3:30–4:30 pm Maria Goeppert-Mayer Lecture Hall
Emergent reliability of neocortical neural coding
Mark Schnitzer, Stanford University
Decades ago, John von Neumann posed the question of how the brain can compute so accurately despite that individual neurons appear to be extremely noisy. For example, sensory perception, which is normally highly reliable, is limited by the information that the brain can extract from the noisy dynamics of sensory neurons. Seminal experiments done ~30 years ago suggested that correlated activity fluctuations within sensory cortical neural ensembles are what limits their coding accuracy. However, without concurrent recordings from thousands of cortical neurons with shared sensory inputs, it remained unknown whether correlated noise actually limits coding fidelity. To address this question, we built a 16-beam two-photon fluorescence mesoscope to monitor neural activity across the mouse primary visual cortex and created analyses to quantify the information conveyed by large neural ensembles. We found that correlated noise-constrained signaling for ensembles of 800–1,300 cortical neurons. Moreover, neural ensemble visual signals were perpendicular to the largest noise mode, which therefore did not limit coding fidelity. The information-limiting noise modes were approximately ten times smaller and concordant with mouse visual acuity. Thus, cortical design principles appear to enhance coding accuracy by restricting ~90% of noise fluctuations to modes that do not limit signaling fidelity, whereas much weaker correlated noise modes bound sensory discrimination. In subsequent experiments, we extended our imaging studies to the entire visual cortex and found that neocortex supports sensory performance through brief elevations in sensory coding redundancy near the start of perception, neural population codes that are robust to cellular variability, and widespread inter-area fluctuation modes that transmit sensory data and task responses in non-interfering channels.