A shared multi-feature population code for sensory reliability across mouse visual cortex

Successful behavior depends on evaluating both what is present in the environment and how reliable that information is. Yet how neural populations represent reliability, and whether this representation is shared across cortical areas, remains unknown. Using two-photon calcium imaging of over 150,000 neurons in mouse visual cortices during presentations of coherent motion stimuli, we find that sensory reliability is encoded through coordinated changes in response amplitude, selectivity, and signal-to-noise ratio. These coherence-dependent response patterns are not fully accounted for by stimulus-based normalization models; rather, they track the distribution of motion energy across directions, while remaining relatively invariant to total motion energy. This multi-feature code is expressed at the single-neuron and population levels and is consistent across visual areas, suggesting a shared encoding scheme rather than area-specific computations. Together, these findings support a common population code for stimulus reliability across visual cortices.