods

Participants

One-hundred twenty healthy participants (40 per condition; 82.5% women) aged between 18 and 34 years (M = 21.8, SD = 2.9) completed Study 1. On the basis of previous reported effect sizes (e.g., [14] reported η2 = .083 for memory emotionality comparing visual noise to a control condition], we determined a priori that a sample size of 120 would be adequate to detect an effect size of f = 0.30 in an omnibus ANOVA with 3 groups with (η2p = .083, α = .05) with a power (1 – β) > .80 (actual power = .84). Eligibility for participation was established using a self-report screening listing the following exclusion criteria: psychological or psychiatric complaints in the past 2 years, severe neurological condition or injury (e.g., epilepsy), pregnancy, current psychoactive medication, alcohol or drug abuse, and having experienced serious traumatic events. Participants were furthermore required to have normal or corrected-to-normal vision and a good understanding of Dutch. All participants gave written informed consent prior to inclusion. Study 1 and Study 2 were approved by the research ethics committee of the Behavioural Science Institute, Radboud University Nijmegen (ECSW2015-1105-310).

Trauma films and PTSD-analogue symptoms

Participants viewed a compilation of traumatic video fragments lasting about 14 min. The fragments depicted medical surgeries, fatal accidents, and war-related atrocities [33]. Participants were instructed to watch the scenes as if they were a witnessing bystander. In order to enhance immersion, reduce distraction, and create a sense of immobility, participants were provided with headphones and asked to place their head on a chin rest at 60 cm distance from the screen, while being instructed to move as little as possible during film viewing [34]. During film viewing, the experimenter remained in the room in order to monitor the participant’s well-being. To avoid memory modulation through social feedback (e.g., [35]), the experimenter was seated outside the participant’s field of view and did not engage in discussions about the content or emotionality of the fragments.

To assess intrusive memories related to the film fragments, participants were supplied with a one-week diary ([8]; translated to Dutch), in which they recorded sudden, involuntary memories as soon as they occurred, or the absence of involuntary memories at least twice a day. For each memory, they were also asked to indicate the content and trigger (for verification), whether it was predominantly based on images, thoughts, or both. Finally, they rated the subjective distress caused by each intrusion on an eleven-point scale (range: 0 = not at all to 10 = extremely). Intrusion frequencies were log-transformed prior to the analyses to correct for their typical right-skewed distribution. Untransformed descriptive statistics are provided in the results section for better readability. Distress scores were averaged across memories, whereby zero was entered if no intrusion had occurred.

We additionally used the Impact of Events Scale (IES; [36]) as a retrospective measure of intrusion-related distress and overall PTSD-analogue symptoms at one-week follow-up. The instructions of the questionnaire were adapted to measure symptoms specifically related to the film fragments. Its Intrusion Symptoms subscale (7 items; e.g., I had waves of strong feelings about it; α = .79) was of particular interest, while the total score (including a subscale on avoidance symptoms; α = .83) served as a measure of overall PTSD symptoms.

Visuospatial rating task

A simple visuospatial rating task was devised to expose participants either to low visually arresting (L-VA) or to high visually arresting (H-VA) displays, depending on their condition, while keeping executive demands to a minimum. For this purpose, participants were instructed to attentively watch a series of images during 15 blocks lasting 30 s. Each block was followed by valence and arousal ratings using Self-Assessment Manikin (SAM; [37]) scales ranging from 1 to 9. Furthermore, participants completed two 100 mm visual analogue scales (VAS) asking ‘How impressive or overwhelming did you find these images?’ and ‘How fascinating, interesting, or original did you find these images?’ (0 = not at all, 100 = very much), in order to assess the ability of each clip to attract the viewer’s attention. For the analyses, each type of rating were averaged per participant across all stimulation periods.

In the H-VA condition, the series of images consisted of 15 video fragments depicting complex abstract moving displays with many visual details (e.g., landscapes built from changing geometrical shapes, viewed from changing perspectives). The fragments were excerpts from 3D fractal animation artworks by Jérémie Brunet ([38]; "Like in a dream II", "Far away", "Weird planet", "Yellow box"). In the L-VA condition, these clips were replaced with 6 still pictures, each shown for 5 s. In order to create a stimulation condition with comparable low-level features (e.g., colours, luminosity, low frequency spatial patterns), the pictures were screenshots of the 3D animations used in the H-VA condition, heavily blurred by means of a 100 px Gaussian filter. In the third, no-task condition, participants were asked to sit quietly for a comparable amount of time.

