Detailed item memory at the cost of spatial context: an fMRI mechanism for intrusive memories after trauma-analog exposure

Background

Traumatic memories are clinically characterised by vivid sensory intrusions alongside fragmented contextual detail. Indeed, influential dual-representation models [1,2] propose that enhanced perceptual/item encoding combined with impaired contextual binding produces involuntary intrusive memories, a hallmark of post-traumatic stress disorder (PTSD). However, this prediction has rarely been tested directly in voluntary free recall and its neural basis during retrieval remains poorly characterised.

Aims

We aimed to (i) quantify how trauma-analogue exposure shifts the composition of voluntary recall between item and spatial-context features; (ii) test whether this shift predicts the frequency of intrusive memories at both between- and within-subject levels; and (iii) identify the neural substrates of item and contextual retrieval, and their trial-level relationship to behavioural memory content.

Methods

Forty-seven healthy participants completed a three-phase trauma-film fMRI paradigm, involving: (1) encoding of 21 trauma-analogue and 21 content-matched neutral video clips; (2) a think-aloud intrusion task, and (3) voluntary spoken free recall in the scanner. Recall transcripts were independently coded by two raters (intraclass correlation coefficients ≥ .80) for items (persons, animals, objects), item details (descriptive features), and spatial-context features. The manual coding was further validated using a representational similarity analysis (RSA) of recall transcripts against ground-truth annotations embedded with a multilingual sentence-transformer model. fMRI analyses comprised parametric univariate general linear models (GLMs) with trial-level memory counts as modulators, and a whole-brain searchlight RSA quantifying event-specific encoding–recall pattern reinstatement. Behavioural effects were tested with 2×2 repeated-measures ANOVAs and follow-up paired t-tests; intrusion-prediction analyses used multiple linear regression (Holm-corrected) and Poisson generalised linear mixed models; reactivation–behaviour links used Tweedie generalised linear mixed models.

Results

Trauma-analogue recall contained more item details and less spatial context than neutral recall (content × condition interaction F(1,46) = 117.33, p < .001), an effect that survived control for between-stimulus differences in content density and was independently confirmed by the embedding-based RSA validation. Between subjects, more recalled details per item together with less spatial context jointly predicted intrusion frequency (both Holm-corrected p ≤ .017). At the trial level, details per item independently predicted per-trial intrusion counts (p = .002). Trauma-analogue recall enhanced activity in item-related regions including bilateral Posterior Medial Cortex (PMC) and angular gyrus, whereas spatial-context retrieval recruited retrosplenial and parahippocampal cortices. The searchlight RSA revealed event-specific PMC reactivation during recall of both trauma-analogue and neutral events, and trial-level PMC reactivation positively predicted item memory (p = .003).

Conclusions

An item-detailed, spatially decontextualised recall profile is a behavioural marker of intrusive-memory frequency in healthy individuals exposed to a trauma analogue, and event-specific PMC reactivation supports the item side of this profile. These findings provide convergent behavioural and neural evidence for the dual-representation account of intrusive memories and identify candidate mechanistic targets for translational research in PTSD and other trauma-spectrum disorders. Generalisation to clinical samples, and to the encoding phase, remain avenues for future research.

References

1. Brewin, C.R., Gregory, J.D., Lipton, M., Burgess, N., 2010. Intrusive images in psychological disorders: characteristics, neural mechanisms, and treatment implications. Psychol. Rev. 117 (1), 210–232.
2. Ehlers, A., Clark, D.M., 2000. A cognitive model of posttraumatic stress disorder. Behav. Res. Ther. 38 (4), 319–345.
3. Chen, J., Leong, Y.C., Honey, C.J., Yong, C.H., Norman, K.A., Hasson, U., 2017. Shared memories reveal shared structure in neural activity across individuals. Nat. Neurosci. 20 (1), 115–125.