The Action Potential

The resting membrane potential

Before a neuron can fire, it must first sit in a poised, ready state. A resting neuron is not electrically neutral: there is a steady voltage across its membrane, with the inside slightly negative compared with the outside. This resting membrane potential is typically around minus 70 millivolts, and it is the baseline from which every action potential begins.

This voltage arises from an uneven distribution of charged particles, or ions, across the membrane. Sodium ions are far more concentrated outside the cell, while potassium ions are more concentrated inside. Two factors set up and hold this arrangement. First, the sodium-potassium pump works continuously, pushing sodium out and drawing potassium in, at the cost of energy. Second, the resting membrane is much more permeable to potassium than to sodium, because potassium leak channels stand permanently open. Potassium therefore drifts outward down its concentration gradient, carrying positive charge with it and leaving the inside of the cell relatively negative, until the growing negativity pulls it back in as fast as the gradient pushes it out. The resting potential is where that balance settles, close to potassium's own equilibrium potential of about minus 90 millivolts, and short of it only because a little sodium leaks the other way.

It is worth pausing on this point: maintaining the resting potential costs energy, and a large share of the brain's fuel is spent simply keeping neurons ready to fire. The resting state is not passive but actively held, a loaded spring waiting to be released.

Why a threshold exists at all

A neuron does not fire in response to just any nudge. Small inputs shift the membrane voltage a little, but the voltage soon drifts back to rest. Only when the inputs together push the membrane up to a critical level, the threshold, usually around minus 55 millivolts, does the neuron fire a full impulse. That much is in every textbook. The interesting question is the one usually skipped: why should there be a critical level at all, and why is it so sharp? Nothing in the membrane is counting millivolts. The threshold is not a rule the neuron obeys; it emerges, and it emerges from a competition between two currents.

Near the resting voltage, two currents pull in opposite directions. The first is the potassium leak. Potassium is far more concentrated inside the cell than outside, and leak channels are permanently open, so potassium drifts steadily outward, carrying positive charge with it and pulling the voltage down. The second is the sodium current. Sodium is concentrated outside the cell, so any voltage-gated sodium channel that happens to open lets sodium in, carrying positive charge with it and pushing the voltage up.

These two currents are not symmetrical, and the asymmetry is the whole story. The potassium leak is steady and indifferent: it does not care what the voltage is doing. The sodium current is self-reinforcing. Voltage-gated sodium channels open when the membrane depolarises, so a little depolarisation opens some sodium channels, which admits some sodium, which depolarises the membrane a little further, which opens more sodium channels, and so on. Sodium entry causes the very condition that produces more sodium entry. It is a positive feedback loop, and positive feedback loops do not respond gently.

Now the two currents can be raced against each other. Below a certain voltage, the sodium feedback loop is too weak to keep up: for every ion of sodium the loop lets in, the potassium leak takes more than that back out, and the disturbance dies away. The neuron returns to rest. But the strength of the sodium loop grows steeply with voltage, while the leak does not. At some particular voltage the loop finally out-runs the leak. From that instant it feeds itself faster than the leak can drain it, and there is no longer anything to stop it. The membrane runs away upward, and it does so whether or not the stimulus that started it continues. That crossover voltage is the threshold.

The all-or-nothing principle follows immediately, and it needs no separate explanation. Once the loop has out-run the leak, the spike is unavoidable, so the neuron cannot fire a little. And the loop cannot run for ever: it stops when sodium's own entry stops driving the voltage further, which happens at a ceiling fixed by the sodium concentration gradient, not by the stimulus. So the neuron cannot fire a lot either. The size of the spike is set by the ion gradients, which are the same whatever provoked it, and a stimulus twice past threshold produces exactly the same spike as one barely past it.

The consequence for the nervous system is deep. A stronger stimulus does not produce a bigger spike; it makes the neuron fire more often. So the strength of a signal, the brightness of a light or the force of a touch, is carried by the rate of firing, the number of spikes per second, rather than by the size of any one spike. The brain gives up amplitude as a channel of information and gets in exchange a signal that cannot fade, blur, or be half-sent, which is exactly what an axon a metre long requires.

The two gates of the sodium channel

The feedback loop above explains how the spike starts. It does not explain how it stops, and left alone it would predict that the membrane, having run away upward, stays there. Something must shut the sodium channels while the membrane is still depolarised, which is exactly the condition that opens them. The resolution is a single, easily missed piece of structure, and it is the most confused point in the whole subject: the voltage-gated sodium channel has two gates, not one, and they respond to the same signal at very different speeds.

Both gates must be open for sodium to flow. At rest, the fast gate is shut and the slow gate is open, so no sodium passes. Depolarise the membrane and both gates begin to respond: the fast one flies open at once, the slow one starts to swing shut but takes its time. For a fraction of a millisecond, both are open simultaneously, and that brief window is when sodium floods in and the upstroke happens. Then the slow gate finishes closing, and the channel shuts itself even though the membrane is still depolarised. The upstroke ends because the channel that produced it has timed itself out.

