The Neuron

What a neuron is

A neuron is a cell, but a highly specialised one. Like other cells it has a membrane, a nucleus, and the usual internal machinery. What sets it apart is that it is electrically excitable: it can generate and conduct electrical signals, and it is shaped to pass those signals on with precision. A neuron is less a general-purpose building block than a wire that can also compute, deciding, moment to moment, whether to fire.

A careful count puts the figure at roughly 86 billion neurons in the adult human brain, each connecting to hundreds or thousands of others. It is not the number of cells that gives the brain its power but the density and pattern of their connections: the wiring, far more than the parts list, is where its capabilities live.

The neuron doctrine, and why it was a fight

That the brain is made of separate cells now seems too obvious to need arguing. It was not: for the closing decades of the nineteenth century the most eminent anatomists in Europe were divided over exactly this question, because the two answers implied entirely different kinds of brain. The obstacle was that nervous tissue stains as an undifferentiated grey smear. The breakthrough was Camillo Golgi's silver impregnation method, which blackens only a small random fraction of the cells in a preparation, but blackens each of those completely. Golgi concluded from his own stain that the axons fused into one continuous network, a reticulum, through which excitation flowed like current through a mesh. This was the reticular theory, and it was the orthodoxy.

Santiago Ramon y Cajal took up Golgi's own stain and turned it against him. Everywhere he looked, the branches of one cell approached the branches of another and stopped; he could find no fusion. The nervous system, he argued, is built from discrete cells, contiguous but not continuous, so the signal must be handed from one to the next across a gap. This is the neuron doctrine, and from it Cajal drew a second claim of equal weight, the law of dynamic polarisation: within each cell, information flows one way, from dendrites through the soma and out along the axon.

Cajal was right, but he could not prove it: the gap he inferred lay far below the resolution of any light microscope. The dispute was settled not by argument but by instrumentation, when electron microscopy was turned on nervous tissue in the 1950s and the junction between two neurons was imaged directly, two membranes separated by a narrow, uniform cleft. The doctrine matters because it fixes what kind of thing the brain is. A continuous net has no natural units and no place where a signal can be selectively strengthened or weakened, so it offers no substrate for learning. Discrete cells joined at modifiable gaps offer both, and the synapse and neuroplasticity both rest on Cajal's answer.

The structure of a neuron

Although neurons vary widely in shape, most share one basic plan: a central cell body with branching inputs on one side and a single long output fibre on the other, information flowing one way, from dendrites through the soma and out along the axon.

At the far end, the axon divides into fine branches ending in swellings called axon terminals, each coming close to, but never touching, a target cell. Between the soma and the axon proper lies a short cone, the axon hillock, which turns out to be the most consequential few micrometres in the cell.

Dendrites: not antennae, computers

The textbook diagram invites a misreading: dendrites drawn as plain lines converging on the soma, described as receivers, and so understood as passive antennae. They are not. On many excitatory neurons the dendrites are studded with thousands of tiny protrusions called dendritic spines, each a stalk with a bulbous head bearing a single excitatory synapse. The narrow neck partly isolates the head, so the chemical events at one synapse, in particular a rise in calcium, stay local rather than flooding the cell. That compartmentalisation is why a neuron can strengthen one of its thousands of inputs without strengthening the rest, which any plausible learning rule requires. Spines are also mobile, growing and shrinking over hours and days, with head size tracking synaptic strength: a dendrite is not fixed wiring but a surface under continuous revision, which is why neuroplasticity is a matter of anatomy as much as chemistry.

The deeper point concerns the membrane. Dendrites were long modelled as passive cables along which a signal spreads and decays. But dendritic membranes also carry voltage-gated channels, so a branch can regenerate a signal rather than merely conduct it: a cluster of inputs arriving together on one branch can trigger a local dendritic spike, delivering a far larger signal to the soma than the same inputs scattered across the tree would have produced. Where an input lands therefore matters, not merely how strong it is, and the tree performs something closer to a two-layer computation, each major branch summing its own inputs before the results combine at the soma. A neuron is not a summing device with a threshold bolted on the end but a small network in its own right, and anyone estimating the brain's computational depth from a count of its cells is working with a floor, not a figure.

The axon hillock: where the decision is made

A cortical neuron may carry something on the order of ten thousand synaptic contacts, excitatory and inhibitory, arriving continuously from many sources at different places on the tree and at different moments. Out of all that the cell must produce a single answer, repeatedly and in real time: fire, or do not fire. Its inputs are combined in two ways at once.

Inhibition is not merely the arithmetic negative of excitation, and its placement is telling: inhibitory synapses are frequently sited on the soma and the hillock itself, close to the point of decision, where they can veto the summed result of a whole tree of excitation arriving from further out.

The tally is read at the axon hillock, and what makes this small cone the decision point is a fact of molecular anatomy: it carries the highest density of voltage-gated sodium channels in the cell. Threshold is a property of that density, so the region with the most channels reaches threshold first, and the impulse is initiated there and nowhere else. Everything the dendrites have done is compressed into one binary comparison: reach threshold and the cell fires a full impulse, fall short and it produces nothing. This is the all-or-nothing principle, and its consequence is that a neuron driven harder does not shout, it repeats itself faster. Signal strength is carried in the rate of firing, not the size of any one spike. The ionic machinery of the spike is set out in full on the page for the action potential.

