Abstract Introduction Sleep reorganizes newly encoded memories into long-term storage through coordinated neural dynamics. In humans, slow oscillation-spindle (SO-SP) coupling during nREM sleep correlates with post-sleep memory improvement, but whether it reflects an active mechanism for long-distance communication or indexes other underlying processes remains unclear. We examined whether the spatiotemporal structure of SO-SP coupling constitutes a computational framework for memory reorganization during sleep. Methods High-density polysomnography dataset 1: young adults (N=15) and older adults (N=15) completed emotional memory tasks on nap and wake days. Dataset 2: young adults (N=35) performed word-pair learning on overnight and wake days. A Convolutional Neural Network with Long Short-Term Memory separated planar, rotating, and local SO-SP traveling waves. Coupling phase was quantified as the absolute distance of the preferred phase from the SO peak, and coupling strength by mean vector length (MVL). Traveling groups were identified when 10% of electrodes within a 200-ms window formed a unidirectional, spatially adjacent cluster. Partial derivatives of phase-gradients estimated propagation direction and speed, phase-gradient directionality (PGD) indexed inter-electrode consistency with the dominant direction defined by the mixture component with the greatest variance, and axial MVL quantified inter-event directional consistency. Statistics were derived from 10,000 permutations. Results In SO-SP coupled events in young adults, 61.5% exhibited simultaneous traveling dynamics. Planar SOs predominantly propagated along the anterior-posterior axis (87.5%), while SPs rotated along a temporal-parietal-frontal pathway across events (72.6%, ps 0.01). In uncoupled events, neither wave showed a preferred direction (ps0.59). Coupled events exhibited greater inter-electrode directional consistency (PGD=0.62) than uncoupled events (PGD=0.42, p 0.001), predicting post-sleep memory benefits (p=0.023). In turn, traveling SO-SPs showed more precise coupling phases (0.04rad) and strength (MVL=0.12) compared to isolated events (0.17rad, MVL=0.02, ps 0.001). Coupling and propagation demonstrate reciprocal scaling, whereby more precise and stronger coupling accompanies more consistent propagation and vice versa. In older adults, traveling wave proportion decreased (40.5%), and propagation range was reduced, disrupting coupling-propagation scaling (ps0.05). Conclusion SO-SP coupling constitutes an active traveling-wave-based computation, in which reciprocal modulation between coupling and propagation implements cross-frequency alignment across cortical spatiotemporal scales, efficiently reorganizing memory representations over long-range during sleep. This computational capacity declines with aging. Support (if any) NIH-R56AG058685, NIH-R01AG040133
Ng et al. (Fri,) studied this question.