R
ESEARCHM
ASTERP
SYCHOLOGYT
HESISAnna Ruth Hudson 10620761
“Future imagery and posttraumatic stress: The role of intrusive future-oriented thoughts in specificity of future visualisations following experience of trauma”
Specialisation: Clinical Psychology Supervisor: Dr. Nexhmedin Morina
External supervisors: Prof. Zachary Steel & Prof. Derrick Silove
Research centre: University of Amsterdam & University of New South Wales
Future imagery and posttraumatic stress:
The role of intrusive future-oriented thoughts in specificity of future visualisations following experience of trauma
Posttraumatic stress disorder (PTSD) has been previously linked to overgeneral autobiographical memory. Lately, studies have explored what effect PTSD has on ability to generate visualisations about one’s future, finding that future imagery appears to be overgeneralised similarly to autobiographical memory. Additionally, individuals with anxiety disorders rate their positive future imagery as less vivid, and less likely to occur. A possible explanation for this relationship lies in intrusive future-oriented thoughts. This study therefore investigated if severity of intrusive future-oriented thoughts mediates the relationship between PTSD symptomatology and future imagery. 93 participants who reported having experienced a traumatic event were asked to generate visualisations about their future in response to positive and negative cue words. Future imagery was assessed for specificity/overgenerality, vividness, participant-rated valence and confidence of occurrence, trauma-relatedness, and response latency. PTSD symptomatology significantly predicted overgenerality and trauma-relatedness of imagery. No significant mediation of severity of intrusive future-oriented thoughts was found. Differences in methodologies to previous studies may explain non-significant findings.
Keywords: PTSD, future imagery, autobiographical memory, overgenerality
1. Introduction
Posttraumatic stress disorder (PTSD) is a psychological disorder which can result following the experience of trauma (Diagnostic and Statistic Manual of Mental Disorders, DSM-5; American Psychiatric Association, 2013). Traumas are defined as exposure to events which include actual or threatened death, serious injury, or sexual violence, and which the individual can have experienced directly, witnessed, or heard about happening to a close associate, such as a friend or family member. Classic symptoms of PTSD include intrusive and avoidance-type symptoms, as well as changes in cognitions, mood, and arousal (APA, 2013). Intrusive symptoms may include persistent re-experiencing of the traumatic event; for example, through unpleasant memories resurfacing unbidden or repeated dreams of the experience. Avoidance-type symptoms may include avoiding memories or external stimuli related to the trauma. Together, these may work to change an individual’s perspective on the world.
Experience of trauma has previously been linked to overgeneral autobiographical memory (OAM; for a review, see Moore & Zoellner, 2007). OAM is a phenomenon where personal memories of one’s past are recalled with less specific detail, particularly when a series of individual memories are blurred together and, instead of one specific event being recalled, multiple events are generalised together into one recollection (Brown et al., 2013; Kleim & Ehlers, 2008; Moore & Zoellner, 2007). Previous studies have found that individuals who have
experienced trauma are more likely to recall memories in an overgeneral fashion in comparison with individuals who have not (e.g. Brown et al., 2013; Henderson, Hargreaves, Gregory, & Williams, 2002; Kangas, Henry, & Bryant, 2005; McNally, Lasko, Macklin, & Pitman, 1995; McNally, Litz, Prassas, Shin, & Weathers, 1994; Stokes, Dritschel, & Bekerian, 2004; see Moore & Zoellner, 2007, for a review). This difference is found even between individuals diagnosed with PTSD and individuals who have experienced trauma (such as combat veterans, e.g. McNally et al., 1994, 1995, Brown et al., 2013, and cancer patients, e.g. Kangas et al., 2005) but have not gone on to develop PTSD, suggesting that it is not necessarily solely the experience of trauma which is associated with OAM, rather the response to the traumatic experience.
Less attention however has been given to how individuals with PTSD are prone to imagining the future. This is surprising considering how one’s experiences shape perception of one’s future. Previous studies suggest that imagining the future requires many of the same neural processes involved in remembering the past (Schachter, Addis, & Buckner, 2007). Tulving (2002) describes the case of K.C., who suffered severe head injury rendering him amnesic. K.C. was not only incapable of recalling any autobiographical memories, but also was unable to imagine his future (Tulving, 2002). There is also considerable evidence for a significant future-oriented component to PTSD. Ehlers and Clark (2000) describe how individuals with PTSD, following their trauma, may develop a “sense of serious current threat” (p.320) and plan their futures accordingly; for example, through avoidance behaviours. Furthermore, Kangas, Henry, and Bryant (2002) note that in individuals who have been diagnosed with PTSD following a medical diagnosis, the relevant stressor is often future-oriented due to fear about the prognosis. In participants diagnosed with PTSD following a multiple sclerosis diagnosis, 74% report distressing future-oriented intrusions related to prognosis concerns and 78% reported that their nightmares were future-oriented (Chalfant, Bryant, & Fulcher, 2004). Lastly, participants diagnosed with PTSD (compared to participants who have experienced trauma but not been diagnosed with PTSD) have been shown to be significantly more likely to report a feeling of foreshortened future (Chalfant et al., 2004; Kleim, Graham, Fihosy, Stott, & Ehlers, 2014).
What little previous literature exists on the influence of PTSD on one’s ability to visualise the future, however, suggests that individuals with PTSD tend to overgeneralise imaginary future events, much in the same way that they recall past memories. Combat veterans diagnosed with PTSD have been found more likely to generate overgeneralised future visualisations to neutral cue words compared to combat veterans without PTSD, and to be additionally more likely to incorporate trauma-related imagery into their visualisations (Brown et al., 2013). Additionally, two other studies found that amongst individuals who had experienced assault or a motor vehicle accident (Kleim et al., 2014) and amongst Vietnam War veterans (McNally et al., 1995), those who reached a clinical diagnosis of PTSD were more likely to produce overgeneralised future imagery, but only in response to positive and not to negative cue words. In both studies, participants with PTSD were again more likely to give trauma-related responses, irrespective of cue valence. McNally and colleagues (1995) further report that participants with PTSD were slower to recall specific memories in response to positive cue words in comparison with participants without PTSD; no other effect of cue valence or participant diagnosis status were found. No attention has as yet been given to the relationship between severity of symptomatology and specificity of future imagery.
Furthermore, little focus has been given to other aspects of imagery, and their relationship with PTSD. Two potential aspects are those of vividness and how confident one is in one’s
visualisations being actualised. Individuals diagnosed with an anxiety disorder have been previously shown to rate positive future visualisations as being less vivid than healthy individuals, yet more vivid for negative visualisations (Morina, Deeprose, Pusowski, Schmid, & Holmes, 2011). Additionally, participants with an anxiety disorder rated negative visualisations as being more likely to occur in comparison with healthy controls, and positive visualisations as less likely to occur. For this study, participants needed to generate visualisations in response to scenario descriptions, whereas other studies ask for spontaneous generalisations to cue words. Importantly, this study did not specifically investigate PTSD; however another study has demonstrated that, in a sample of Palestinian youths, level of traumatic exposure is negatively correlated with positive future expectations (Lavi & Solomon, 2005).
