SPATIAL NAVIGATION ENGAGES A DISTRIBUTED NEURAL REPRESENTATION

Navigation is a goal-directed behaviour that requires integrating sensation, reward, movement, and internal spatial representations. Navigation-related signals have not only been reported in the hippocampus and entorhinal cortex, which are central to spatial coding, but also more widely across the cerebral cortex. How distributed are navigation-related signals across the brain, and are spatial representations separable from signals encoding sensation, reward, and movement?
Using Neuropixels probes, we recorded over 20,000 neurons across the neocortex, hippocampus, thalamus, striatum, and midbrain while mice navigated an audiovisual virtual corridor with two sensory-identical halves. To decouple virtual position from external inputs and physical movement, we varied landmark contrast, sound intensity, and wheel gain across trials.
In most regions, many neurons showed a preferred virtual position, firing more at that position than at the sensory-identical position. This spatial preference was strongest in the primary motor cortex, primary somatosensory cortex, and midbrain. Consistent with this, a Bayesian decoder predicted position with an accuracy above chance from population activity in most regions. Decoding performance depended strongly on visual landmark contrast but not on the presence of auditory cues.
To test whether activity reflects spatial representations beyond non-spatial variables (sensation, reward, movement), we fit linear models to single-neuron activity. Across areas, 72 ± 2% of neurons were well fit. Adding virtual-position predictors significantly improved variance explained in 23 ± 1% of neurons, indicating that spatial representations separable from signals encoding sensation, reward, and movement are present across the brain.