Slow oscillations provide the spatio-temporal framework for long-range neural communication during sleep
Slow oscillations (SOs, <1Hz) during non-rapid eye movement sleep are thought to reflect sleep homeostasis and support memory consolidation. Yet, the fundamental properties of SOs and their impact on neural network communication are not understood. We used effective connectivity to estimate causal information flow across the electrode manifold during SOs and found two peak of information flow in specific phases of the SO. We show causal communication during non-rapid eye movement sleep peaks during specific phases of the SO, but only across long distances. We confirmed this prediction by cluster analysis demonstrating greater flow in global, compared with local, SOs. Finally, we tested the functional significance of these results by examining which SO properties supported overnight episodic memory improvement, with the underlying assumption that memory consolidation would engage global, long-range communication. Indeed, episodic memory improvement was predicted only by the SO properties with greatest causal information flow, i.e., longest distances between sinks and sources and global, but not local, SOs. These findings explain how NREM sleep (characterized as a state of low brain connectivity) leverages SO-induced selective information flow to coordinate a wide network of brain regions during memory formation.