Principal study findings

This study examined whether different social cognitive indices were associated with trauma and PTSD symptoms. Data were collected at a single time point, from a clinical and non-clinical sample.

We found that our case group reported greater difficulties in mentalising on the RFQ-8. In the logistic regression analyses, there was weak evidence of an association between self-report of hypomentalisation measured using the RFQ-U and experience of a trauma. We found evidence that hypomentalisation measured using the RFQ-8 was associated with significantly increased odds of PTSD symptoms.

We also found an association between greater use of mentalising language on the naturalistic Modified-STOMP task of spontaneous social cognition and PTSD symptoms. This finding, however, did not persist in a sensitivity analysis restricted to the Prolific sample only.

We found no evidence of any significant association between the three performance-based measures of facial emotion processing and mental state attribution (SST, Oddity, RMET) and history of trauma or PTSD symptoms.

Comparison with previous studies

Previous reviews (including quantitative meta-analyses) have found that social cognition, particularly mentalisation, is impaired in people with PTSD compared with controls. ^{Reference Stevens and Jovanovic10–Reference Janssen, van Est, Hilbink, Gubbels, Egger and Cillessen12} We found that mentalisation differed between the cases and controls using the RFQ-8 and the Modified-STOMP, but not with more constrained performance-based measures of social cognition focused on emotion recognition and attribution.

We found limited intercorrelations between most of the social cognition measures. Interrelations between social cognition measures have typically been found to be modest in previous research, ^{Reference Yeung, Apperly and Devine38} consistent with the idea that social cognition reflects a range of distinct, domain-specific abilities. We found strong correlation between the RFQ-U and the RFQ-C measures. The RFQ-U and RFQ-C are mainly derived from coding of the same set of questions. Previous work has found that a general mentalising difficulty could result in both hyper- and hypomentalising. ^{Reference Weinstein, Whitmore and Mills39} However, confirmatory factor analysis has also shown that the RFQ-8 is a unidimensional measure that most accurately captures hypomentalising. ^{Reference Woźniak-Prus, Gambin, Cudo and Sharp40}

We found an association between greater use of ‘Insight’ words in the written descriptions of a social movie scene from the Modified-STOMP task and increased odds of PTSD, potentially reflecting hypermentalising. Perhaps consistent with this, previous work found that a greater proportion of ‘Insight’ words in a written trauma account at baseline predicted less trauma symptom reduction over time in a population exposed to a natural disaster. ^{Reference Marshall and Venta41}

A recent meta-analytic study has observed an association between mentalising deficits and psychopathology more broadly. ^{Reference Kivity, Levy, Johnson, Rosenstein and LeBreton42} It is possible our findings reflect other psychopathology more common in a PTSD sample rather than PTSD per se. This study also found that explicit mentalising, such as self-report rather than task-based measures, had a greater correlation with psychopathology, as we observed. ^{Reference Kivity, Levy, Johnson, Rosenstein and LeBreton42}

Implications and areas for future research

The social–cognitive hypothesis of PTSD theorises that difficulties with social cognition increase the likelihood of PTSD developing following a traumatic life event. ^{Reference Sharp, Fonagy and Allen3} This study has shown an association between mentalising difficulties measured via the RFQ-8 and PTSD. We also found that greater proportions of ‘Insight’ words (reflecting mental state terms like ‘think’ and ‘feel’) used to describe a movie scene (Modified-STOMP) was associated with PTSD on the main analysis, but not the sensitivity analysis restricted to the Prolific sample. This finding initially appears to be the opposite of what we had hypothesised: that lower proportions of ‘Insight’ words would be associated with PTSD. However, use of mental state terms may not necessarily reflect mentalising accuracy, and the finding could reflect hypermentalising: that is, there may have been greater use of mentalising language, but this could have been inaccurate or excessive, given that previous work suggests that increased use of ‘Insight’ words indicate attempts to make sense of events, ^{Reference Tausczik and Pennebaker43} which may be more difficult or less automatic in individuals with lower social cognitive ability. The inconsistency between the findings of the main and sensitivity analyses could be attributable to reduced power as a result of either a smaller sample size in the sensitivity analysis, a type 1 error or because the difference in social cognitive ability between the clinical cases and Prolific controls is greater than that between the Prolific cases and Prolific controls.

Another aspect to consider is the difference between reflective functioning/mentalising and theory of mind. These are often conflated concepts in social cognition, but mentalising (as measured using the RFQ-8) looks at awareness of subjective states and mental processes in oneself and others (a meta-social–cognitive ability), whereas theory of mind is the awareness of the experiences of others’ through perspective taking (a social–cognitive ability). ^{Reference Bora44} In borderline personality disorder, another psychiatric condition that often includes a significant traumatic component, reflective functioning is impaired to a much greater degree than theory of mind. ^{Reference Bora44} There is only weak correlation between self-report and measured social cognitive abilities. ^{Reference Murphy and Lilienfeld8}

We did not find an association between performance-based measures of social cognition (which covered primarily other aspects of social cognition such as face emotion recognition and emotion attribution to animated shapes) and PTSD. These findings support the hypothesis that specific difficulties in mentalising (reflective functioning), rather than broad social cognition impairments, are associated with PTSD. However, because of the cross-sectional nature of the study, it is hard to determine the direction of causality. Further work is needed in a well-powered longitudinal cohort study to determine if mentalising difficulties measured before trauma exposure are associated with increased risk of developing PTSD. Additional research could examine why mentalisation difficulties are associated with increased likelihood of trauma exposure and PTSD, the latter ideally measured by a clinical interview. If the association between mentalising tendencies and PTSD is found to be robust, this could allow early identification of vulnerable individuals in populations with a heightened risk of exposure to trauma, such as the military or emergency service workers. This could be assessed efficiently in clinical settings with the addition of the RFQ-8 tool.

