Cortical feedback to V1 and V2 contains unique information about high-level scene structure

CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not. Abstract: Early visual cortical neurons receive highly selective feedforward input, which is amplified or disamplified by contextual feedback and lateral connections. A significant challenge for systems neuroscience is to measure the feature space that drives these feedback channels. We occluded visual scenes and measured non-feedforward stimulated subregions of V1 and V2 using fMRI and multi-voxel pattern analyses. We found that response patterns in these subregions contain two high-level scene features, category and depth information. Responses in non-feedforward stimulated V1 and V2 differed from each other, suggesting that feedback to these two areas has unique information content. Further, we reveal that computational models of visual processing inadequately describe early visual cortex because they do not account for the brain's internal modeling of the world.