A neuron's parts and the resting membrane potential

In S9.1 you built a battery. A pump burned ATP to stack ions on the wrong side of a thin greasy membrane, and that separated charge sat there as stored voltage, a charged capacitor waiting for a path to open. A neuron is what happens when a cell puts that battery to work. It is an ordinary cell in most respects, same membrane, same pumps, same channels, but its shape and its channels are tuned for one job: take signals in at one end, integrate them, and send a signal out the other. This lesson gives the neuron its anatomy, then derives the one number every brain-computer interface is ultimately trying to read, the resting voltage of about -70 millivolts. We build that number from the rule you already have. We do not just assert it.

Four parts, one direction of flow

Strip away the biology vocabulary and a neuron resolves into a signal path with four stages, laid out in the order the signal travels.

Dendrites are the input structures. They are branches, often thousands of them, spreading out from the cell body like the crown of a bare tree. Their job is to receive signals arriving from other neurons and funnel them inward. A single neuron can collect input from thousands of other cells across this branching surface, so most of what a neuron "hears" lands on its dendrites.

The soma, or cell body, is the hub the dendrites feed into. It holds the nucleus and keeps the cell alive, and electrically it is where all those incoming signals pile up together. The soma is the integrator: it sums the nudges arriving from every dendrite into one running electrical total.

The axon is the single output cable. Usually a neuron has exactly one, a smooth fiber leaving the soma that carries the neuron's decision away from the cell body. Axons can be long. A motor neuron's axon runs from your spinal cord to a muscle in your foot, roughly a meter of cable on a cell you could not see without a microscope.

The axon terminals are the branched far end of the axon. This is where the axon reaches the next cell and hands the signal off, at a specialized junction called a synapse. What happens across that junction is the whole subject of BCI-2.1. For now, just fix the role: terminals are the output ports where one neuron passes its signal to the next.

So the logic reads cleanly in one direction: dendrites receive, the soma integrates, the axon transmits, the terminals hand off. Input, decide, output, forward. That one-way flow is not an accident of drawing. The axon fires in one direction and the synapse passes signal one way only, so a neuron is a directional device, not a two-way wire.

A pipeline, until you look closer

A programmer can hold the whole layout as a pipeline. The dendrites are the many input wires feeding in. The soma is an accumulator that sums those inputs into a single value. The axon is the output bus that carries the result onward. The terminals are the fan-out, wiring that result into the inputs of the next stage. Input ports, accumulator, output bus, fan-out. It is a clean mental model and it gets the direction of flow exactly right.

Now the failure edge, because an analogy without its limit is a bug, and this one breaks in a way that matters for everything a BCI does. A neuron is not a clean digital latch. The soma does not read its inputs on a clock edge, store a crisp bit, and hold it. It sums analog voltages that arrive continuously in real time, and it leaks. An input that is not reinforced drains away within a few milliseconds, so the accumulator is constantly forgetting. There is no defined moment when it samples. It integrates continuously and fires whenever the running total happens to cross a threshold, then resets to rest. And the sum is not plain addition. Some inputs push the total up (excitation) and some pull it down (inhibition), and where an input lands on the dendrites changes how much it counts. Think leaky analog integrator, not register. The pipeline picture gives you the shape. It badly undersells how analog, weighted, and lossy the middle really is, which is exactly why decoding what a neuron "meant" is hard rather than a matter of reading a bit.

The four ions, and the pumps that sort them

The anatomy runs on the electrical trick from S9.1, so zoom back to the membrane. Signaling in a neuron is carried by four ions, and the first thing to know about each is which side of the membrane it is piled on, because that stored imbalance is the fuel for everything.

Potassium (K+) is packed inside the cell and sparse outside, more than tenfold denser inside (roughly thirtyfold). This is the ion that sets the resting voltage, as you are about to derive.

Sodium (Na+) is the mirror image, piled up outside and sparse inside, roughly tenfold. This is the ion that floods in when a neuron fires.

Chloride (Cl-) is a negative ion, piled outside. It is mostly the currency of inhibition, of damping a neuron down and making it harder to fire.

Calcium (Ca2+) is kept almost absent inside the cell, a startlingly steep gradient of about ten-thousandfold. Because the inside baseline is nearly zero, even a tiny puff of calcium entering is a loud, unambiguous signal, which is why cells use calcium as a trigger. Among many jobs, a calcium influx at the terminals is what tells them to release their signal to the next cell.

Who builds and holds all four gradients? Pumps, burning ATP (S9.2), exactly as in S9.1. The workhorse is the sodium-potassium pump you already met: three Na+ out and two K+ in per ATP, cycling without pause. It builds the two big gradients, potassium in and sodium out, at the same time. Dedicated calcium pumps bail Ca2+ back out to keep the interior near zero, and chloride's level is set by still other transporters riding on those gradients. The headline is uniform: every gradient here is bought and maintained by a pump spending ATP against constant leak. Cut the ATP and, as the S9.1 poison-the-pump case showed, the gradients slowly bleed away.

Notice what we have and have not built yet. We have gradients, the deposits, the charged battery. We do not yet have a voltage. The voltage appears only when we let the membrane feel one of these gradients selectively. That is the next step.

