Does the direction of hippocampal replay during sleep determine whether memories are strengthened or weakened?

Replay is the brain rehearsing a route it already took

During sleep, the hippocampus — a brain region central to forming new memories — reactivates patterns of activity that match paths the animal walked while awake. Researchers call this replay. It is widely thought to support the conversion of fragile new experiences into lasting memories [1].

Replay has a direction: the reactivated sequence can run forward, roughly the order the animal experienced it, or backward. A review of how replay starts argues that the head-direction network, which tracks which way the animal is facing, influences hippocampal activity at the moment replay begins, and that replay in turn resets that network to pick a direction for the next event [1]. That describes how replay is steered, not what steering accomplishes for memory.

Direction is measured, but not linked to remembering or forgetting

The idea that replay direction sorts memories into keepers and discards is appealing, but the studies that examine direction do not test that link. The review of replay initiation examines direction and its upstream control without addressing whether direction determines memory strengthening or weakening [1]. A computational model of replay dynamics during sharp-wave ripples — brief, coordinated bursts of hippocampal activity — also treats directionality but does not examine the relationship between direction and memory outcomes [4].

So the honest answer to the direction question is that it remains untested in this evidence. Direction is a real feature of replay, but no result here shows that flipping it would flip a memory from strengthened to weakened.

What actually separates new memories from old ones: pupil substates

The clearest result comes from mice whose large hippocampal ensembles were recorded while their sleep was tracked through pupil size. Oscillating pupil fluctuations revealed a previously unknown microstructure within non-REM sleep. Replay of recent experiences dominated during sharp-wave ripples in contracted-pupil substates, while replay of previous memories occurred preferentially during dilated-pupil substates [3].

The behavioural test is what makes this more than a correlation. Disrupting sharp-wave ripples during contracted-pupil non-REM sleep impaired recall of recent memories; the same disruption during dilated-pupil substates had no behavioural effect [3]. In other words, the substate — not the direction of the replayed sequence — was what mattered for whether a recent memory survived.

The two substates also differ in their wiring. Contracted-pupil substates showed stronger input from outside the hippocampus, while dilated-pupil substates showed greater recruitment of local inhibition, the hippocampus's own braking circuitry [3]. The authors suggest this temporal segregation lets the brain handle new and old memories in separate slots, avoiding interference while learning continuously [3]. That is a proposed function, not a demonstrated purpose.

REM sleep and the bigger picture

Replay is not confined to non-REM sleep. A study of hippocampal assembly reactivation and sequence replay during REM sleep found that novelty and anxiety shape these processes, and proposed mechanisms by which REM sleep could contribute to memory consolidation [2]. It did not test whether replay direction influences memory strengthening or weakening.

Taken together, the evidence supports a picture in which sleep's fine structure — which substate the brain is in — organizes which memories get replayed and which ones depend on that replay to be recalled later [3]. Direction of replay is a genuine object of study [1][4], but the claim that it determines strengthening versus weakening is not established by these studies.

Sources used in this answer

Related research questions