Intrusive memories of trauma: A target for research bridging cognitive science and its clinical application.

Intrusive memories of a traumatic event can be distressing and disruptive, and comprise a core clinical feature of post-traumatic stress disorder (PTSD). Intrusive memories involve mental imagery-based impressions that intrude into mind involuntarily, and are emotional. Here we consider how recent advances in cognitive science have fueled our understanding of the development and possible treatment of intrusive memories of trauma. We conducted a systematic literature search in PubMed, selecting articles published from 2008 to 2018 that used the terms "trauma" AND ("intrusive memories" OR "involuntary memories") in their abstract or title. First, we discuss studies that investigated internal (neural, hormonal, psychophysiological, and cognitive) processes that contribute to intrusive memory development. Second, we discuss studies that targeted these processes using behavioural/pharmacological interventions to reduce intrusive memories. Third, we consider possible clinical implications of this work and highlight some emerging research avenues for treatment and prevention, supplemented by new data to examine some unanswered questions. In conclusion, we raise the possibility that intrusive memories comprise an alternative, possibly more focused, target in translational research endeavours, rather than only targeting overall symptoms of disorders such as PTSD. If so, relatively simple approaches could help to address the need for easy-to-deliver, widely-scalable trauma interventions.

Introduction: Reviewing cognitive science for its clinical implications in the area of intrusive memories of trauma

In this review, we examine specific advances in cognitive science that pertain to the topic of intrusive memories of trauma. This review is part of a special issue, led and edited by Professor Bethany Teachman, which illustrates how advances in cognitive science have improved our understanding of the aetiology, maintenance, and reduction of psychopathology, while also highlighting the many questions that remain unanswered. Here, we demonstrate how combining research from different fields is starting to improve our understanding of the development and reduction of intrusive memories.

What are intrusive memories of trauma?

After suffering psychological trauma, people can repeatedly experience sensory-perceptual impressions of the event, which intrude involuntarily into consciousness (Ehlers & Steil, 1995; Michael, Ehlers, & Halligan, 2005). These “intrusive memories” typically take the form of visual images (e.g., pictures in the mind's eye), but can also include sounds, smells, tastes and bodily sensations (Ehlers, Hackmann, & Michael, 2004), and come with a range of negative emotions associated with the “hotspots” in the trauma memory (Grey & Holmes, 2008). For example, after having a head-on collision in their car, a person may describe intrusive images of seeing “dust and smoke, debris everywhere” and hearing the “bang” of the airbags (Iyadurai et al., 2017, Supplementary information).

Why are we interested in intrusive memories of trauma?

Over the course of their lifetime, most people will experience or witness a traumatic event (estimated up to 70%; Benjet et al., 2016), in which they are exposed to “actual or threatened death, serious injury, or sexual violence” (Diagnostic and Statistical Manual of Mental Disorders, 5th edition; DSM-5; American Psychiatric Association, 2013, p. 271). Intrusive memories are common following trauma exposure, especially in the first few days and weeks. For most people, they abate over time. For example, in a study of motor vehicle accident patients, 76% had intrusive memories in the first few weeks, dropping to 25% at 3 months, and 24% at 1 year (Mayou, Bryant, & Duthie, 1993). Therefore, the presence of early intrusive memories does not necessarily indicate that they will persist; for some they will, for others they will not.

Regardless of their time course, intrusive memories may in and of themselves be associated with clinical levels of distress and perceived functional impairment, either early after trauma or later. In other words, they may cause distress in their own right - even if they occur outside of a diagnosed psychological disorder. One example is in the case of sub-threshold Post-Traumatic Stress Disorder (PTSD), in which symptoms are associated with clinical impairment, even though the full disorder is not diagnosed (Zlotnick, Franklin, & Zimmerman, 2002). However, intrusive memories will not be associated with clinical impairment for everyone. Indeed, intrusive memories are considered to lie on a continuum with non-clinical involuntary memory experiences in daily life (Berntsen, 2009).

For some people, intrusive memories from as early as the first few days are not only associated with clinical distress and functional impairment, but are also recognised as a symptom of a psychological disorder. Specifically, “recurrent, involuntary, and intrusive distressing memories of the traumatic event(s)” are symptoms of both Acute Stress Disorder (ASD) from 3 days to 1 month post-trauma and PTSD from one month post-trauma (American Psychiatric Association, 2013). While intrusive memories occur across many other psychological disorders (Brewin, Gregory, Lipton, & Burgess, 2010; Holmes & Mathews, 2010), and hence hold transdiagnostic relevance to clinical psychology, in PTSD they comprise a core clinical feature.

Cognitive models of PTSD place intrusive memories of trauma at the hub of symptomatology (Brewin, 2014; Brewin & Holmes, 2003; Ehlers & Clark, 2000) – potentially acting as a driver of the other three symptom clusters (avoidance, negative alterations in cognitions and mood, and hyperarousal). For example, when they spring to mind, intrusive memories elicit high levels of emotional distress (Ehlers & Steil, 1995; Foa & Riggs, 1993), and can be “attention hijacking” (Clark & MacKay, 2015), with the potential to disrupt concentration and cause impairment in day-to-day functioning (Holmes et al., 2017). Network models of PTSD have demonstrated that intrusive memories are centrally connected to and likely to activate other symptoms of the disorder (Bryant et al., 2017; see also Haag, Robinaugh, Ehlers, & Kleim, 2017). This has raised the suggestion that “clinically targeting an important core symptom, such as intrusive re-experiencing in PTSD… can produce downstream beneficial effects on other symptoms in a kind of therapeutic cascade” (McNally, 2012, p. 225).

However, as mentioned earlier, not all intrusive memories of trauma are associated with psychopathology; nor are all early intrusive memories associated with longer-term symptoms. One view is that intrusive memories after trauma may, at times, serve some adaptive function in the survival of the physical and psychological self (Krans, Näring, Becker, & Holmes, 2009). For example, they may act as “warning signals” to help prevent future harm (Ehlers et al., 2002) or may aid protection of self-coherence in autobiographical memory (Conway & Pleydell-Pearce, 2000; Conway, Singer, & Tagini, 2004). As yet, we are unable to accurately predict whether early intrusive memories following trauma are indicative of long-term poor prognosis or not. Such issues are common to secondary prevention endeavours and are discussed further in Section 5.2 on “Prevention and early intervention soon after trauma”.

What is the focus of this review?

In this review, we focus on intrusive memories as one key symptom that can arise after trauma. Specifically, we discuss how cognitive science has made notable theoretical contributions to our understanding of intrusive memory development and has opened up possible new intervention approaches. Throughout the review, we consider intrusive memories as a clinical phenomenon of interest in its own right, as well as an important symptom of psychopathology following trauma, i.e., in the broader context of ASD and PTSD.

Treatments following trauma need to be progressed in a number of ways. First, we currently lack mechanistic insights into the effectiveness of existing treatments (Kazdin, 2007), including those to address psychopathology after trauma; such mechanistic insights are crucial, given that not everyone benefits from existing treatments (Hoge & Chard, 2018; Holmes, Craske, & Graybiel, 2014). Second, preventive interventions after trauma are lagging behind (Roberts, Kitchiner, Kenardy, & Bisson, 2009; Sijbrandij, Kleiboer, Bisson, Barbui, & Cuijpers, 2015). Third, interventions that can be scaled up and adapted for global use are much-needed. A recent editorial has called for “a more substantive evolution in PTSD treatments” and suggested that “novel treatment approaches need to be aggressively pursued.” (Hoge & Chard, 2018, p. 344).

We suggest that focusing on intrusive memories of trauma may help in developing new approaches to therapeutic interventions through a targeted (precision) approach. Further, focusing on the single process of intrusive memory, rather than the whole disorder, allows us to bridge between cognitive science research and psychopathology research with greater precision, as cognitive science tends to examine specific processes in isolation under controlled conditions.

In this review we describe some key theories, methods and findings from cognitive science that have offered insights into the development and reduction of intrusive memories of trauma, rather than PTSD as a whole. We refer the reader to other excellent reviews on cognitive factors in PTSD (Elwood, Hahn, Olatunji, & Williams, 2009; Hayes, Vanelzakker, & Shin, 2012) and on intrusive memories in the context of PTSD specifically (Brewin, 2011, Brewin, 2014; Brewin et al., 2010; Ehlers, 2010; Ehlers et al., 2004).

Our review focuses on recent studies (within the last decade) that have investigated causal associations, and that highlight distal and proximal internal factors, such as neural, hormonal, psychophysiological and cognitive processes. First, we discuss studies that investigated processes that contribute to intrusive memory development, followed by studies that targeted these processes using behavioural/pharmacological interventions to reduce intrusive memories. Next, we consider possible clinical implications of this work and highlight some emerging research avenues, which are supplemented by new data to examine some unanswered questions. We finish with a discussion of some of the merits and limitations of the research reviewed. In conclusion, we raise the possibility that intrusive memories comprise an alternative, possibly more focused, target in translational research endeavours, rather than only targeting overall symptoms of PTSD.

Method: systematic literature search

A search of PubMed was conducted to identify articles published from 2008 to 2018 (for pragmatic reasons we chose the last decade), using the terms “trauma” AND (“intrusive memories” OR “involuntary memories”) in their abstract or title. The 112 articles resulting from this search, as well as relevant references cited in those articles, were selectively reviewed. Titles and abstracts were used to screen articles suitable for inclusion. Articles were selected if they were peer-reviewed; published in English; presented data obtained from human participants; and intrusive memories of (experimental) trauma was one of the outcome measures and reported separately from other PTSD symptoms (i.e., not only as part of the PTSD re-experiencing symptom cluster score, as this encompasses other symptoms such as nightmares and physiological reactivity). Furthermore, we only selected studies that either used prospective correlational designs or experimental manipulations to obtain insights into factors that might causally contribute to the development and change in intrusive memories. We excluded genetic, cross-sectional correlational, qualitative, uncontrolled pilot studies and case studies. Of the 112 articles resulting from the search, 52 articles met the inclusion criteria. An additional 24 eligible articles were identified via cross-references, yielding 76 articles in total.