One possible explanation for the relationship between posttraumatic stress disorder (PTSD) and future imagery could be in the presence of involuntarily generated imagery, or intrusive future-oriented thoughts. Intrusive thoughts are a classic symptom of PTSD (APA, 2013). These can be retrospective (for example, intrusive memories traditionally associated with PTSD) or they can also be prospective in nature, comprising distressing and unwanted images of possible future events (Deeprose & Holmes, 2010). It has been previously suggested that OAM in PTSD occurs as a result of avoiding painful emotions associated with traumatic memories and has a protective function (Conway & Pleydell-Pearce, 2000; Kleim et al., 2014; Moore & Zoellner, 2007; Sumner, 2012; Williams, 1996; Williams, 2006; Williams et al., 2007). In support of this, research has showed that frequent occurrence of intrusive memories predicted OAM in a sample of trauma-exposed participants (Wessel et al., 2002). Furthermore, the CaR-FA-X model of OAM (Williams, 2006; Williams et al., 2007) suggests that OAM is influenced by three factors: capture and rumination, functional avoidance, and impaired executive control. Sumner (2012) reviews this model, reporting that OAM appears to be a cognitive avoidance strategy, and that avoidance of traumatic memories is related to reduction of psychological distress following trauma. Prospective intrusive thoughts, or intrusive future-oriented thoughts, include unwanted cognitions about one’s future, and can be as distressing as unwanted traumatic memories if they are negative or stressful in character (Deeprose & Holmes, 2010). Additionally, prospective negative thoughts are already thought to contribute to the maintenance of anxiety disorders (Morina et al., 2011). In a sample of traumatised Iraqi refugees, intrusive fears for family left behind significantly predicted higher levels of PTSD (Nickerson, Bryant, Steel, Silove, & Brooks, 2010). Therefore, it could be reasonable to suggest that overgeneralised and less vivid future imagery may play a similar role in the avoidance of pain associated with intrusive future-oriented thoughts. Thus, severity of intrusive future-oriented thoughts could explain the relationship between PTSD symptomatology and more overgeneralised and less vivid future imagery.
This study aimed to investigate how severity of PTSD symptomatology affects future imagery. Specifically, the study sought to examine the relationship between severity of PTSD symptomatology and the overgenerality (or lack of specificity) of future imagery, and furthermore to what extent this relationship is explained by the severity of intrusive future-oriented thoughts. This study further intended to investigate the relationship between severity of PTSD symptomatology and other aspects of future imagery. These aspects were vividness, valence, and trauma-relatedness of visualisations, how confident participants were that their visualisation could one day be actualised, and the speed with which participants generated such visualisations. The study also investigated whether these relationships differ depending on the valence of cue words used to generate future visualisations. In addition, the study was interested
in investigating the correlation between characteristics of future imagery and of memory, as this has never been studied before. Finally, the study aimed to explore the relationship between aspects (vividness, etc.) of future imagery generated in response to cue words and those generated in a more open-ended fashion.
It was hypothesised that there would be a negative relationship between PTSD symptomatology and: specificity, vividness, participant-rated likelihood of future visualisations, and valence. That is, individuals with more severe PTSD symptomatology were predicted to generate less specific (or more overgeneralised) visualisations, and to rate their imagery as less vivid, less likely to occur, and more negative. Additionally, participants with more severe PTSD symptomatology were predicted to take longer to generate visualisations and to be more likely to incorporate trauma-related imagery. It was hypothesised that cue word valence would moderate these relationships with effects only seen in visualisations cued by positive words. The relationship between severity of symptomatology and future imagery was predicted to be partly mediated by the severity of intrusive future-oriented thoughts. Lastly, it was hypothesised that there would be a positive relationship between future imagery and memory on: specificity, vividness, valence, trauma-relatedness, and response latency.
2. Methods 2.1. Participants
An a priori power analysis indicated that 61 participants should be recruited for the experiment in order to detect a medium-large effect (f 2 = 0.25, α = 0.05, power = 0.90). This medium-large
effect was based on the findings of Brown and colleagues (2013) and Kleim and colleagues (2014) who found medium to large effect sizes for differences in specificity of future imagery. The final sample was composed of 93 participants, who were between 18-66 years of age (M = 29.5 years, SD = 9.76 years) and of whom 50.5% were female (see Table 1 for further sample information). Two additional participants were excluded; one due to less than 50% of trials completed and the other due to poor English. Participants were recruited online through the Prolific Academic website, and received £7.50 for their participation. To prevent participants from attempting to complete the study more than once, the experiment was coded so as to only allow single attempts.
In order to take part in the study, participants had to have experienced a traumatic event meeting the DSM-5 Criterion A for PTSD (APA, 2013). Participants were pre-screened and not permitted to take part if they did not rate their level of English as either “native or bilingual” or “fluent or near-fluent”, or if they suffered from psychosis, severe head trauma or injury, or substance/alcohol abuse. Participants were required to be 18 years or older.
Participants were categorised as “full”, “subthreshold”, or “non-PTSD” (see Table 1). Full PTSD was defined as meeting DSM-5 Criteria B-E for PTSD (symptoms) and Criterion F (duration). Subthreshold PTSD was defined as meeting two or three criteria from Criteria B-E, as well as Criterion F (see McLaughlin et al., 2015, for a explanation and review of “subthreshold PTSD”) or meeting all Criteria B-E but not Criterion F (acute stress disorder). Non-PTSD was defined as meeting none or only one criterion from Criteria B-E. Participants cannot be clinically
diagnosed with PTSD due to the lack of complete assessment for Criteria G (functional significance) and H (exclusion), but it is expected that these groups will be acceptable.
Table 1. Sample characteristics.
Total sample (n = X) Variable n % Sex Male 46 49.5 Female 47 50.5 Employment status Employed or student 79 84.9 Unemployed or retired 14 15.1 Level of education
Did not graduate high school 0 0
High school diploma 12 12.9
Some university/college education 25 26.9
Bachelor’s degree 46 49.5
Master’s degree or doctorate degree (PhD) 10 10.8
First language
English 79 84.9
Other 14 15.1
Meets DSM-5 criteria for a diagnosis of PTSD
Clinical PTSD 11 11.8
Subthreshold PTSD 22 23.7
Non-PTSD 60 64.5
Traumatic exposure level
Single traumatic exposure 45 48.4
Multiple traumatic exposures 48 51.6
Sought treatment as a result of trauma 20 21.5
Has received behavioural or psychopharmacological
therapy for PTSD 3 3.2
M SD
Age 29.5 years 9.76 years
Time since worst traumatic experience 7.8 years 8.36 years
PCL-5 score 24.8 17.40
IFES score 37.2 15.84
Intrusion subscale score 14.4 6.34
BDI-II score 16.9 10.73
FAS score 37.9 15.81
Note: PCL-5 = PTSD Checklist for DSM-5 (PTSD symptomatology); IFES = Impact of Future Event
Scale (severity of intrusive future-oriented thoughts); BDI-II = Beck Depression Inventory II (depression symptomatology); FAS (verbal fluency).
2.2. Measures
The study was a correlational design. The three independent variables were: severity of PTSD symptomatology (correlational), severity of intrusive future-oriented thoughts (correlational), and valence of cue words (within-subjects). The six main dependent variables were: specificity,
vividness, participant-rated likelihood, participant-rated valence, and trauma-relatedness of future visualisations, and response latency. The secondary dependent variables were: specificity, vividness, participant-rated valence, and trauma-relatedness of past visualisations (memories) and response latency. Covariates were: depression symptomatology and verbal fluency scores. The experiment was conducted online and was approved by the Ethics Review Board of the University of Amsterdam (Faculty of Social and Behavioural Sciences).
2.2.1. Questionnaires
PTSD Checklist for DSM-5 (PCL-5)
PTSD symptomatology was assessed using the PCL-5 (Weathers et al., 2013). For this scale, participants are required to respond on the basis of their worst event, defined as “the event that currently bothers you the most”. They are first asked questions to ascertain the type of traumatic event that occurred (e.g. “Briefly identify the worst event”, “How did you experience it?”), before indicating on a 0-4 scale how frequently in the past month they experienced 20 different symptoms corresponding to the DSM-5 definition of PTSD (e.g. “Repeated, disturbing, and unwanted memories of the stressful experience”, “Feeling jumpy or easily startled”). Stating that they had experienced a symptom “Quite a bit” was interpreted as endorsement of that symptom. The PCL has previously been shown to have acceptable test-retest reliability (r = 0.66), good internal consistency (a meta-review found out of nineteen studies all report scores above 0.75), and high convergent validity with the Clinician-Administered PTSD Scale (Wilkins, Lang, & Norman, 2011).
