An engram is the physical trace a memory leaves in the brain. When you learn something new, a specific subset of neurons becomes active, and those cells undergo lasting chemical and structural changes that allow the experience to be stored and later recalled. The concept dates back more than a century, but only in the past fifteen years have researchers gained the tools to tag, observe, and even artificially reactivate individual engram cells in living animals. What has emerged is a picture of memory that is far more dynamic, distributed, and manipulable than most people imagine.
How Scientists Find Engram Cells
The biggest practical challenge in studying engrams has always been telling which neurons, out of the billions in a brain, actually participated in encoding a particular memory. The breakthrough came from a class of genes called immediate early genes (IEGs). These genes are normally quiet, but when a neuron fires intensely during an experience, IEGs switch on rapidly. Researchers can detect this activity after the fact, which effectively gives them a snapshot of which cells were engaged during learning.1PubMed Central. Retrosplenial cortex in spatial memory: focus on immediate early genes mapping One particularly well-studied IEG, c-Fos, has become a workhorse for engram research. Scientists can engineer mice so that neurons expressing c-Fos during a specific experience also produce a fluorescent protein or a light-sensitive channel, permanently tagging those cells for later identification or reactivation.2Current Biology. Engram stability and maturation underlie the transition from recent to remote memory
This tagging approach lets researchers do something remarkable: they can let an animal form a memory, label the neurons involved, and then come back days or weeks later to ask whether those same neurons reactivate when the animal remembers. The ability to visualize and manipulate memory traces at the level of individual cells has transformed engram research from a theoretical idea into an experimental science.3Molecular Brain. Catching the engram: strategies to examine the memory trace
Proving That Engrams Are Real
Seeing which neurons are active during learning is suggestive, but it does not prove those cells actually hold the memory. The definitive test came from optogenetics, a technique that uses pulses of light to switch genetically modified neurons on or off. In a landmark experiment, researchers trained mice to fear a particular environment by pairing it with a mild foot shock, then tagged the hippocampal neurons that were active during that fear learning with a light-sensitive protein called channelrhodopsin-2. Later, when the mice were placed in a completely different, safe environment and the tagged neurons were activated with light, the animals froze in fear, as if they were reliving the original frightening experience.4PubMed Central. Optogenetic stimulation of a hippocampal engram activates fear memory recall
The freezing only happened when the light was on, and only in mice whose fear-learning neurons had been tagged. This demonstrated that reactivating a sparse, specific set of hippocampal neurons was enough to trigger recall of a particular memory. It was the first time anyone had shown that stimulating an engram, rather than just observing it, could produce a behavioral response indistinguishable from natural remembering.
How the Brain Picks Which Neurons Store a Memory
Only a tiny fraction of available neurons end up in any given engram. In the hippocampal dentate gyrus, for example, only a few percent of cells are recruited for a single memory. This raises the question of what determines which neurons get chosen. Research points to two factors working together: neurons that happen to be more excitable at the moment of learning are more likely to be recruited, and neurons with higher levels of a protein called CREB (which boosts excitability) have a competitive advantage in being included in the engram.5PubMed Central. Which Neurons Will Be the Engram – Activated Neurons and/or More Excitable Neurons?
Keeping engrams sparse turns out to be critical for distinguishing one memory from another. If too many neurons were recruited every time, different memories would overlap extensively, making them hard to tell apart. The brain enforces sparsity through a form of lateral inhibition: when a granule cell in the dentate gyrus fires during learning, it activates nearby inhibitory interneurons, which suppress the dendrites of surrounding cells and prevent them from joining the engram. Experiments have shown that when this inhibitory brake is weakened, the engram expands, and contextual fear memories become less precise.6Neuron. Inhibitory Interneurons Regulate Engram Size and Prevent Over-Activation So memory specificity depends not just on which neurons fire, but on which neurons are actively prevented from firing.
What Changes Inside Engram Cells
Being recruited into an engram is not just about firing once. The neurons undergo structural changes that physically encode the memory. Dendrites, the branching input-receiving extensions of neurons, are increasingly recognized as fundamental units of memory storage. They can integrate incoming signals in complex, nonlinear ways and support local protein production right at the site where connections between neurons are strengthened or weakened.7PubMed Central. The dendritic engram Dendritic spines, the tiny protrusions where synapses form, grow, shrink, or stabilize depending on the memory being stored. These physical modifications mean that the engram is not just a pattern of activity. It is a structural rearrangement of the cell itself.