Insights from cognitive science on the development of intrusive memories

In this section we aim to present an overview of studies that built on insights from cognitive science to study individual differences in the development of intrusive memories of trauma. Most of these studies used prospective correlational designs in non-clinical populations (using experimental analogues of trauma); a few studies used such designs in at-risk or clinical populations.

As an experimental model of exposure to psychological trauma, most of these studies employed variations of the trauma film paradigm (Holmes & Bourne, 2008; Horowitz, 1969; James et al., 2016). This is a widely-used method for inducing analogue trauma symptoms, by asking participants to view video footage with traumatic content. Numerous studies have demonstrated that film stimuli with traumatic content can induce intrusive memories that share characteristics with those experienced following actual trauma (Holmes & Bourne, 2008; Horowitz, 1969; James, Lau-Zhu, Clark, et al., 2016). However, there are limitations of the trauma film paradigm, as with any experimental model, which we raise later in the discussion (Section 6).

Assessments of intrusive memories varied across studies. Most studies used ambulatory assessments over multiple days after the experimental trauma using, for example, a daily diary or an app in which participants recorded their intrusive memories as they occurred. Some studies additionally measured intrusions in the laboratory (Lau-Zhu, Holmes, & Porcheret, 2018) using tasks designed to evoke intrusions (e.g., presenting blurred still images from the film). Other studies only used retrospective measures in which participants indicated retrospectively how often they had experienced intrusions, or how much distress their intrusions had caused, over a certain period (days or weeks). These measures included the Clinician Administered PTSD Scale-IV (Blake et al., 1995), the intrusion subscale of the Impact of Event Scale—Revised (Weiss & Marmar, 1997), and the PTSD Checklist for DSM-5 (Weathers et al., 2013).

In this section, on the development of intrusive memories, we first describe brain imaging studies examining neural mechanism, followed by studies investigating the role of hormones. We then describe studies looking at psychological and physiological factors, and finally studies examining the role of cognitive processing styles.

Neural mechanisms underlying the development of intrusive memory

While brain imaging studies have captured the involuntary recall of a traumatic event in real time (e.g., in patients with PTSD), investigating the neural mechanisms during encoding of a real-life traumatic event (i.e., peri-traumatic mechanisms that could be involved in intrusive memory development) is impossible. Such investigations therefore rely on using experimental trauma analogues. Despite the wealth of brain imaging research in the broader field of emotional memory (Labar & Cabeza, 2006), only a few studies identified here (all using functional magnetic resonance imaging, fMRI) looked at brain mechanisms underlying the formation of involuntary intrusive memories of experimental trauma. In this section we only discuss prospective studies that focused on predicting intrusive memory recall, excluding studies that focused on neural correlates of involuntary memory recall, or brain abnormalities in established PTSD, as they can de facto provide little information about development.

In one study (Bourne, MacKay, & Holmes, 2013), 22 non-clinical volunteers viewed a traumatic film during fMRI scanning and then filled in a week-long intrusion diary. For each individual, the film was subsequently divided into scenes that resulted in later intrusions (‘intrusive scenes’); intrusions for other participants, but not that particular individual (‘potential scenes’); and scenes from neutral clips that did not cause intrusions in any of the participants (‘control scenes’). Clusters of brain activation in response to ‘intrusive scenes’ compared to ‘potential scenes’ during film viewing were identified in the left inferior frontal gyrus and bilateral middle temporal gyrus. This finding was recently replicated (Clark, MacKay, Woolrich, & Holmes, 2016), suggesting that distinct neural traces for intrusive memories are already (partially) formed during exposure to experimental trauma. In a re-analysis of the data, these activation patterns were used in a machine learning approach to predict which moments in the film would come back as intrusions in the subsequent week, which was possible with high accuracy both within individuals (97.3% accuracy), and across individuals (68.0% accuracy) (Clark et al., 2014).

Another study (Battaglini et al., 2016) used visual stills (e.g., of mutilated bodies and neutral objects) as analogue trauma, with a concurrent task to induce either visual (matching the target still to one of two smaller images) or verbal (semantic judgment of the target still as “natural” or “artificial”) processing of the pictures. Brain activation was compared between individuals with a subsequent low (0–1) versus high (>5) intrusion score assessed two days later using the Impact of Event Scale. In the high versus low intrusion group, verbal processing of negative stills revealed clusters of activation in the left amygdala, bilateral hippocampal gyrus, dorsal anterior cingulate cortex, and bilateral inferior frontal gyrus, while visual processing of negative stills only showed activation in the bilateral inferior frontal gyrus. This finding suggests there may be different neural mechanisms underlying verbal and visual processing during trauma that make an individual susceptible to developing intrusions.

In the final study (Gvozdanovic, Stämpfli, Seifritz, & Rasch, 2017), using a between-subjects design, participants were presented either with a neutral film clip or an aversive film clip (a rape scene). Within three hours after encoding, neural responses to implicit memory stimuli (unattended neutral cues from both films) and explicit memory stimuli (an audiotape of someone retelling the traumatic scene) were assessed. In the group who viewed the aversive film, greater activity in the retrosplenial cortex during the implicit task, and in the posterior cingulate cortex during the explicit task, predicted fewer subsequent intrusions in a diary. These findings indicate that neural responses to early retrieval in the immediate aftermath of experimental trauma can predict the formation of intrusive memories.

The role of stress and sex hormones

The role of hormones in memory, and emotional memory in particular, has received considerable attention over the past few decades. For example, research suggests that the stress hormone cortisol may impair the recall of declarative (non-intrusive) emotional memories, while both cortisol and noradrenaline (also involved in the stress response) are thought to enhance the encoding and consolidation of emotional events (e.g., Holz, Lass-Hennemann, Streb, Pfaltz, & Michael, 2014; Schwabe, Joëls, Roozendaal, Wolf, & Oitzl, 2012). Extending these insights to involuntary memory, several studies have specifically addressed the relationship between hormonal changes at the time of (experimental) trauma and subsequent intrusive memories. The studies described below examined the role of the stress hormone cortisol, as well as the female sex hormones estradiol (a form of estrogen) and progesterone.

In a sample of healthy participants, salivary cortisol was collected pre-, peri- and post-viewing of an aversive film (Chou, La Marca, Steptoe, & Brewin, 2014a). Lower cortisol levels post-film predicted greater vividness of intrusions (recorded in a seven-day diary). However, higher cortisol levels peri-film predicted greater frequency of intrusions, but only among individuals with greater sympathetic activation (i.e., those who showed heart rate acceleration in response to acoustic startle probes). This suggests that the effect of cortisol on intrusive memory development may be dependent on time of assessment (or of cortisol administration in pharmacological intervention studies) relative to the aversive learning experience (Joëls, Fernandez, & Roozendaal, 2011; Schwabe & Wolf, 2014). Another study (Cheung, Chervonsky, Felmingham, & Bryant, 2013) similarly found that lower cortisol levels after viewing negative images during a cold water pressor test (immersion of hand in ice cold water) predicted more subsequent intrusive memories, but only in men. Meanwhile, higher general levels of estradiol predicted more intrusive memories in women.

A further study (Wegerer, Kerschbaum, Blechert, & Wilhelm, 2014) looked at the role of estradiol in women with a natural menstrual cycle (during early follicular or luteal cycle phase), in a conditioned-intrusion paradigm. The paradigm included a differential fear conditioning procedure with short violent film clips as unconditioned stimuli (US), followed by a fear extinction phase. Women with lower levels of estradiol showed stronger intrusive memories (greater number, duration and distress) of the violent clips, assessed over 3 days, and this relationship was at least partially explained by the conditioned responding (skin conductance) observed during fear extinction. The different findings in relation to estradiol may be attributed to whether the study sample included women on hormonal contraceptives (Cheung et al., 2013) or not (Wegerer et al., 2014).

Some studies found an association between progesterone and intrusive memories. Women reported having had more intrusive memories of an emotional film (measured using a retrospective estimate) if they viewed it during the luteal phase relative to the non-luteal phase of their menstrual cycle, i.e., when progesterone levels are high (Ferree & Cahill, 2009; Ferree, Kamat, & Cahill, 2011), and progesterone level was positively correlated with intrusive memory frequency (Ferree et al., 2011). However, other studies, discussed above in the context of estradiol, failed to find an association between progesterone and intrusive memories of experimental trauma (Cheung et al., 2013; Wegerer et al., 2014). It should be noted that studies on sex hormones are complicated by the fact that influences of estradiol and progesterone can be hard to disentangle in naturally cycling women (high progesterone levels mainly occur in the presence of increased estradiol levels). One study even found that it was the ratio of estradiol-to-progesterone, rather than estradiol or progesterone alone, that was negatively correlated with the frequency of intrusive memories (Soni, Curran, & Kamboj, 2013).

No studies resulting from our literature search used a prospective design to examine the role of sex hormones on intrusive memory development after real-life trauma. However, we found it noteworthy that one cross-sectional study suggested that women who were exposed to real-life trauma during the mid-luteal phase of their menstrual cycle (again, when progesterone levels are high) were more than three times more likely to experience intrusive memories in the acute period (7–10 days) after trauma than women in the non-luteal phase (Bryant et al., 2011).

