What are the functions of primary visual cortex (V1)?

Li Zhaoping's paper, accepted for publication in Current Opinions in Neurobiology, presents three recently discovered functions of primary visual cortex (V1). First, V1 acts as a motor cortex for exogenously guiding saccades by constructing a bottom-up saliency map of the visual field. This means V1 directly controls eye movements to focus on salient features, a role previously attributed to higher motor areas. Second, V1 initiates a processing bottleneck: a massive reduction of visual information begins at its output to downstream areas. This bottleneck filters out most visual data, leaving only a fraction for higher-level processing. Third, downstream recognition is limited by this impoverished information, so V1 supports ongoing recognition by providing additional information queried by top-down feedback from downstream areas, directed predominantly to central visual field representations.

These functions underpin a framework where vision is mainly 'looking' and 'seeing' through the bottleneck. Looking selects a fraction of visual information into the bottleneck, largely by saccades that center selected contents at gaze. Seeing recognizes the selected contents. Looking and seeing rely mainly on processing in the peripheral and central visual fields, respectively. This challenges traditional views that V1 is solely a sensory processing area, highlighting its role in active vision and attention. The paper synthesizes decades of research into a cohesive model, offering new insights for neuroscience and AI vision systems.