Deriving the resting -70 millivolts

Here is the resting potential, built from scratch, using three ingredients already on the table.

First, the gradients exist, because the pump built them: potassium high inside, sodium high outside. Second, at rest the membrane is not equally leaky to every ion. It is studded with potassium leak channels, channels that sit open even at rest, so a resting membrane is far more permeable to K+ than to anything else and barely lets sodium through. Third, an open channel lets its ion flow passively downhill, down its gradient, spending no energy (S9.1).

Now run it forward. Potassium is crowded inside and the leak channels are open, so K+ does what any gradient does through an open path: it flows out, from crowded to sparse, down its concentration gradient. But every potassium ion carries a positive charge, and every one that leaves takes that positive charge with it, stranding an unbalanced negative charge behind inside the cell (large negatively charged proteins and other anions that cannot follow it out). So the inside of the cell starts going negative.

Here is the self-limiting heart of it. As the inside grows more negative, it begins to pull the positive potassium back in, because opposite charges attract. So potassium now feels two forces at once. There is a chemical push outward, since it is still crowded inside and wants to spread out. And there is a growing electrical pull inward, since the negative interior tugs the positive ion back. At first the chemical push wins and K+ keeps leaving, but every ion that leaves makes the inside more negative, which strengthens the inward electrical pull. At some voltage the inward electrical pull exactly cancels the outward chemical push. Net flow stops. That balance point is an equilibrium, and its voltage is called the equilibrium potential for potassium, the one voltage at which potassium's electrical and chemical forces perfectly cancel.

For potassium's roughly thirtyfold gradient, that balance lands around -90 millivolts, inside negative. If the membrane were perfectly and only permeable to potassium, a resting neuron would sit right there.

A real neuron rests a little higher, at about -70 millivolts, and the reason is honest and simple: the membrane is not perfectly potassium-selective. A small trickle of sodium leaks in. Sodium is piled outside and the inside is negative, so both forces drive sodium inward, and that small inward flow of positive charge nudges the voltage up from potassium's -90 toward zero, settling near -70. The resting potential is essentially the potassium equilibrium, pulled slightly positive by a minor sodium leak. Potassium wins the tug-of-war because the membrane is so much more permeable to it, so the resting voltage sits close to potassium's number and far from sodium's.

That is the resting membrane potential: about -70 mV, inside negative, a steady voltage that exists because selective potassium permeability lets the membrane settle near potassium's equilibrium. The capacitor from S9.1 is charged, and now it is idling at a defined -70.

A charged cell, idling

So a resting neuron is not doing nothing. It is holding about -70 millivolts across its membrane, a charged capacitor kept topped up by tireless pumps, its voltage pinned near potassium's equilibrium by selective potassium leak. Every part of the anatomy is wired into that voltage. The dendrites collect inputs that nudge the local voltage up or down, the soma sums those nudges into one running total, and the axon stands loaded and ready.

The entire point of holding a steady -70 is to have something to swing away from. When enough excitatory input pushes the voltage at the start of the axon past a threshold, voltage-gated sodium channels snap open, the small sodium leak becomes a flood, and the membrane voltage rockets from -70 up toward positive before other channels and the pumps reset it. That fast, all-or-nothing swing, traveling down the axon, is the action potential, the neuron's actual output signal. It is the battery you just charged, spent in one controlled discharge.

Key terms

dendrite

A branching input structure of a neuron that receives signals from other cells and funnels them toward the soma.

soma

The cell body of a neuron, holding the nucleus and acting as the integrator where incoming signals are summed.

axon

The single output cable of a neuron that carries its signal away from the soma, sometimes over long distances.

axon terminal

The branched end of an axon where it hands its signal off to the next cell at a synapse.

resting membrane potential

The steady voltage across a resting neuron's membrane, about -70 millivolts inside negative, set by selective potassium permeability acting on ion gradients.

equilibrium potential

For a given ion, the one membrane voltage at which its chemical push and electrical pull exactly cancel, so its net flow through open channels stops.

potassium leak channel

A channel that sits open at rest, making the membrane far more permeable to K+ than to other ions and pinning the resting voltage near potassium's equilibrium.

sodium-potassium pump

An ATP-burning pump that moves three Na+ out and two K+ in per cycle, building and maintaining the gradients that the resting potential later reads.

Resting is not free: the -70 costs ATP every second

It is tempting to read "resting potential" as the neuron idling at no cost. It is the opposite. The gradients leak constantly, potassium trickling out and sodium trickling in, so the sodium-potassium pump has to run without pause just to hold the line, and every cycle spends an ATP (S9.2). Holding a steady -70 is a continuous metabolic bill. This is a large part of why your brain, about two percent of your body weight, burns something like a fifth of your resting energy, and why neurons are so fragile the moment their fuel is cut. Stop delivering oxygen and glucose, ATP runs out within minutes, the pump stalls, the gradients bleed away, and the voltage collapses. The resting state is a battery held charged by a pump that is never allowed to rest. That metabolic reality also puts hard limits on brain-computer interfaces, which BCI-4.1 returns to: the living tissue you record from is expensive, delicate, and reacts to being probed.