Emergence of a time-independent population code in auditory cortex enables sound categorization and discrimination learning

The temporal structure of sensory inputs contains essential information for their interpretation by the brain1-9. Sensory systems represent these temporal cues through two codes: the temporal sequences of neuronal activity and the spatial patterns of neuronal firing rate3,7,10-20. However, it is still unknown which of these two coexisting codes causally drives sensory decisions3,10,20,21. To separate their contributions, we designed an optogenetic stimulation paradigm in the mouse auditory cortex to generate neuronal activity patterns differing exclusively along their temporal or spatial dimensions. Training mice to discriminate these patterns shows that they efficiently learn to discriminate spatial but not temporal patterns, indicating that spatial representations are necessary for sensory learning. In line with this result, we observed, based on large-scale neuronal recordings of the auditory system, that the auditory cortex is the first region in which spatial patterns efficiently represent temporal auditory cues varying over several hundred milliseconds. This feature is shared by the deep layers of neural networks trained to categorise time-varying sounds. Therefore, the emergence of a spatial code for temporal sensory cues is a necessary condition to associate temporally structured stimuli to decisions. We expect this constraint to be crucial for re-engineering perception by cortical stimulation.