Bayesian surprise intensifies pain in a novel visual-noxious association

Pain perception is not solely determined by noxious stimuli, but also varies due to other factors, such as beliefs about pain and its uncertainty. A widely accepted theory posits that the brain integrates prediction of pain with noxious stimuli, to estimate pain intensity. This theory assumes that the estimated pain value is adjusted to minimize surprise, mathematically defined as errors between predictions and outcomes. However, it is still unclear whether the represented surprise directly influences pain perception or merely serves to update this estimate. In this study, we empirically examined this question using virtual reality. In the task, participants reported felt pain via VAS after their arm was stimulated by noxious heat and thrusted into by a virtual knife actively. To manipulate surprise level, the visual threat suddenly disappeared randomly, and noxious heat was presented in the on- or post-action phases. We observed that a transphysical surprising event, created by sudden disappearance of a visual threat cue combined with delayed noxious heat, amplified pain intensity. Subsequent model-based analysis using Bayesian theory revealed significant modulation of pain by the Bayesian surprise value. These results illustrated a real-time computational process for pain perception during a single task trial, suggesting that the brain anticipates pain using an efference copy of actions, integrates it with multimodal stimuli, and perceives it as a surprise.