CORTEX-WIDE REPRESENTATIONAL DRIFT WITHIN DIFFERENT LAYERS

A longstanding assumption in systems neuroscience is that although single neurons fluctuate over time, population-level sensory representations remain stable. Consequently, representational drift has been viewed as a microscopic phenomenon that averages out at the population scale. Here, we challenge this view by demonstrating robust, cortex-wide population-level representational drift, even under tightly controlled sensory input and stable behavior. Using widefield calcium imaging across 25 cortical areas in excitatory layer 2/3 (L2/3) and layer 5 (L5) neurons during a whisker-based texture discrimination task, we tracked sensory-evoked population activity over five consecutive days. Despite stable task performance, both layers exhibited pronounced multi-day reorganization of population activity patterns, indicating that cortical population codes drift rather than remain invariant. Strikingly, this drift was strongly structured by cortical layer, revealing a previously unrecognized laminar axis of differentiation. L5 showed a global, monotonic suppression of responses across nearly all cortical regions, whereas L2/3 exhibited spatially localized, heterogeneous, and stimulus-dependent changes, giving rise to distinct drift motifs across cortex. Representational drift extended beyond the sensory epoch and spanned the entire trial, with L5 displaying more prolonged temporal engagement than L2/3. Together, our findings provide the first evidence for cortex-wide population-level representational drift and demonstrate that drift is a distributed, layer-dependent property of cortical circuits, even when sensory input and behavior remain stable.