Introduction 167

The period following a traumatic event is believed to play a role in PTSD symptoms over the long term. Accordingly, attempts to offer preventive interventions aimed to reduce development of PTSD are often offered following the traumatic experience. For example, psychological debriefing interventions in the immediate aftermath of a traumatic event have been used widely to help victims following trauma (McNally et al., 2003). Although studies have shown that these interventions might in fact exacerbate subsequent adverse reactions to the trauma (Mayou et al., 2000) this only demonstrates that processing in the immediate aftermath of a trauma is a crucial factor and is fragile to external influences.

Experimental studies have established that the time during encoding of emotionally negative events is important to subsequent memory intrusions. For instance, studies using the trauma paradigm (described in Chapter 1.2.5) generally demonstrate that a visuospatial task (such as the computer game Tetris, modelling clay or complex finger tapping) administered during encoding reduces the number of subsequent memory intrusions (Bourne et al., 2010; Krans et al., 2009; Holmes et al., 2004;

168

Logan and O’Kearney, 2012). In contrast, carrying out a verbal task (e.g. counting backwards in threes) during encoding has been shown to increase the number of subsequent intrusions (Bourne et al., 2010; Krans et al., 2009; Nixon et al., 2007). Likewise, studies investigating neutral deliberate memory have shown that carrying out a task during encoding can impair subsequent memory performance due to divided attention. For instance, Craik et al. (1996) showed that carrying out a distracting task while encoding information impaired subsequent recognition, cued recall and free recall memory performance. For emotionally negative material, deliberate memory have been assessed along with studies investigating memory intrusions. These studies showed more mixed results for deliberate memory, with some showing an effect of peri-traumatic manipulations (Bourne et al., 2010; Krans et al., 2009; Holmes et al., 2004) while others found no effect (Das et al., 2016; Krans et al., 2010). Taken together, these studies demonstrate that modulating processing during trauma film encoding plays an important role in how the ‘event’ is subsequently remembered.

Theories on memory consolidation also implicate the time following encoding of an event as critical in memory formation. In relation to memory intrusions, several studies have shown that manipulating processing after encoding of a trauma film can affect the subsequent number of reported intrusions (Deeprose et al., 2012; Holmes et al., 2009; Das et al., 2016). Many of these studies used designs similar to the studies investigating intrusions and the relationship with peri-traumatic processing and have shown that intrusion development can be affected by tasks carried out immediately after encoding (Deeprose et al., 2012) and 30 minutes post- encoding (Holmes et al., 2009; Deeprose et al., 2012; Green and Bavelier, 2003).

Furthermore, a recent study tested the use of a visuo-spatial task in preventing memory intrusions in people who had just been exposed to trauma (traffic accidents) and found that carrying out a visuospatial task in the aftermath of the traumatic experience did indeed reduce memory intrusions for real-life events (Iyadurai, et al., 2017).

169 These findings are consistent with studies showing that neural processing both during and after an event influences subsequent memory. For neutral information, studies have found that spontaneous post-encoding fluctuations in neural activity during rest are similar to activity patterns during encoding. It is believed that this reinstatement is comparable to the post-encoding replay processes observed in animals, which are believed to reflect memory consolidation (Bird et al., 2015; Staresina et al., 2013; Tambini et al., 2010; Tambini and Davachi, 2013, Foster and Wilson, 2006; Marr, 1971; McClelland et al., 1995, Ben-Yakov et al., 2011, 2013).

Staresina et al. (2013) investigated post-encoding offline processing of learned information in a study where participants were presented with object-scene pairs during encoding followed by an active delay period in which participants carried out an odd/even number judgement task. This period was immediately followed by a source memory test, in which participants were cued with either an object or a scene presented during encoding and were instructed to retrieve its paired associate. Findings from this study showed that spontaneous offline reinstatement of activity in the entorhinal cortex in the MTL and in the retrosplenial cortex that was present during encoding of an object-scene pair predicted whether or not the pair was subsequently retrieved in the memory test.

In a more recent study, Bird et al. (2015) demonstrated that the above findings could be extended to include naturalistic scenes. In this study, participants were presented with a series of short audio-visual video clips and were subsequently asked to actively rehearse silently most of these videos while still in the MRI scanner, while some were used as controls with no rehearsal. Analysis of these data showed that rehearsed videos were remembered better than videos that were not rehearsed and that the similarity in neural activity in the posterior cingulate cortex between encoding and rehearsal predicted subsequent memory performance on a free recall test (Bird et al., 2015). Similarly, Ben-Yakov et al. (2011) showed that hippocampal and caudate nucleus activity time-locked to the offset of an audiovisual video clip correlated with subsequent memory performance for gist in the scenes, suggesting that the offline activity might play a role in memory

170

formation, perhaps by binding together the individual elements of the video clips over time (Ben-Yakov et al., 2011).

