Memories, we’ve long believed, live in synapses — the connections between neurons that strengthen when we learn and weaken when we forget. Lose the synapses, lose the memory.
It’s one of the most foundational ideas in neuroscience, and it underpins much of what we think we know about conditions like Alzheimer’s disease.
A new study from Japan has just complicated that picture considerably.
The Textbook Version
The standard account goes something like this: when you learn something new, certain neurons in your brain fire together repeatedly, causing the connections between them to physically strengthen.
Tiny protrusions on the neurons — called dendritic spines — grow larger, forming more robust synapses. These physical structures are where memories live. When they’re damaged or destroyed, as happens in Alzheimer’s, memories fade.
It’s a tidy story. The new study suggests it’s incomplete.
Hibernation as a Natural Experiment
Researchers at Japan’s Okinawa Institute of Science and Technology took an unorthodox approach to testing how memories actually survive. Rather than studying disease, they induced artificial hibernation in mice — a state that dramatically lowers body temperature and metabolism, causing the brain to shed synapses rapidly in order to conserve energy.
First, the mice were trained on two standard memory tasks: learning to associate a particular chamber with a mild electric shock, and navigating a maze to find a sugary reward. Then the team triggered artificial hibernation for two days, tracking synapses throughout using fluorescent proteins.
The results were stark. Within minutes, spines began to shrink and disappear. Within 24 hours, more than half of all synapses were gone — including the large, supposedly stable ones considered most important for long-term memories.
And yet, when the mice woke up, their memories were intact. They froze with fear upon re-entering the shock chamber. They navigated the maze as fluently as before.
“It was astonishing,” said the study’s lead author. “Logically, if all our engram synapses were essential for memory retention as traditionally thought, memory should have massively deteriorated.”
The Rebuilding Act
What happened next was equally surprising. The spines that had vanished during hibernation grew back — and around 80% of them regrew in exactly the same locations they had previously occupied. The brain appeared to be rebuilding dismantled circuits from a blueprint it had somehow preserved.
To test whether this recovery was specific to hibernation, the team ran a comparison group: mice given anaesthesia plus a drug that blocks synaptic changes — a combination known to cause amnesia. These animals also lost large numbers of synapses, but never recovered their memories. The difference pointed to something unique about how hibernation handles memory preservation.
A Hidden Architecture
Careful analysis revealed the answer: a small cluster of unusually resilient synapses that survived hibernation intact. These clusters formed a distinctive network structure — one neuron connecting to multiple neighbours simultaneously, like a hub in a transport network. Critically, the more of these surviving clusters a mouse retained, the better it performed on the memory tasks after waking.
The clusters appear to act as anchors — scaffolding that preserves the essential pattern of a memory even as the surrounding connections dissolve. When the brain reawakens and rebuilds, those anchors guide the reconstruction, drawing new synaptic growth back to the right places.
“This suggests that for long-term memory, only particular clusters of synapses matter — the rest may be dispensable,” said study author Kazumasa Tanaka.
What This Could Mean
The implications are significant, and cautiously hopeful. If memories don’t depend on every individual synapse but rather on a higher-level architecture of key clusters, it raises the possibility that memories lost in conditions like Alzheimer’s — where synapse loss is extensive — might not be truly gone. They might simply be unreachable, their anchor clusters damaged or destroyed before the rest of the circuit could be reconstructed.
That’s a meaningful distinction. A memory that is erased is gone. A memory that is structurally preserved but inaccessible might, in theory, be recovered — if the right conditions could be restored.
The researchers are now working to understand what makes these anchor clusters so resilient at a molecular level. Identifying those mechanisms could eventually lead to interventions targeting the early stages of Alzheimer’s, potentially preserving the scaffolding that allows memories to be rebuilt before it’s too late.
The study is a reminder that the brain continues to surprise us — and that even what we thought were its most settled principles may turn out to be only the beginning of a much more interesting story.