Psychological and psychophysiological predictors

A number of studies have identified ‘psychological’ or ‘affective’ factors associated with intrusive memory development. Both stable (trait) as well as transient (state) factors have been examined. Investigation of traits reveals information about individual differences in susceptibility for developing intrusive memories (i.e., more distal factors), while state factors reveal information about circumstances at a particular moment that increase the likelihood of experiencing intrusive memories in the future (i.e., more proximal factors). As will be discussed, certain states (e.g., high autonomic arousal in response to stressful events) may be associated with certain traits (e.g., high trait anxiety), and the distinction may therefore be somewhat arbitrary. However, whereas trait factors may be difficult to influence, state factors may point to malleable targets for interventions early after trauma.

Several studies using experimental models of trauma have found a clear association between ‘negative emotionality’ traits and intrusive memory development. For example, greater general disgust propensity predicted a higher number of intrusive memories of graphic film footage (Bomyea & Amir, 2012). In another study, trait anxiety, depression and trait dissociation were all positive correlated with intrusion frequency in the week after trauma-film viewing (Laposa & Alden, 2008). Further, this study found that film-induced changes in anxiety, as well as post-film state anxiety, were associated with intrusion frequency, and state anxiety mediated the relationship between the trait factors (anxiety, depression and dissociation) and intrusions. In a follow-up study, post-film anxiety levels again predicted intrusions (Laposa & Rector, 2012). Feelings of guilt (experimentally manipulated using a staged computer crash and related loss of data) also predicted the number of intrusive memories of the event and intrusion-related distress (Bub & Lommen, 2017).

Complementing these findings, a meta-analysis of 16 experiments using the trauma film paradigm (n = 458, all no-task control condition; Clark, MacKay, & Holmes, 2015) showed that an absence of intrusions was predicted by a low self-reported emotional response to the trauma film (composite negative mood change score, including emotions such as fear and sadness), as well as low trait anxiety and low depression levels. In line with this, one study using structural equation modelling (Regambal & Alden, 2009) found that greater emotional reactivity during a distressing film was associated with greater intrusive memory frequency over the subsequent week. Interestingly, non-clinical participants who scored high, compared to low, on measures of vulnerability for hypomanic experience (e.g., elated, energetic), also showed elevated numbers of intrusive memories after viewing a film with traumatic content (Malik, Goodwin, Hoppitt, & Holmes, 2014), suggesting that perhaps not negative emotionality per se, but any type of altered emotional processing may influence intrusive memory development.

Finally, changes in state psychological factors such as those discussed above are usually accompanied by physiological stress responses (e.g., changes in skin conductance). For example, both greater mood change and higher physiological reactivity have been found to predict more intrusions immediately after an aversive film (Ripley, Clapp, & Beck, 2017) and also in the 24 h after a virtual reality analogue trauma (an unexpected emergency situation simulated in an underground parking lot; Schweizer et al., 2017). In women, lower heart rate variability (regarded as a sign of less adaption to environmental changes) prior to watching a stressful film predicted more intrusive memories of the film (Rombold-Bruehl et al., 2017), while peri-film decreases in heart rate predicted the frequency of image-based (rather than verbal) intrusive memories (Chou, La Marca, Steptoe, & Brewin, 2014b) (see also Holmes, Brewin, & Hennessy, 2004). Relatedly, two studies showed that enhanced conditionability to neutral cues presented in between (Wegerer, Blechert, Kerschbaum, & Wilhelm, 2013) or during (Streb, Conway, & Michael, 2017) distressing film clips predicted subsequent intrusive memories (a combination of number, duration and distress). Finally, skin conductance responses to a distressing film have been found to moderate the relationship between psychological traits (anxiety sensitivity) and intrusive memories (Olatunji & Fan, 2015). That is, the number of intrusions significantly increased with greater skin conductance responses at high levels of general anxiety sensitivity, but not at low and moderate levels of anxiety sensitivity.

Together, these studies suggest a complex interplay between psychological and physiological processes in response to analogue trauma, which result in the occurrence of intrusive memories following these events. The next section will focus on cognitive processes during (analogue) trauma and how they may contribute to the formation of intrusive memories.

The role of cognitive processing styles

A number of cognitive processes have been implicated in intrusive memory development. The first two studies examined the role of mental imagery. A mental image can be defined as “the same sort of representation that is produced during the first phases of perception, but is created from stored representations in memory not from on-line sensory input.” (Kosslyn, 2005, p. 852). Intrusive memories are mental imagery-based (Pearson, Naselaris, Holmes, & Kosslyn, 2015). In one study, the vividness of general mental imagery, assessed before analogue trauma, predicted the number, vividness and emotional distress associated with image-based intrusive memories of the analogue trauma, recorded over five days (Morina, Leibold, & Ehring, 2013). However, another study found a negative relationship between a questionnaire measure of spontaneous use of (non-emotional) imagery in daily life and intrusions of analogue trauma (Krans, Näring, Speckens, & Becker, 2011). Potentially, general spontaneous use of imagery and the vividness of imagery rely on different mechanisms (Pearson, Deeprose, Wallace-Hadrill, Burnett Heyes, & Holmes, 2013).

A number of studies examined the link between the development of intrusive memories after a trauma film and memory contextualisation, i.e., the binding of items in memory with each other and with their surrounding context (Bisby & Burgess, 2017). Across various studies, weaker memory contextualisation was associated with more subsequent intrusive memories (Marks, Steel, & Peters, 2012; Meyer et al., 2013; Meyer, Krans, van Ast, & Smeets, 2017). Further, indirect evidence for the association between weaker memory contextualisation and increased intrusions came from two studies that investigated the acute effects of alcohol (low and high doses) versus placebo, given prior to viewing a trauma film, on subsequent intrusive memories. Both studies found an inverted ‘U’ dose–response on the number of intrusive memories, with a low dose of alcohol increasing intrusions relative to placebo, while a high dose decreased intrusions (Bisby, Brewin, Leitz, & Curran, 2009; Bisby, King, Brewin, Burgess, & Curran, 2010). These findings were interpreted within a dual representation model of intrusive memory development (Brewin et al., 2010), whereby the low dose of alcohol impaired the contextual memory system but not the image-based memory system, while the high dose was hypothesised to impair both memory systems.Conversely, one study found that greater contextual memory was associated with more intrusions of experimental trauma (D. G. Pearson, 2012). This inconsistency seems to be best explained by how different studies defined ‘context’, as ‘context’ may be operationalised as primarily perceptual (e.g., physical environment), or conceptual (meaning of an experience through verbal information; e.g., Krans, Pearson, Maier, & Moulds, 2016). With this in mind, findings regarding memory contextualization seem to indicate that intrusive memory development is associated with poor perceptual integration, and increased (negative) conceptual processing (D. G. Pearson, Ross, & Webster, 2012; Zhang, van Ast, Klumpers, Roelofs, & Hermans, 2018). As an example of how negative conceptual processing may enhance intrusive memory development, a number of experimental studies have found that rumination (versus, for example, distraction) after a trauma film increased the experience of intrusions and intrusion-related distress (Ball & Brewin, 2012; Ehring, Fuchs, & Klasener, 2009; Hawkins & Cougle, 2013; Zetsche, Ehring, & Ehlers, 2009).

Attentional processes also seem to be an important factor in intrusive memory development. Studies have shown that biased attentional processing of trauma-related cues after exposure to analogue trauma predicted higher subsequent intrusive memory frequency (Schäfer, Zvielli, Höfler, Wittchen, & Bernstein, 2018). That is, reaction times in a dot-probe task were slower on trials with trauma-related cues compared to other threat-related and neutral cues, and a stronger bias predicted more intrusions. The general ability to control thoughts (suppress memory retrieval or resist pro-active interference in working memory, i.e., ignore irrelevant information) has been associated with fewer (Nixon, Cain, Nehmy, & Seymour, 2009; Verwoerd, Wessel, de Jong, Nieuwenhuis, & Huntjens, 2011) or less distressing (Streb, Mecklinger, Anderson, Lass-Hennemann, & Michael, 2016) intrusive memories following a trauma film. It is notable that the general ability to suppress memory retrieval may differ from thought suppression as a coping strategy in response to intrusive memories – the latter of which has been associated with more subsequent intrusive memories (Regambal & Alden, 2009; mentioned in Section 3.3 above). Although no evidence for a direct relationship between self-reported attentional control and intrusive memories has been found (Hagenaars & Putman, 2011), attentional control did moderate the relationship between self-assessed tonic immobility (i.e., ‘freezing’ in response to an aversive film) and intrusive memories. The degree to which spontaneous and/or deliberate ‘thought suppression’, ‘memory control ability’ and ‘attentional control’ rely on overlapping or distinct mechanisms may be an interesting topic for future research.

One further study examined other cognitive factors highlighted in cognitive models of PTSD (e.g., Ehlers & Clark, 2000), and found that peri-traumatic dissociation, data driven processing, and lack of self-referent processing were all positively associated with the development of intrusive memories (Laposa & Rector, 2012).