Impact of Future Events Scale (IFES)
Frequency of IFT was assessed using the IFES (Deeprose & Holmes, 2010). Participants are asked to state three future events that they have been imagining lately, and to indicate if they are positive or negative, before rating 24 items for frequency on a 5-point Likert scale from “Not at all” to “Extremely”. Examples of items are: “I thought about the future when I didn’t mean to” and “My feelings about the future were kind of numb”. This scale has been shown in two nonclinical samples to have good test-retest reliability, with high correlation between IFES scores between tests (r = 0.73, p < 0.001) and no significant difference in number of negative events stated (χ2 (9, N = 48) = 6.89, p = 0.65), and also good internal consistency (Cronbach’s
alpha = 0.87) (Deeprose, Malik, & Holmes, 2011). Beck Depression Inventory II (BDI-II)
Depression symptomatology was measured using the BDI-II (Beck, Steer, & Brown, 1996). This is a 21-item scale commonly used to assess depression. Participants are presented with groups of statements evaluating different symptoms of depression (for example, feelings of worthlessness) and select the one that has most applied to them over the past two weeks. Scores of 14-28 are used to indicate mild to moderate depression; scores higher than 29 indicate severe depression. The inventory has high reliability (coefficient alpha estimates of around 0.92; Beck et al., 1996) and good validity (correlation of r = 0.71 with the Hamilton Psychiatric Rating Scale Depression-Revised in a psychiatric outpatient sample; Brown et al., 1996).
Verbal fluency was measured using the FAS (Lezak, 1976). This task requires participants to think of as many words as possible beginning with the letters F, A, and S in turn, and is scored by summing the number of acceptable words produced. This is a widely used, standard task, and was chosen due to its ease of integration with an online study.
Future-Self Questionnaire (FSQ)
Another, exploratory, measure of participants’ future visualisations was obtained using the FSQ (Morina, 2013). This questionnaire asks participants to imagine themselves in 1 year and in 5-10 years. Participants then complete 5 sentences beginning “I will...” for each time point regarding how they imagine themselves. Participants also rate each image for vividness and how confident they feel that it will occur. An additional variable was added to the FSQ and participants were therefore also asked to rate their image for valence. The FSQ is a novel questionnaire and no psychometric data about it exist.
Demographics
Participants lastly were asked to complete a demographics questionnaire, asking them: age, sex, time since last traumatic event, nationality, first language, if they had ever sought treatment as a result of the traumatic experience, and if they had received behavioural or psychopharmacological treatment for PTSD.
2.2.2. Visualisation task AMT and AMT-Future
In the AMT and AMT-Future (Williams & Broadbent, 1986), participants are presented with a cue word on screen and asked to generate either a memory (AMT) or a possible future event (AMT-Future) associated with this word. For the version of these tasks used in this experiment, once they had generated a visualisation, participants pressed a button to proceed to the next screen where they could freely describe the memory or future event that they had generated by typing it out. Participants had a maximum of three minutes to generate a visualisation before the experiment continued automatically to the next trial. Participants were additionally asked to rate on three Likert/Likert-type scales how vivid their visualisation was, the valence of their visualisation, and how likely they thought it was that this visualisation could one day occur (for future imagery only). The list of cue words used in this task was taken from the study of Kleim and colleagues (2014; for details on generation see Kleim & Ehlers, 2008). This list consists of six positive (e.g. “pleased”) and six negative (e.g. “hopeless”) words. The same words were used in the generation of both past and future visualisations.
Vividness, participant-rated likelihood, and valence scales
Vividness and participant-rated likelihood of visualisations were measured through two one-item 5-point Likert-type scales. Valence of visualisations was measured with a 7-point Likert scale. Participants were asked “How vivid was your visualisation?” (from “no image at all” to “very vivid”), “How likely do you think it is that your visualisation will one day happen to you in real life?” (from “very unlikely / never” to “very likely / certain”), and “Do you feel that visualisation you have is a positive or negative one?” (from “very negative” to “very positive”).
2.3. Procedure
Before beginning the experiment, potential participants underwent the screening procedure. This consisted of questions ascertaining that they had experienced a traumatic event, were not receiving treatment for PTSD, did not suffer from psychosis, head trauma or injury, or substance/alcohol abuse, spoke an appropriate level of English, and were 18 years of age or above. Potential participants who responded in such a way that they did not meet the inclusion criteria were then redirected to an exit page. Participants who did meet the inclusion criteria then read an information brochure and signed an electronic consent form.
Participants then completed five scales or tasks to assess PTSD symptomatology, severity of intrusive future-oriented thoughts, image of future self, depression, and verbal fluency. These were presented in randomised order. Following this, they completed the visualisation task (AMT and AMT-Future).
The visualisation task was first explained to the participant including examples of appropriate and inappropriate visualisations of past and future events. Appropriate past visualisations had to be of any personal memory of an event which had occurred on a specific day at some point in their past. Appropriate future visualisations had to be of a personally relevant future event which could occur on a specific day in the future. For future visualisations, events could be already scheduled to occur (e.g. their birthday) or a fantasy (e.g. buying a house), but had to be physically possible to occur. Participants were encouraged to give as much detail as possible (where, when, who, etc.). The experiment consisted of four practice trials (neutral cue words) followed by 24 experimental trials (positive and negative cue words; both past and future visualisations). All trials were presented in randomised order. At the end of the experiment, participants completed the demographics questionnaire.
2.4. Scoring
Participants’ responses to each cue word were scored by the experimenter for specificity and trauma-relatedness. Specificity was scored on a 0 (no response) to 3 (specific visualisation) scale taken from the study of Brown and colleagues (2013). Specificity was scored with a 3 if participants gave a visualisation of an event which could take place within a 24 hour timeframe and included details such as people, location, or time. Specificity was scored with a 2 if participants gave an event which took place over more than 24 hours or which was a repeated event. Specificity was scored with a 1 if they responded with something other than a future event or memory (depending on type of trial), or if their visualisation was especially vague and/or lacking details. Omissions were scored with a 0. Trauma-relatedness was scored as absence or presence of “worst trauma”-related content, or presence of “other trauma”-related/pathological content. “Worst trauma”-related content was defined as content in the visualisation referring to either the trauma reported by that participant in the PCL-5 as most bothering, or its consequences. An example of “worst trauma”-related content is “I am supposed to be helping my Mum to clear out my Nan’s house next week” (trauma: grandmother dying from cancer). “Other trauma”-related/pathological content was defined as references to other events which would meet the DSM-5 Criterion A as a traumatic event, but which was not reported by the participant as their most bothering experience (for example, death of family members, severe medical illness), or references to depression or extreme hopelessness.
2.5. Statistical analyses
The Statistical Package for the Social Sciences (SPSS version 22) was used for all analyses. Repeated measures ANOVAs were used to investigate if PTSD symptomatology (operationalised as PCL-5 scores; between-subjects factor) and cue valence (within-subjects factor) predicted characteristics of future imagery including specificity, vividness, participant-rated likelihood of occurrence, valence, trauma-relatedness, response latency, and number of failures to generate a visualisation (omissions). For each characteristic, moderation by cue valence was also investigated. Characteristics of future imagery (except for trauma-relatedness and omissions) were operationalised as the average score for each; failures to generate a visualisation in response to a cue word were treated as missing data and thus excluded, except for specificity where they were scored with a 0 following the approach of Brown and colleagues (2013). Trauma-relatedness was operationalised as the sum total of images containing trauma-related content. Omissions were operationalised as the sum total of trials where the participant did not provide any detail about their visualisation, and presumably failed to generate an image. Simple linear regressions were then used to investigate if PTSD symptomatology predicted these characteristics both for images generated in response to positive cue words and to negative cue words. Repeated measures ANOVAs were again used to investigate if severity of intrusive future-oriented thoughts (operationalised as IFES scores; between-subjects factor), depression symptomatology (operationalised as BDI scores; between-subjects factor), and verbal fluency (operationalised as FAS scores; between-subjects factor) separately predicted characteristics of future imagery found to be significantly related to PTSD symptomatology. Moderation by cue valence was also investigated, and simple linear regressions were again performed.
