Recurrent pattern completion drives the neocortical representation of sensory inference
Hyeyoung Shin, Mora B. Ogando, Lamiae Abdeladim, Uday K. Jagadisan, Séverine Durand, Ben Hardcastle, Hannah Belski, Hannah Cabasco, Henry Loefler, Ahad Bawany, Josh Wilkes, Katrina Nguyen, Lucas Suarez, Tye Johnson, Warren Han, Ben Ouellette, Conor Grasso, Jackie Swapp, Vivian Ha, Ahrial Young, Shiella Caldejon, Allison Williford, Peter A. Groblewski, Shawn R. Olsen, Carly Kiselycznyk, Jérôme Lecoq, Hillel Adesnik
Abstract
When sensory information is incomplete, the brain relies on prior expectations to infer perceptual objects. Despite the centrality of this process to perception, the neural mechanisms of sensory inference are not understood. Here we used illusory contours (ICs), multi-Neuropixels measurements, mesoscale two-photon (2p) calcium imaging and 2p holographic optogenetics in mice to reveal the neural codes and circuits of sensory inference. We discovered a specialized subset of neurons in primary visual cortex (V1) that respond emergently to illusory bars but not to component image segments. Selective holographic photoactivation of these 'IC-encoders' recreated the visual representation of ICs in V1 in the absence of any visual stimulus. These data imply that neurons that encode sensory inference are specialized for receiving and locally broadcasting top-down information. More generally, pattern completion circuits in lower cortical areas may selectively reinforce activity patterns that match prior expectations, constituting an integral step in perceptual inference.