Beyond structural changes, epigenetic mechanisms also play a role. Chemical modifications to DNA and the proteins that package it can alter how genes in engram neurons are read out over time, translating a brief learning experience into a long-lasting change in cell function.8PubMed. Engrampigenetics: Epigenetics of engram memory cells This layer of regulation helps explain how a memory that was formed in seconds can persist for years: the epigenetic marks act as a kind of molecular bookmark, keeping the engram neurons in a state that supports future reactivation.
How Memories Move From the Hippocampus to the Cortex
A new conscious memory initially depends on both the hippocampus and the neocortex. Over time, through a process called systems consolidation, the hippocampus gradually guides a reorganization so that the cortical networks can support the memory on their own.9PubMed Central. Memory consolidation This is why damage to the hippocampus often wipes out recent memories but leaves older ones largely intact: the older memories have already been “transferred” to cortical storage.
Engram research has revealed that this consolidation is not a simple copy-and-paste. When researchers label engram cells at the time of learning and then check back weeks later, the engram has matured. The ensemble of tagged cells in the hippocampus becomes less critical for recall, while cortical engram cells that were subtly tagged from the start become more important. Studies tracking engram stability over time have shown that the cortical engram is not created from scratch during consolidation; traces of it exist almost immediately, but they require time and repeated reactivation (much of which happens during sleep) to become fully functional.2Current Biology. Engram stability and maturation underlie the transition from recent to remote memory
Silent Engrams and the Nature of Amnesia
One of the most surprising discoveries in engram research is that a memory can physically exist in the brain even when the animal cannot recall it. Researchers found that when they blocked protein synthesis shortly after learning, mice developed amnesia and could not retrieve the memory through normal cues. But the engram cells were still there. When those cells were artificially reactivated with light, the mice showed full recall, behaving as if they had never lost the memory.10PubMed Central. Silent memory engrams as the basis for retrograde amnesia
The researchers coined the term “silent engrams” for this phenomenon. The implication is profound: at least some forms of amnesia may not involve the destruction of a memory but rather a failure of the retrieval pathway. The information is encoded in the cells, but the connections needed to access it through natural recall cues have been disrupted. This reframes the traditional view that amnesia equals memory loss, suggesting instead that it can be a problem of access rather than storage.
Forgetting as an Active Process
Forgetting has traditionally been thought of as the passive decay of unused memories. Engram research challenges that idea. Evidence now suggests that forgetting involves active biological mechanisms that gradually rewire engram circuits, making certain memories harder to access over time.11PubMed. Neurobiological mechanisms of forgetting across timescales Rather than a flaw in the system, this appears to be a feature: the brain seems to prioritize memories that are currently relevant while reducing access to those that are not. Active forgetting helps prevent the accumulation of outdated or irrelevant information that could interfere with present decision-making.
The connection between silent engrams and active forgetting raises an interesting question: if forgotten memories still exist physically in the brain, how many of your “lost” memories are actually silent engrams waiting for the right reactivation? The honest answer is that nobody knows. Most of this work has been done in mice under very controlled conditions, and it remains unclear how broadly the concept applies to the complex, everyday forgetting that humans experience.
Memories Are Not Fixed Records
Even after a memory has been consolidated and stabilized, recalling it opens a window during which it can be changed. This process, called reconsolidation, was demonstrated in a breakthrough experiment where rats were trained to fear a sound, then given a drug that blocks protein synthesis in the amygdala immediately after they were reminded of the fear memory. The drug erased the fear response, suggesting that the act of recall had temporarily destabilized the memory, and without the protein synthesis needed to re-stabilize it, the memory was lost.12PubMed Central. Memory engram stability and flexibility
At the molecular level, reconsolidation involves two stages. First, when a memory is recalled and new information is present, the engram is destabilized through protein degradation. Then, it is re-stabilized through new protein synthesis, which can incorporate updated information.13PubMed Central. Molecular Mechanisms of Reconsolidation-Dependent Memory Updating This means your memory of an event is not a fixed recording. Every time you recall it, there is a brief period where the content can shift. This has obvious implications for eyewitness testimony and therapeutic approaches to trauma, though translating the precise pharmacological interventions used in animal studies to human clinical practice remains a major challenge.
