Effective Connectivity Reveals Dual-Route Mechanism of Visual Prediction Precision via Insula and Pulvinar

The brain's ability to weight predictions by their precision is a central mechanism in predictive processing, enabling optimal integration of prior expectations with incoming sensory input. Despite its theoretical significance, the neural circuitry that implements precision-weighted prediction remains unclear. Using 7-Tesla fMRI and dynamic causal modelling (DCM), this study investigated how the brain encodes the precision of predictions during a visual cueing task with high- and low-precision conditions. We focused on the key regions implicated in predictive processing: the insular cortex, the pulvinar nucleus of the thalamus and primary visual cortex (V1). Behaviourally, participants showed significantly greater accuracy in the high-precision condition (p < 0.001), confirming effective task manipulation. DCM analyses revealed that high-precision predictions elicited excitatory modulation of connectivity from the insula to V1 (P_p = 0.95), alongside inhibitory influences from the insula to the pulvinar (P_p = 0.99) and from the pulvinar to V1 (P_p = 0.89). Furthermore, leave-one-out cross validation revealed that individual differences in behavioural sensitivity to precision were positively predicted by pulvinar-to-insula connectivity (r = 0.36, p = 0.026) and negatively predicted by the connectivity between pulvinar and V1 (pulvinar to V1: r = 0.35, p = 0.033; V1 to pulvinar: r = 0.37, p = 0.026), highlighting the behavioural relevance of these pathways. Together, these findings suggest a dual-route mechanism whereby the insula directly enhances top-down predictions in V1 while indirectly dampening bottom-up sensory input via the pulvinar. This mechanism may facilitate Bayesian integration under uncertainty and offers new hypotheses into how precision weighting may be disrupted in neuropsychiatric conditions.