How Memories Form: Synaptic Plasticity (LTP)

A memory is not stored inside a single cell — it lives in the strength of the connections between cells. This is one synapse in your hippocampus, the brain’s memory-forming hub. Press play and watch Hebb’s rule in action: when the sending neuron and the receiving neuron fire together, a gatekeeper called the NMDA receptor spits out its magnesium plug, calcium pours in, and the synapse wires itself stronger by pulling in more AMPA receptors. That strengthening is long-term potentiation (LTP) — the cellular act of learning.

Try this: start on Weak input and watch nothing stick. Then switch to Coincident strong input and see the magnesium blown out of the NMDA channel, calcium flood the spine, and new AMPA receptors slot into the membrane. Finally hit NMDA blocked (alcohol) to watch a memory fail to form — the blackout.

The Science in Plain Language

1. A memory is a change in a connection

Your brain has roughly 86 billion neurons, and each one reaches out to thousands of others across tiny junctions called synapses — something like a hundred trillion connections in all. Learning does not add new facts to a hard drive. It re-weights these connections: some get stronger, some get weaker. The psychologist Donald Hebb captured the idea in 1949 in a single sentence, now paraphrased as “neurons that fire together, wire together.” In 1973, Terje Lømo and Tim Bliss found the physical proof in the rabbit hippocampus: a brief high-frequency burst of activity made a synapse fire bigger for hours afterward. They called it long-term potentiation, or LTP — and it is still the best cellular model we have for how a memory is written.

2. The NMDA receptor is a coincidence detector

At the heart of the animation is the NMDA receptor, and it is genuinely clever. It is a channel that will only open when two things happen at the same time. First, the sending neuron must release glutamate (the brain’s main “go” signal) onto it. Second, the receiving spine must already be depolarised — electrically excited. Why both? Because at rest the NMDA pore is physically corked by a magnesium ion (Mg²⁺). Glutamate alone cannot push it out. Only when the membrane voltage climbs — here, above roughly −35 mV — does the positive magnesium get electrically shoved out of the pore. Glutamate and voltage together: that is the molecular version of “fire together.” The neighbouring AMPA receptors do the depolarising, letting sodium (Na⁺) in the instant glutamate arrives.

3. Calcium is the “make this stronger” signal

The magic of the NMDA channel is what it lets through: calcium (Ca²⁺). A resting spine holds calcium at only about 50–100 nanomolar — vanishingly little. When a coincident signal blows the magnesium out, calcium floods in and can briefly spike into the micromolar range, a hundred-fold jump, right at that one spine. That calcium pulse is the switch. It activates an enzyme called CaMKII (calcium/calmodulin-dependent protein kinase II), which is so central that neuroscientists nickname it the “memory molecule.” A key rule — the calcium-control hypothesis — is that a big, brief calcium spike says “strengthen” (LTP), while a small, prolonged rise says “weaken” (LTD). Watch the calcium meter: only the coincident scenario pushes it past the LTP line.

4. AMPA receptors: turning up the volume

How does a spine actually get “stronger”? It adds more AMPA receptors to its surface. CaMKII, switched on by calcium, drives extra AMPA receptors (built from subunits called GluA1) out of a reserve pool inside the spine and into the membrane. Now the very same puff of glutamate lands on more receptors and produces a bigger electrical response — the synapse has literally turned up its own volume. In the lab this shows up as the field response climbing to 150–250% of baseline. That is what you see in the animation when new teal receptors slide into the membrane and the strength readout climbs.

5. From minutes to years: protein synthesis and sleep

The first phase of LTP (early-LTP) just reshuffles receptors you already had — it fades in an hour or two. To make a memory last, the spine must build new parts. Repeated, spaced activity switches on genes: transcription factors like CREB and growth signals like BDNF trigger fresh protein synthesis, and the spine physically grows a bigger head and a wider contact. This is late-LTP, and it is why a drug that blocks protein synthesis (anisomycin, in animal studies) can erase a long-term memory while leaving a short-term one intact. Much of this consolidation happens while you are asleep: during deep sleep the hippocampus “replays” the day’s activity in fast bursts (sharp-wave ripples), stabilising the important synapses and letting go of the noise.

6. Long-term depression: use it or lose it

Strengthening is only half the story. The opposite process, long-term depression (LTD), actively weakens synapses that fire weakly or out of sync — the small, prolonged calcium signal from section 3. LTD removes AMPA receptors from the membrane, and unused spines can shrink and disappear entirely (synaptic pruning). This is not a bug; it is essential. A brain that only ever strengthened would saturate into noise. Forgetting the trivial is what lets the important stand out. In the animation, leave it on Weak input and watch the strength drift down: that is your brain pruning what you do not practise.

7. What this means for learning (and an honest myth)

Every study trick that actually works maps onto this biology. Active recall (testing yourself) forces strong, coincident firing — real LTP — whereas passively re-reading barely nudges the synapse. Spaced repetition beats cramming because repeated, separated activations are exactly what flips early-LTP into durable late-LTP; cramming saturates a synapse that then decays overnight. And sleep is not optional downtime — it is when consolidation runs. The honest myth-correction: you cannot “learn in your sleep” by playing audio while unconscious — the hippocampus cannot encode new facts then. What sleep does, powerfully, is lock in what you already learned while awake. Re-reading your notes at midnight feels productive; a quiz plus a good night’s sleep is what your synapses actually respond to.

8. When memory breaks: Alzheimer’s, alcohol, ketamine

Because this one mechanism underlies memory, damaging it produces disease. In Alzheimer’s disease, sticky amyloid-β fragments collect at synapses, block LTP, push synapses toward LTD, and ultimately cause the synapse loss that tracks best with the actual memory decline — try the Amyloid-β button and watch strengthening fail. Alcohol and the anaesthetic ketamine both block the NMDA receptor; with the coincidence detector shut, no calcium enters and no new memory is written. That is what an alcoholic blackout actually is — not forgetting, but a failure to record in the first place, because LTP never happened. (The Alzheimer’s drug memantine is a gentle NMDA blocker — it dampens the toxic background calcium noise without fully closing the channel.)

↑ Back to the animation

Connections

• All Interactive Visualizations
• Nerve Impulse
• Dopamine & Reward
• Alzheimer’s Disease
• PTSD