Operation span

To assess individual differences in WMC, we used an automated version [32] of the working operation span task (OSPAN; [39]). This task requires participants to remember a sequence of letter stimuli, next to verifying a mathematical equation before seeing each letter. In 4 practice trials, participants initially learned to memorize and recite short letter sequences by selecting the letters in correct order from a 4×3 letter matrix using the computer mouse. Then, they performed 15 math solving trials, in which participants were asked to solve a simple equation, and to verify a solution that was provided on the next screen. Mean reaction time on these trials plus 2.5 SD subsequently served as a time limit for math problems in the remainder of the task. In the final practice phase, participants performed 3 trials of letter recall with a set size of 2, whereby each letter was preceded by a math problem. Finally, the real testing phase continued with set sizes increasing from 3 to 7, with 3 sets of each size. An absolute OSPAN score was calculated as the sum of the sizes of all perfectly recalled sets. Data from three participants were removed because they remembered exceptionally few letters (<20) or committed too many math errors (>40). On average, all other participant remembered 60.1 words (SD = 10.3) and committed 4.1 math errors (SD = 2.6).

Affective responses

Mood responses to trauma film viewing were monitored using the Dutch Positive and Negative Affect Schedule (PANAS), state version [40], consisting of two 10-item subscales for negative affect (NA; αs > .80) and positive affect (PA; αs > .84). The PANAS ratings were complemented by four 100 mm visual analogue scales (VAS) measuring current levels of specific negative emotions (sad, fearful, shocked, angry).

Individual differences in involuntary cognition

To quantify baseline individual differences in the frequency of involuntary memories, we administered the Involuntary Autobiographical Memory Inventory in English (IAMI; [41]). The IAMI consists of two 10-item subscales measuring the frequency of involuntary thoughts on 5-point scales (0 = never, 4 = once an hour or more) separately for future (α = .90) and past events (α = .90; total score α = .94). The IAMI is supplemented by two similar 5-item subscales measuring voluntary thoughts about future (α = .81) and past events (α = .88; total score α = .91).

Procedure

Participants were invited to two individual laboratory sessions separated by a one-week interval, during which they completed the intrusion diary. In each session, they were seated in front of a computer screen in a sound-isolated testing cubicle. They initially completed the IAMI and performed the OSPAN task. Next, they viewed the trauma film fragments, preceded and followed by current mood assessments. Directly afterwards, they were randomly assigned to either the H-VA, L-VA, or no-task condition and performed the respective task. Next, they were provided with the diary and given extensive instructions on its use. Upon their return to the laboratory one week later, participants handed in their diary and the experimenter briefly checked all responses for readability and adherence to the instructions. Participants then completed the IES. Finally, exit questions assessed participants’ own judgement of how accurate their diary was, as well as demand characteristics. Finally, they were debriefed, compensated, and dismissed.

Statistical analyses

The main hypotheses were tested by means of ANOVA and multiple regression analyses. In case of violations of the sphericity assumption, we report Greenhouse-Geisser epsilon and corrected p-values along with uncorrected degrees of freedom. Alpha was set at 0.05 (two-tailed) for all tests. Standard frequentist models were computed in SPSS (version 25) and complemented with Bayes factors (BF) computed in JASP (version 0.9; [42]) using default priors. BF₁₀ reflects the likelihood of the alternative (H1) over the null hypothesis (H0) given the data, whereby the prior assumption was that H1 and H0 are equally likely. Main effects were tested against a null with participants as a sole predictor. Interaction effects were tested against models including the respective main or lower order interaction effects. Evidence is conventionally considered “anecdotal” with BFs > 1, “moderate” with BF > 3, “strong” with BF > 10, and “very strong” with BF > 30. Either BF₁₀ or BF₀₁ is reported, whichever is greater than 1. To render extreme BFs more readable, we provide natural log-transformed values in some analyses. The analysed dataset can be obtained via the Open Science Framework using the following link: https://osf.io/f38cd/.

Results

Baseline differences between conditions

Participants in the three experimental conditions did not differ in mean age, F (2,117) = 0.62, p = .54, BF₀₁ = 10.1, or sex distribution, χ² (2) = 0.3, p = .84, BF₀₁ = 7.6. The groups also did not differ in mean OSPAN scores (overall M = 44.2; SD = 15.0), F (2,114) = 0.85, p = .43, BF₀₁ = 6.1, or on IAMI total scores (overall M = 2.1; SD = 0.7), F (2,117) = 0.16, p = .85, BF₀₁ = 11.0.

