The Physics of Memory: How Your Brain Stores Information

Your Hard Drive Is Made of Meat

I want to be upfront about something. The physics of how your brain stores memories is not fully understood. Not even close. We know a lot of the pieces. We know the key molecular mechanisms. We can watch neurons fire in real time and map synaptic connections with electron microscopes. But the complete story — how a subjective experience becomes a stable physical trace in neural tissue that can be recalled years later — remains one of the hardest open problems in science.

That said, what we do know is remarkable. And it’s physics all the way down. Ion gradients, electric fields, diffusion kinetics, protein folding, and thermodynamics. Your memories are not stored in some ethereal medium. They’re stored in the physical structure of your brain — in the geometry and chemistry of connections between cells. Let me walk through how.

The Neuron: An Electrical Machine

Your brain contains roughly 86 billion neurons. Each one is, at its most basic, an electrochemical device that receives inputs, integrates them, and occasionally fires an output signal. The physics is membrane biophysics — the behaviour of a thin lipid membrane studded with protein channels that selectively allow ions to pass.

At rest, a neuron maintains a voltage difference of about -70 millivolts across its cell membrane — negative inside relative to outside. This resting potential is maintained by ion pumps (sodium-potassium ATPase) that actively transport 3 Na⁺ ions out for every 2 K⁺ ions in, creating concentration gradients. Potassium is concentrated inside the cell; sodium is concentrated outside. Leak channels allow some potassium to diffuse out, making the interior more negative. The equilibrium is described by the Nernst equation — straightforward thermodynamics applied to ion distributions across a semi-permeable membrane.

When the neuron receives enough excitatory input to push its membrane potential above a threshold (roughly -55 mV), voltage-gated sodium channels open explosively. Na⁺ ions rush into the cell, driven by both the concentration gradient and the electrical gradient. The membrane potential spikes to about +40 mV in less than a millisecond. This is the action potential — the neuron’s fundamental signal.

The action potential is all-or-nothing. It fires at full amplitude or not at all. There’s no half-signal, no weak spike. This digital character is important — it means signals can propagate over long distances (down axons up to a metre long in motor neurons) without degrading, because each segment of the axon regenerates the full spike independently. It’s like a fuse burning along a wire — each section ignites the next at full intensity.

After firing, potassium channels open, K⁺ flows out, and the membrane repolarises. The whole cycle takes about 1–2 milliseconds. A neuron can fire up to several hundred times per second, and the firing rate encodes information — more excitation means more spikes per second.

The Synapse: Where Memory Lives

The action potential races down the axon and arrives at a synapse — the junction between one neuron and the next. There are roughly 100 trillion synapses in your brain. And this is where memory physically happens.

At a chemical synapse, the arriving action potential triggers calcium ions (Ca²⁺) to flow into the presynaptic terminal through voltage-gated calcium channels. The calcium influx causes vesicles — tiny membrane-bound sacs containing neurotransmitter molecules — to fuse with the cell membrane and release their contents into the synaptic cleft, a gap of about 20 nanometres. The neurotransmitter molecules (glutamate, for most excitatory synapses) diffuse across the cleft and bind to receptor proteins on the postsynaptic neuron.

This entire process — electrical signal to chemical signal back to electrical signal — takes about 0.5–1 millisecond. It’s astonishingly fast for a biochemical process, and the speed is limited largely by diffusion physics: how quickly neurotransmitter molecules random-walk across the 20 nm gap.

Now here’s the key. The strength of a synapse — how effectively an action potential in the presynaptic neuron excites the postsynaptic neuron — is not fixed. It’s adjustable. And adjusting synaptic strength is, as far as we understand it, the fundamental mechanism of memory.

Long-Term Potentiation: The Memory Mechanism

In 1973, Terje Lømo and Tim Bliss discovered that brief, high-frequency stimulation of synapses in the hippocampus (a brain region critical for forming new memories) produced a long-lasting increase in synaptic strength. They called it long-term potentiation — LTP. It was, and remains, the best candidate mechanism for how memories are physically encoded.

Here’s how it works at the molecular level. And it’s gorgeous physics.

The postsynaptic membrane contains two types of glutamate receptors that matter here: AMPA receptors and NMDA receptors. AMPA receptors are the workhorses — when glutamate binds, they open and allow Na⁺ to flow in, depolarising the postsynaptic neuron. Standard excitatory transmission.

NMDA receptors are different. They’re coincidence detectors. The channel has a magnesium ion (Mg²⁺) physically blocking its pore at resting potential. Glutamate can bind to the receptor, but no current flows because the magnesium plug is in the way. Only when the postsynaptic membrane is already depolarised — by simultaneous AMPA activation from strong or repeated input — does the electrical field dislodge the magnesium ion and allow the NMDA channel to open.

When NMDA channels open, they admit calcium ions. And calcium is the trigger for everything that follows.

The calcium influx activates a cascade of intracellular enzymes — CaMKII, PKC, and others — that phosphorylate existing AMPA receptors (making them more conductive) and trigger the insertion of additional AMPA receptors into the postsynaptic membrane. More receptors means the synapse responds more strongly to the same amount of neurotransmitter. The synapse has been strengthened. This is early LTP, and it can happen within minutes.

