Inference, validation and predictions about statistics and propagation of cortical spiking in vivo

Abstract

Electrophysiological recordings of spiking activity can only access a small fraction of all neurons simultaneously. This spatial subsampling has hindered characterizing even most basic properties of collective spiking in cortex. In particular, two contradictory hypotheses prevailed for over a decade: the first proposed an asynchronous irregular, the second a critical state. While distinguishing them is straightforward in models, we show that in experiments classical approaches fail to infer them correctly, because subsampling can bias measures as basic as the correlation strength. Deploying a novel, subsampling-invariant estimator, we find evidence that in vivo cortical dynamics clearly differs from asynchronous or critical dynamics, and instead occupies a narrow \"reverberating\" regime, consistently across multiple mammalian species and cortical areas. These results enabled us to predict cortical properties that are difficult or impossible to obtain experimentally, including responses to minimal perturbations, intrinsic network timescales, and the strength of external input compared to recurrent activation.