Vision as looking and seeing through a bottleneck

Li Zhaoping's new paper, published in Current Opinion of Neurobiology, challenges decades of vision research by highlighting a fundamental bottleneck: only a tiny fraction of retinal input is ever recognized by the brain. The author argues that traditional frameworks have largely overlooked this constraint, leading to slower progress downstream of primary visual cortex (V1). The proposed model reframes vision as two distinct processes: 'looking,' where the peripheral visual field selects information via gaze shifts to center content at the fovea, and 'seeing,' where the central visual field recognizes that content.

Converging evidence suggests V1 initiates the bottleneck by generating a bottom-up saliency map that guides saccades exogenously. Meanwhile, top-down feedback along the visual pathway, targeting mainly the central visual field representation, refines seeing. The paper emphasizes that progress will accelerate through falsifiable theories that explicitly link behavior with neural substrates, and by experimental designs that avoid forced fixation and precisely track gaze. This shift could impact fields from computer vision to neuroscience, offering a more biologically plausible foundation for understanding how we perceive the world.