Sensory and action neural tuning explains how priors guide human visual decisions

Abstract

Prior expectations bias how we perceive the world. Despite well-characterized behavioral effects of priors, such as confirmation bias, their neural mechanisms remain unclear. Contemporary theories postulate conflicting predictions: does the brain enhance expected sensory information (sharpening the expected representation) or rather prioritize unexpected information (dampening the expected representations)? Here, we combined reversal learning with a noisy motion discrimination task to investigate how priors impact sensory and action information processing. Using behavioral modeling to infer participants latent priors and EEG to track neural dynamics, we demonstrated that priors differentially impact sensory and action representations over time. While priors introduced long-lasting biases on action coding irrespective of their validity, sensory information was selectively enhanced with confirmatory evidence. Critically, dampening of action representations predicted confirmation biases, whereas sensory tuning dynamics tracked speed-accuracy trade-offs. These findings reveal a dissociable and temporally dynamic influence of priors in visual decisions, reconciling competing theories of predictive perception. Significance statementPrior expectations, abstract rules, and recent decisions impact perception and ensuing actions. While priors are beneficial in most circumstances, they might be detrimental in other scenarios, as exemplified by confirmation biases, serial response biases, or switch costs. Albeit well-characterized at the behavioral level, the neural mechanisms underlying these biases remain undetermined. By tracking participants latent priors and neural dynamics with high-density EEG, the authors demonstrate that priors differentially impact sensory and action information, uncovering the neural mechanisms underlying confirmation biases and speed-accuracy trade-offs in predictive perception.