Summary and theoretical links with PTSD research

The studies reviewed in this section provided numerous insights into the neural, hormonal, psychophysiological and cognitive factors that contribute to intrusive memory development. Neuroimaging studies have shown that neural factors during encoding of an event, not solely factors during consolidation and retrieval of a memory, are important for the future occurrence of intrusive memories. This is consistent with the finding that peri-traumatic processes at the time of trauma are an important predictor of subsequent PTSD symptoms (Ozer, Best, Lipsey, & Weiss, 2003). While current evidence suggests that the inferior frontal gyrus, retrosplenial cortex and the posterior cingulate cortex may be involved in intrusive memory formation, further studies are needed. In addition, whilst there have been a number of imaging studies looking at consolidation processes occurring in the immediate aftermath of an emotional experience, such as post-encoding rest-periods in fear conditioning paradigms (e.g., Hermans et al., 2016), this review identifies a gap in the literature regarding the relationship between neural processes occurring early after encoding and subsequent intrusive memories.

Research reviewed on the effects of hormones on the development of intrusive memories focused on cortisol and sex hormones. Studies reveal a complex picture of the interplay between cortisol and both internal factors (e.g., sympathetic activation) and external factors (e.g., time of assessment). With regard to sex hormones, studies in women suggest that lower estrogen levels are associated with stronger intrusive memories (assessed by frequency, distress and duration), and that the luteal phase of the menstrual cycle (when progesterone levels are high) is associated with a higher frequency of intrusive memories. These findings have important implications for understanding sex differences in propensity to develop intrusive memories of trauma, and possibly anxiety and stress disorders in general. For example, higher rates of PTSD are seen in women compared to men (Kessler, Sonnega, Bromet, Hughes, & Nelson, 1995) and a meta-analysis showed that women are more vulnerable to re-experiencing symptoms (Tolin & Foa, 2006).

Research into psychological and psychophysiological factors involved in intrusive memory development show that negative emotionality (both as a trait and a state); heightened emotional arousal at the time of trauma; and lower heart rate variability (regarded as a sign of less adaption to environmental changes) are all associated with the formation of intrusive memories. These findings appear consistent with what is already understood about pre- and peri-traumatic risk factors for PTSD; in particular that emotional state during and immediately after trauma, and psychiatric history, are predictors of later PTSD symptoms (Brewin, Andrews, & Valentine, 2000; Ozer et al., 2003).

Finally, in terms of cognitive processing styles, studies suggest that enhanced perceptual and data-driven processing, accompanied by poorer contextual memory integration, play a crucial role in the development of intrusions. This again seems consistent with the wider PTSD literature, in which a hypothesised shift from contextual towards perceptual-based processing during trauma is a key feature of cognitive models of PTSD (Brewin et al., 2010; Ehlers & Clark, 2000). Findings also suggest that individuals who engage more in vivid mental imagery may do so because their cognitive processing style is highly perceptual (as opposed to semantic/conceptual), leading to a stronger consolidation of sensory information and a higher likelihood of experiencing intrusions of an event.

Insights from cognitive science on the reduction of intrusive memories

In this next section we highlight recent studies that have capitalised on insights from cognitive science to reduce the distress, vividness and/or frequency of intrusive memories. These studies typically exploit principles of memory malleability, by interfering with memory during the transient time windows in which memory is labile and vulnerable to alteration (McGaugh, 2000; Nader, 2003). We start the section with a brief summary of basic principles of memory plasticity, and the neurobiological processes underlying so-called memory consolidation and reconsolidation. In the subsequent sections we discuss different approaches that directly or indirectly target these putative processes. We discuss pharmacological and endogenous, behavioural (sleep and cognitive task interference), and cognitive manipulations to reduce intrusive memories of experimental and/or real-life trauma.

Basic principles of memory plasticity

Most of the knowledge about the malleability of memory accumulated in the last two decades comes from experimental research in non-human animals (Josselyn, Köhler, & Frankland, 2015; Tonegawa, Liu, Ramirez, & Redondo, 2015). Following acquisition, memory is thought to undergo a time-dependent process of stabilisation necessary for the memory to persist, referred to as memory ‘consolidation’ (McGaugh, 1966, McGaugh, 2000), which may be further divided into ‘synaptic’ and ‘systems’ consolidation. ‘Synaptic consolidation’ refers to a strengthening of local neural circuits via a cascade of molecular processes involving protein synthesis and the formation of new synaptic connections, and is thought to occur minutes to hours after information encoding; ‘systems consolidation’ refers to the reorganization of long-term memory over distributed brain circuits, occurring over days to years after initial encoding (Dudai, 2012). In the current review, most work has focused on the time window that is thought to overlap mainly with synaptic consolidation, in which memories are susceptible to interference (McGaugh, 1966, McGaugh, 2000).

So far, only one study has directly investigated the molecular processes underlying memory consolidation (Das et al., 2016), by targeting N-Methyl d-Aspartate (NMDA) receptors, which play an important role in synaptic consolidation. Nitrous oxide, also known as ‘laughing gas’, blocks NMDA receptors, thus potentially interfering with synaptic consolidation. Das et al. (2016) found that individuals who inhaled nitrous oxide for 30 min immediately after trauma film viewing, relative to those who inhaled medical air (control condition) for the same duration, showed a steeper (quicker) reduction in intrusive memory frequency over one week. While the applicability of laughing gas as an intervention soon after trauma may require further research, this study provides important evidence that the mechanisms underlying memory stabilisation may be disrupted using relatively simple procedures, in ways that are quite similar across different types of memory and even different species (see Visser, Lau-Zhu, Henson, & Holmes, 2018 for further detail).

Rather than there being a ‘one-off’ opportunity in which memory is malleable, relatively recent research indicates that even consolidated memories are not fixed. Upon retrieval, memories can re-enter a transient labile state, where they are again vulnerable to modification (e.g., Alberini & Ledoux, 2013; Nader & Hardt, 2009; Sara, 2000). The re-stabilisation of a memory is a putative process termed memory ‘reconsolidation’, and requires de-novo protein synthesis (Nader, Schafe, & LeDoux, 2000). This suggests that there may be an additional window of opportunity to interfere with the storage of trauma memories. Indeed, it has been suggested that “novel pharmacological and psychotherapeutic approaches that target memory reconsolidation…need to be moved toward the top of the priority list” (Hoge & Chard, 2018, p. 344).

Pharmacological and endogenous manipulations

To influence intrusive memories, a number of studies have targeted brain processes via indirect manipulations, for example by administering the stress hormone cortisol, or by increasing endogenous stress hormones using behavioural manipulations. In a randomised double-blind placebo-controlled study (Graebener, Michael, Holz, & Lass-Hennemann, 2017), female participants received either a low dose of cortisol (20 mg) or placebo for three days following negative film viewing (i.e., outside the consolidation window). Contrary to the prediction that cortisol may inhibit memory retrieval, the cortisol group did not report fewer subsequent intrusions than the placebo group. Similar null findings were obtained in a randomised, double-blind, placebo-controlled crossover study in 30 female inpatients with PTSD (Ludäscher et al., 2015): hydrocortisone administration for a week did not show any effect on the frequency or intensity of intrusive memories (but see Sijbrandij et al., 2015, discussed in Section 4.6 below).

Administration of cortisol (20 mg) prior to film viewing in a sample of 60 women also had no effect on the frequency or vividness of intrusive memories (Rombold et al., 2016). However, another study by the same group found that activity of the noradrenergic system prior to film viewing did have an effect on intrusive memories (Rombold et al., 2016). In this study, participants received either 10 mg of yohimbine (which stimulated noradrenergic activity), 0.15 mg of clonidine (which inhibited noradrenergic activity), or placebo, 60 min prior to viewing a distressing film. The yohimbine group showed a delayed decline in intrusive memories in the four days after film viewing. Interestingly, yohimbine increased endogenous cortisol levels, providing indirect evidence for the hypothesised role of cortisol in the consolidation of intrusive memory.

Another study aimed to increase endogenous stress hormones using a behavioural manipulation (Bryant, McGrath, & Felmingham, 2013). Participants experienced either a cold pressor test (immersion of their hand in ice cold water for 3 min; high-stress group) or submerged their hand in warm water for the equivalent time (low-stress group) immediately after viewing aversive and neutral images. Findings showed relatively enhanced free recall and greater intrusions two days later (using the Impact of Event Scale intrusion subscale) in the high-stress group. Further, in women, but not men, an increase in adrenergic and glucocorticoid response – from baseline to the behavioural manipulation – predicted more intrusive memories (Bryant et al., 2013). A follow-up study (Cheung, Garber, & Bryant, 2015) used a similar approach, but induced stress immediately before memory reactivation two days after the trauma film. Results showed that the ‘stress + reactivation’ group had higher intrusion scores than both the ‘reactivation only’ or ‘stress only’ groups, suggesting that acute stress enhances intrusive memories only when the memory trace is reactivated shortly afterwards. Only in the ‘stress + reactivation’ group did higher cortisol level predict a higher intrusion score. As the stress induction occurred before the reactivation of the memory, it is unclear whether reactivation acted to destabilise the memory, or instead served as a new learning experience in a stressful context. Thus, whether the effects can be explained by reconsolidation-update mechanisms or some other account remained unclear.

