Representational drift influences learning of a sensory cortical microstimulation task

Abstract

Sensory cortical representations can be highly dynamic, raising the question of how representational stability impacts perception. We used volumetric two-photon calcium imaging to track neural activity as mice learned an optical microstimulation task. Mice were required to report the number of photostimulation pulses delivered to opsin-expressing neurons in layer 2/3 of primary vibrissal somatosensory cortex. We found that the amount of photostimulus-evoked activity could predict animal choice. Over the course of training, this photostimulus-evoked activity rapidly declined, particularly among highly photoresponsive neurons. Population activity levels across training were similar between animals that learned and those that did not. However, the photoresponsive population exhibited turnover both within and across behavioral sessions that was elevated among animals that failed to learn. Animals that failed to learn also exhibited a faster deterioration in stimulus decoding. Our results show that the rate at which sensory cortical representations change constrains learning and impacts behavior.