← Reference library

REFERENCE LIBRARY

Neuron & Action PotentialNeuron & Action Potential

The only "letter" the brain writes with: a single all-or-none electrical pulse. This page covers the parts it lives on, how it gets triggered, why it has to be all-or-none — and one fact that usually gets left out: large parts of the nervous system don't use that letter at all, and finish their computing in smooth, continuous voltage instead.

Four parts: receive · decide · send · hand off

A neuron isn't mysterious. Functionally it's four stages:

Dendrites

That big bush of branches is the inbox. Signals from thousands of upstream neurons land here, some pushing toward "fire," some toward "don't."

Soma (cell body)

Sums all those pushes continuously. This stage is fully analog — bigger and smaller, no pulses anywhere yet.

Axon

One long output wire; the longest in your body runs from the spine to the toe (close to a metre). Action potentials happen only here.

Terminals

The far end, where the electrical signal is converted into a chemical one (neurotransmitter release) and handed to the next cell's dendrites.

The action potential: one all-or-none flip

At rest there's a voltage difference across the membrane, inside about −70 millivolts relative to outside. Input from the dendrites nudges that number upward, and once it crosses a line (about −55 mV, the threshold) a fixed routine fires off. Whether you barely cross or blow past it, the flip has the same height, shape and duration:

① Threshold

A population of membrane "gates" (sodium channels) is waiting for that voltage. At the line they open together.

② Depolarization

Positively charged sodium ions pour in and the voltage shoots to around +40 mV. This step drives itself: the more gates open the more ions flow, and the more ions flow the more gates open — which is exactly where "all-or-none" comes from. Once started, it can't stop halfway.

③ Repolarization

Sodium gates slam shut, potassium channels open, positively charged potassium leaves, and the voltage drops back — usually dipping slightly below rest.

④ Refractory period

Sodium gates need a millisecond or two to reset, and during that window no amount of input produces another spike. This hard limit sets the neuron's maximum firing rate (a few hundred per second for most cells) and keeps pulses travelling one way instead of doubling back.

Why it has to be all-or-none

Because it has to travel far. A smoothly varying voltage (an analog signal) decays within about a millimetre of membrane. An all-or-none pulse doesn't: it is regenerated from scratch at every step along the way, so a pulse arriving after a metre looks identical to one arriving after a millimetre. The price is that the amplitude dimension is given up entirely — all the information has to be squeezed into when it fires, how densely, and which cell fired.

Myelin is the matching accelerator: an insulating sheath that glial cells wrap around the axon (glial cells). Along the wrapped stretches the pulse isn't rebuilt point by point; it hops, refreshing only at the bare gaps (nodes of Ranvier), which takes conduction speed from around a metre per second to over a hundred. Multiple sclerosis is what happens when the immune system strips that sheath: signals arrive late and out of order.

Counterintuitive: plenty of places never spike at all

Textbooks can leave the impression that nervous system = spikes. Not so. The first cell layers in the retina — photoreceptors and bipolar cells — produce no action potentials whatsoever. They work entirely in continuously varying voltage, computing contrast and edges along the way, and only the last layer (ganglion cells) converts the result into pulses for the trip to the brain (visual pathway).

The rule is clean: spike when you need distance, stay analog when you're computing locally. Summation inside dendrites, synaptic strength, small local circuits — all analog. So "a neuron is a switch" is a convenient simplification; a large share of the actual computation happens in the continuous territory before the switch.

There's also a bill to pay. Spikes are expensive: after every one, ion pumps have to haul the displaced sodium and potassium back where they belong, and those pumps are a major share of the brain's energy budget. So evolution kept firing frugal — only a small fraction of neurons are spiking at any given moment, while the rest sit quiet. That constraint shapes how the brain encodes everything.