Perception of visual noise patterns depends on correlations produced in V1: A simulation study

Visual noise is widely used as a stimulus in vision science, yet the neural representation of noise itself remains poorly understood. We investigated how external visual noise is encoded in the primary visual cortex (V1) using a biologically inspired neural network model (VOneNet). We simulated V1 responses to noise patterns and analyzed both the response of individual model neurons and the correlations among them. As expected, simulated neurons with smaller receptive fields showed stronger responses. Importantly, spatially independent noise stimuli elicited correlations across the population. These correlations were critical for reconstructing the original noise pattern in a reconstruction analysis that served as a probe of perceptual representation: decorrelating modeled neural activity substantially degraded reconstruction quality. Correlations among simulated simple cells, rather than complex cells, played a dominant role in encoding external noise. Our findings suggest that the perception of visual noise depends not only on local neural activity but also on population-level correlations in V1. Implications for visual snow syndrome, a condition in which illusory external noise is perceived, are discussed.