Posteromedial cortical networks encode visuomotor prediction errors.

Abstract

Predicting future events based on internal models is essential for animal survival. Predictive coding postulates that errors between prediction and observation in lower-order areas update predictions in higher-order areas through the hierarchy. However, it is unclear how predictive coding is implemented in the hierarchy of the brain. Herein, we report the neural mechanism of the hierarchical processing and transmission of bottom-up prediction error signals in the mouse cortex. Ca2+ imaging and electrophysiological recording in virtual reality revealed responses to visuomotor mismatches in the retrosplenial, dorsal visual, and anterior cingulate cortex. These mismatch responses were attenuated when mismatches became predictable through experience. Optogenetic inhibition of bottom-up signals reduced a behavioral indicator for prediction errors. Moreover, cellular-level mismatch responses were modeled by Bayesian inference using a state-space model. This study demonstrates hierarchical circuit organization underlying prediction error propagation, advancing the understanding of predictive coding in sensory perception and learning in the brain. O_FIG O_LINKSMALLFIG WIDTH=188 HEIGHT=200 SRC="FIGDIR/small/504075v2_ufig1.gif" ALT="Figure 1"> View larger version (82K): org.highwire.dtl.DTLVardef@1f03960org.highwire.dtl.DTLVardef@46df08org.highwire.dtl.DTLVardef@dc97ccorg.highwire.dtl.DTLVardef@7200ae_HPS_FORMAT_FIGEXP M_FIG C_FIG