ANOMALOUS RELAXATION OF THE COLLECTIVE ACTIVITY IN THE CORTEX

In the absence of stimuli, neuronal activity in cortical circuits exhibits rich dynamics, structured across multiple spatial and temporal scales. In spite of these observations, a common assumption both in experimental and theoretical work is that there exists a well-defined ‘equilibrium state’, identified with the ‘spontaneous state’, to which the network dynamics relax back following a perturbation. To investigate this assumption, here we study experimentally the relaxation of the network dynamics to this putative spontaneous state. To that end, we take advantage of the naturally occurring alternation between ‘running’ and ‘non-running’ episodes in head-restrained mice implanted with Neuropixel probes. By averaging across multiple episodes, we find that the network activity, after substantial modulation during ‘running’, relaxes back to a stationary level during ‘non-running’. If there really is an equilibrium state however, repeating the same analysis in single episodes should lead to the same results. This is not what we find; instead, we see that estimates of the network activity obtained from different episodes can be substantially different. These differences are not explained by finite-sampling effects, by the statistics of inter-spike intervals or by the autocorrelation structure of the single-neuron spiking processes. Our results show that cortical networks, at least in the so-called spontaneous state, exhibit some typical signature of non-equilibrium dynamics. These findings call for a departure from ‘equilibrium’-based modeling and pave the way for more realistic descriptions of the cortical dynamics.