Hot metacognition: poorer metacognitive efficiency following acute but not traumatic stress

Abstract

Aberrations to metacognition-the ability to reflect on and evaluate self-performance-are a feature of poor mental health. Theoretical models of post-traumatic stress disorder propose that following severe stress or trauma, maladaptive metacognitive evaluations and appraisals of the event drive the development of symptoms. Empirical research is required in order to reveal whether disruptions to metacognition cause or contribute to symptom development in line with theoretical accounts, or are simply a consequence of ongoing psychopathology. In two experiments, using hierarchical Bayesian modelling of metacognition measured in a memory recognition task, we assessed whether distortions to metacognition occur at a state-level after an acute stress induction, and/or at a trait-level in a sample of individuals experiencing intrusive memories following traumatic stress. Results from experiment 1, an in-person laboratory-based experiment, demonstrated that heightened psychological responses to the stress induction were associated with poorer metacognitive efficiency, despite there being no overall change in metacognitive efficiency from pre- to post-stress (N = 27). Conversely, in experiment 2, an online experiment using the same metamemory task, we did not find evidence of metacognitive alterations in a transdiagnostic sample of patients with intrusive memory symptomatology following traumatic stress (N = 36, compared to 44 matched controls). Our results indicate a relationship between state-level psychological responses to stress and metacognitive alterations. The lack of evidence for pre- to post-stress differences in metamemory illustrates the importance for future studies to reveal the direction of this relationship, and consequently the duration of stress-associated metacognitive impairments and their impact on mental health.

Fig. 1. Metamemory task design.

In the encoding phase of the task, participants were presented with neutral-neutral, negative-negative, and neutral-negative image pairs and asked whether they could vividly imagine a link between them. A four-second fixation cross was displayed between the presentation of each pair. After a 10-min break, participants returned for a surprise memory test. They were shown one image from each image pair or a new ‘lure’ image and were instructed to judge whether it was new or old. They then rated their confidence from 1 to 3 in their new/old judgement. If the image was selected as old, participants were then asked to select the image pair from a set of 6 images. Again, participants rated their confidence in the selection from 1 to 3. Images used in the task were selected from the International Affective Pictures System (IAPS) and are not presented in Fig. 1.

Fig. 2. Modified raincloud plots illustrating memory performance in Experiment 1.

Memory performance is measured for (A) item memory trials, quantified as d-prime (Z(hits) – Z(false alarms)), and (B) associative memory trials (Z(hits)), at pre- and post-stress. Individual scatter points show subject-level memory performance estimates for each task, kernel density plots represent the distribution of values and box plots depict medians and interquartile ranges.

Fig. 3. Metacognitive performance estimated by a group-level Bayesian hierarchical model (Fleming).

A Metacognitive efficiency (µ meta-d’/d’) did not differ between pre- and post-stress, as demonstrated by the posteriors of the group-level estimates. B The difference in group posteriors of metacognitive efficiency between pre- and post-stress encompasses zero (fixed line) consistent with no evidence of a relationship. Dashed lines represent the 95% HDI (95% HDI = (−0.2156, 0.2729)). C Using the RHMeta-d model, higher standardised subjective stress scores predicted poorer logMratio at post-stress. Dashed lines represent the 95% confidence intervals. D The distributions of samples over the regression beta parameter, with dashed lines representing the 95% HDI which does not encompass zero (fixed line), is consistent with relatively strong evidence for a relationship between subjective stress and metacognitive efficiency (95% HDI = (−0.3438, −0.0289)).

Fig. 4. Modified raincloud plots illustrating memory performance in Experiment 2.

Memory performance is measured for (A) item memory trials, quantified as d-prime (Z(hits) – Z(false alarms)), and (B) associative memory trials (Z(hits)), for the control and intrusive memories groups. Individual scatter points show subject-level memory performance estimates for each task, kernel density plots represent the distribution of values and box plots depict medians and interquartile ranges.

Fig. 5. Metacognitive efficiency estimated by a group-level Bayesian hierarchical model (Fleming).

A Metacognitive efficiency (µ meta-d’/d’) was unaffected by clinical intrusive memories, as demonstrated by the posteriors of the group-level estimates for the controls and intrusive memories groups. B The difference in group posteriors of metacognitive efficiency between controls and individuals with intrusive memories encompasses zero (fixed line), consistent with no evidence for a relationship between clinical status and metacognitive efficiency (95% HDI = (−0.1988, 0.1192)). Dashed lines represent the two-tailed 95% HDI. C Using the RHMeta-d model, no relationship was observed between severity of intrusive memories and logMratio. Dashed lines represent the 95% confidence intervals. D The distributions of samples over the regression beta parameter, with dashed lines representing the 95% HDI which encompasses zero (fixed line), demonstrates no relationship between standardised intrusive memory scores and metacognitive efficiency (95% HDI = (−0.0897, 0.0521)), controlling for age and sex.