Strengths and limitations

A key strength of this study design is the breadth and variety of social cognition measures used, including both self-report questionnaires and constrained and open-ended performance-based tasks. We used a larger sample size than many other studies in the field, which will have increased the power and precision of the findings. Logistic regression allowed us to adjust for several confounder variables.

However, the study must be viewed in the context of several limitations. First, as this is a case–control study, it is problematic to infer causal relationships as the temporality of any relationships present cannot be determined from data collected at a single time point. Therefore, even though our results are consistent with mentalisation difficulties resulting in greater likelihood of PTSD, it is also plausible that trauma exposure and PTSD result in greater difficulties mentalising. Other researchers have suggested that trauma exposure can result in impaired mentalising, ^{Reference Fonagy, Campbell and Luyten45} and that this relationship is complex and multifaceted. ^{Reference Nietlisbach, Maercker, Rossler and Haker46} Previous qualitative research we have conducted has also hinted at a bidirectional relationship between social cognition and PTSD. ^{Reference Wiseman, Lawrence, Bisson, Hotham, Karl and Zammit15}

In addition, there is a difference between ability and propensity in social cognitive research. ^{Reference Guazzelli Williamson and Mills7} Most of the tasks used in this study would measure mentalizing ability, or the accuracy of mentalising, rather than an individual’s likelihood to mentalise (propensity). Performance on the measures used can be situation-dependent and laboratory tasks may lack ecological validity. ^{Reference Yeung, Apperly and Devine38} There is also only weak correlation between self-report and task-based measures of social cognition. ^{Reference Murphy and Lilienfeld8}

Our cases were selected both from NHS services and from the Prolific sample screened using the PCL-5, and therefore are somewhat heterogenous. We are less confident that the Prolific cases would meet the threshold for a clinical diagnosis of PTSD. More detailed information on trauma histories and chronicity of symptoms for the clinical and Prolific participants would have added to the study analyses. However, as data were collected remotely during the COVID-19 pandemic, we limited data collection on traumatic events and disability to minimise distress. We completed sensitivity analyses restricted to the Prolific sample to manage the heterogeneity.

There are also issues with the data collection techniques. Controlled laboratory conditions were not used for data collection for most participants because of the COVID-19 pandemic restricting face-to-face research. Most study participants were recruited via the online Prolific platform. We chose to recruit from this platform as it has a good reputation and study participants are more honest and naïve then other recruitment platforms. ^{13,Reference Peer, Brandimarte, Samat and Acquisti47} However, social science studies repeated online during the pandemic were found to have a reduced effect size compared with pre-pandemic findings. ^{Reference Peyton, Huber and Alexander48} Participants completing the study online may have had divided attention or been doing the bare minimum to complete the study and obtain the financial reward (satisficing). ¹³ We attempted to manage this by embedding three attention checks into the Prolific tasks and questionnaires. All participants passed at least two of these. However, participants may have been aware of these and looked out for them. Careless responding would likely result in an underestimate of the true effect size.

We had planned to use both subscales of the RFQ-8 to examine the association between hypomentalisation (RFQ-U) and hypermentalisation (RFQ-C) with the outcome variables. However, because of high collinearity between the subscales (rho = 0.93), we decided to only include the RFQ-U scale. Recent confirmatory factor analysis has demonstrated the RFQ-8 is unidimensional and best captures hypomentalising, so we have used it post hoc in this context. ^{Reference Woźniak-Prus, Gambin, Cudo and Sharp40}

The use of multiple measures of social cognition is both a strength and weakness in our study. Looking at multiple measures means we are more likely to have found a type 1 error. The RFQ-8 is also a self-report measure and some of the association seen between could be a result of capturing general psychological distress and symptom severity rather than mentalising difficulties specifically. However, our main association between the RFQ-8 and PTSD has a moderately large effect size and a small p-value in both our main and sensitivity analyses, providing evidence of an association (p < 0.001).

This study found higher rates of trauma exposure and PTSD prevalence than in the general population. Eighty-five per cent of our Prolific participants reported a trauma exposure compared with a 70% worldwide trauma prevalence. ^{Reference Kessler, Aguilar-Gaxiola, Alonso, Benjet, Bromet and Cardoso49} Thirty per cent of participants screened positive for PTSD in the Prolific sample, higher than the lifetime prevalence of 7%. ^{Reference de Vries and Olff50} It could be that Prolific participants with trauma exposure or PTSD were particularly interested in completing this study, or that the PCL-5, which is a screening tool, may have detected generic psychiatric distress as well as PTSD (i.e. low specificity). This would be consistent with a recent meta-analytic finding that mentalisation difficulties are correlated with psychopathology more broadly. ^{Reference Kivity, Levy, Johnson, Rosenstein and LeBreton42} We did not examine type or number of traumas in detail in our analyses, which may have affected the results.

We only adjusted for four variables (age, gender, verbal IQ, autism status) as confounders in our analyses. Other unmeasured confounders, such as ethnicity and socioeconomic status, could also have influenced the results.

In conclusion, this study used data collected from a clinical and an online sample to examine the relationship between social cognition and PTSD symptoms. We found a significant association between a self-report measure of mentalisation difficulties and PTSD symptoms. We also found some association between an unconstrained movie-based task of spontaneous mentalisation and PTSD symptoms, but this finding was not repeated in a sensitivity analysis restricted to the non-clinical sample. We did not find an association between more constrained performance-based measures of social cognition and the outcome measures. Our findings, despite their limitations, inform social cognitive theories of PTSD.