Mediation analyses were conducted using Process Macro (version 2.13) to investigate if the relationships between PTSD symptomatology and characteristics of future images were partly mediated by severity of intrusive future-oriented thoughts. The model was tested as in Figure 1, with cue valence as a moderator on both the predictor and the mediator. Direct and indirect effects were tested using bootstrapping procedures and a 95% confidence interval was computed. Mediation analyses were only performed where both the predictor (PTSD symptomatology) and mediator had been shown to be significantly related (p < 0.05) to the dependent variable.
Pearson correlations and repeated measures ANOVAs were performed to analyse the relationship between past and future visualisations. For these analyses, specificity, vividness, valence, trauma-relatedness (worst and all), response latency, and number of omissions of visualisations were investigated. ANOVAs were performed using time orientation (past and future) and cue word valence (positive and negative) as within-subjects factors, and PTSD status (clinical/subthreshold and non-PTSD) as a between-subjects factor. Correlation analyses were performed separately for positive and negative cue words.
For the exploratory analyses investigating characteristics of imagery generated with the FSQ, repeated measures ANOVAs were performed using PTSD symptomatology as a between-subjects factor and time point as a within-between-subjects factor. Dependent variables were vividness, participant-rated likelihood of occurrence, valence, and trauma-relatedness of imagery. These were operationalised in the same way as above. Further analyses on the FSQ were also performed using repeated measures ANOVAs with PTSD symptomatology as a
between-subjects factor and image valence as a within-between-subjects factor. Image data generated with the FSQ was compared to image data generated with the AMT-f using Pearson correlations (separately for positive and negative imagery) and repeated measures ANOVAs, using PTSD symptomatology as a between-subjects factor and task type and image type (positive or negative) as within-subjects factors. Final analyses on relationships between the different characteristics of future imagery (vividness, likelihood of occurrence, valence, and “worst trauma”-relatedness) generated with the AMT-f were conducted with Pearson correlations.
Figure 1. Model of the relationship between PTSD symptomatology and future imagery, as mediated by severity of intrusive future-oriented thoughts, depression symptomatology, and verbal fluency.
3. Results
All participants met the DSM-5 Criterion A for PTSD (traumatic event). Of the sample of 93 participants, 11 met DSM-5 Criteria B-E for PTSD (symptoms) as well as Criterion F (duration). These participants will be referred to as having “full PTSD”. A further 21 participants met two or three criteria out of Criteria B-E, as well as Criterion F, and were therefore categorised as having subthreshold PTSD. One participant met Criteria B-E but did not meet Criterion F, and was therefore also categorised as having “subthreshold PTSD”. Participants with full or subthreshold PTSD will be referred to “participants with PTSD” for the sake of brevity. The remainder (n = 60) were categorised as “non-PTSD participants”.
Participants with full or subthreshold PTSD did not significantly differ from participants without PTSD in: age(PTSD: M = 30.3 years, SD = 9.32 years; non-PTSD: M = 29.0 years, SD = 10.04 years; t(91) = 0.599, p = 0.550, d = 0.130, n.s.), sex (PTSD: N female = 18, % = 54.5; non-PTSD: N female = 29, % = 48.3; χ2
(1) = 0.329, p = 0.566, d = 0.121, n.s.), education level
Severity of intrusive future-oriented thoughts
PTSD
symptomatology Future imagery
Depression symptomatology Verbal fluency β = 0.628 β = 0.583 β = -0.450 Cue valence
(χ2
(3) = 5.167, p = 0.160, n.s.), employment status (PTSD: N employed/student = 30, % = 90.9;
non-PTSD: N employed/student = 49, % = 81.7; χ2
(1) = 1.422, p = 0.233, d = 0.252, n.s.), or
English as a first language (PTSD: N English = 27, % = 81.8; non-PTSD: N English = 52, % = 86.7; χ2
(1) = 0.391, p = 0.532, d = 0.131, n.s.). Participants also did not differ in their level of
traumatic exposure (PTSD: N multiple traumatic events = 17, % = 51.5; non-PTSD: N multiple traumatic events = 31, % = 51.7; χ2
(1) = 0.000, p = 0.989, d = 0.003, n.s.). Additionally, there was
no significant difference in the length of time that had elapsed since their worst traumatic event (PTSD: M = 7.3 years, SD = 9.27 years; non-PTSD: M = 8.0 years, SD = 7.87 years; t(91) =
-0.384, p = 0.702, d = 0.083, n.s.) nor their last traumatic event (PTSD: M = 4.5 years, SD = 5.70 years; non-PTSD: M = 4.32 years, SD = 5.57 years; t(84) = 0.143, p = 0.887, n.s.) between
participants with and without PTSD. Participants with PTSD were not more likely to have sought treatment as a result of their traumatic experience (PTSD: N sought treatment = 10, % = 30.3; non-PTSD: N sought treatment = 10, % = 16.7; χ2
(1) = 2.345, p = 0.126, d = 0.325, n.s.);
however participants with full or subthreshold PTSD were more likely to have received behavioural or psychopharmacological treatment for PTSD (PTSD: N received treatment = 3, % = 9.1; non-PTSD: N received treatment = 0, % = 0; χ2
(1) = 5.636, p = 0.018, d = 0.514). All
three participants who had received treatment for PTSD were participants categorised as currently having full PTSD.
Participants with full or subthreshold PTSD had a mean PCL-5 score of 43.2 (SD = 14.00), whereas non-PTSD participants had a mean PCL-5 score of 14.8 (SD = 8.69). Participants with full or subthreshold PTSD experienced additionally significantly higher levels of depression symptoms (PTSD: M = 22.9, SD = 10.47; non-PTSD: M = 13.6, SD = 9.41; t(91) =
4.392, p < 0.001, d = 0.952) and intrusive future-oriented thoughts (PTSD: M = 49.2, SD = 12.38; non-PTSD: M = 30.7, SD = 13.61; t(91) = 6.460, p < 0.001, d = 1.400) than non-PTSD
participants. When asked to name three future events that they had been thinking of lately, participants with PTSD did not differ in the number of positive future events named compared to non-PTSD participants (PTSD: M = 1.7, SD = 0.98; non-PTSD: M = 1.9, SD = 0.82; t(91) =
-1.065, p = 0.290, d = 0.231, n.s.). This indicates that participants with PTSD were not more likely to have been thinking of negative future events lately compared to participants without PTSD. Finally, participants with full or subthreshold PTSD were less verbally fluent in comparison with participants without PTSD (PTSD: M = 30.3, SD = 17.23; non-PTSD: M = 41.9, SD = 13.54; t(91) = -3.473, p = 0.001, d = 0.753); verbal fluency between native and
non-native English speakers, however, did not differ (non-native English speakers: M = 38.7, SD = 15.95; non-native English speakers: M = 33.1, SD = 14.63; t(91) = 1.214, p = 0.228, d = 0.352, n.s.). 3.1. PTSD symptomatology and future imagery
The first two hypotheses concerned the relationships between PTSD symptomatology and various characteristics of future imagery, and whether these relationships were moderated by the valence of the cue word used to generate images. Characteristics of future imagery included: specificity, vividness, participant-rated likelihood of occurrence, valence, trauma-relatedness, and response latency (see Table 2 for mean scores for each characteristic).
