A cortical output channel for perceptual categorization

Perceptual categorization allows the brain to transform diverse sensory inputs into discrete representations that support flexible behavior [1–7]. Auditory cortex (ACtx) has been implicated in this process [8–14], but the cell-type-specific circuits that implement category learning remain unknown. We trained head-fixed mice to categorize the temporal rate of amplitude-modulated noise while performing longitudinal two-photon imaging of layer (L)5 extratelencephalic (L5 ET) neurons alongside comparison populations of L2/3 and L5 intratelencephalic (L5 IT) neurons. With learning, L5 ET neurons underwent pronounced tuning modifications and developed robust, categorical responses, whereas L2/3 and L5 IT neurons did not. This categorical code was task engagement-dependent: it was present during behavior and absent during passive listening in the same neurons on the same day, indicating context-gated expression. Using a generalized linear model to dissociate stimulusfrom choice-related signals, we confirmed that categorical selectivity in L5 ET neurons reflected sensory encoding rather than motor confounds. All three populations carried choice signals, but these were strongest in L5 ET neurons, suggesting a role in linking sensory categorization to action selection. These findings identify a projection-specific, deep-layer cortical output channel in which L5 ET neurons acquire categorical representations and selectively propagate behaviorally relevant signals to downstream targets.