PREDICTIVE PROCESSING OF TACTILE SENSORY INFORMATION IN MICE ENGAGED IN A WHISKER-GUIDED LOCOMOTION TASK

The predictive processing framework posits that the brain actively predicts sensory inputs rather than passively processing them. How this principle operates during naturalistic, active sensation, particularly in the tactile domain, remains largely unexplored. Here, we investigate the mesoscale neural signatures of predictive processing in the mouse whisker system during a self-paced, whisker-guided locomotion task. We combined fibroscopic wide-field calcium imaging of cortical excitatory neurons (EmxCre x floxed GCaMP6f mice) with high-resolution behavioural tracking, using a novel motorized optical rotary joint to achieve stable recordings in freely-moving animals. To manipulate tactile expectations, we trained mice to navigate a linear track where an obstacle was consistently present on one side in 90% of trials, creating a strong sensory prediction. This expectation was violated in rare "omission" trials (10%), where the obstacle was absent.
Our results show that mice exhibit robust predictive behavior, pre-emptively steering away from the expected obstacle location and modulating their whisking strategy on omission trials. Critically, these prediction violations evoked a spatiotemporally structured cortical activation. After subtracting out global, non-specific activity fluctuations, we found that unexpected omissions of the obstacle elicited a localized "ghost" representation in the corresponding barrel-related columns of the primary somatosensory cortex (wS1). This neural signature was timed to when the actual touch would have occurred and its spatial emergence in wS1 depended on the animal’s position within the track. The omission-related signal subsequently propagated from wS1 to the posterior parietal cortex (PPC), sequentially engaging the rostrolateral (RL) and posterior parietal (PtP) areas.