Targeting sleep

Sleep is thought to enhance memory consolidation, particularly for emotional relative to neutral materials (Nishida, Pearsall, Buckner, & Walker, 2009; Payne, Chambers, & Kensinger, 2012; Payne, Stickgold, Swanberg, & Kensinger, 2008; Wagner, Gais, & Born, 2001). Thus, the prevention of sleep (i.e., sleep deprivation) has been postulated as a tool for weakening the initial consolidation of unwanted (intrusive) memories. Three experimental studies directly tested this hypothesis using the trauma film paradigm followed by a prolonged period of wake before sleep. One study employed total overnight sleep deprivation for 24 h in a hospital environment (Porcheret, Holmes, Goodwin, Foster, & Wulff, 2015), and showed that the manipulation reduced the number of intrusive memories reported in the initial few days (n = 44, ~75% female per group). In contrast, another study employed day-wake of at least 8 h (Kleim, Wysokowksy, Schmid, Seifritz, & Rasch, 2016), and instead reported the day-wake group to have more intrusive memories towards the end of the week (n = 71, 100% female). Further, a recent study (Woud et al., 2018) found that after viewing a trauma film, participants who had a 90 min nap rather than watching a neutral film experienced fewer intrusive memories later as indexed in a daily intrusion diary and by the Impact of Event Scale intrusion subscale (n = 105, 69% female). To clarify these findings, future research should systematically investigate the impact of the duration of wake and whether or not natural sleep cycles are disrupted (as is the case for full sleep deprivation, but not day-wake), as well as the impact of gender differences.

Cognitive task interference

A series of studies have investigated the use of cognitive task interference as a strategy to reduce intrusive memories. This rationale is based on three key insights from neuroscience and cognitive psychology: 1) memory can be disrupted shortly after an event, or upon retrieval (see section on ‘Basic principles of memory plasticity’, and Visser et al., 2018); 2) working memory has limited capacity (Baddeley, 2003); and 3) intrusive memories of an event include mental imagery, which requires working memory resources (Andrade, Kavanagh, & Baddeley, 1997; Baddeley & Andrade, 2000). Therefore, visuospatial tasks, which occupy working memory, may compete for working memory resources with visual imagery. The effects of such “dual-task” or “concurrent task” interference on intrusive memory development have been demonstrated in a number of experiments for time periods when memories are assumed to be labile, that is, either during trauma film viewing (Holmes et al., 2004), or within minutes to hours after film viewing. In the latter, film reminder cues preceded the cognitive task to orientate the participant to film content (Holmes, James, Coode-Bate, & Deeprose, 2009; Holmes, James, Kilford, & Deeprose, 2010).

Within the last decade, the use of computer game play has been examined as a form of cognitive interference task for intrusive memories of trauma. A series of studies have used the computer game ‘Tetris’, which has been shown to be associated with visuospatial working memory function (Lau-Zhu, Holmes, Butterfield, & Holmes, 2017). In one experiment, participants who completed a brief reminder task for a trauma film and then played Tetris, either 30 min (Experiment 1) or 4 h (Experiment 2) after viewing the film, reported fewer intrusive memories in the subsequent week (in a daily diary) compared to participants in a reminder-only group without Tetris game-play (Holmes, James, et al., 2010). The same study showed that, compared with individuals in the visuospatial Tetris and reminder-only groups, individuals who played a verbal-based computer game called ‘Pub Quiz’ reported comparable (Experiment 2) or even increased (Experiment 1) numbers of intrusive memories, despite both games being rated by participants as equally enjoyable/distracting (Holmes, James, et al., 2010). In both these experiments, interference tasks were administered within hours of the film, during the purported time-window of memory consolidation (McGaugh, 2000). Before film viewing, when the memory is not yet established, a similar procedure does not appear to be effective in reducing intrusions (James, Lau-Zhu, Tickle, Horsch, & Holmes, 2016).

Tetris game play may also be used to reduce intrusive memories for established (24 h old) memory of experimental trauma (James et al., 2015), based on the hypothesis that this could enable interference with memory reconsolidation (Nader & Hardt, 2009). In this study, individuals who underwent a memory reactivation procedure for the film 24 h after trauma film viewing and then played Tetris had fewer intrusive memories than a no-reactivation/no-Tetris group (Experiment 1). A second experiment replicated these findings (James et al., 2015), and further showed that neither memory reactivation nor Tetris game-play alone were sufficient to reduce intrusive memory frequency (Experiment 2). This finding is in line with the idea that both memory reactivation (hypothesised to initiate reconsolidation processes and render memory malleable, in this case showing still pictures for the film) and Tetris game play (to interfere with memory for the film) were needed to reduce the number of intrusive memories of a previously consolidated (analogue) trauma memory. The notion of memory reconsolidation has proved useful in inspiring new avenues for intervention for older memories of trauma, even though some potential limitations have been raised (Treanor, Brown, Rissman, & Craske, 2017).

While multiple types of visuospatial game (not just Tetris) may hold potential to reduce visual image-based intrusions, selecting tasks that are effective may not be straightforward. A GamePlay app thought to tax visuospatial working memory, compared to a no-task control condition, did not show any effects on intrusions, regardless of whether or not trauma retrieval cues were presented prior to game play (Asselbergs et al., 2018). However, it is unclear whether such an app sufficiently taxed working memory. Other tasks that at face value do not appear to be visuospatial have been found to be effective in reducing intrusions too, such as a word game (Hagenaars, Holmes, Klaassen, & Elzinga, 2017) and counting backwards in 3′s (Krans, Näring, & Becker, 2009). Further research is needed to disentangle which elements of a task are crucial to observe cognitive interference.

Findings using derivations of the trauma film paradigm add to our understanding of less common but equally distressing forms of intrusive memories of trauma. Tabrizi and Jansson (2016) investigated cognitive task interference for auditory intrusive memories of experimental trauma. The frequency of auditory intrusive memories of an auditory (analogue) trauma was compared in four groups, each testing different hypothesised parts of working memory interference: visuospatial sketchpad (modelling specific shapes in clay); phonological loop (verbal; counting backwards from 10 to 1); executive processes (verbal; counting backwards in 3′s from specified numbers) and no-task. Results were in line with a modality-specific view of memory processing: individuals in the two verbal groups reported fewer auditory intrusive memories in the subsequent week compared with visuospatial and no-task groups.

Researchers have started to translate these experimental findings to clinical settings. An early translational study with PTSD patients indicated that concurrent eye movements, relative to counting or no task, reduced the vividness and emotionality of established trauma memories in PTSD patients (Lilley, Andrade, Turpin, Sabin-Farrell, & Holmes, 2009). More recent studies have focused on early intervention by administering a task within 6 h after a real-world trauma (Horsch et al., 2017; Iyadurai et al., 2017). One proof-of-principle randomised controlled study investigated mothers who had undergone an emergency caesarean section. Women who received usual care plus a cognitive task procedure involving 15 min of Tetris game-play – whilst still in their hospital bed (i.e., the same context in which the trauma occurred) – reported fewer intrusive memories and a trend towards reduced re-experiencing symptoms after one week, compared to those who received usual care only (using intention-to-treat analysis). At one month, groups differed in PTSD diagnosis (intervention group: 4.2%, versus control group: 30.4%; using per protocol analysis) and avoidance symptoms, although these were not primary outcomes, and the small sample is noted.

Another proof-of-principle randomised controlled study tested a similar behavioural intervention with individuals who had experienced a traumatic motor vehicle accident (Iyadurai et al., 2017). Seventy-one patients waiting in a hospital emergency department were randomised to either the behavioural intervention (a trauma reminder cue – briefly recalling the worst moments of the accident – and then 20 min of Tetris game-play) or an attention-placebo control (filling in a simple written activity log for an equivalent length of time). Intention-to-treat analyses showed medium effect sizes for the reduction in the number of intrusive memories in the week after the accident (primary outcome: 62% reduction) and in intrusion symptom score at one week (secondary outcome) in the intervention compared to control group, but small to negligible effect sizes for other PTSD symptoms (and anxiety and depression symptoms) at one week and one month. Again, tests of replication are required, as well as studies designed to test longer-term outcomes.

Changing cognitive processing styles

Instead of interfering with the memory by, for example, using competing tasks, other studies have examined whether the experience of the memory itself may be changed. In two studies, a computerized programme was used to train participants to re-evaluate intrusive memory symptoms after (Woud, Holmes, Postma, Dalgleish, & Mackintosh, 2012) or prior to (Woud, Postma, Holmes, & Mackintosh, 2013) viewing a trauma film. The training involved participants viewing scripted vignettes that either resolved positively (positive training, i.e., “intrusive memories are normal”) or negatively (negative training, i.e., “having intrusions means I am going mad”). Positive reappraisal training led to fewer intrusive memories (Woud et al., 2012) and less intrusion-related distress (Woud et al., 2013). Relatedly, a recent study (Krans, Brown, & Moulds, 2018) found that participants who were instructed to recall autobiographical memories of success (high self-efficacy condition) relative to memories of failure (low self-efficacy condition) reported fewer intrusive memories of a distressing film in the following week, regardless of whether the manipulation was administered before or after film viewing.

Two studies showed beneficial effects of training participants to process information in a concrete (“How did it happen?”) versus abstract (“Why did it happen?”) style prior to traumatic film viewing: training a concrete processing style resulted in fewer intrusive memories over the subsequent week (White & Wild, 2016) and reduced the relationship between trait rumination and intrusive memory frequency (Schaich, Watkins, & Ehring, 2013). However, training a concrete versus abstract thinking style after analogue trauma did not affect intrusion frequency (Ehring, Szeimies, & Schaffrick, 2009). This suggests that certain cognitive styles can be trained, and that such training, when delivered preventively, may protect against developing intrusions.

Other post-trauma cognitive processes may also be targeted, via techniques that have either been developed in the laboratory or derived from established treatments. For example, directing attention away from trauma reminders after a trauma film, using a form of attention bias modification training, resulted in fewer film-related intrusive memories reported in a diary over the subsequent three days (Verwoerd, Wessel, & de Jong, 2012). Further, changing the valence of distressing memories, through either imaginal exposure or autobiographical memory elaboration, also neutralised the valence of trauma-associated cues (i.e., evaluative conditioning effects) soon after traumatic picture stories, and reduced subsequent intrusive memory frequency (Ehlers, Mauchnik, & Handley, 2012). Finally, imagery rescripting, believed to promote change via re-evaluating the meaning of the trauma memory, when applied 30 min after a trauma film, resulted in fewer intrusive memories over the subsequent week compared to two control interventions: imagery re-experiencing and positive imagery (Hagenaars & Arntz, 2012).

