Trial-by-trial variability in cortical responses exhibits scaling in spatial correlations predicted from critical dynamics

Simple sensory stimuli or motor outputs engage large populations of neurons in the mammalian cortex. When stimuli or outputs repeat, the robust population response contrasts with fluctuating responses of individual neurons, known as trial-by-trial variability. To understand this apparent discrepancy, a detailed identification of the underlying spatiotemporal correlations is required. Here, we analyze spatial correlations in the instantaneous fluctuations between neurons relative to the neuronal population. Using 2-photon imaging of visual and auditory responses in primary cortices of awake mice, we show that these correlations grow linearly with the size of the observed cortical area. We extend these observations to the cortical mesoscale by analyzing local field potentials in behaving nonhuman primates. In network simulations, we demonstrate this linear growth in spatial correlation to emerge at criticality. Our findings suggest that trial-by-trial variability is a signature of critical dynamics in cortex maintaining robust, long-range spatial correlations among neurons.