Severity of PTSD symptomatology significantly predicted specificity of visualisations (F(1,91) = 9.927, p = 0.002, η2 = 0.098). Cue valence was also a significant predictor in the model
ones, although there was no significant moderation of cue valence on PTSD symptomatology (F(1,91) = 0.840, p = 0.362, η2 = 0.009, n.s.). In simple linear regressions, PTSD symptomatology
negatively predicted the specificity of both positive images (R2 = 0.107, F
(1,91) = 10.903, β =
-0.327, p = 0.001) and negative images (R2 = 0.060, F
(1,91) = 5.802, β = -0.245, p = 0.018), with
higher scores of symptomatology being associated with less specific images.
There was no significant effect of PTSD symptomatology on vividness of visualisations (F(1,91) = 0.558, p = 0.457, η2 = 0.006, n.s.). Again, vividness differed depending on the cue
valence (F(1,91) = 8.606, p = 0.002, η2 = 0.102) with positive images being rated as more vivid, but
again there was no significant moderation by cue valence (F(1,91) = 0.757, p = 0.387, η2 = 0.008,
n.s.). Simple linear regressions also revealed no relationship between PTSD symptomatology and image vividness for both positive images (R2 = 0.001, F
(1,91) = 0.119, β = 0.036, p = 0.730, n.s.)
and negative images (R2 = 0.012, F
(1,91) = 1.072, β = 0.108, p = 0.303, n.s.).
PTSD symptomatology did not predict the confidence that participants had in their visualisations occurring one day (F(1,91) = 0.106, p = 0.745, η2 = 0.001, n.s.). Participants were
significantly more likely to think that their positive visualisations would occur one day than their negative visualisations (F(1,91) = 8.533, p = 0.004, η2 = 0.086) but there was no significant
moderation effect (F(1,91) = 0.160, p = 0.690, η2 = 0.002, n.s.). Simple linear regressions also
revealed no relationship between PTSD symptomatology and participant-rated confidence in images for both positive visualisations (R2 = 0.000, F
(1,91) = 0.004, β = -0.006, p = 0.953, n.s.) and
negative visualisations (R2 = 0.003, F
(1,91) = 0.280, β = -0.055, p = 0.598, n.s.).
PTSD symptomatology also did not predict how positive or negative participants rated their visualisations (F(1,91) = 0.263, p = 0.609, η2 = 0.003, n.s.). Participants rated their imagery
generated in response to positive cue words as significantly more positive than that generated in response to negative cue words (F(1,91) = 309.746, p < 0.001, η2 = 0.773) but there was no
significant moderation effect (F(1,91) = 0.177, p = 0.675, η2 = 0.002, n.s.). PTSD symptomatology
was not related to valence of either positive images (R2 = 0.000, F
(1,91) = 0.012, β = -0.011, p =
0.914, n.s.) or negative images (R2 = 0.003, F
(1,91) = 0.271, β = -0.054, p = 0.604, n.s.).
However, PTSD symptomatology did significantly predict both the number of “worst trauma”-related images (F(1,91) = 17.490, p < 0.001, η2 = 0.161) and the total number of
trauma-related images (F(1,91) = 5.430, p = 0.022, η2 = 0.056) generated by participants. For the number
of “worst trauma”-related images there was no significant effect of cue valence (F(1,91) = 0.064, p
= 0.801, η2 = 0.001, n.s.) but there was a trend towards a significant moderation effect between
cue valence and PTSD symptomatology (F(1,91) = 3.456, p = 0.066, η2 = 0.037). Simple linear
regressions revealed that higher scores of PTSD symptomatology were significantly associated with more “worst trauma”-related images for both positive imagery (R2 = 0.159, F
(1,91) = 17.148,
β = 0.398, p < 0.001) and negative imagery (R2 = 0.131, F
(1,91) = 13.740, β = 0.362, p < 0.001).
For the total number of trauma-related images (“all trauma”-related images) there was a significant effect of cue valence (F(1,91) = 20.131, p < 0.001, η2 = 0.181) but no significant
moderation (F(1,91) = 0.123, p = 0.726, η2 = 0.001, n.s.). However, simple linear regressions
revealed a significant positive relationship between PTSD symptomatology and number of trauma-related images for positive cue words (R2 = 0.091, F
(1,91) = 9.059, β = 0.301, p = 0.003),
yet not for negative cue words (R2 = 0.020, F
(1,91) = 1.843, β = 0.141, p = 0.178, n.s.).
PTSD symptomatology also did not predict participants’ response latency in generating visualisations (F(1,91) = 0.808, p = 0.371, η2 = 0.009, n.s.). Positive images were generated
was no significant moderation effect (F(1,91) = 1.400, p = 0.240, η2 = 0.015, n.s.). PTSD
symptomatology did not significantly predict response latency for either positive (R2 = 0.001,
F(1,91) = 0.124, β = -0.037, p = 0.726, n.s.) or negative images (R2 = 0.015, F(1,91) = 1.427, β =
-0.124, p = 0.235, n.s.).
Finally, severity of PTSD symptomatology was not associated with the number of failures to generate an image, or number of omitted trials (F(1,91) = 0.014, p = 0.905, η2 = 0.000,
n.s.). There was a trend towards more omissions on negative trials (F(1,91) = 3.826, p = 0.054, η2 =
0.040) but again, no significant moderation by cue valence (F(1,91) = 0.426, p = 0.515, η2 = 0.005,
n.s.). In simple linear regressions, there was no relationship for either positive imagery (R2 =
0.001, F(1,91) = 0.073, β = 0.028, p = 0.787, n.s.) or negative imagery (R2 = 0.002, F(1,91) = 0.150, β
= -0.041, p = 0.699, n.s.).
Table 2. Characteristics of past and future visualisations in the visualisation task. Future visualisations (future events)
PTSD (full/subthreshold) Non-PTSD
Positive Negative Positive Negative
Variable M SD M SD M SD M SD Specificity 1.9 0.66 1.8 0.55 2.3 0.55 2.0 0.65 Vividness 4.0 0.78 3.8 0.73 3.8 0.68 3.5 0.69 Likelihood of occurrence 3.9 0.60 3.5 0.51 3.9 0.63 3.5 0.55 Valence 2.4 0.55 -1.5 1.00 2.3 0.51 -1.5 1.10 All trauma-relatedness 0.7 1.04 1.3 1.53 0.2 0.50 1.4 1.20 Worst trauma-relatedness 0.3 0.92 0.6 1.39 0.0 0.13 0.2 0.68 Response latency 16.5 22.64 20.5 25.85 16.2 16.14 22.0 21.89 Number of omissions 0.3 0.72 0.48 0.83 0.4 0.69 0.57 0.96 Past visualisations (memories)
PTSD (full/subthreshold) Non-PTSD
Positive Negative Positive Negative
Variable M SD M SD M SD M SD Specificity 2.1 0.63 2.2 0.53 2.5 0.51 2.5 0.50 Vividness 4.4 0.58 4.3 0.69 4.5 0.37 4.3 0.47 Valence 2.3 0.64 -1.8 0.93 2.1 0.61 -1.5 0.89 All trauma-relatedness 0.6 0.87 1.9 1.66 0.5 0.80 1.4 1.38 Worst trauma-relatedness 0.3 0.64 1.3 1.68 0.51 0.51 0.6 1.14 Response latency 19.5 30.59 18.7 25.11 17.7 15.60 19.5 15.92 Number of omissions 0.5 0.87 0.1 0.33 0.3 0.62 0.3 0.65
Note: For this table, individual participant scores for variables were calculated by averaging scores for each
visualisation, and then averaged: specificity (0-3); vividness and likelihood of occurrence (1-5); valence (-3 to +3 where a negative score indicates negative valence and a positive score indicates positive valence). All trauma-relatedness was calculated by the number of trauma-related (worst/other) visualisations per participant (0-6); worst trauma-relatedness was calculated by the number of visualisations referring to the participant’s self-reported “worst traumatic experience” (0-6). Average response latency was calculated across participants excluding omissions. Number of omissions was calculated by the number of trials per participant where they failed to generate an image (0-6).
