Self-Other Neural Representational Distinction is Associated with Distorted Social Predictions in Social Anxiety

Background: Social anxiety (SA) is characterized by distorted negative expectations about the future. Because optimism bias is inherently comparative, the extent to which these predictions reflect a fundamental disruption in self- versus other-referential processing remains poorly understood. Methods: Across two experiments, we investigated the mechanisms underlying social predictions in SA. In Experiment 1, individuals with high ( n = 52, LSAS > 65) and low SA ( n = 51, LSAS < 35) rated the future likelihood of positive and negative social outcomes for themselves or for a protagonist. In Experiment 2, 35 adults spanning a continuous spectrum of SA completed the task during fMRI. Behavior was analyzed using linear mixed-effects models. Neurally, multivariate pattern analysis (MVPA) quantified self-other representational distinctions, while a parametric general linear model (GLM) tracked neural sensitivity to likelihood predictions. Results: Behaviorally, higher SA was characterized by an attenuated optimism bias: individuals predicted a lower likelihood of positive outcomes, but a higher likelihood of negative outcomes, for themselves compared to others. Neurally, higher trait SA correlated with greater self-other representational distinction within the left pregenual anterior cingulate cortex (pgACC). Concurrently, the precuneus exhibited stronger parametric tracking of self-negative predictions versus others. Crucially, pgACC representational distinction during positive outcome processing significantly mediated the association between SA severity and heightened prediction-tracking in the precuneus. Conclusions: These findings offer a novel mechanistic framework, suggesting that an exaggerated neural segregation of the self from others drives hyper-vigilant, risk-sensitive social predictions in SA. This candidate circuit-level dynamic warrants further replication in clinical cohorts.