How cortical circuits build representations of complex objects is poorly understood. The massive dimensional expansion from the thalamus to the primary sensory cortex may enable sparse, comprehensive representations of higher order features to facilitate object identification. To generate such a code, cortical neurons must integrate broadly over space, yet simultaneously obtain sharp tuning to specific stimulus features. The logic of cortical integration that may synthesize such a sparse, high dimensional code for complex features is not known. To address this question, we probed the integration and population coding of higher order stimuli in the somatosensory and visual cortices of awake mice using two-photon calcium imaging across cortical layers. We found that somatosensory and visual cortical neurons sum highly specific combinations of sensory inputs supra-linearly, but integrate other inputs sub-linearly, leading to selective responses to higher order features. This integrative process generates a sparse, but comprehensive code for complex stimuli from the earliest stages of cortical processing. These results from multiple sensory modalities imply that input-specific supra-linear summation may represent a widespread cortical mechanism for the synthesis of higher order feature codes. This new mechanism may explain how the brain exploits the thalamocortical expansion of dimensionality to encode arbitrary complex features of sensory stimuli.

Data collected from mice using in vivo 2-photon calcium imaging. Mice transgenically or virally expressed GCaMP6s in pyramidal cells, and recordings were targeted to either layer 4 or layer 2/3 of either S1 or V1. Responses to visual or vibrissal stimulation were recorded. Putative single cells have been segmented, calcium fluorescence traces extracted, and event rates deconvolved. Meanwhile, behavioral parameters of the mouse were collected (eye movements in the case of V1 data, running speed).