Neural mechanisms of individual differences in prior weight during scene recognition

Human perception relies on integrating sensory input with prior knowledge, yet individuals differ substantially in how they weigh these information sources. While previous research has identified cortical regions implicated in prior-guided visual recognition, a network-level understanding of individual variability in prior utilization remains lacking. Here, to uncover how task-evoked functional connectivity predicts the degree to which individuals rely on prior knowledge, we reanalyzed a previously published fMRI dataset from a scene-recognition task with parametrically controlled image naturalness. We combined generalized psychophysiological interaction (gPPI) analyses with connectome-based predictive modeling (CPM) to model inter-regional connectivity modulations specific to prior-related conditions. These connectivity patterns served as input features for connectome-based predictive models to predict individual prior weights estimated via a Bayesian behavioral model. Our findings demonstrate that the connectivity patterns among predefined cortical/subcortical nodes, spanning multiple large-scale functional networks, reliably predict individual differences in prior use during image-scene recognition. This integrative framework offers a mechanistic account of perceptual variability and provides a foundation for individualized markers of cognitive strategy in uncertain environments.