Bench modeSteps, parts, and safety only. Big type for a phone at the bench.

Synapses and plasticity

How one neuron changes another, how the connection strengthens or weakens with use, and why that is the mechanism behind a BCI user learning to control a cursor and a stroke patient recovering.

AssumesNeuron biophysicsSpineComputational neuroscience

You are skimming: the title, the first figure, and the short version. Switch to Read in the header for the full page, or Deep to open every deep dive.

A spike arriving at the end of an axon opens calcium channels, vesicles of fuse with the membrane, the transmitter diffuses across a 20-nanometre gap and opens channels on the next cell, and that cell’s membrane potential moves by a fraction of a millivolt. That is a . There are perhaps a hundred trillion of them in your head, and each one’s strength changes with use. The changing is , it is the physical basis of learning, and it is also the reason a person can learn to drive a cursor with a decoder that was fit badly.

Excitation and inhibition

Glutamate is the main excitatory transmitter: it opens channels that let sodium in and depolarize the next cell by a millivolt or less per synapse. GABA is the main inhibitory one: it opens chloride channels that hold the cell near rest. A cortical pyramidal neuron receives tens of thousands of synapses, roughly 80 percent excitatory, and it fires when enough excitatory inputs arrive close together in time and space to overcome the leak, the inhibition, and the threshold. Cortex is a balance: excitation and inhibition track each other closely, and nearly every network rhythm in the next explainer comes from the interplay.

The coincidence detector

One glutamate receptor, the NMDA receptor, only passes current when the transmitter is present and the membrane is already depolarized (a magnesium ion blocks it at rest and is expelled by depolarization). So it opens when the presynaptic neuron fired and the postsynaptic neuron was active at the same time. It passes calcium, and calcium is the signal that changes the synapse. This one molecule implements Hebb’s rule from 1949: cells that fire together wire together.

Long-term potentiation and depression

A brief high-frequency burst of input to a synapse can double its strength for hours, days, or longer: long-term potentiation, discovered in the hippocampus in 1973 and found nearly everywhere since. Low-frequency input weakens it: long-term depression. The mechanisms involve inserting or removing receptors from the postsynaptic membrane and, over longer times, growing or shrinking the synapse itself. is the temporal version: if the input spike arrives a few milliseconds before the output spike, the synapse strengthens; a few milliseconds after, it weakens. Causality gets rewarded.

Predict before you look

A BCI decoder is fit once and then left fixed for weeks while a person practices with it daily. The decoder is imperfect. What happens to performance?

Improves, often a lot. Studies in monkeys and humans show that with a fixed decoder, the recorded neurons’ tuning shifts over days toward what the decoder rewards, and performance climbs. The brain treats the decoder as a new limb and learns it. This is plasticity in service of an engineering shortcut, and it is why the field talks about co-adaptation: a learning user and a learning decoder, ideally not fighting.

Homeostasis

If potentiation ran unchecked, every synapse would saturate. Neurons scale all their synapses up or down to keep their average firing in a working range, over hours to days. This homeostatic plasticity is why a network can learn without exploding and why, after a sensory loss, the deprived cortex becomes more excitable and takes on new inputs, which is the basis of sensory substitution and of the cross-modal reorganization seen in blind and deaf people.

Why an engineer needs this

Three reasons. Learning to use a device. Cochlear implant users take months to a year to reach their best speech understanding; the implant does not change, the auditory cortex does. Every prosthesis and every BCI depends on the user’s plasticity, and the rate of learning is part of the design. Rehabilitation. Stroke recovery is plasticity: undamaged cortex taking over functions. Rehabilitation technology, including BCI-driven robotics and vagus nerve stimulation paired with therapy, works by promoting it, and the pairing timing matters for exactly the STDP reasons above. Neuromodulation mechanisms. Whether deep brain stimulation works by silencing, by driving, or by inducing plasticity is still argued; the answer decides how to program it.

Deep dive Neuromodulators and the third factor 3 min

Dopamine, acetylcholine, noradrenaline, and serotonin are released broadly and change how plastic synapses are, rather than carrying specific signals. A synapse that would not change on its own will potentiate if dopamine arrives shortly after, which is how reward teaches: three-factor learning, with the third factor saying “that was worth remembering.” Vagus nerve stimulation paired with rehabilitation exploits this by triggering noradrenaline and acetylcholine release at the moment of a movement, and it was approved for stroke rehabilitation on that basis.

Deep dive Plasticity and the foreign body response 2 min

An implanted array records fewer neurons each year as scar tissue grows and neurons near the electrodes die or retreat. Meanwhile the surviving neurons’ tuning drifts. Decoders that were fit to a stable population must be refit, and the user re-learns. Whether the drift is the tissue, the electrodes, or the brain’s own plasticity is an open question that matters enormously for chronic BCIs; Phase 6 returns to it.

Recall
How does the NMDA receptor implement Hebb's rule?
It opens only when glutamate is present (presynaptic firing) and the postsynaptic membrane is depolarized (postsynaptic activity), passing calcium that triggers synaptic strengthening. Coincident activity strengthens the connection.
Recall
What does spike-timing-dependent plasticity reward?
Causal timing: an input spike just before the output spike strengthens the synapse; just after weakens it.
Recall
Why do cochlear implant users improve for months after activation even though the device does not change?
Auditory cortex reorganizes through plasticity to make use of the new sparse input; the learning is in the brain, and its rate is part of the device's effective performance.
Explain it to your roommate

Explain what this page was about to your roommate in three sentences. No jargon they would not know.