Those three things are worth setting out together, because textbooks usually present them as unrelated facts about the impulse.

1. Why the rising phase stops

   The inactivation gate closes on its own schedule, cutting off the sodium inflow in mid-spike. Without it, the positive feedback loop would have nothing to end it.

2. Why there is an absolute refractory period

   Immediately after the spike the sodium channels are inactivated, not merely closed. No stimulus of any size can reopen them, because reopening requires repolarisation first, and repolarisation takes time. For that interval the neuron simply cannot fire, whatever is done to it.

3. Why the impulse travels one way

   Current from an active patch of membrane spreads in both directions, forwards into fresh membrane and backwards into the membrane the spike has just left. But the membrane behind is inactivated and cannot respond. Only the membrane ahead can. The impulse does not choose a direction: it is the only direction available to it.

One mechanism, three phenomena. This is why the two-gate structure is worth the effort of getting straight. Nearly everything peculiar about the nerve impulse, its brevity, its recovery time, its stubborn one-way travel, falls out of a single slow gate that shuts when the membrane goes up and can only be reset by bringing the membrane back down.

The phases of the action potential

Once threshold is crossed, the action potential unfolds as a fast, stereotyped sequence of ion movements. The whole event lasts only a millisecond or two. Following it phase by phase shows how the neuron's voltage swings from negative to positive and back again.

1. Resting state

   The membrane sits at about minus 70 millivolts. Voltage-gated sodium and potassium channels are closed, and the ion gradients built by the sodium-potassium pump stand ready. The neuron is poised but quiet.

2. Depolarisation, the upstroke

   When the voltage reaches threshold, the fast activation gates of the voltage-gated sodium channels snap open while the slow inactivation gates are still open. Sodium ions, driven by both concentration and charge, rush into the cell. Their positive charge drives the inside voltage upward fast, past zero and on toward roughly plus 40 millivolts. This rapid reversal is the sharp rising edge of the spike.

3. Repolarisation, the downstroke

   The rise is short-lived. The slow gates finish closing and the sodium channels inactivate, cutting off the inflow, while slower voltage-gated potassium channels now open fully. Potassium ions flow out of the cell, carrying positive charge with them, and the voltage falls back down toward its resting level.

4. After-hyperpolarisation, the undershoot

   Potassium channels are slow to close, so a little too much potassium leaves. For a brief moment the inside becomes even more negative than the resting value, dipping below minus 70 millivolts. This undershoot is called after-hyperpolarisation.

5. Return to rest

   The potassium channels finally close and the membrane settles back to its resting potential. As the voltage returns to negative values, the inactivation gates on the sodium channels are released and swing open again, and the neuron becomes able to fire once more.

Three definitions are worth fixing here, as they name the parts of the spike and recur throughout neuroscience.

Why the peak is where it is

The figure of roughly plus 40 millivolts looks arbitrary, and readers usually accept it as a measured fact with no explanation attached. It has one. Every ion has a voltage at which the electrical pull on it exactly cancels the chemical push of its concentration gradient, so that opening a channel for it produces no net flow. That voltage is its equilibrium potential, and it is set purely by how unequally the ion is distributed across the membrane. For sodium, which is roughly ten times more concentrated outside the cell than inside, the equilibrium potential is about plus 60 millivolts. The full derivation is on the page for ion channels and electrolytes, and the rule it yields is the one that makes everything on this page fall into place: an open channel drags the membrane voltage towards the equilibrium potential of whatever ion it passes.

So when the sodium channels open, the membrane is not driven to some arbitrary positive value. It is dragged towards plus 60 millivolts, sodium's own equilibrium potential, and that is the ceiling the runaway feedback loop is running at. It never quite arrives, for two reasons: the inactivation gates shut the sodium channels partway through the climb, and the delayed potassium channels open, dragging in the opposite direction, towards potassium's equilibrium potential of about minus 90 millivolts. The peak of the spike is where those two pulls happen to cross, which lands at around plus 40 millivolts rather than at plus 60. The same logic explains the undershoot: with the extra potassium channels still open at the end of the spike, the membrane is briefly pulled closer to potassium's minus 90 than the resting leak alone would take it.

The refractory period

Immediately after firing, a neuron enters a brief recovery window called the refractory period, during which it cannot readily fire again. It has two parts, and both follow directly from the two gates.

In the absolute refractory period, no new action potential is possible at all, at any stimulus strength. The reason is inactivation: the slow gates on the sodium channels are shut, and no amount of depolarisation will lift them, because depolarisation is the very thing that shut them. They are released only by repolarisation. Until the membrane has come back down, the cell's sodium channels are not shut-and-ready but shut-and-unavailable, and a neuron with no available sodium channels has no positive feedback loop to run.

In the following relative refractory period, firing is possible but harder, needing a stronger than usual stimulus. Here two things are working against the neuron at once: only some of the inactivation gates have reset, so the sodium loop is weaker than usual, and the delayed potassium channels are still open, so the leak pulling the voltage down is stronger than usual. Both effects move the crossover point, which is to say they raise the threshold. The neuron can fire, but the input has to be bigger to win the race.

