The cost of scrutiny: social anxiety constrains frontal-midline theta-dependent neural network reconfigurations during probabilistic learning

Social anxiety is reliably characterized by biases toward avoidance and aversive learning. Frontal-midline theta (FM-theta) oscillations have been implicated as a key neural mechanism underlying these learning biases. Here we combined dynamic network neuroscience with computational modeling to investigate how FM-theta coordinates large-scale functional network interactions on a trial-to-trial basis for feedback-guided learning, and how these neural dynamics are altered in social anxiety across different social contexts. A sample of 146 individuals (mean age = 20.4 years) with varying levels of social anxiety completed a probabilistic selection task while performing alone and under social observation. Results indicated that FM-theta contributed to two theta-dependent functional subgraphs, each differently supporting probabilistic learning. Frontal-parietal functional connectivity increased following aversive outcomes and scaled with prediction error, consistent with error-monitoring and feedback-updating. In contrast, social anxious individuals exhibited heightened context-sensitive frontal-occipital connectivity for positive feedback across contexts and for negative feedback during performance observation. Social scrutiny amplified integration between these neural network patterns, impairing learning performance, particularly in high socially anxious individuals. Together, these findings show that social anxiety is associated with dysregulated, context-sensitive FM-theta neural network dynamics, reflecting reduced neural flexibility for adaptive learning under scrutiny.