For the memory to last longer than a few hours, protein synthesis is required. The calcium signal activates transcription factors (CREB is the most studied) that travel to the cell nucleus and trigger the production of new proteins. These proteins are shipped back to the specific strengthened synapses — a logistical challenge that isn’t fully understood, since a single neuron can have 10,000 synapses and needs to deliver new proteins to only the ones that were active. The new proteins restructure the synapse physically: the dendritic spine (the small protrusion where the synapse sits) grows larger, new scaffolding proteins stabilise the extra receptors, and sometimes entirely new synaptic connections form.

This is your memory, physically. A pattern of synapses, across a network of neurons, that have been structurally remodelled at the molecular level. The memory is the structure. Destroy the structure, and the memory is gone.

The Hippocampus: Temporary Storage

Not all brain regions handle memory the same way, and this is where the physics of information transfer between regions matters.

The hippocampus — a seahorse-shaped structure deep in the temporal lobe — is essential for forming new declarative memories (facts and events). Damage to both hippocampi, as in the famous case of patient H.M. (Henry Molaison), destroys the ability to form new long-term memories while leaving old memories largely intact.

The current model is that the hippocampus acts as a temporary, fast-learning storage system. New experiences are rapidly encoded in hippocampal synapses through LTP. Over time — days to weeks to months — the memories are gradually transferred to the neocortex (the brain’s outer layer) through a process called systems consolidation. This transfer happens primarily during sleep, especially during slow-wave sleep, when the hippocampus “replays” recent experiences in compressed form, driving synaptic changes in cortical networks.

Why the two-stage system? Speed versus capacity. The hippocampus can learn quickly — a single dramatic event can create a strong hippocampal memory trace immediately. But hippocampal capacity is limited. The neocortex has vastly more synapses and can store far more information, but it learns slowly — it needs repeated exposure (or repeated hippocampal replay during sleep) to form stable connections.

This is why sleep deprivation destroys memory formation so effectively. It’s not just that you’re tired and inattentive. You’re physically preventing the hippocampal-to-cortex consolidation process. The memories encoded during the day never get properly transferred, and they fade.

Forgetting: Controlled Demolition

Forgetting is not a failure of the system. It’s a feature.

If you remembered everything with perfect fidelity, you’d be overwhelmed. Every face on every commute, every word of every conversation, every flicker of every screen. The cognitive load would be paralysing. Borges wrote a story about this — “Funes the Memorious” — about a man who remembered everything and found it unbearable. The physics of synaptic plasticity includes mechanisms for weakening connections, not just strengthening them.

Long-term depression (LTD) is the reverse of LTP. Low-frequency, weak stimulation of synapses — the kind that occurs when a connection is rarely used — triggers a cascade that removes AMPA receptors from the postsynaptic membrane and shrinks dendritic spines. The synapse weakens. The memory trace fades. The molecular machinery that builds memories can also disassemble them.

There’s also active forgetting — recently discovered mechanisms in which specific molecular signals tag synapses for weakening. The protein Rac1, for example, promotes the active erasure of fear memories in animal models. This isn’t passive decay. It’s the brain deliberately clearing information that it evaluates as no longer useful. The evaluation criteria aren’t fully understood, but emotional significance, repetition, and sleep-dependent consolidation all seem to play roles.

How Much Can You Store?

The storage capacity of the human brain is genuinely difficult to estimate, because the brain doesn’t store information the way a computer does. There’s no bit, no byte, no memory address. Information is distributed across networks, encoded in continuous synaptic weights, and tangled up with processing in ways that make clean capacity estimates almost meaningless.

That said. The Salk Institute published a study in 2016 estimating that each synapse can store about 4.7 bits of information, based on evidence that synapses have roughly 26 distinguishable strength levels. With about 100 trillion synapses, that gives a raw storage capacity of roughly 1 petabyte — a million gigabytes. Other estimates, using different assumptions, range from 100 terabytes to 2.5 petabytes.

For context, 1 petabyte is roughly the storage capacity of 500 million floppy disks, or about 13 years of continuous HD video. It’s a lot. But the comparison is misleading because the brain’s “storage” is nothing like a hard drive. You can’t write arbitrary data to it. You can’t read it back bit by bit. It degrades, transforms, and reconstructs every time you access it. And the same physical substrate that stores information also processes it — there’s no separation between memory and computation.

Your brain is not a computer. It’s something else entirely — something that uses physics (electricity, diffusion, thermodynamics, mechanical forces) to do something that no human-built machine yet replicates: turn experience into lasting structure, and structure back into experience.

The Hard Problem Hiding in the Physics

I want to end with an honest admission. Everything I’ve described — action potentials, synapses, LTP, protein cascades, systems consolidation — is mechanism. It’s the how. It doesn’t touch the why. Why does a pattern of strengthened synapses feel like a memory? Why does reactivating a neural circuit produce the subjective experience of remembering your grandmother’s kitchen?

This is the hard problem of consciousness, and it sits stubbornly at the intersection of physics, neuroscience, and philosophy. The physics of ion channels and synaptic weights is well-characterised. The emergence of subjective experience from those physical processes is not understood at all. Not even a little.

Every memory you have is, at one level, a collection of modified protein structures at specific synapses in a network of electrically excitable cells. At another level, it’s your past. Your identity. The thing that makes you you rather than a collection of atoms that happens to be shaped like a person.

Both descriptions are true. How they connect is the deepest open question in science. And it starts with the physics of a 70-millivolt potential across a lipid membrane, eight billion times over.