Half the connections went, and the fear stayed

What the hibernation took

Start with the subtraction, because it is larger than it sounds. Driven into deep hypothermia by a drug acting on a small cluster of cells in the hypothalamus, the mice spent about 48 hours with their body temperature down and their hippocampal neurons barely firing. Electron microscopy through the CA1 region afterwards put the fall in synapse density at 52.5 per cent against animals that never hibernated. The spines that remained were no larger than usual, which removes the tidiest explanation: that the brain keeps its strongest contacts and lets the weak ones go. Five days after arousal the animals were returned to the chamber where they had been shocked, and they froze as they had before.^[1]

It would be neat to conclude that the memory had simply moved out of the hippocampus while the lights were off, the way older memories are thought to migrate towards the cortex. The team tested that directly. After arousal they lesioned the hippocampus, and the freezing disappeared. Whatever carried the fear through those two days was still sitting in the same structure, in a network stripped to roughly half its connections.^[1]

What it left behind

The survivors had a shape. When the researchers lit up the connections running between the memory-carrying neurons themselves, the contacts that held their numbers through hibernation were the ones sitting in clusters, defined as two labelled contacts within 5 micrometres of each other on the same dendrite. Isolated contacts fell away. The clustered survivors also ran small, smaller than their neighbours on the same branch, and 33.33 per cent of them met a presynaptic bouton that was touching a spine on another neuron as well, against 3.33 per cent of spines picked at random in animals that never hibernated.^[1]

Put those two details together and they point at an arrangement rather than an inventory. A bouton that touches several spines at once binds separate neurons into one group; a cluster of such contacts along a single dendrite is a piece of wiring geometry that no individual synapse holds. Hold the reading lightly, though. Hibernation might protect these clusters for metabolic reasons with no bearing on memory at all, and the comparison the team ran, a long anaesthesia combined with a drug that blocks actin polymerisation, cut synapse density, wrecked the memory and removed the clustered contacts instead of the isolated ones, which changes several things at once. The link between clustering and retention is strong, and it remains a correlation.^[1]

How far does this reach?

There is an old intuition that a person is the sum of their connections, and that losing the connections means losing the person. This experiment settles nothing about that, and was never going to: one fear memory, one region, a rodent held in a state no mouse enters in the wild. What it supplies is a worked example of continuity surviving demolition, a case where the material inventory changed enormously and the thing that inventory was supposed to carry came through. Should that generalise, then what makes you continuous with the person who woke yesterday sits closer to a pattern of relationships than to a list of parts. That is a large should, and closing it belongs to the specialists.^[1]

The claim turns testable the moment someone can take a cluster apart without touching anything else. If a group manages selective de-clustering, leaving synapse counts intact while breaking the spatial grouping, and retention falls with it, the architecture reading holds. If retention survives, the clusters are tracking something else. Watch for that experiment, in a second laboratory and ideally a second species, before the end of 2027. Until then the honest position is that a brain gave up more than half of a structure and kept what the structure was for, which is already enough to sit with for a while.^[1]