SCALE-FREE HIPPOCAMPAL NETWORK DYNAMIC IN SLEEP PREDICTS FUTURE RETENTION OF SPATIAL MEMORIES

The role of the hippocampal reactivation during sharp wave-ripples (SWRs) in memory consolidation has been well understood. However, it is less well understood how memory is related to the coding-independent properties of hippocampal activity. To better understand these aspects, we applied the Phenomenological Renormalization group to CA1 spiking data recorded from rats. The scaling exponent of activity variance α, measured during the sleep/rest session that followed learning, predicted memory retention during the subsequent recall session. Further analysis showed that 1) The strength of reactivation during SWRs correlated with both memory retention and α. Nevertheless, α predicted memory retention even after controlling for reactivation, while reactivation no longer predicted memory retention when α is controlled for. 2) Excluding SWR periods from the data did not affect results, as α still predicted memory retention. Interestingly, α measured during the sleep/rest period, preceding learning, also predicted memory retention performance. We also found that α is a transferable quantity that can predict memory retention across animals: training a model in a subset of animals produced a significant prediction of the memory retention of the excluded animals. Finally, we linked α to functional characteristics of individual cells, their intrinsic timescale and their burst propensity. We also showed how these characteristics predict memory retention, and that α retained its predictive power after controlling for these functional characteristics. In summary, we uncovered aspects of hippocampal activity that are behaviorally-relevant and predict memory retention independent of spatial coding and reactivation during SWRs.