Affective responses to the trauma films

A 2 (Time: pre-film, post-film) by 3 (Condition) mixed ANOVA revealed a main effect of Time for NA, F (1,117) = 96.47, p < .001, η²p = .45, log(BF₁₀) = 32.5 with scores increasing in response to film viewing from 12.5 (SE = 0.3) to 17.4 points (SE = 0.6). Similarly, PA decreased with time, F (1,117) = 78.78, p < .001, η²p = .40, log(BF₁₀) = 27.3, from 27.5 (SE = 0.5) to 23.1 points (SE = 0.6). Also scores on each of the four VAS scores increased with time, all Fs > 37.94, ps < .001, η²ps > .24, log(BF₁₀) > 13.7 (anxious: M_difference = 13.2, SD = 23.5; shocked: M_difference = 36.8, SD = 29.6; angry: M_difference = 13.0, SD = 21.6; sad: M_difference = 20.7, SD = 24.4). For none of the affective responses were there any effects involving condition, Fs < 1.67, p > .19, η²p < .03, BFs₀₁ > 2.0.

Visuospatial rating task

Table 1 summarizes the mean subjective ratings from the visual rating task, separately for the H-VA and L-VA conditions (note that these data were not collected in the resting condition). As expected, participants in the H-VA condition rated the visual displays as visually more impressive and more interesting, compared to participants in the L-VA condition. Furthermore, there were no differences between conditions in the valence ratings. Participants in the L-VA condition tended to rate the displays as slightly more arousing, although the corresponding BF is small, providing only “anecdotal” evidence (see Table 1). Notably, the H-VA group rated the stimuli not only as more impressive and interesting on average, but also used a wider range of scores for the individual stimuli than the L-VA group, both for impressiveness (Range M_Difference = 20.8; SE = 5.4), t(78) = 3.83, p < .001, d = 0.86, BF₁₀ = 95.89, and for interest (Range M_Difference = 17.0; SE = 4.9), t(78) = 3.83, p < .001, d = 0.78, BF₁₀ = 35.97. Consequently, when examining the maximum rather than averages scores across stimuli, the mean differences between the conditions become even more pronounced for both impressiveness (H-VA: M = 74.4; SD = 23.1; L-VA: M = 48.0 SD = 28.4), t(78) = 4.55, p < .001, d = 1.02, BF₁₀ = 934.92, and interest (H-VA: M = 77.6; SD = 15.0; L-VA: M = 57.6 SD = 28.7), t(78) = 3.89, p < .001, d = 0.87, BF₁₀ = 117.66.

Download:

Table 1. Mean ratings of the visual displays in the H-VA and L-VA condition.

https://doi.org/10.1371/journal.pone.0228416.t001

PTSD analogue symptoms

Means, standard deviations, as well as comparison statistics for intrusive memories and related symptoms are summarized in Table 2. As can be seen, our main analyses on weekly intrusion numbers indicated no differences between the three groups, and there also were no differences in other intrusion-related symptoms. Instead, we found moderate to strong support for the null hypotheses.

Download:

Table 2. PTSD analogue symptoms per experimental condition in Study 1.

https://doi.org/10.1371/journal.pone.0228416.t002

In addition to these outcomes, we explored potential group differences in intrusion development over the one-week period by subjecting log-transformed intrusive memories to a 7 (Time: 7 days) by 3 (Condition) mixed ANOVA (see Fig 1). There only was a Time main effect, F (6,702) = 33.57, p < .001, η²p = .22, log(BF₁₀) = 75.2. Meanwhile, the data strongly favoured the absence of a main effect for Condition, F (2,117) = 0.39, p = .68, η²p = .01, BF₀₁ = 19.3, as well as an absence of a Condition by Time interaction, F (12,702) = 0.94, p = .50, η²p = .01, BF₀₁ = 174.2. When entering IAMI total scores and Negative Affect increase as additional covariates (i.e., factors potentially accounting for individual differences in intrusions), additional main effects emerged only for IAMI total scores, F (1,115) = 12.44, p = .001, η²p = .10, BF₁₀ = 200.8, and NA increase, F (1,115) = 5.94, p = .016, η²p = .05, BF₁₀ = 11.8.

Download:

Fig 1. Intrusive memories (log-transformed) per condition per day.

Error bars indicate 95% confidence intervals.

https://doi.org/10.1371/journal.pone.0228416.g001

OSPAN

To test for linear associations between OSPAN performance and intrusive memories and potential moderation effects of our experimental manipulation, we performed hierarchical linear regression analyses with three steps. First, OSPAN scores were entered as predictors of intrusion symptoms, followed by two dummy variables representing the L-VA and H-VA groups, respectively (i.e., no-task served as the reference group). In the third and final step, we entered interaction terms for L-VA and H-VA (i.e., z-transformed OSPAN scores × group dummy).