Taken together, these studies show that increases in pattern similarity between encoding and post-encoding phases as a result of either conscious rehearsal or offline processing predict subsequent memory performance for both object scene pairs and naturalistic information (Staresina et al., 2013; Bird et al., 2015; Tambini and Davachi, 2013)

Recently, studies have begun to elucidate the neural correlates of intrusion development. Bourne et al. (2013) conducted the first prospective study investigating the neural basis of intrusive memory formation. Here, participants encoded emotionally negative videos while in the MRI scanner and activity during encoding was subsequently matched with scenes that intruded, potential intruding scenes (scenes where others had intrusions but not the participants) and control scenes. This study found that intruding scenes were related to increased neural activity in a wide range of regions such as the amygdala and anterior cingulate cortex (ACC). Both of these regions are known to be involved in processing of negative emotion, including a critical role of the amygdala in fear memory (Davis, 1992) and a role of the ACC in self-regulation of emotion and threat perception qua the role of the ACC in error and conflict detection (Allman et al., 2001; Kalish et al., 2005; Carter et al., 1998). Furthermore, comparing encoding activity for intruding scenes and potential scenes showed an increase in activity in the left inferior frontal gyrus and in the bilateral middle temporal gyrus for intruding scenes (Bourne et al., 2013). In a more recent study by Clark and colleagues (2016) another imaging phase of the above study was added immediately after encoding where participants reported spontaneous memory intrusions for the scenes they had just encoded, allowing the investigation of neural signatures of both encoding and retrieval of intrusive scenes. Consistent with the study by Bourne et al. (2013), this replicated their finding of increased encoding activity in the left inferior frontal gyrus for subsequently intruding scenes. Moreover, this study showed that this structure is also involved in the involuntary retrieval of the traumatic scenes (Bourne et al., 2013). Hence, both of these studies indicate a role of activity in the inferior frontal gyrus (IFG) during

171 encoding for subsequent intrusions, and more generally demonstrate the use of the experimental trauma paradigm in neuroimaging studies. The left IFG is involved in sentence comprehension and semantic processing, indicating a role for this structure in verbal and conceptual processing of information (Friederici et al., 2003). This finding is consistent with unitary accounts of trauma memory, proposing that memory intrusions arise as a result of particularly strong encoding and rehearsal (Rubin et al., 2008a; 2008b).

The studies by Bourne et al. (2013) and Clark et al. (2016) investigated the role of neural activity during encoding in subsequent memory intrusions. However the relationship between post-encoding neural activity and subsequent memory intrusions is unknown. Hence, this chapter aimed to investigate the role and neural correlates of both peri- and post-encoding in relation to subsequent memory intrusions. Furthermore, this study builds on studies investigating the role of post- encoding processing to subsequent deliberate memory performance and investigates if previous findings generalise to include emotionally negative information.

For this study, the following analyses were planned: 1) GLM analyses of intruding versus non-intruding videos and also of remembered versus forgotten videos. 2) Fit finite impulse response models to the data to include a time series analysis which allows for more detailed analysis of neural responses to the presented stimuli and 3) an RSA analysis which allows the comparative analysis of the similarity of activity patterns in the encoding and subsequent 30 sec rest phases. The latter two analyses were planned but were not included in this thesis.

Hypotheses for the GLM analyses were formulated based on previous empirical studies investigating intrusion development and on theory-driven hypotheses in terms of the dual representation theory and on unitary accounts of intrusion development outlined in previous chapters. Based on the unitary account, it would be expected that intruding clips and remembered clips would both be associated with greater amygdala and hippocampal activity than non-intruding and forgotten clips as both memory types would depend on hippocampal processing and be facilitated by increased amygdala activity.

172

In contrast, from the dual representation theory it could be predicted that remembered video clips would be associated with greater hippocampal activity compared to forgotten clips, and this contrast was not expected to be significant for intruding versus non-intruding clips. Conversely, intruding clips would be associated with more amygdala activity than non-intruding clips and this contrast was not expected to be significant for remembered versus forgotten clips.

Also, based on previous experiments investigating intrusion development (Bourne et al., 2013; Clark et al., 2016), it was expected that intruding clips would be associated with greater activity in the amygdala and the ACC. The present study did not distinguish between intruding and potential scenes, but as a wide range of videos intruded across participants, it was also expected that intruding scenes would be associated with greater activity in the left inferior frontal gyrus and in the bilateral middle temporal gyrus.