How Does the Sleeping Brain Separate Old and New Information?

Adding a file to an archive seems straightforward: find a space, insert the document, and leave earlier records where they belong. Applied to the brain, the analogy immediately encounters a problem. Newly learned information needs to be retained, but learning also requires keeping what was learned before. Activity associated with both earlier and recent experiences may reappear during sleep. Could such a system organize not only what to preserve, but also when to attend to different records? A study in mice approached this question through changes in pupil size.

This is not a story about assigning more work to sleep and turning every night into a productivity project. Researchers investigated whether treating sleep as a uniform interval conceals important details. In an apparently resting animal, different memory processes may take prominence at different times. To recognize that possibility, a record saying only “asleep” may not be enough. A marker capable of tracking changes within sleep becomes useful. In this investigation, the alternating contraction and dilation of the pupil provided one way to study that temporal detail.

Changing states within a shared sleep category

A research highlight published by the US National Institute of Mental Health on April 2, 2025 describes Cornell University researchers examining pupil changes during sleep in mice. Pupils contracted during REM sleep, whereas contraction and expansion alternated during non-REM sleep. Researchers used these changes to distinguish different brain states. The memory findings discussed here concern substates within non-REM sleep. Combining REM and non-REM into one unchanging mechanism would lose the distinction the study investigated. [1]

Pupil size functions as a window in this account, not as a label displaying the contents of a memory. A small pupil does not let an observer directly read which experience an animal is recalling. The researchers considered pupil state together with recordings of brain activity. The marker's significance comes from its relationship with other measurements. For the same reason, this experiment does not establish that looking into someone's eyes reveals the memory currently being processed. A visible indicator cannot be reduced to the process it indicates.

Two records, from the eye to the hippocampus

The team examined neuronal activity in the hippocampus, a brain region involved in memory formation. They asked whether activity associated with older and newer memories was linked to pupil changes. According to the NIMH account, neurons associated with older memories were activated more frequently during large-pupil states, while those associated with new memories were activated more frequently during small-pupil states. A pattern of timing within sleep therefore became visible. This was not a demonstration that memories could be seen in the eye; two kinds of recorded activity showed different temporal preferences. [1]

The words “old” and “new” need care as well. In ordinary conversation, an old memory might mean something experienced years ago. The distinction in this research account concerns previously learned information and recently acquired information. The mouse experiment should not be narrated as a direct comparison between a person's childhood recollections and yesterday's conversation. Without inventing such an age boundary, the scientific question remains understandable: how can a system strengthen recent records while preserving earlier ones? Novelty and continuity are two sides of the same problem.

We can return to the archive analogy without treating it as a physical model of the brain. Imagine working on different groups of documents at different times rather than opening every record simultaneously. This illustrates why temporal separation might matter. It does not assert that the brain contains literal shelves, that memories remain as unchanging packages, or that researchers decoded every detail of their contents. The analogy's limited purpose is to make visible a coordination problem: adding a record and preserving an earlier one are related tasks, but not identical tasks.

Observing a state and intervening at a chosen time

Researchers also used optogenetics, a method that uses light to control neuronal activity. The NIMH account reports that disruption while pupils were small prevented mice from retaining newly learned information without affecting previously learned information. Disruption during large-pupil states did not affect the retention of the new information. The timing of the intervention was consequently as important to the comparison as its general description. Under the broad heading of sleep, there were intervals with different behavioral consequences. [1]

The scientific paper's accessible abstract specifies selective disruption of hippocampal sharp-wave ripples. Disrupting those events during contracted-pupil non-REM substates impaired recall of recent memories; the same manipulation during dilated-pupil substates had no behavioral effect. This was not an experiment showing that making the pupil smaller strengthens memory. It tested disruption of brain activity at a time identified by the pupil-related state. [2]

That distinction explains why a marker can be useful without being the mechanism itself. A change in the pupil can help researchers select an interval; the intervention targets neural activity, not the eye. An illuminated indicator in a building might identify which section is open, but the indicator does not perform the work taking place inside. This is an analogy separating a sign from the activity it marks. Turning the study into an “eye exercise” recommendation would misidentify what was manipulated in the experiment.

The condition without a detected effect also matters

The absence of an effect on new-information retention after disruption during large-pupil states is not merely a footnote. It enables a comparison based on timing. The change observed after intervention in the small-pupil state does not resemble a finding in which every intervention produces the same outcome. Nevertheless, failing to detect a behavioral effect under one condition is not proof that the brain is doing nothing during that interval. The measured behavior and selected intervention test a defined question; they do not constitute an exhaustive test for the absence of every possible function.

Likewise, finding more old-memory-related activity during large-pupil periods does not by itself explain every mechanism maintaining earlier information. The temporal separation offers an important clue about the organization of different memory processes. It cannot supply the outcome of an experiment that was not performed under another condition. A study need not become a complete theory of memory to be valuable. Its strength here is the combination of observation and selective intervention around the same timing question.

There are therefore several distinct links in the account. Pupil changes mark different intervals; neural recordings connect those intervals to different patterns; an intervention tests consequences at selected times. Collapsing these links into the sentence “the pupil stores memories” would make the story shorter but less accurate. Keeping them separate preserves the route from a visible physiological change to a question about the organization of learning. It also helps a reader see why an apparent detail of timing can be scientifically consequential.

Evidence-based conclusion: Temporal organization, not a human prescription

The NIMH account and the abstract of Chang and colleagues' 2025 paper support the presence of memory-related substates within non-REM sleep in mice. Reactivation associated with earlier and recent information took prominence in different pupil states, and disrupting neural activity in a particular state affected the retention of new learning. This article is based on the accessible institutional account and paper abstract. It does not claim independent reanalysis of raw recordings or a full audit of every experimental method. [1][2]

The distance between species and experimental conditions must remain visible when moving toward human memory, sleep difficulties, or treatment. NIMH discusses how these findings might eventually contribute to understanding memory problems; that is not confirmation of an available treatment. Reading particular memories by monitoring a person's pupils, or targeting a specific pupil size to strengthen new information, is not established by these sources. The more defensible conclusion is that learning involves more than adding records. The temporal organization of a system preserving both earlier and recent information deserves investigation in its own right.

Sources

1. National Institute of Mental Health, “How the Brain Creates New Memories While Maintaining Old Ones,” April 2, 2025. Institutional account of NIH-supported mouse research, describing pupil states, hippocampal activity, and optogenetic comparisons.
2. Chang et al., “Sleep microstructure organizes memory replay,” Nature 637, 1161–1169, 2025; PubMed record and abstract. DOI: 10.1038/s41586-024-08340-w. Narrow verification against the directly accessed scientific abstract; full data analysis is not claimed.