Research reviewed in this section capitalised on insights from cognitive science to reduce the frequency, distress, and/or duration of intrusive memories. These studies typically exploited principles of memory malleability, by interfering with memory during the transient time windows in which memory is labile and vulnerable to alteration (McGaugh, 2000; Nader & Hardt, 2009). However, the level at which they interfered differed widely (Visser et al., 2018). One study showed molecular processes underlying memory consolidation can be targeted at the receptor level, using laughing gas to block NMDA receptors, leading to a steeper decline in intrusive memories over the week. Other studies have used hormonal manipulations, such as administrations of hydrocortisone to reduce intrusive memory recall. Whilst hormonal manipulation studies reviewed here did not reveal any effect on intrusive memories, a previous meta-analysis of randomised controlled trials did find a small beneficial effect of hydrocortisone on overall PTSD diagnosis (Sijbrandij et al., 2015).

Studies that used sleep manipulations shortly after (analogue) trauma have revealed mixed results. The possible differential impact of wake versus (partial or full) sleep deprivation on intrusive memory frequency, as well as why the effects of staying awake become apparent at different times, is an important area for further investigation, as these studies are clearly at an early stage. Possible factors to explore are the duration of wake and whether natural sleep cycles are disrupted (as is the case for full sleep deprivation, but not day-wake).

A number of studies used cognitive task interference to reduce the frequency of intrusive memories. Questions remain as to whether such interventions depends on modality-specific interference specifically (as suggested in Deeprose, Zhang, Dejong, Dalgleish, & Holmes, 2012; Holmes et al., 2009; Holmes, James, et al., 2010) or generally taxing working memory (as suggested in Engelhard, van den Hout, & Smeets, 2011; Tadmor, McNally, & Engelhard, 2016; Van den Hout & Engelhard, 2012), and what other characteristics of the task are crucial to obtain effects. Finally, a number of studies suggest that it is possible to influence characteristics of intrusive memories by changing cognitive styles and appraisals about trauma-related symptoms, or by directly changing the valence of trauma-related cues.

Most of the studies described in this section involved non-clinical samples following experimental trauma. However, within studies investigating clinical samples following real trauma there was some heterogeneity (e.g., female inpatients with PTSD, patients in a hospital emergency department after a motor vehicle accident). Future studies may examine generalisability of findings across different samples.

Clinical implications and future directions

We have discussed a broad spectrum of research in the last decade in which cognitive science has shed light on the specific symptom of intrusive memories of a traumatic event: how they develop and some strategies by which they can be reduced. In this section we take a step back to consider possible implications of this research for clinical practice in the field of psychological trauma. As the possibility of reducing intrusive memories after trauma raises novel questions about how individuals experience intrusive memories, this section is supplemented by brief survey data, which may help open up questions and further thinking in this area to link cognitive science to clinical developments.

We select three key areas that we consider as priorities for the future: a) understanding the mechanisms of trauma treatments, b) prevention and early intervention following trauma, and c) supporting vulnerable populations of trauma survivors on a large scale. These are highlighted in line with a recently commissioned paper on the future of psychological treatments research (Holmes et al., 2018). In each of these three areas we highlight emerging lines of research, as informed by the literature review, that have the potential to move the field forwards.

Mechanisms of treatment

Understanding treatment mechanisms is critical to target treatments more efficiently: indeed, this goal has been identified as a priority for psychological treatment research in general (Holmes et al., 2018; Kazdin, 2007). For trauma in particular, there has been a call to better understand the essential components of effective trauma-focused treatments, and to explore modularity of treatments whereby specific techniques can be employed rather than a full manualized protocol (Hoge & Chard, 2018). Leading evidence-based psychological interventions for PTSD include trauma-focused cognitive behavioural therapy, prolonged exposure therapy, and eye movement desensitization and reprocessing (American Psychological Association, 2017; National Institute for Health and Clinical Excellence, 2005; World Health Organization, 2013a). What these have in common is that the treatment targets problematic aspects of the trauma memory, whether by techniques such as exposure (Foa & Kozak, 1986), restructuring (Ehlers & Clark, 2000), or dual-task interference (Van den Hout & Engelhard, 2012). A clinically successful outcome is not for someone to forget or erase the trauma memory (Holmes, Sandberg, & Iyadurai, 2010) but to be able to deliberately recall the traumatic event without being overwhelmed, and without having frequent, distressing intrusive memories of the trauma.

In this review we have highlighted how cognitive science studies have manipulated key characteristics of intrusive memories (frequency, vividness, and distress) by intervening at multiple levels of mechanism: from interventions that target molecular processes (Section 4.1), through pharmacological interventions (Section 4.2) and behavioural interventions (Section 4.4), all the way to cognitive/affective manipulations such as imagery rescripting (Section 4.5). These studies point to multiple mechanistic pathways through which established trauma interventions might take effect. For example, trauma-focused cognitive behavioural therapy may at once reduce endogenous stress hormones through techniques such as imaginal exposure, whilst also changing cognitive processing of trauma memories using techniques such as imagery rescripting. Critically, a cognitive science approach suggests that we may be able to employ different intervention techniques to target the same processes, with implications for flexible use of specific treatment components rather than entire manualized treatments. Such patient-tailored treatment approaches offer an alternative to a one-size-fits-all approach.

The studies reviewed here have predominantly tested manipulations derived from laboratory research, with a minority testing those derived from therapeutic interventions (e.g., imagery rescripting). Understanding treatment mechanisms requires bi-directional translation between laboratory and clinic. For example, a series of rigorous laboratory studies have examined the effects of eye movements, a key technique in eye movement desensitization and reprocessing, on the vividness and emotionality of aversive (non-intrusive) memories. These experiments have demonstrated that the efficacy of such techniques can be explained in the context of working memory theory, whereby eye-movements occupy limited cognitive resources during memory retrieval, resulting in the storage of less vivid/emotional memory representations (see Van den Hout & Engelhard, 2012 for an overview).

In addition to understanding treatment mechanisms, we need to understand individual differences in treatment effect. Trials of PTSD treatments are beginning to examine potential moderators of treatment efficacy, such as social problems or emotion regulation (Cloitre, Petkova, Su, & Weiss, 2018; Ehlers et al., 2013). However, there is a great deal yet to be understood regarding the influence of individual differences and demographic factors (such as age, gender and ethnicity) on treatment outcome. Taking a symptom-focused approach might help to hone precision approaches. For example, trait vividness of general mental imagery has been associated with the occurrence of intrusive memories (Morina et al., 2013). Thus, by focusing on the impact of potential moderators on single symptoms in the laboratory, we may be able to test factors that are not routinely collected in clinical trials, but that may be useful in understanding why symptoms arise or treatments work for some but not others.

Prevention and early intervention soon after trauma

Prevention of mental illness is high on the mental health agenda internationally (World Health Organization, 2013b) yet progress has been slow. A unique feature of trauma-related disorders ASD and PTSD is that, unlike other mental disorders in DSM-5, they are linked to a specific onset event (Criterion A; American Psychiatric Association, 2013). This opens up the opportunity to employ prevention strategies soon after the occurrence of the traumatic event (known as “secondary prevention”). Previously, a prominent approach to secondary prevention was psychological debriefing (also known as critical incident stress debriefing): a structured intervention designed to target PTSD as a whole, delivered within the first few days of a trauma. This approach typically includes recalling the trauma and talking about it in detail with a therapist or counsellor (Rose, Bisson, Churchill, & Wessely, 2002). However, this intervention has been shown to be ineffective, and possibly even harmful (Rose et al., 2002), and therefore it is not recommended in evidence-based clinical practice.

Another psychological intervention – a prolonged exposure therapy based on fear extinction training (involving imaginal and behavioural exposure) – has shown preventive effects on PTSD in an early stage trial (Rothbaum et al., 2012) and future research is awaited. Data are not described for intrusions specifically; rather, the focus is on PTSD as a whole. This intervention also involved talking through the trauma in detail, which does not hold appeal to some people. The feasibility and practical implications of administering such therapist-delivered interventions soon after trauma requires further empirical examination. A number of pharmacological candidates have also been tested, with only hydrocortisone showing promise in randomised controlled trials (Sijbrandij et al., 2015). There remains a need for safe, effective and feasible preventive treatments that can be administered in the early aftermath of trauma.

Cognitive science research opens some new potential avenues for prevention, by focusing on intrusive memories of trauma (rather than several symptom clusters at once) and seeking to precisely modify this target outcome. By intervening in the first few hours after trauma and targeting memory consolidation, some studies suggest it may be possible to reduce the frequency of intrusive memories over the subsequent days. The wide range of interventions reviewed here include nitrous oxide (Section 4.2), sleep deprivation (Section 4.3), competing cognitive tasks (Section 4.4), and imagery rescripting (Section 4.6), and we suspect many more are possible, targeting different mechanisms at different levels. Further, those studies that provided early tests of interventions following real (rather than experimental) trauma, showed moderate effect sizes on intrusive memory frequency in the first week post-trauma (e.g., Cohen's d of 0.65 in Horsch et al., 2017; Cohen's d of 0.67 in Iyadurai et al., 2017). These effect sizes are encouraging, but as the reproducibility literature cautions, are likely to be high given they are the first studies of their kind. As the pool of translational studies in this area increases, we will be able to better evaluate the potential clinical impact of these interventions.