The first step did not yield any model explaining a significant proportion of variance, r²s < .02, Fs (1,115) < 2.2, ps > .14, BF₀₁ of coefficients ranging between 1.90 and 5.08. In line with the group analyses in the previous section, the group dummy variables did not add any explained variance either, r² changes < .032, Fs (2,113) < 1.9, ps > .16, BF₀₁ range: 1.00–8.45. Finally, adding the two group by OSPAN score interaction terms also failed to explain additional variance, r² changes < .046, Fs (2,111) < 1.5, ps > .23, with BFs₀₁ 1.22 and 3.39 in favour of the null hypotheses and largely in the “anecdotal” range. We additionally performed similar regression models to directly compare the H-VA and the L-VA groups (H-VA serving as the reference group). These analyses did not reveal any additional effects.

Demand effects and diary accuracy

On average, participants in all three conditions did not assume that performing a visuospatial task following viewing of the trauma film had (or would have had) a notable influence on intrusive memories (anchors: -10 = extreme decrease; 10 = extreme increase), M = -0.4, SD = 3.6, with no differences between conditions, F (2,117) = 1.42, p = .25, η²p = .02, BF₀₁ = 3.9. Moreover, participants generally indicated that their diary was fairly accurate (anchors: 0 = not accurate at all; 10 = very accurate), M = 8.4, SD = 1.4, again, with no differences between conditions, F (2,117) = 1.50, p = .23, η²p = .03, BF₀₁ = 3.7. When omitting 8 participants with accuracy ratings below 7 from the analyses, all conclusions drawn from the group analyses reported above remain virtually unchanged.

Summary and discussion

Overall, Study 1 provides moderate to strong evidence that performing a simple visual rating task directly after viewing a trauma film has no beneficial effect on the development of intrusive memories, as compared to a no-task control condition. Furthermore, it was irrelevant whether participants viewed highly arresting, impressive, and interesting images, as opposed to low visually and less interesting arresting displays. Our data also provide anecdotal to moderate evidence against a direct or moderated association of OSPAN scores with intrusive memories. Taken together, our findings seem to suggest that the build-up of intrusive memories is not directly dependent on WMC, while engaging with visuospatially arresting images in a relatively passive rating task does not mimic the beneficial effects that have been reported using more active visuospatial tasks, such as pattern tapping and the computer game Tetris [4–6].

Study 2

Based on the findings reported above, Study 2 addressed two cognitive components that may be required for a reduction of intrusive memories, in addition to the ability of the H-VA task to capture attention and interest; (a) high working memory load and (b) concurrent activation of the aversive memories. That is, a working memory account predicts that interference with intrusive memories is a function of load on working memory resources, in particular on visual short-term memory (VSTM) resources [8, 14, 19]. Meanwhile, some studies suggest that the modulation of intrusive memories requires active competition for working memory resources between trauma film memories and the concurrent task. For instance, an eye-movement intervention was found to effectively reduce the emotionality and vividness of aversive memories, but only when these memories were actively held in mind [25, 43]. Similarly, once the memories of a trauma film have been consolidated (i.e., after a longer interval), playing the video game Tetris has been found to reduce intrusive memories of trauma film only when memories of the film had been reactivated prior to gameplay [44, 45]. Although reactivation may not be required in the period following film viewing (e.g., [7]), many studies have still included a brief reminder task (e.g., [9]).

To test these possibilities, we devised an imagery condition, in which we instructed participants to memorize and to actively imagine the H-VA displays from Study 1 in order to maximize working memory load. We expected that performing this task after trauma film viewing would reduce intrusive memories compared to a no-task comparison group. Next, in order to test the role of concurrent activation of trauma film memories, a second experimental group performed the same task, while being confronted repeatedly with trauma-film reminder pictures (i.e., imagery + reminder condition). Notably, it is conceivable that this additional rehearsal of reminder pictures would lead to more intrusive memories, potentially cancelling out beneficial effects of working memory engagement. However, the degree to which participants would engage in imagery should be negatively associated with intrusive memories. Therefore, we included a recognition test for the H-VA displays at the end of the experiment as a proxy to active working memory engagement during the task. We expected better recognition performance to correlate with fewer intrusive memories, and particularly so in the imagery + reminder condition.