3.2. PTSD symptomatology, intrusive future-oriented thoughts, and future imagery The third hypothesis was that the relationships between the severity of PTSD symptomatology and characteristics of future imagery would be partly mediated by the severity of intrusive oriented thoughts. PTSD symptomatology significantly predicted severity of intrusive future-oriented thoughts (R2 = 0.394, F
(1,91) = 59.202, β = 0.628, p < 0.001), with higher numbers of
PTSD symptoms predicting more frequent intrusive future-oriented thoughts. For these analyses, only specificity, number of “worst trauma”-related images, and total number of trauma-related images (“all trauma”-trauma-related images) were investigated. This is due to all other characteristics of future imagery not being significantly related to PTSD symptomatology and therefore inappropriate to enter into a mediation analysis1.
There was a trend towards severity of intrusive future-oriented thoughts predicting the specificity of images (F(1,91) = 2.881, p = 0.093, η2 = 0.031) although cue valence did not
moderate this relationship (F(1,91) = 0.131, p = 0.719, η2 = 0.001, n.s.). Simple linear regressions
revealed a trending relationship between intrusive future-oriented thoughts and specificity of positive images (R2 = 0.031, F
(1,91) = 2.918, β = -0.176, p = 0.091), with more severe thoughts
predicting less specific images, yet no relationship with the specificity of negative images (R2 =
0.020, F(1,91) = 1.903, β = -0.143, p = 0.171, n.s.).
Severity of intrusive future-oriented thoughts did not predict the number of “worst trauma”-related images (F(1,91) = 2.042, p = 0.156, η2 = 0.022, n.s.) and nor was this moderated by
cue valence (F(1,91) = 0.010, p = 0.919, η2 = 0.000, n.s.). Additionally, severity of intrusive
future-oriented thoughts did not have a significant relationship with the total number of trauma-related images (F(1,91) = 0.238, p = 0.627, η2 = 0.003, n.s.); again this was not moderated by cue valence
(F(1,91) = 1.033, p = 0.312, η2 = 0.011, n.s.). In simple linear regressions, there was a trend
towards higher severity of intrusive future-oriented thoughts predicting more “worst trauma”-related images for positive cue words (R2 = 0.034, F
(1,91) = 3.182, β = 0.184, p = 0.078) yet not
for negative cue words (R2 = 0.013, F
(1,91) = 1.167, β = 0.113, p = 0.283, n.s.). There was no
significant relationship between severity of intrusive future-oriented thoughts and the number of “all trauma”-related images for either positive imagery (R2 = 0.021, F
(1,91) = 1.939, β = 0.144, p =
0.167, n.s.) or negative imagery (R2 = 0.000, F
(1,91) = 0.024, β = -0.016, p = 0.878, n.s.).
It was then decided to examine more carefully severity of intrusive future-oriented thoughts by investigating the relationships between characteristics of future imagery and scores on the intrusion subscale of the IFES. PTSD symptomatology significantly positively predicted intrusion subscale scores (R2 = 0.244, F
(1,91) = 29.335, β = 0.494, p < 0.001). Repeated measures
ANOVAs were used to investigate associations between intrusion subscale scores and specificity, “worst trauma”-relatedness, and “all trauma”-relatedness of imagery.
Intrusion subscale scores significantly predicted the specificity of images (F(1,91) = 4.436, p = 0.038, η2 = 0.046) and was not moderated by cue valence (F
(1,91) = 0.000, p = 0.983, η2 = 0.000,
1 Baron and Kenny (1986) describe the necessary prerequisites to a mediation analysis as follows: firstly, there must
be a significant relationship between the independent variable (in this case, PTSD symptomatology) and the mediator (in this case, severity of intrusive future-oriented thoughts); secondly, the independent variable must significantly predict the dependent variable (in this case, the different characteristics of future imagery); and thirdly, there must be significant relationships between the mediator and the dependent variables. Due to the conditions of the second prerequisite failing to be met in this case, and thus a mediation analysis being inappropriate (and a mediational relationship being the focus of this paper), it was decided not to report the relationships with the dependent variables that had already been found to be not significantly related to the independent variable.
n.s.). Simple linear regressions however only found trending negative relationships, with intrusion subscale scores tending to predict less specific images for both positive (R2 = 0.039,
F(1,91) = 3.660, β = -0.197, p = 0.059) and negative (R2 = 0.038, F(1,91) = 3.642, β = -0.196, p =
0.059) cue words.
Furthermore, participants’ scores on the intrusion subscale significantly predicted the number of “worst trauma”-related images generated (F(1,91) = 4.098, p = 0.046, η2 = 0.043),
although this was not moderated by cue valence (F(1,91) = 0.911, p = 0.342, η2 = 0.010, n.s.).
Intrusion subscale scores significantly positively predicted number of “worst trauma”-related images in response to positive cue words (R2 = 0.042, F
(1,91) = 4.010, p = 0.048) but only trended
towards predicting number in response to negative cue words (R2 = 0.035, F
(1,91) = 3.315, β =
0.187, p = 0.072). However, intrusion subscale scores were not found to be associated with the total number of trauma-related images (F(1,91) = 0.224, p = 0.637, η2 = 0.002, n.s.) and nor was
this moderated by cue valence (F(1,91) = 1.261, p = 0.264, η2 = 0.014, n.s.). Additionally, simple
linear regressions revealed no significant relationships for either positive cue words (R2 = 0.023,
F(1,91) = 2.141, β = 0.152, p = 0.147, n.s.) or negative cue words (R2 = 0.001, F(1,91) = 0.046, β =
-0.022, p = 0.831, n.s.).
To check that neither depression nor verbal fluency explained the relationship between PTSD symptomatology and future imagery instead of severity of intrusive future-oriented thoughts, these were added as additional mediators. PTSD symptomatology significantly predicted both depression symptomatology (R2 = 0.340, F
(1,91) = 46.797, p < 0.001; β = 0.583, p
< 0.001) and verbal fluency (R2 = 0.203, F
(1,91) = 23.116, p < 0.001; β = -0.450, p < 0.001), with
higher numbers of PTSD symptoms predicting more depression symptoms and lower verbal fluency scores.
Depression symptomatology was not found to be significantly associated with the specificity of visualisations (F(1,91) = 0.001, p = 0.977, η2 = 0.000, n.s.) nor was this moderated by
cue valence (F(1,91) = 0.042, p = 0.837, η2 = 0.000, n.s.). Simple linear regressions equally revealed
no relationship on either positive trials (R2 = 0.000, F
(1,91) = 0.003, β = -0.006, p = 0.953, n.s.) or
negative trials (R2 = 0.000, F
(1,91) = 0.012, β = 0.012, p = 0.912, n.s.). There was however a
significant relationship between depression symptomatology and the number of “worst trauma”-related images generated by participants (F(1,91) = 6.427, p = 0.013, η2 = 0.066), with higher
depression symptomatology significantly predicting more “worst trauma”-related images in response to positive cue words (R2 = 0.102, F
(1,91) = 10.332, β = 0.319, p = 0.002) but only
trending towards predicting more images in response to negative cue words (R2 = 0.038, F
(1,91) =
3.588, β = 0.195, p = 0.061). There was no significant moderation effect (F(1,91) = 0.028, p = 0.868, η2 = 0.000, n.s.). As well as this, there was a significant association with the total number
of trauma-related images (F(1,91) = 6.372, p = 0.013, η2 = 0.065) which was not moderated by cue
valence (F(1,91) = 0.152, p = 0.697, η2 = 0.002, n.s.). Higher depression symptomatology
significantly predicted higher numbers of trauma-related images for positive cue words (R2 =
0.106, F = 10.776, β = 0.25, p = 0.001) but not for negative cue words (R2 = 0.023, F
(1,91) =
2.132, β = 0.151, p = 0.148, n.s.).