As discussed in the introduction, not all intrusive memories of trauma are associated with psychopathology; nor are all early intrusive memories associated with longer-term symptoms. Therefore, as is the case for most areas of secondary prevention in health, we may not know if a given individual would have needed the preventive treatment, as only a subsample would have developed longer-term difficulties. Such observations raise the issue of whether preventive approaches should be offered to all individuals (“universal prevention”), even though some people will not “need” it, or only be applied to those at high risk after screening (“selective prevention” or “indicated prevention”; World Health Organization, 2002). In the area of trauma, we would need to develop more precise predictive tools to aid screening for risk factors, and attempts to do this are underway (e.g., Carlson, Palmieri, & Spain, 2017; Russo, Katon, & Zatzick, 2013). From a pragmatic and clinical point of view, the prevention of acute or early-term distress (even if it did not have additional, longer term benefits), may be an approach valued by some patients, particularly if it is brief, cost-effective, and without negative side effects. However, further research on this perspective is warranted.

There is the possibility that for some people, intervening early may not only be unnecessary, but could even lead to adverse outcomes. It could be the case that early interventions may compromise natural resilience after trauma. This is of particular relevance given the important finding, as mentioned above, that the talking therapy “critical incident stress debriefing” was ineffective and could even lead to poorer symptom outcome (Rose et al., 2002). Indeed, critical incident stress debriefing has been identified on a provisional list of potentially harmful therapies (Lilienfeld, 2007). Assessment of negative or iatrogenic effects in psychological therapy is often neglected (Barlow, 2010; Holmes et al., 2018). In the area of trauma, the public demand for prompt intervention that can arise after a crisis must be balanced against issues related to treatment effectiveness and the risk of unintended iatrogenic harm. Examining whether any new treatments, such as those to reduce intrusive memories, have adverse effects (and for whom) should therefore be part of the wider research agenda.

Functional impact of early intrusive memories of real-life trauma: A network perspective and new data

Intrusive memories can be associated with functional impairment. For example, we know that older intrusive memories can be experienced as highly distressing by PTSD patients (e.g., Ehlers & Steil, 1995), can be associated with avoidance of situations that induce intrusions (e.g., Ehlers & Clark, 2000), and interfere with concentration (e.g., in refugees: Holmes et al., 2017). We also know that the frequency of early intrusive memories can be associated with later emotional adjustment. For example, following childbirth the frequency of involuntary memories, and whether they were experienced as distressing or enjoyable, was associated with the presence of post-traumatic stress symptoms at six weeks (Briddon, Isaac, & Slade, 2015).

Interestingly, emerging network studies have examined the links between early intrusive memories and other early and late post-traumatic stress symptoms. Network models view mental disorders, including PTSD, as dynamic systems with multiple links between symptoms (e.g., Borsboom & Cramer, 2013). Recent network analysis of acute and chronic post-traumatic stress symptoms has found early intrusive memories (as well as physiological reactivity) to be centrally linked to other PTSD symptoms in injured, hospitalized trauma survivors (Bryant et al., 2017) and to be a strong predictor of later PTSD in assault survivors (Haag et al., 2017).

A commentary on Bryant et al. (2017) summarised that “intrusions and physiological reactivity to reminders of the trauma scored high on centrality metrics, indicating that activation of these 2 symptoms are especially likely to activate other symptoms in the network. Conversely, successful early intervention targeting these symptoms would likely prevent the full syndrome of PTSD from emerging.” (McNally, 2017, p. 125). However, we note that these findings may not necessarily generalise to all individuals exposed to trauma (including, for example, those who are not physically injured). It remains to be tested whether reducing intrusive memories in the early phase post-trauma can have longer-term effects, and even impact on other PTSD symptoms.

What is more, little data seems available regarding the functional impact of early stage intrusive memories outside of a clinical diagnosis of ASD or PTSD. Presented for the first time here in this review, we conducted a pragmatic survey of 350 individuals at a local university (see also Table 2 in Section 5.3). We note that limitations to this pragmatic survey study include retrospective self-report assessment and a small sample size, but hope the data might open further research nearer the time of traumatic events. One hundred and eighteen respondents to the survey indicated that they had experienced a traumatic event, 81 (68.6%) of whom reported they had experienced intrusive memories in the first week after their traumatic event (see Table 1).

Table 2

Online survey questions and responses regarding participants' willingness to receive novel treatment approaches soon after a traumatic event (N = 350).

Survey question	Responses
	Yes	No
Would you be willing to receive a treatment delivered via mobile phone?	80.2%	19.8%
Would you be willing to play a computer game for 10–20 min as part of a treatment to reduce intrusive memories?	85.7%	14.3%
Would you be willing to stay awake on the first night after a traumatic experience if it could reduce your intrusive memories over the course of a week?	67.1%	32.9%

Note. Participants were university staff and students who received the online survey via an email from their department/college. See Appendix A for further details.

Table 1

Online survey questions and responses regarding experiences of intrusive memories in the first week after a traumatic event (N = 81).

Survey question	Responses
How distressing were your intrusive memories in the first week after the traumatic event?	No distress	2.5%
	Minimal distress	13.9%
	Distress clearly present	40.5%
	Pronounced distress	26.6%
	Extreme distress	16.5%
How much difficulty did you have in dismissing the memories i.e. putting them out of your mind and thinking about something else?	No difficulty	3.7%
	Some difficulty	40.7%
	Considerable difficulty	43.2%
	Extreme difficulty	12.4%
How much did your intrusive memories in the first week after the traumatic event affect your functioning (social, occupational, or other important areas e.g. relationships with other people, social life, work, parenting, schoolwork, housework, volunteer work etc.)?	No adverse impact	9.9%
	Mild impact	25.9%
	Moderate impact	30.9%
	Severe impact	23.5%
	Extreme impact	9.9%

Note. Participants were university staff and students who received the online survey via an email from their department/college. Results are from a subset of respondents (23.1%) who indicated that they had experienced intrusive memories in the first week after a traumatic event (N = 81, out of a total sample of N = 350). See Appendix A for methodology.

Results from the survey suggest that in the first week after a trauma, >80% of respondents reported that their intrusive memories were clearly distressing and > 60% of respondents indicated they had moderate to extreme impact on everyday functioning (see Table 1, and see Appendix A for methodology). We hope these findings stimulate further research to answer questions such as whether reducing intrusive memory frequency soon after trauma may be a means to reducing subjective distress and improving functioning in this early period. For example, some people may suffer negative effects of intrusive memories to the extent that they are compromised in their ability to engage in activities that might be helpful for their recovery, such as seeking social support, accessing medical care, engaging in breastfeeding after a traumatic childbirth, or filing a car insurance claim. For these people, reducing the frequency of associated early intrusive memories may provide some benefit and even allow them to recover more effectively. Others, for whom there is neither distress nor functional impairment, would not be predicted to benefit from a reduction in intrusion frequency. As discussed earlier, potential adverse effects should also always be monitored.

Scalable interventions: Trauma is a global issue

Trauma is ubiquitous worldwide, and traumatic events such as war, terrorist attacks, and natural disasters feature frequently in the media. The sheer scale requires new and innovative strategies to treat and mitigate the potential consequences for psychopathology. We need interventions that are low-intensity, require few resources to deliver, are culturally adaptive, and are accessible to many (Holmes et al., 2018).

Therapist-delivered psychological treatments are difficult to deliver to large numbers of people in different parts of the world. Innovative strategies for disorders such as anxiety and depression involve training community health workers (Patel, Chowdhary, Rahman, & Verdeli, 2011), and this remains to be done following trauma. As a case in point, the refugee crisis worldwide is expanding (Abbott, 2016; United Nations High Commissioner for Refugees, 2015). Refugee host countries (even in Europe) lack the capacity and infrastructure to deploy evidence-based face-to-face psychological treatments at the scale needed, potentially alongside interpreters (Ullmann et al., 2015). The number of psychological treatments tested with refugee populations is low (Nickerson, Bryant, Silove, & Steel, 2011), and while initial evidence indicates that trauma-focused approaches work across cultures (Nose et al., 2017), cultural acceptability of these interventions is important (Hassan et al., 2015). As another example, workers who experience regular exposure to traumatic incidents as part of their jobs, such as emergency department staff, crisis responders, paramedics, firefighters and the police, are often unable to access time-intensive psychological treatments. Here there is a need for brief, low-intensity interventions that are feasible and acceptable alongside a pressured working environment.

Cognitive science-informed interventions may offer one way forwards, among other approaches. From the studies reviewed here, a few approaches in particular hold potential to be disseminated widely and speedily by those who are not specialised in trauma-focused care, and in a non-invasive (e.g., not pharmacological) manner. For example, simple cognitive task interventions (Horsch et al., 2017; Iyadurai et al., 2017), or possibly even sleep deprivation (Porcheret et al., 2015), may be feasible to use in the early aftermath of trauma. However, whether people would be willing to undertake such interventions is an important question. Findings from two proof-of-concept trials reviewed earlier (Section 4.4) provided some support for the acceptability of a computer game intervention following a motor vehicle accident (Iyadurai et al., 2017) and emergency caesarean section (Horsch et al., 2017). Table 2 below shows new survey data, which suggests that such novel approaches may be acceptable for many people (see Appendix A for methodology).

