Scientists Discover a Neural Switchboard That Stores New Memories Without Erasing Old Ones

Memory consolidation during sleep

The same CA1 hub neurons that handle daytime communication don’t clock out at night. During sleep, they remain highly active inside sharp-wave ripples—brief, high-frequency bursts of neural activity—replaying the firing patterns from waking behavior . This nighttime replay loop is central to memory consolidation, the process by which fragile new memories are solidified into stable, long-term storage.

Previous research supports the idea that sleep is when the brain sorts and stabilizes memories. A 2025 NIH-funded study found that new and old memories are reactivated during sleep through distinct physiological states, helping keep them separate . The NYU Langone study adds a circuit-level explanation: the switchboard mechanism keeps the pathway from the hippocampus to the cortex open during sleep, ensuring that replay consolidates new information without scrambling older memory traces.

Implications for Alzheimer’s disease

The CA1 region is known to be one of the earliest brain areas affected in Alzheimer’s disease . In fact, studies have shown that synapse organization in the hippocampal formation is vulnerable early in the disease, with differences in postsynaptic targets and synaptic shapes appearing even when overall synaptic density looks normal .

Dr. Zhe S. Chen, a co-senior author of the NYU Langone study, noted that the newly discovered switchboard mechanism "may provide clues as to how memory circuits fail in Alzheimer's disease and other conditions that affect the brain's ability to recall events and find places" .

If the CA1 hub cells lose their ability to maintain separate channels for incoming and outgoing signals, the brain could begin mixing up new and old information—or fail to store new memories altogether—producing the type of memory impairment seen in Alzheimer’s . The hippocampus also contains distinct layers of CA1 neurons with unique molecular signatures that may be differentially vulnerable in conditions like Alzheimer’s and epilepsy, adding another layer of complexity to understanding how memory circuits degrade .

A blueprint for better artificial intelligence

Beyond neuroscience and medicine, the discovery holds lessons for artificial intelligence. Current AI systems suffer from a well-documented problem called catastrophic forgetting: when a neural network is trained on a new task, it often overwrites the weights it learned for previous tasks. The mammalian brain, by contrast, can learn continuously without losing old knowledge.

The NYU Langone study suggests that the brain achieves this through architectural separation of input and output streams within shared circuitry—a design principle that could be translated into next-generation AI systems . Rather than retraining entire networks on new data, AI architectures might incorporate analogous "switchboard" modules that route new information through dedicated channels while preserving existing representations.

The researchers described their findings as a potential "biological blueprint" for designing AI that updates continuously, a holy grail in the field .

Caveats and next steps

It is important to note that this study was conducted in mice navigating a controlled laboratory environment. While the hippocampal circuit organization is conserved across mammals, firm conclusions about the human brain or more naturalistic memory behaviors will require further research .

The NYU Langone team plans to investigate whether similar switchboard-like channels exist in other memory circuits beyond the CA1-to-cortex pathway. Understanding whether this mechanism generalizes could broaden both neuroscientific insights and applications for treating memory disorders.