Creating False Memories by Manipulating Engrams
Perhaps the most dramatic demonstration of what engrams are came from experiments in which researchers created entirely false memories in mice. In one study, mice were allowed to explore a safe environment (Context A), and the hippocampal neurons active during that exploration were tagged. The next day, those tagged Context A neurons were artificially reactivated with light while the mouse simultaneously received a foot shock in a different environment (Context B). The result: the mice later showed fear when returned to Context A, the safe environment where they had never been shocked. Their brains had linked the artificial reactivation of the Context A engram with the shock, producing a fear memory for an event that never happened.14PubMed. Creating a false memory in the hippocampus
The false memory was not a vague unease. It was context-specific, it activated the same brain regions as a natural fear memory, and it drove robust fear behavior.15PubMed Central. Inception of a false memory by optogenetic manipulation of a hippocampal memory engram Follow-up work showed that the timing of engram manipulation matters: silencing engram neurons shortly after fear conditioning could prevent a false memory from forming, while activating them later could create one.16PubMed Central. The role of neuronal excitability, allocation to an engram and memory linking in the behavioral generation of a false memory in mice These experiments show that memories are constructed from the reactivation of cell ensembles, and if you manipulate the ensemble, you manipulate the memory, regardless of whether the experience it represents actually occurred.
Emotion and Engrams
Engrams do not just store what happened. They encode how you felt about it. Research in the amygdala, a brain region central to emotional processing, has shown that fear memories and the memories formed during fear extinction (learning that something once feared is now safe) are stored in genetically distinct populations of neurons. Strikingly, the neurons that store fear extinction memories overlap significantly with neurons that respond to natural rewards. Activating one set can drive the behavior associated with the other, suggesting that learning that something is safe is, at the neuronal level, a form of reward learning that actively opposes the original fear engram.17PubMed. Amygdala Reward Neurons Form and Store Fear Extinction Memory
This finding has implications for how we think about anxiety disorders and exposure therapy. If fear extinction works by building a new reward-like engram that competes with the old fear engram, then the fear is not erased but rather overridden. It also means the original fear engram may linger and potentially resurface, which matches clinical experience: phobias and PTSD symptoms sometimes return after apparently successful treatment.
Infantile Amnesia and Engram Expression
Most mammals, including humans, cannot recall events from very early life. This phenomenon, infantile amnesia, has puzzled researchers for decades. Engram labeling technology has offered a new angle on it. In mouse experiments, infant animals were shown to form engrams during early experiences, but those engrams were effectively silent: they could not be retrieved through natural cues. When researchers artificially reactivated the tagged engram cells, however, the memories came back, suggesting the memories were encoded but inaccessible.18PubMed Central. Immune activation state modulates infant engram expression across development
Intriguingly, the same study found that immune activation during development altered whether infantile amnesia occurred. Male mice born to mothers with immune activation (used as a model for autism spectrum disorder) did not experience infantile amnesia in the way typical mice did, and their engram ensembles showed differences in size and dendritic spine structure. This hints that the developmental switches governing whether early memories become accessible are not purely about brain maturation. They may be influenced by immune and inflammatory states during early life, opening a surprising link between the immune system and memory expression.
What We Know About Engrams in Humans
Nearly all of the engram manipulation work described above has been done in mice, for the obvious reason that it requires genetic engineering and invasive procedures. But recordings from human brains, typically in epilepsy patients who have electrodes implanted for clinical purposes, provide complementary evidence. Single-neuron recordings from the human medial temporal lobe, including the hippocampus, have revealed cells with remarkably selective response profiles: some fire for a specific concept or person regardless of how that concept is presented, while others track whether a stimulus is familiar or is being recalled from an episode.19PubMed Central. The Architecture of Human Memory: Insights from Human Single-Neuron Recordings These findings are consistent with the engram framework established in animal research, showing that specific human neurons are tied to specific memories and concepts.
One insight from human recordings is that semantic information (the general meaning of something) is activated before episodic information (the specific experience of encountering it), and that memory content and memory strength appear to be carried by different cell populations. Another is that some visually selective cells remain active for several seconds after a stimulus disappears, providing a cellular basis for working memory in humans. These observations map onto the broader engram picture: memory is not one thing happening in one place but a distributed, multi-layered process in which different aspects of an experience are handled by different neural populations, even within the same brain region.
New Tools for Mapping Engrams Across the Whole Brain
Early engram studies focused on one region at a time, often the hippocampus or amygdala. But memories, especially rich real-world memories, are almost certainly distributed across many brain areas. Recent advances in tissue clearing, which makes brain tissue transparent so it can be imaged in three dimensions, combined with genetic labeling of engram cells, are now allowing researchers to map engrams across the entire brain at once.20PubMed Central. Brain-wide immunolabeling and tissue clearing applications for engram research These brain-wide maps are starting to reveal how different regions contribute to a single memory and how those contributions change as a memory ages and consolidates. The technology is still mostly limited to animal models, but it represents a shift from studying engrams as local phenomena to understanding them as whole-brain network events. How those networks reorganize over weeks and months, and how failures in that reorganization might contribute to memory disorders, is one of the most active frontiers in the field.