Predictive Coding: How Many Faces?

Since von Helmholtz's (1867) supposition that the basis of perception is anchored in unconscious inference, it has been widely accepted that the brain uses a generative model to predict sensory input using learned statistical regularities about the world. Any discrepancy between these predictions and the sensory input produces a prediction error signal, which is then passed up the cortical hierarchy to update future expectations. In this way, the brain can highlight novel or surprising stimuli and efficiently allocate more resources when needed. Theoretically, our understanding of this predictive coding framework is extensive. However, empirical evidence highlighting the neural mechanisms that underlie the integration of prior information with sensory evidence is limited. In a recent study, Brodski, Paasch, et al. (2015) identified a neurophysiological correlate of prediction error during visual processing using a paradigm in which prior information was developed through lifelong experience.