This longitudinal investigation followed eighty-nine individuals across a 20-year span to probe neurodevelopmental effects of early life stress on mentalizing. At young adulthood, participants completed the Reading the Mind in the Eyes Test (RMET) and a matched control task during functional MRI. The study examined behavioral performance and multiple neural outcome measures to characterize how early caregiving-related risk relates to the development of neural representations supporting mentalizing.
Two indices of early life stress (ELS) were assessed separately. The first was a prospective cumulative risk score derived from caregiver wellbeing assessed during pregnancy and the first year of the participant’s life. The second was a retrospective self-report measure of adverse experiences collected from participants in late adolescence. The authors analyzed these indices independently to compare developmental timing and measurement approach.
Participants performed the RMET, a task commonly used to probe mentalizing and theory of mind, alongside a matched control condition designed to control for non-social perceptual and lexical demands. Behavioral accuracy on the RMET was measured and related to ELS indices to test whether early exposure to caregiving-related risk predicts adult social-cognitive performance.
During RMET and control conditions, functional MRI data were collected and analyzed with multiple approaches. Neural outcomes included standard univariate activation analyses as well as multivariate representational techniques. The analyses tested regional effects across a parcellation of 360 cortical regions, enabling a broad, region-wise search for ELS-related differences in neural responses.
The study applied representational similarity analysis (RSA) to characterize condition-specific neural representational geometry within regions. In addition, inter-subject RSA (IS-RSA) was used to evaluate how similar participants’ neural representational geometries were to one another and whether that similarity related to ELS. IS-RSA was computed using two pairwise approaches: an absolute pairwise difference metric and a pairwise average metric, enabling tests of whether ELS related to divergence from a common representational pattern or to graded shifts in representational geometry across participants.
Analyses revealed associations between the prospective caregiver-based ELS index and both behavior and neural representations. Prospectively assessed higher ELS was associated with better RMET performance in young adulthood. At the neural level, only a single cortical region survived correction across the 360 regions examined in RSA and IS-RSA analyses: the left anterior inferior frontal sulcus. In this region, IS-RSA results conformed to the pairwise average pattern: participants with low ELS exposure shared a more typical or canonical neural representational geometry, while those with higher prospectively assessed ELS showed increasingly idiosyncratic neural representational geometries.
Additionally, within the same left anterior inferior frontal sulcus region, greater prospective ELS was associated with weaker separation between RMET and the matched control condition. This indicates a reduced condition-specific distinction in representational geometry for participants exposed to higher early caregiving risk, suggesting altered differentiation of social-cognitive processing relative to non-social or lexical-semantic demands.
In contrast to the prospective index, the retrospective self-report of adverse experiences collected in late adolescence did not show significant associations with RMET behavior or with neural measures reported in this study. The authors explicitly note the absence of behavioral or neural effects for the retrospective ELS measure.
The pattern observed—prospective ELS predicting more idiosyncratic neural representational geometry in the left anterior inferior frontal sulcus and reduced RMET-versus-control separation—suggests that early caregiving-related risk may alter how mentalizing-related information is represented in specific frontal regions. The finding that low-exposure participants converged on a more typical representational geometry while high-exposure participants diverged implies increased heterogeneity of neural coding for mentalizing after early stress exposure.
The reduced representational separation between RMET and control further raises the possibility that ELS affects the developmental differentiation of social-cognitive computations from overlapping lexical-semantic or semantic-control processes in the same region. The authors highlight that future work should carefully disentangle mentalizing-specific representations from lexical-semantic and semantic-control demands when interpreting RMET-related neural activity.
This study provides longitudinal evidence linking early caregiving-related risk to altered neural representations supporting mentalizing in adulthood, with a focal effect in the left anterior inferior frontal sulcus. The dissociation between prospective and retrospective ELS measures underscores the importance of measurement timing and method when probing developmental effects of early adversity. Future research should replicate these representational findings, investigate causal mechanisms, and examine how altered representational geometry relates to social functioning across contexts. The authors also recommend explicit control for lexical-semantic and semantic-control demands in mentalizing tasks to clarify the specificity of observed neural effects.
Note: This report summarizes the results as presented in the preprint. The article is a preprint and has not been peer reviewed.