Types of neuron

Neurons are classified by what they do and by their shape, and both matter, because a neuron's job and its form are closely linked.

By function

That split is quietly misleading, because it presents the three classes as though they were comparable in size. They are nothing of the kind. The overwhelming majority of neurons in the human brain are interneurons, with no direct contact with the world at either end. The brain is not principally a device for converting stimulus into response. It is a device for talking to itself, and any account of cognition that treats it as an elaborate reflex arc has misjudged the proportions.

By structure

By shape, neurons are grouped by how many projections leave the cell body. Multipolar neurons, one axon and many dendrites, are the commonest in the brain. Bipolar neurons, one axon and one dendrite from opposite poles of the soma, occur in some sensory systems, notably the retina. Unipolar or pseudounipolar neurons have a single projection that splits, one branch running to the periphery and one into the spinal cord: the form taken by the sensory neurons carrying touch, temperature, and pain from the body.

Three cells worth knowing by name

How a neuron signals

The whole sequence, from arriving input to outgoing message, uses two different kinds of signalling in turn.

1. Integration and threshold

   Excitatory and inhibitory inputs are summed across the dendrites and soma, and the result is tested against threshold at the axon hillock.

2. The electrical impulse

   If threshold is reached, the cell fires an action potential, a brief, self-propagating spike of voltage that races down the axon without weakening.

3. Passing the message on

   At the axon terminals the impulse triggers release of neurotransmitters across the synapse, converting the electrical signal into a chemical one.

This two-part design, electrical along the axon, chemical across the synapse, is fundamental. The electrical impulse is fast and reliable over distance; the chemical step is slower but far more flexible, letting the brain adjust the strength of each connection. That flexibility is the physical basis of learning.

Myelin and the economics of speed

Not all axons carry signals at the same speed, and the difference is largely down to myelin, which the neuron does not make itself. It is built by glial cells: in the brain and spinal cord by oligodendrocytes, each of which can wrap segments of many axons at once, and in the nerves of the body by Schwann cells, each wrapping a single segment of one axon. The peripheral arrangement is less economical, but it leaves those nerves far better able to repair themselves after injury.

In both cases the sheath is discontinuous, laid down in segments separated by short bare gaps, the nodes of Ranvier, and it is essentially only at the nodes that voltage-gated sodium channels are concentrated. The insulated stretches cannot regenerate the impulse; they carry it passively, fast and with little leakage, so the depolarisation at one node arrives at the next still strong enough to push it past threshold. The impulse is therefore rebuilt only at the nodes and appears to leap between them: saltatory conduction, from the Latin for jumping. The gain is two orders of magnitude. An unmyelinated fibre conducts at roughly one metre per second; the fastest myelinated fibres reach up to about 100.

The importance of all this is clearest when it fails. In demyelinating conditions such as multiple sclerosis the sheath around central axons is damaged, and the consequence follows from the mechanism above: an axon whose channels sit only at widely spaced nodes, stripped of its insulation, is worse off than an unmyelinated fibre, because the bare stretches between nodes have no channels with which to regenerate anything. The signal leaks away and may never reach the next node. Conduction slows, becomes unreliable, or blocks outright, which is why the disease disturbs movement, sensation, vision, and coordination. Myelin is not an optional refinement but a condition of the system working.

The cost of being excitable

The brain is a small fraction of adult body weight and yet consumes a strikingly large share of the body's energy at rest. What is all that fuel spent on? Not thinking, at least not in any sense that would satisfy the asker, but on maintaining ion gradients.

A neuron signals by letting ions move down their concentration gradients: sodium falls into the cell, potassium leaves it. Every impulse spends a little of a stored gradient, and the gradient is the battery. To fire again the cell must put the ions back, which is thermodynamically uphill and must be paid for. The payment is made by the sodium-potassium pump, a membrane enzyme that consumes ATP to expel sodium and import potassium, and it runs continuously, in every neuron, for as long as the cell is alive. It is the brain's single largest consumer of energy. The resting potential is not a resting state at all; it is a state actively held against a constant leak, at a constant price. See ion channels and electrolytes and brain energy metabolism.

The consequence is severe. Because neurons store almost no fuel and the pump cannot stop, the brain has essentially no reserve. Interrupt the blood supply and ATP fails within a minute or two: the pump falters, the gradients collapse, the membrane depolarises uncontrollably, and calcium floods in to trigger the cascades that kill the cell. This is why ischaemic stroke destroys tissue in minutes while other organs tolerate the same insult far longer. The vulnerability is not incidental to what a neuron is. It is the price of being electrically excitable, and it is also why so much of the nervous system's architecture, myelin included, is quietly organised around saving energy.

Neurons by the numbers

These figures are approximate, drawn from modern counts rather than older estimates. The 86 billion figure comes from the isotropic fractionator, which dissolves the brain into a uniform suspension of free nuclei and counts them directly.

Myths worth correcting

Three claims about the neuron circulate widely and are wrong. Each, if believed, distorts the picture built above.

Related topics