Behavior-relevant top-down cross-modal predictions in mouse neocortex

Abstract

AbstractAnimals rely on predicting their environment and the consequences of their actions to adapt to a constantly changing world. The predictive coding hypothesis proposes that the brain generates predictions and continuously compares them with bottom-up sensory inputs to guide behavior. However, how the brain reconciles conflicting top-down predictions and bottom-up sensory information during behavior remains unclear. To address this question, we simultaneously imaged neuronal populations in the mouse somatosensory cortex and the posterior parietal cortex during an auditory-cued texture discrimination task. After mice learnt the task with fixed tone-texture matching, mismatched pairing caused conflicting tone-based texture predictions and actual texture inputs. When top-down interaction was dominant, texture representations in both areas were modified and mice decided based on the predicted rather than actual texture, whereas dominant bottom-up interaction corrected the representations as well as behavioral choice. Our findings provide evidence for hierarchical predictive coding in the mouse neocortex and open new avenues for understanding higher cognitive functions.