Sekėjai

Ieškoti šiame dienoraštyje

2025 m. lapkričio 2 d., sekmadienis

Why don’t new memories overwrite old ones? Sleep science holds clues

 

“Research in mice points towards a mechanism that avoids ‘catastrophic forgetting’.

 

New clues have emerged in the mystery of how the brain avoids ‘catastrophic forgetting’ — the distortion and overwriting of previously established memories when new ones are created.

 

A research team has found that, at least in mice, the brain processes new and old memories in separate phases of sleep, which might prevent mixing between the two. Assuming that the finding is confirmed in other animals, “I put all my money that this segregation will also occur in humans”, says György Buzsáki, a systems neuroscientist at New York University in New York City. That’s because memory is an evolutionarily ancient system, says Buzsáki, who was not part of the research team but once supervised the work of some of its members.

 

Scientists have long known that, during sleep, the brain ‘replays’ recent experiences: the same neurons that were involved in an experience fire in the same order. This mechanism helps to solidify the experience as a memory and prepare it for long-term storage.

 

To study brain function during sleep, the research team exploited a quirk of mice: their eyes are partially open during some stages of slumber. The team monitored one eye in each mouse as it slept. During a deep phase of sleep, the researchers observed the pupils shrink and then return to their original, larger size repeatedly, with each cycle lasting roughly one minute. Neuron recordings showed that most of the brain’s replay of experiences took place when the animals’ pupils were small.

 

Sleep loss impairs memory of smells, worm research shows

 

That led the scientists to wonder whether pupil size and memory processing are linked. To find out, they enlisted a technique called optogenetics, which uses light to either trigger or suppress the electrical activity of genetically engineered neurons in the brain. First, they trained engineered mice to find a sweet treat hidden on a platform. Immediately after these lessons, as the mice slept, the authors used optogenetics to reduce bursts of neuronal firing that have been linked to replay. They did so during both the small-pupil and large-pupil stages of sleep.

 

Once awakened, the mice had completely forgotten the location of the treat — but only if firing had been reduced during the small-pupil stage. “We wiped out the memory,” says Wenbo Tang, a co-author of the Nature paper and a systems neuroscientist at Cornell University in Ithaca, New York.

 

By contrast, when the team reduced bursts of neuronal firing during the large-pupil phase shortly after a lesson, the mice went straight to the treat — making clear that their fresh memories were intact.

Blast from the past

 

Other experiments by the team showed that the large-pupil phase of sleep has its own function: it helps to process established memories, which in mice means those that formed in the few days before a snooze, rather than those from the same day.

 

The neurobiologist studying how the brain changes as it ages

 

“The brain was preserving the older memories during this large-pupil sub-state, but incorporating new memories during the small-pupil sub-state,” says co-author Azahara Oliva, a physicist at Cornell University. This two-phase system is a “possible solution to this problem of how the brain can incorporate new knowledge but also maintain the old knowledge intact”.

 

The paper takes “a very important step”, says Maksim Bazhenov, a systems neuroscientist at the University of California, San Diego, who was not involved in the research. It shows that the handling of established memories and new memories “is not all mixed up, which could potentially lead to interference, but instead [is] nicely separate in time”.

 

Catastrophic forgetting also affects artificial neural networks, which are algorithms modelled on the brain and underlie many of today’s artificial intelligence (AI) tools. Insights into how the brain avoids this problem might inspire algorithms that can be used to help AI models avoid it as well, Tang says.” [1]

 

1. Nature 637, 524-525 (2025) Traci Watson

 

Komentarų nėra: