Replay articles from across Nature Portfolio

Artificial neural networks are known to perform well on recently learned tasks, at the same time forgetting previously learned ones. The authors propose an unsupervised sleep replay algorithm to recover old tasks synaptic connectivity that may have been damaged after new task training.

Sleep is known to promote memory consolidation, but the extent to which this is dependent on the memory’s relevance remains unclear. Here, the authors use a brain decoding approach to show that neural representations of rewarded experiences undergo a privileged reactivation during sleep, favouring their consolidation.

Sleep after learning helps to strengthen new memories. Here, the authors link this memory benefit to the reactivation of learning experiences when two endogenous sleep rhythms—slow oscillations and sleep spindles—coincide.

Learning-activated engram neurons play a critical role in memory recall but the role of these neurons in offline memory consolidation is unclear. The authors show that sleep-associated reactivation of learning-activated sensory neurons is necessary for memory consolidation.

Schechtman et al. examine whether the effect of cued reactivation during sleep on memory depends on the amount of information being reactivated via cues during sleep. Their results show that multiple memories can be consolidated in parallel, suggesting brain capacity for reactivation is not limited by separate resources needed for individual memories.

Thalamic head direction (HD) cells are necessary to establish spatial maps in the hippocampus. Here, the authors show that HD cells tuned to a particular direction are coupled to individual hippocampal ripple events during sleep, suggesting an influence of the replay of specific trajectories during sleep memory consolidation.