Predictive coding in the primate brain: From visual to fronto-limbic systems

Predictive coding theory has emerged as a powerful framework for understanding neural information processing, offering a unified account of perception, cognition, and behavior through hierarchical minimization of prediction error. Notably, recent studies emphasize that the anatomical hierarchy and neurophysiological dynamics of the primate visual cortex are closely aligned with this framework. In this review, we examine the neural implementation of hierarchical information processing in the primate cerebral cortex from multiple perspectives, focusing on predictive coding theory. We aim to integrate findings from mathematical modeling, cortical microcircuits, laminar structure, neural oscillations, and clinical research. One of the central themes emerging from this integrative perspective is that interareal communication in the cortex is characterized by frequency-specific neural dynamics. Beta-band oscillations support top-down signals, whereas gamma-band oscillations convey bottom-up signals. These patterns closely reflect the directionality of anatomically defined projections. We also highlight that neural networks trained to predict upcoming visual input can develop functional properties observed in the primate visual system without these properties being explicitly supervised. Finally, we argue that hierarchical interaction provides a valuable perspective for understanding large-scale network dysfunctions associated with disorders such as depression and anxiety. These abnormalities may reflect disruptions in predictive signaling across functional domains rather than isolated regional deficits. ∙ Predictive coding provides a unified account of perception, cognition, and emotion. ∙ Neural dynamics in primate visual areas align with predictive coding principles. ∙ Oscillatory hierarchies match the anatomical organization of cortical circuits. ∙ Anxiety-related fronto-limbic networks can be framed within predictive coding. ∙ Predictive coding characterizes cognitive-limbic interactions in depression.