The refractory period also sets an upper limit on how fast a neuron can fire. Because each spike must be followed by this recovery, there is a ceiling on the number of action potentials per second, which in turn bounds how much information a single neuron can carry through its firing rate. In practice this ceiling is high enough for most purposes, allowing hundreds of spikes each second in many cells, but it is a real limit, and it is set directly by how long the channels take to reset after each impulse.

Propagation along the axon

An action potential is not much use if it stays in one spot. Its purpose is to travel the length of the axon and reach the terminals, and the way it does so explains one of the most important features of nerve signalling: the impulse does not weaken with distance.

The key is that the action potential is regenerated at every step rather than simply passed along. When one patch of membrane depolarises, the inflow of sodium there spreads a little current to the neighbouring membrane, nudging it up to threshold. That neighbour then runs the same positive feedback loop and fires its own full-sized action potential, which in turn triggers the next patch, and so on down the axon. Each point rebuilds the signal to its full strength, so the spike that arrives at the far end of a metre-long axon is exactly as large as the one that set out.

Note that the current spreads both ways from the active patch, forwards and backwards alike; the axon does not know which way it is meant to be sending anything. The signal goes forwards only because the membrane behind has just fired and its sodium channels are inactivated, so it cannot respond to the current arriving at it. Direction is not built into the axon. It is a by-product of the fact that a patch of membrane needs time to recover.

This is the answer to why the signal does not fade. A passive electrical signal would die away over distance, as a voltage does along a wire. The action potential escapes this fate by actively remaking itself at every point, trading a little speed for perfect reliability over long distances. It is a striking design: the neuron gives up the simplicity of passing charge along passively, and in return gains a signal that arrives at the far terminal every bit as sharp and full as it began, no matter how long the journey.

Saltatory conduction and myelin

Regenerating the impulse at every single patch of membrane is reliable but slow. Many axons speed things up dramatically by wrapping themselves in myelin, a fatty insulating sheath laid down in segments along the fibre by glial cells. Between these segments lie short bare gaps, the nodes of Ranvier, about a millimetre apart, and it is essentially only at these nodes that the voltage-gated sodium channels are concentrated.

It is often said that the insulation lets the signal skip or jump from one node to the next. That is a mnemonic, not a mechanism, and it puts the conclusion where the cause should be. Nothing jumps. The impulse still has to get from node to node, and it does so through the axon in the ordinary way, by current spreading along a cable. What myelin changes is not the route but the quality of the cable, and it does so in two ways at once.

Put those together. Less charge is spent charging the membrane and less leaks away through it, so the current entering the axon at an active node spreads down the core further, and it spreads there faster. It arrives at the next node, a millimetre away, still strong enough to push that node past threshold. The node fires a full-sized action potential, which regenerates the current, and the process repeats.

So the action potential is regenerated only at the nodes, and for a plain reason: the nodes are where the voltage-gated sodium channels are. The internodes have almost none, and could not fire an action potential if they wanted to. They are crossed passively, but very fast. Because fresh spikes appear at the nodes and nowhere in between, the impulse appears to leap along the fibre, and the process is named saltatory conduction, from the Latin for leaping. Nothing has actually leapt. The regeneration is discontinuous while the current spread is not.

The gain in speed is two orders of magnitude. A myelinated fibre conducts at up to about 100 metres per second, while an unmyelinated fibre of similar diameter manages roughly one. But speed is only half of what myelin buys, and the other half is easy to overlook.

The mechanism also explains why losing myelin is so much worse than never having had it. In a demyelinating disease such as multiple sclerosis, the sheath around central axons is stripped away, and what is left is not an ordinary bare fibre. It is a fibre whose sodium channels sit only at nodes a millimetre apart, with long stretches of naked membrane in between that have almost no channels at all. The current now leaks out of those bare internodes as it crosses them, arrives at the next node too weak to reach threshold, and the signal simply dies. An unmyelinated axon at least has channels everywhere and can rebuild the spike continuously. A demyelinated one has the worst of both designs: the sparse channel distribution of a fast fibre and the leaky cable of a slow one. Conduction slows, becomes unreliable, or blocks outright, which is why the disease disturbs movement, sensation, vision, and coordination.

The action potential by the numbers

These values are typical figures rather than fixed constants; the exact numbers vary between neurons and conditions, but they capture the scale of the event. The quantitative account of how sodium and potassium currents produce these voltages was worked out by Alan Hodgkin and Andrew Huxley in the early 1950s, in experiments on the giant axon of the squid. Their achievement was not to observe the spike, which was already known, but to derive it: they measured how the sodium and potassium conductances depend on voltage and on time, wrote those dependencies into equations, and found that the equations produced an action potential of the right size, shape, and speed with nothing else added. The threshold, the all-or-nothing behaviour, and the refractory period all emerged from the model rather than being put into it, which is the strongest possible demonstration that they are consequences of the channels and not separate rules. That model remains a foundation of neuroscience and earned Hodgkin and Huxley a share of the Nobel Prize.

Related topics