For roughly 40 years, the working assumption in neuroscience went something like this: the hippocampus, the brain’s primary memory hub, starts out mostly blank. Neurons sit idle, waiting for experience to wire them together. A baby’s first sights, sounds, and touches gradually build the circuitry that will one day let an adult recall a phone number or a childhood birthday party.
A study published in May 2026 in Nature Communications argues the opposite is true, at least in mice. Researchers at the Institute of Science and Technology Austria (ISTA) found that the hippocampal CA3 region, a circuit at the core of memory formation, begins life densely wired and tangled. Over the following weeks, the brain strips away most of those connections, sculpting the lean, organized network that adults depend on. The memory system does not start empty and fill up. It starts overcrowded and trims down.
A full slate, not a blank one
The ISTA team, working in the laboratory of neuroscientist Peter Jonas, used a technique called multicellular patch-clamp circuit mapping to record electrical activity from groups of CA3 neurons simultaneously. By measuring which cells were functionally connected to which, they could build a wiring diagram of the circuit at three developmental stages: postnatal days 7 through 8 (roughly equivalent to a newborn), postnatal days 18 through 25 (a juvenile), and postnatal days 45 through 50 (a young adult).
At the earliest time point, connections between CA3 neurons were local, dense, and apparently random. By the later stages, the network had reorganized into something distributed, sparse, and structured. The transformation was dramatic enough that ISTA’s own description of the work frames it as a shift from “tabula plena” (a full slate) to a refined circuit, directly challenging the old “tabula rasa” (blank slate) model that had dominated thinking about how memory circuits develop.
In plain terms: a newborn mouse’s memory circuit looks less like an empty notebook and more like a notebook in which every page has been scribbled on. The brain’s job during early development is not to start writing but to start erasing, keeping only the marks that matter.
Why this matters for memory
The finding offers a plausible biological explanation for a familiar puzzle: why are infant memories so unreliable? Adults rarely recall anything before age three or four, a phenomenon psychologists call infantile amnesia. If the CA3 circuit is cluttered with indiscriminate connections early in life, it would struggle to separate similar experiences. Two trips to the park, a dog barking, a red ball rolling across grass: in a densely wired network, those events could produce overlapping neural patterns that blur together rather than forming distinct, retrievable memories.
As pruning proceeds and the network becomes sparser, individual experiences could be encoded more distinctly, supporting the sharper recall that emerges in older children. That logic is consistent with the data, but it remains an inference drawn from connectivity patterns, not a direct behavioral demonstration. No one has yet shown that blocking or accelerating pruning in CA3 changes how well a mouse remembers a specific event.
How the hippocampus differs from the cortex
The pruning model is not entirely new to neuroscience. Researchers have long known that the cerebral cortex undergoes synaptic pruning during development, trimming back an initial surplus of connections. But earlier work on the neocortex, described in a study published in the Proceedings of the National Academy of Sciences, found something subtly different: neighboring cortical pyramidal neurons share extensive physical “touches” (structural appositions) that can be converted into functional synapses by activity. The specific authors and publication year of that study are not confirmed in the sources reviewed here, so readers should treat the reference as contextual rather than precisely cited. In that model, the raw material exists, but the connections are not yet switched on. It is closer to a blank slate with pencil marks lightly sketched in.
The ISTA findings push further. In hippocampal CA3, the early connections are not just structural potential waiting to be activated. They are functional synapses, already transmitting signals, that must be selectively removed. The distinction matters because it implies different developmental rules for different brain regions. The cortex may build up from a scaffold of latent contacts. The hippocampus, or at least its CA3 circuit, may carve down from a thicket of active ones.
A separate line of research, published in Nature Neuroscience, adds another wrinkle. That work found that adult hippocampal network dynamics can be preconfigured by the timing of embryonic neurogenesis, meaning some aspects of how the circuit will eventually behave are set before the animal is born and before it encounters any sensory experience. As with the neocortical study, the specific authors and publication year are not confirmed in the sources reviewed here. The finding supports the broader idea that hippocampal circuits are not blank at birth, but it does not specify whether the mechanism involves dense-then-pruned wiring or some other form of pre-patterning. The relationship between embryonic neurogenesis timing and the postnatal pruning the ISTA group measured has not been tested directly.
The limits of what we know
Several important caveats apply. All of the primary data comes from mice. No equivalent mapping of CA3 connectivity across human developmental stages exists, and the ISTA team does not claim direct human applicability. Mouse and human hippocampal circuits share broad organizational features, but the timelines and specific mechanisms of synaptic pruning differ across species. Whether the same dense-to-sparse trajectory occurs in a human infant’s brain is an open question that will require different experimental approaches, likely involving postmortem tissue analysis or advanced neuroimaging, to answer.
The study also cannot yet say how much of the pruning is driven by experience. The ISTA team measured connectivity at fixed developmental time points but did not manipulate the animals’ early environments. Classic work in other brain regions suggests that both spontaneous neural activity and sensory input shape pruning, but whether CA3 follows the same rules is unknown. Experiments that combine circuit mapping with controlled changes in early-life experience, such as sensory enrichment or deprivation, will be needed to clarify how flexible the process is.
There are also technical limits to consider. Multicellular patch-clamp mapping reveals which neurons are functionally connected under laboratory conditions, but it samples only a subset of cells in each recording session. It cannot capture every synapse in the circuit. Complementary approaches, including large-scale anatomical reconstructions and in vivo imaging of activity patterns during behavior, will be important for confirming and refining the picture.
What the pruning model could eventually explain
If the dense-to-sparse trajectory holds up across further experiments and species, it could reshape how scientists think about several longstanding questions. Neurodevelopmental conditions in which synaptic pruning is thought to go awry, including autism spectrum disorder and schizophrenia, have been studied primarily in cortical circuits. Extending that lens to hippocampal CA3 could open new lines of investigation into memory-related symptoms in those conditions. But that possibility remains speculative; no data yet links the specific pruning process described by the ISTA team to any clinical diagnosis.
More immediately, the work challenges researchers to update a metaphor that has shaped how they design experiments and interpret results. If the hippocampus does not wait passively for experience to build its wiring from scratch, then studies of early memory formation need to account for what is already there. The starting state of the circuit is not silence. It is noise, and the developmental task is not construction but editing.
That reframing carries a certain elegance. The brain, it seems, does not learn to remember by slowly accumulating connections. It learns by figuring out which of its many early connections to throw away. Understanding how it makes those choices, and what happens when the process goes wrong, may eventually clarify not only how memories form but why some never take hold at all.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
