Frontal-midline theta promotes context-dependent risk aversion in social anxiety

Social anxiety is reliably characterized by biases toward avoidance and aversive learning. Here we examined the neurocomputational processes underlying these biases and examined whether these biases persist across different social contexts. A sample of 154 participants (84% female, mean age = 20.42) with low to subclinical levels of social anxiety completed a probabilistic selection task in two contexts: performing alone and under social scrutiny. We analyzed frontal-midline theta (FM-theta) activity to uncover neurocomputational processes underlying aversive learning in social anxiety. Results showed that participants performed more accurately alone and predominantly preferred a win-stay strategy. Social anxiety led to an increased use of both win-stay and lose-switch strategies. In response to negative feedback, FM-theta power predicted post-error slowing, an effect heightened in socially anxious participants. Punishment-based learning was stronger in all participants but social anxiety amplified this effect, particularly when participants performed alone. Our findings suggest that social anxiety bolsters the role of FM-theta in aversive control, promoting reactive avoidance and risk-averse decision-making. This mechanistic insight links social anxiety to a cascade of orienting, control, and learning biases, and positions FM-theta as a potential neural target for interventions aimed at reducing risk aversion and enhancing adaptive learning in socially anxious individuals.