Every time you pick up a new skill, whether it is navigating an unfamiliar city or finally nailing a tricky chord on the guitar, your brain has to change. For decades, neuroscientists assumed that meant building fresh connections between neurons, a slow and metabolically expensive process. But a striking discovery from MIT suggests the brain may have been playing a longer game all along: roughly one in four synapses in the adult mouse cortex is already wired up and ready to go, just waiting for the right signal to switch on.
A hidden reserve of ready-made connections
The finding comes from a 2022 study led by Dimitra Vardalaki and colleagues in Mark Harnett’s lab at MIT. Using super-resolution protein imaging and electrophysiology in the adult mouse visual cortex, the team discovered that approximately 25% of excitatory synapses lacked AMPA receptors, the molecular gatekeepers that allow a synapse to pass along a signal under normal conditions. These synapses still carried NMDA receptors, a different type of glutamate receptor that only opens under specific voltage conditions. In practical terms, the connections had all the structural hardware of a working synapse but stayed functionally mute during ordinary brain activity.
Neuroscientists call these “silent synapses.” The concept is not new. Researchers first described connections with NMDA-mediated but no AMPA-mediated responses in the mid-1990s, in foundational work on glutamatergic synaptic plasticity. Those early experiments also showed that long-term potentiation, or LTP, the cellular process most closely linked to learning and memory, could “unsilence” these synapses by triggering the insertion of AMPA receptors into the postsynaptic membrane. Separate studies in developing animals found that many thalamocortical synapses start out silent in early life and convert to active connections through exactly this mechanism.
What changed in 2022 was the prevailing assumption about timing. Most neuroscientists believed silent synapses were a feature of the developing brain, prominent during the critical periods of early childhood and then largely pruned away as circuits matured. The Harnett lab’s data showed that was wrong, at least in the mouse visual cortex. Adult brains retained a dense population of these dormant connections.
Why earlier methods missed them
The structural home for most of these silent synapses turned out to be filopodia: thin, finger-like protrusions that extend from dendrites. Filopodial synapses carried NMDA receptors but not AMPA receptors, making them a physical platform for silence. The problem is that filopodia are tiny and fragile. Earlier light-microscopy techniques and standard electrophysiological recordings had largely missed them or dismissed them as immature structures in the process of being pruned. The super-resolution imaging used by Vardalaki’s team resolved individual synaptic proteins at a scale fine enough to distinguish AMPA-containing from AMPA-lacking contacts, revealing a dense population of these small structures studding the dendritic trees of mature neurons.
A subsequent review in the Annual Review of Neuroscience by the same research group synthesized the broader evidence, arguing that newly formed glutamatergic synapses can begin life in a silent state even in adulthood. The review cataloged what is and is not known about silent synapse prevalence across brain regions, species, and measurement methods, establishing a framework for the field going forward. Its central argument: AMPA-lacking synapses are not developmental relics but appear dynamically throughout life, potentially offering a flexible substrate for experience-dependent rewiring.
About that “30%” number
The headline figure of 30% has circulated widely in popular science coverage, but the peer-reviewed data are more specific. Vardalaki and colleagues measured roughly 25% of excitatory synapses as silent in the adult mouse visual cortex. The higher number likely reflects rounding or informal extrapolation in secondary reporting. No published study has measured silent synapse prevalence across the entire brain, so any whole-brain percentage remains an estimate.
That distinction matters because the 25% figure comes from a single cortical region. No primary research has yet quantified silent synapses in the prefrontal cortex, which governs decision-making and working memory, or in sensory areas beyond vision. Whether the proportion holds steady, rises, or falls in other regions is an open question as of mid-2026.
The gap between mice and humans
All primary physiological measurements of adult silent synapses come from mouse tissue. The assumption that human cortex harbors a similar reservoir is reasonable given the conserved molecular machinery of glutamatergic signaling across mammals, but no direct study has confirmed it with human neurons in situ. Bridging that gap will likely require either rare neurosurgical tissue samples or the development of noninvasive biomarkers that can be tied back to synaptic receptor states. Some researchers have begun exploring human iPSC-derived neurons as a proxy, but those lab-grown cells do not replicate the full complexity of intact cortical circuits.
Can silent synapses actually drive learning?
This is the question that matters most to anyone reading beyond the headline, and it is the one with the least direct evidence so far. The mechanistic case is strong: LTP can convert silent synapses into active ones in brain-slice preparations, and the sheer number of dormant connections suggests a large capacity for rapid circuit remodeling without the need to grow entirely new synapses. Research has shown that this postsynaptic conversion affects synaptic gain, transmission dynamics, and the probability that a given input will drive a neuron to fire.
But no published study has yet tracked silent synapse activation in a living animal performing a specific learning task. The leap from “these synapses can be unsilenced in a dish” to “they drive real-time skill acquisition in a behaving animal” still lacks direct in vivo behavioral evidence. The claim that silent synapses “sit waiting until you need to learn something new” is a reasonable inference from the molecular and circuit data, not a directly observed behavioral sequence.
The relationship between silent synapses and disease is similarly preliminary. Reviews have discussed AMPA-silent synapses in the context of addiction, stress-related disorders, and neurodegeneration, but those discussions rely on indirect evidence and theoretical frameworks rather than longitudinal patient data. No primary study has tracked how aging or specific diseases alter the dynamics of adult silent synapses over time.
What researchers are trying to figure out next
Even at the cellular level, several basic questions remain unanswered. Scientists do not yet know how stable filopodial silent synapses are over weeks or months, or how frequently they transition into fully active, AMPA-containing synapses under naturalistic patterns of brain activity. The molecular triggers that determine whether a given filopodium will stabilize, unsilence, or retract are still being mapped. The relative contributions of presynaptic versus postsynaptic changes during unsilencing remain debated, as does the influence of neuromodulators like dopamine and acetylcholine on the process.
The most anticipated experiments will combine in vivo imaging, genetic tagging of silent synapses, and carefully designed behavioral tasks to test whether specific learning episodes activate specific dormant connections. That kind of study would move the field from a compelling mechanistic story to a demonstrated causal chain. Until then, the safest interpretation is that adult cortex contains a substantial pool of structurally identifiable, AMPA-lacking synapses that can, in principle, be converted into active connections, offering a powerful candidate mechanism for how mature brains learn quickly without wholesale rewiring. The full picture of what these synapses actually do in a thinking, learning brain is still coming into focus.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