Verbal fluency showed a significant association with specificity of visualisations (F(1,91) =
12.465, p = 0.001, η2 = 0.120) but was not moderated by cue valence (F
(1,91) = 0.024, p = 0.878,
η2 = 0.000, n.s.). Significant relationships were found for both positive (R2 = 0.105, F
(1,91) =
10.732, β = 0.325, p = 0.001) and negative (R2 = 0.095, F
(1,91) = 9.519, β = 0.308, p = 0.003) cue
significantly associated with the number of “worst trauma”-related images generated (F(1,91) =
2.095, p = 0.151, η2 = 0.023, n.s.), which was not moderated by cue valence (F
(1,91) = 0.329, p =
0.567, η2 = 0.004, n.s.), nor the total number of trauma-related images generated (F
(1,91) = 0.000,
p = 0.988, η2 = 0.000, n.s.), again not moderated by cue valence (F
(1,91) = 1.970, p = 0.164, η2 =
0.026, n.s.). Verbal fluency did however predict the number of “worst trauma”-related images generated in response to positive cue words (R2 = 0.053, F
(1,91) = 5.074, β = -0.230, p = 0.027)
but not in response to negative cue words (R2 = 0.008, F
(1,91) = 0.727, β = -0.089, p = 0.396, n.s.)
in simple linear regressions. Verbal fluency was not related to the total number of trauma-related images for either positive cue words (R2 = 0.017, F
(1,91) = 1.568, β = -0.130, p = 0.214, n.s.) or
negative ones (R2 = 0.006, F
(1,91) = 0.559, β = 0.078, p = 0.456, n.s.).
Mediation analyses were intended to be conducted to investigate if the relationships between PTSD symptomatology and characteristics of future images were partly mediated by severity of intrusive future-oriented thoughts. However, due to the lack of significant relationships found between IFES scores and future imagery, it was considered inappropriate to conduct these analyses2. Instead, mediation analyses were run using intrusion subscale scores of
the IFES as the main mediator.
The relationship between PTSD symptomatology and specificity of future visualisations was not mediated by intrusive future-oriented thoughts. The conditional direct effect of PTSD symptomatology on specificity in this model was significant for positive images (effect = -0.011, CI [-0.019, -0.003], p = 0.010), but not for negative images (effect = -0.007, CI [-0.015, 0.001], p = 0.091). The conditional indirect effect was not significant for either positive (effect = -0.001, CI [-0.005, 0.003]) or negative (effect = -0.002, CI [-0.007, 0.002]) images. The relationship between PTSD symptomatology and specificity of future visualisations was mediated by verbal fluency, but only for negative imagery. The conditional direct effect in this model was again significant for positive images (effect = -0.008, CI [-0.016, -0.0003], p = 0.042), but not for negative images (effect = -0.005, CI [-0.013, -0.003], p = 0.225, n.s.). The conditional indirect effect was significant for negative images (effect = -0.004, CI [-0.009, -0.0001]) but not for positive images (effect = -0.004, CI [-0.008, 0.001]). However, the index of moderated mediation was overall not significant (index = 0.0004, CI [-0.006, 0.007]). Thus, verbal fluency significantly mediatedfor the specificity of negative visualisations, but not overall.
The relationship between PTSD symptomatology and the number of “worst trauma”-related visualisations was not mediated by intrusive future-oriented thoughts. The conditional direct effect of PTSD symptomatology was significant in this model for both positive (effect = 0.013, CI [0.003, 0.023], p = 0.015) and negative (effect = 0.021, CI [0.010, 0.031], p < 0.001) visualisations. Conditional indirect effects however were not significant for either positive (effect = 0.0002, CI [-0.003, 0.002]) or negative (effect = 0.0003, CI [-0.005, 0.006]) visualisations. The relationship between PTSD symptomatology and the number of “worst trauma”-related visualisations was also not mediated by depression symptomatology. In this model, the conditional direct effect of PTSD symptomatology was significant for negative visualisations (effect = 0.022, CI [0.011, 0.033], p < 0.001), yet only trending for positive visualisations (effect = 0.011, CI [-0.001, 0.022], p = 0.063). Conditional indirect effects were not significant for either positive visualisations (effect = 0.003, CI [0.002, 0.011]) or negative visualisations (effect = -0.001, CI [-0.010, 0.011]). Finally, the relationship between PTSD symptomatology and the
number of “worst trauma”-related visualisations was not mediated by verbal fluency. Conditional direct effects in this model were significant for both positive visualisations (effect = 0.012, CI [0.002, 0.022], p = 0.019) and negative visualisations (effect = 0.023, CI [0.013, 0.034], p < 0.001). Conditional indirect effects were not significant for either positive (effect = 0.001, CI [-0.002, 0.005]) or negative (effect = -[-0.002, CI [-0.009, 0.003]) visualisations.
The relationship between severity of intrusive future-oriented thoughts and the total number of trauma-related (“all trauma”-related) visualisations was not significant, thus this was not tested in a mediation analysis. The relationship between PTSD symptomatology and the total number of trauma-related visualisations was not found to be mediated by depression symptomatology. Conditional direct effects of PTSD symptomatology were not significant for either positive visualisations (effect = 0.007, CI [-0.008, 0.023], p = 0.343, n.s.) or negative visualisations (effect = 0.006, CI [-0.009, 0.022], p = 0.441, n.s.). Additionally, conditional indirect effects of depression symptomatology were not significant for either positive visualisations (effect = 0.006, CI 0.001, 0.015]) or negative visualisations (effect = 0.005, CI [-0.007, 0.017]).
3.3. Future imagery and memory
The fourth hypothesis was that there would be a significant positive relationship between characteristics of future imagery and memory, including specificity, vividness, valence, trauma-relatedness and how fast participants were to generate visualisations. The data of one participant (non-PTSD) was lost due to computer error. Table 2 displays the means of characteristics of memories.
Participants who generated more specific future imagery were more likely to also generate more specific past imagery, or memories (positive cue words: r = 0.704, p < 0.001; negative cue words: r = 0.621, p < 0.001). Participants who rated their future visualisations as more vivid were also more likely to rate their memories as vivid as well (positive cue words: r = 0.566, p < 0.001; negative cue words: r = 0.550, p < 0.001). Participants who rated their imagery as more positive (in response to positive cue words) were also more likely to report that the memories they generated were positive as well (r = 0.438, p < 0.001), whereas for negative cue words participants who reported their imagery as more negative were also more likely to generate negative memories (r = 0.400, p < 0.001). Furthermore, participants who incorporated content relating to their worst trauma in their future visualisations were also likely to do so in their past visualisations (positive cue words: r = 0.260, p = 0.012; negative cue words: r = 0.551, p < 0.001); however, the predisposition to incorporate any sort of trauma-related content (“all trauma”-related) into future visualisations was only associated with the same predisposition in negative past visualisations and not memories generated in response to positive cue words (positive cue words: r = 0.122, p = 0.248, n.s.; negative cue words: r = 0.414, p < 0.001). Participants who had faster response latencies for generating future visualisations were also more likely to have faster response latencies when generating past visualisations (positive cue words: r = 0.588, p < 0.001; negative cue words: r = 0.813, p < 0.001). Lastly, participants who failed to generate future images were also likely to fail to generate past images, both for positive cue words (r = 0.596, p < 0.001) and negative cue words (r = 0.325, p = 0.002).