Discussion: Linking cognitive science and implications for psychopathology in the area of intrusive memories of trauma

In the context of this special issue on how recent advances in cognitive science have contributed to our understanding of the aetiology, maintenance and reduction of psychopathology, we have reviewed how cognitive science research in the last 10 years has informed our understanding of the development and reduction of intrusive memories of trauma. We have chosen to focus on intrusive memories as a clinical phenomenon of interest in its own right, as well as a core clinical feature of PTSD and ASD. We note, as a limitation, that there may be other relevant papers that were not identified by our search terms. Given the complexity and heterogeneity of a full mental health diagnosis such as PTSD, a precise mapping between cognitive processes and broad symptom clusters can be a challenge. Instead, most experimental studies (such as those discussed here) use a reductionist approach to obtain mechanistic insights into the cognitive processes that contribute to a single symptom, in this case intrusive memories of a traumatic event. In this discussion we consider a number of issues that cut across the studies reviewed, with implications for translation between laboratory-based experimental paradigms and the clinical context.

Translational merit and limitations of experimental models of trauma

The vast majority of studies in this review used an experimental analogue of real-life trauma, such as film material, static pictures, virtual reality, and laboratory scenarios. A benefit of experimental models is that they enable variables around the time of trauma to be examined and manipulated under controlled conditions, which is rarely possible in the case of real-life trauma. However, there are clear limitations to the ecological validity of this approach. For example, viewing a trauma film (such as those used in Holmes, James, et al., 2010; James et al., 2015) is by no means equivalent to experiencing real-life trauma, as some individuals may, for instance, choose to view distressing films recreationally (i.e., a non-traumatic experience). Thus, experimental findings should be corroborated with prospective studies following real-life trauma. Reporting in prospective clinical studies on intrusive memories specifically would help towards this aim (in this review, only 4 out of the 76 studies identified followed real-life trauma). Early-stage translational work moving similar protocols used in the laboratory to the clinic are starting to show promise. An example since this review was conducted is for older intrusive memories after complex trauma (Kessler et al., 2018).We note that repeated or extreme indirect exposure to aversive events in the line of work (such as police having to view electronic media or pictures of real murder or abuse) is now included as a diagnostic criterion for a traumatic event in the DSM5 (American Psychiatric Association, 2013, p. 271). Relatedly, studies looking at media viewing of real trauma have shown that repeated and enduring viewing of scenes in media footage, for example of the Boston Marathon bombings, is associated with acute stress symptoms (Holman, Garfin, & Silver, 2014). Such studies highlight the importance of further understanding this form of media-based exposure.

Assessment of intrusive memories

Across the studies reviewed, assessment of intrusive memories was predominantly self-report, for example, using a daily diary or retrospective questionnaire. Self-report measures may be susceptible to demand characteristics, in that participants may respond based on how they believe they are expected to respond. However, some studies reviewed here have indicated that expectations did not differ between experimental groups (e.g., Holmes, James, et al., 2010), and/or that expectations did not predict intrusion outcomes (Iyadurai et al., 2017; James et al., 2015). Further, routine clinical assessments that serve the basis for diagnosis and treatment (including for psychopathology where intrusive memories are implicated, such as PTSD) are also based on self-report, sometimes months or years after the trauma. Intrusive memories entail conscious awareness of the memory springing back to mind – a form of involuntary explicit memory (Berntsen, 2009; Visser et al., 2018) – and thus by definition require self-report.

Nevertheless, experiential phenomena such as intrusive memories can be measured in more than one way, for example by ‘provoking’ them in the laboratory so that they are at least reported under more controlled conditions (Lau-Zhu et al., 2018). These assessments may also be complemented with physiological outcomes, such as heart rate and skin conductance response to trauma cues (e.g., Kunze, Arntz, & Kindt, 2015; Meyer et al., 2013; Streb et al., 2017; Wegerer et al., 2013), performance-based tasks that assess their putative underlying cognitive processes, such as perceptual priming (Holz et al., 2014; Sündermann, Hauschildt, & Ehlers, 2013) or memory/attention biases (Nixon et al., 2009; Schäfer et al., 2018; Verwoerd et al., 2012; Verwoerd, Wessel, de Jong, & Nieuwenhuis, 2009).

The studies discussed within this review typically measured intrusive memories in terms of their frequency, either using ambulatory measures, such as reporting the occurrence of an intrusion when it happens in a diary, or retrospective clinical measures of PTSD symptom frequency and severity, such as the Clinician Administered PTSD Scale-IV or variations of the PTSD Checklist for DSM-5. Compared to ambulatory measurements, these retrospective questionnaires provide relatively coarse information and do not take into account that the occurrence of intrusive memories may be highly context dependent (Kleindienst et al., 2017); however, their use in clinical practice as well as research permits bi-directional translation. Only a few studies examined other characteristics of intrusive memories, such as distress or duration (e.g., Streb et al., 2017; Wegerer et al., 2013). Given that intrusion distress and “nowness” (the extent to which they are experienced as occurring in the present moment) have been indicated as predictors of later PTSD (Michael, Ehlers, Halligan, & Clark, 2005), assessing various characteristics of intrusive memories as outcomes in experimental studies may be fruitful for clinical translation.

Finally, the majority of studies assessed intrusive memories in the first few days after (real or experimental) trauma. It may be useful to examine intrusive memories after longer time periods, given that we know they naturally decline over time for most people. For example, many participants in experimental studies report zero intrusive memories after a few days (e.g., Butler, Wells, & Dewick, 1995; Das et al., 2016; James et al., 2015). However, intrusive memories do not decline within a week for everyone; neither after analogue trauma (e.g., see James et al., 2015, Fig. 2), nor after real-life trauma (e.g., see Horsch et al., 2017; Iyadurai et al., 2017, Figs. 3 and 2 respectively). Thus, longer time periods are needed to better understand their time course, and what factors contribute to them persisting or not. This is critical for considering translational implications of these studies for PTSD treatment and prevention.

Reproducibility of research

The cognitive science of intrusive memories is at an early stage, and there is still a relatively small body of work in this area compared to other forms of memory. Throughout the work reviewed above, as in all psychological science, it remains important to consider reproducibility. Open science will clearly help in this regard (Nosek et al., 2015). Studying intrusive memories in the laboratory requires detailed experimental psychopathology paradigms, often requiring multiple laboratory sessions, for which extensive training is typically required to reproduce findings. Suggestions to aid researchers in their pursuit of adequate tests of replication include: sharing protocols which are accompanied by an agreed set of minimum training standards in their use (e.g., for trauma film viewing, any intervention procedure, intrusive memory monitoring, and data checking); training of the experimenter involving a scientist on the original study, as well as pre-specified protocol-adherence checks, in order to monitor adherence to the original protocol of the study; and training in how to score and interpret measures.

Issues of individual difference and diversity

From the studies reviewed, it is hard to draw any strong conclusions about the impact of diversity (which includes but is not limited to age, gender, ethnicity, socioeconomic status and so forth) on the aetiology and reduction of intrusive memories. The majority of studies were conducted in Western European, American and Australian regions, and many used student samples. Furthermore, studies looking at the role of hormones in intrusive memory development often use all-male or all-female samples in order to reduce variance, as some hormones (e.g., cortisol) have been shown to have differential effects in men or women (e.g., Bryant et al., 2013). Therefore, embracing the complexity of individual differences is a critical area for future research. We also need to question our basic assumptions about the way we study experimental trauma in cognitive science, for example, the personal relevance of the trauma film footage to different cultural groups, ages, genders, etc., and its impact in generating intrusive memories. This may be particularly relevant for studies where only a low rate of intrusive memories is observed, rendering them liable to floor effects; a film that worked to generate intrusive memories in one country or time may not transfer to another. Careful work to develop appropriate stimulus materials needs to be undertaken in consultation with specific study populations.

Conclusion

Research and clinical practice in the field of psychological trauma continues to focus predominantly on full mental health disorders, such as PTSD. This review has demonstrated how focusing on a single core symptom - here intrusive memories of a traumatic event - can offer novel opportunities for bridging the gap between cognitive science and therapeutic interventions. With the expanding experimental literature, the time is now ripe for accelerating translational links, and building on emerging real-world data to help address the global burden of trauma.

Funding

This work was supported by a Wellcome Trust Strategic Award [098461/Z/12/Z]; the European Union's Horizon 2020 Research and Innovation Programme under the Marie Skłodowska-Curie agreement [No 705641]; the Medical Research Council Cognition and Brain Sciences Unit, University of Cambridge [SUAG/013/RG91365]; the Cambridge International Scholarship Scheme [the Cambridge Commonwealth Trust and the Cambridge Overseas Trust]; the Swedish Research Council [2017-00957]; the Lupina Foundation; and the Swiss National Science Foundation [SNF; grant 32003B_172982]. Funding to pay the Open Access publication charges for this article was provided by the Medical Research Council Cognition and Brain Sciences Unit, University of Cambridge [SUAG/013/RG91365].

Contributors

Author RMV conducted the systematic literature search and LI, ALZ, KP, AH and ELJ provided summaries of research studies. Authors LI, RMV, ALZ, EAH and ELJ designed the online intrusive memory survey. Authors LI and ELJ analysed the results of the survey. Authors LI and RV wrote the revised draft of the manuscript and all authors contributed to and have approved the final manuscript.

Conflict of interest

ALZ, AH, ELJ, KP, LI and RMV declare that they have no conflicts of interest. EAH is an Honorary Professor of Clinical Psychology at the University of Oxford, Department of Psychiatry, holds an honorary contract at the Medical Research Council (MRC) Cognition and Brain Sciences Unit, University of Cambridge, UK and serves on the Board of the Charity “MQ; transforming mental health”; she receives no remuneration for these roles.