ANOVAs conducted revealed a significant difference in the specificity of past and future visualisations (F(1,90) = 22.876, p < 0.001, η2 = 0.203), with memories being more specific than
future imagery, as well as in the specificity of positive and negative imagery (F(1,90) = 9.693, p =
0.002, η2 = 0.097). PTSD symptomatology was also associated with the specificity of
visualisations (F(1,90) = 15.015, p < 0.001, η2 = 0.143) and was moderated by cue valence (F(1,90) =
4.302, p = 0.041, η2 = 0.046). There were no other significant interactions.
There was a significant difference in vividness ratings with past visualisations being rated as more vivid than future ones (F(1,90) = 68.061, p < 0.001, η2 = 0.431); furthermore there was a
significant interaction between time orientation and severity of PTSD symptomatology (F(1,90) =
4.118, p = 0.045, η2 = 0.044) with participants with fewer PTSD symptoms rating their future
imagery as less vivid. As expected positive images were also rated more vividly than negative ones (F(1,90) = 13.896, p < 0.001, η2 = 0.134) but there was no association between the vividness
ratings and PTSD symptomatology of participants (F(1,90) = 0.000, p = 0.989, η2 = 0.000, n.s.).
No other interactions were significant.
Participants rated their images generated in response to positive cue words significantly more positively than their images generated in response to negative cue words (F(1,90) = 411.180,
p < 0.001, η2 = 0.820). Participants did not rate their future imagery as any more positive than
their past visualisations (F(1,90) = 1.910, p = 0.170, η2 = 0.021, n.s.). Again there was no significant
with severity of PTSD symptomatology (F(1,90) = 0.542, p = 0.464, η2 = 0.006, n.s.) and no
significant interactions between variables.
Participants did not incorporate any more content related to their worst trauma for past visualisations in comparison with future visualisations (F(1,90) = 2.185, p = 0.143, η2 = 0.024, n.s.)
nor in response to positive cue words compared to negative cue words (F(1,90) = 0.479, p = 0.491,
η2 = 0.005, n.s.). There was however a significant association with severity of PTSD
symptomatology (F(1,90) = 24.374, p < 0.001, η2 = 0.213), and a significant interaction between
severity of PTSD symptomatology and cue word valence (F(1,90) = 9.917, p = 0.002, η2 = 0.099).
No other interactions were significant. Furthermore, participants did not incorporate more content related to any trauma in their past compared to future visualisations (F(1,90) = 0.010, p =
0.922, η2 = 0.000, n.s.). Participants were however more likely to incorporate “all trauma”-related
content in response to negative cue words (F(1,90) = 21.581, p < 0.001, η2 = 0.193) or if they had
higher numbers of PTSD symptoms (F(1,90) = 9.429, p = 0.003, η2 = 0.095).
There were no significant differences in response latency in generating a past or future visualisation (F(1,90) = 0.110, p = 0.741, η2 = 0.001, n.s.), nor a significant association with severity
of PTSD symptomatology (F(1,90) = 0.676, p = 0.413, η2 = 0.007, n.s.), although as expected there
were differences depending on the valence of the cue word (F(1,90) = 6.007, p = 0.016, η2 =
0.063). There were no significant interaction effects.
Lastly, participants showed no difference in number of omissions of past visualisations compared to future visualisations (F(1,90) = 1.839, p = 0.179, η2 = 0.020, n.s.), yet a trend towards
more omissions in response to negative cue words (F(1,90) = 3.758, p = 0.056, η2 = 0.040, n.s.).
There was again no association with PTSD symptomatology (F(1,90) = 0.000, p = 0.997, η2 =
0.000, n.s.).
3.4. Exploratory analyses
Alongside studying characteristics of future imagery generated in response to cue words, this experiment also explored characteristics of future imagery produced in response to more open-ended questions. As part of the FSQ, participants were asked to generate images of themselves
in the future at two time points (“how do you see yourself in one year?” and “how do you see yourself in 5-10 years?”), and to rate these images for vividness, likelihood of occurrence, and valence.
PTSD symptomatology was associated with how positively or negatively participants rated their images (F(1,91) = 6.213, p = 0.014, η2 = 0.064) (see Table 3 for average valence ratings at
both time points). Furthermore there was a difference in ratings between time points (F(1,91) = 5.866, p = 0.017, η2 = 0.061) and a significant interaction between PTSD diagnosis and time
point (F(1,91) = 6.140, p = 0.015, η2 = 0.063). There was no significant association between
numbers of PTSD symptoms and the rated valence of images one year in the future (R2 = 0.015,
F(1,91) = 1.359, β = -0.121, p = 0.247, n.s.), but participants with higher numbers of PTSD
symptoms had significantly more negative images of themselves 5-10 years in the future (R2 =
0.092, F(1,91) = 9.255, β = -0.304, p = 0.003). Furthermore, when grouped, participants without
PTSD rated their images of themselves 5-10 years in the future as significantly more positive than their earlier images (t(59) = 2.427, p = 0.018, d = 0.271), whereas participants with full or
subthreshold PTSD rated their images similarly at both time points (t(32) = -1.263, p = 0.216, d =
0.209, n.s.).
These images were then split into positive images and negative images according to how participants themselves rated the image (see Table 3 for mean number of images generated). Neutral images were discarded. Participants generated a different number of positive images of themselves one year in the future compared to 5-10 years (F(1,91) = 4.070, p = 0.047, η2 = 0.043);
additionally, PTSD symptomatology predicted number of positive images (F(1,91) = 4.420, p =
0.038, η2 = 0.046). Furthermore, there was a significant interaction between severity of PTSD
symptomatology and time point (F(1,91) = 5.465, p = 0.022, η2 = 0.057). Simple linear regressions
discovered no significant association between PTSD symptomatology and the number of positive images of themselves one year in the future (R2 = 0.006, F
(1,91) = 0.543, β = -0.077, p =
0.463, n.s.) but found that participants with more severe PTSD symptomatology were significantly less likely to have positive self-images for 5-10 years in the future (R2 = 0.082, F
(1,91)
= 8.095, β = -0.286, p = 0.005). For the number of negative images generated, there was a significant difference for time point (F(1,91) = 4.300, p = 0.041, η2 = 0.045); severity of PTSD
symptomatology was a significant predictor (F(1,91) = 8.700, p = 0.004, η2 = 0.087) and again,
there was a significant interaction between PTSD symptomatology and time point (F(1,91) = 6.386,
p = 0.013, η2 = 0.066). Again, there was no significant difference in the number of negative
images generated at the first time point (one year in the future) (R2 = 0.015, F
(1,91) = 1.353, β =
0.121, p = 0.248, n.s.) but for 5-10 years in the future, participants with PTSD were significantly more likely to generate negative self-images (R2 = 0.128, F
(1,91) = 13.319, β = 0.357, p < 0.001).
Lastly, there was no significant difference in the total number of images generated at either time point (F(1,91) = 0.557, p = 0.457, η2 = 0.006, n.s.). There was a trend towards a significant
difference depending on PTSD symptomatology (F(1,91) = 3.853, p = 0.053, η2 = 0.041); simple
linear regressions revealed that participants with more severe PTSD symptomatology were less likely to generate images of themselves 5-10 years in the future (R2 = 0.048, F
(1,91) = 4.614, β =
-0.220, p = 0.034) but that there was only a trending difference for this for one year in the future (R2 = 0.030, F
(1,91) = 2.798, β = -0.173, p = 0.098). There was no significant interaction between
