The Motor System

The problem an intention has to solve

The standard shorthand for the motor system is a chain of command: the brain decides, sends a signal down the spinal cord, and the muscle obeys. It is a comfortable description because it matches the experience of moving, which is that you want the cup and then you have the cup, with nothing apparently in between.

The shorthand is wrong in a specific and instructive way. It treats the gap between the intention and the contraction as empty, as though the difficult part were the deciding and the rest were delivery. In fact almost the entire difficulty is in that gap, and the Russian physiologist Nikolai Bernstein was the first to state clearly why.

And the choosing is only the first difficulty. Four more sit behind it.

Muscles are not motors. A muscle can only pull, never push, so every joint requires opposing muscles, and their forces must be balanced against each other in every instant. A muscle's force also depends on its current length and on the speed at which it is already shortening, so the same neural command produces a different force depending on where the limb already is.

Loads are unknown. Lifting a cup requires a different force depending on whether it is full or empty, and the brain does not know which until it has begun to lift. Everyone has experienced the violent overshoot of picking up a box expected to be heavy that turns out to be empty. That overshoot is direct evidence that the command was issued on the basis of a prediction about the load, and that the prediction was wrong.

Feedback is slow. Sensory signals from the arm take tens of milliseconds to reach the brain, and the return command takes tens more. A fast reach is over in a few hundred milliseconds. A control system that waited for feedback before correcting would always be correcting an error that had already changed. Fast movement cannot be run on feedback alone, and this fact will turn out to determine the architecture.

The body changes. Limbs grow, muscles fatigue and strengthen, a heavy coat alters the arm's inertia. Any fixed mapping from command to consequence goes stale, so the system must continuously recalibrate itself.

Everything that follows on this page, the hierarchy, the tract, the reflexes, the two subcortical modulators, is machinery for solving those five problems. Read the anatomy as a set of answers and it stops being a list of names.

A hierarchy, and a division of labour

The motor system is organised as a hierarchy, and the useful way to describe it is not by listing the levels but by asking what question each level answers. Each answers a different one, and each hands a more specified problem downwards.

1. Whether, and what: the prefrontal and association cortex

   Before any movement is planned, something must decide that a movement is warranted at all, and which goal it serves. That is the business of the prefrontal cortex and the parietal association areas: weighing the goal against competing goals, holding the intention in mind, and suppressing the action if the situation changes. See executive function, and, for the anatomy, the frontal lobe. Nothing here is yet a movement. It is a decision to have one.

2. How, and in what order: the premotor and supplementary motor areas

   Immediately in front of the primary motor cortex lie two planning regions. The supplementary motor area (SMA), on the medial surface, is engaged most strongly by movements generated internally: sequences produced from memory, from an intention, without an external prompt. The premotor cortex, more lateral, is engaged most strongly by movements guided by external cues: reaching towards a seen object, shaping the hand to the shape of a handle. Both regions are active before movement begins, and both encode the movement in terms of its goal and its sequence rather than its muscles.

3. Now: the primary motor cortex

   M1, on the precentral gyrus, is where the command is issued. Its output cells project down the corticospinal tract to the cord. If the planning areas answer "what and in what order", M1 answers "go, in this direction, with this force, now".

4. Do it: the spinal cord and the motor neuron

   The command arrives at the spinal cord, where it meets the motor neurons, the local reflex circuitry, and the pattern generators. The cord is not merely the last wire. It contributes a great deal of its own, as a later section will argue. But it is where the neural signal finally becomes a contraction.

The evidence for this division is not merely anatomical. The planning stages are visible physiologically. Recordings from the SMA and premotor cortex show activity building well before movement onset, and scalp recordings show a slow negative potential, the readiness potential, that develops over hundreds of milliseconds before a self-initiated movement begins. There is a preparation phase, and it is measurable. And it can be dissociated clinically: damage to the SMA can leave a person able to move perfectly well in response to an external cue while impairing the ability to initiate the same movement spontaneously, which is precisely the dissociation the internal versus external distinction predicts.

The final common path

All of that structure narrows, at the end, to a single cell type.

The alpha motor neuron sits in the ventral horn of the spinal cord, or in a motor nucleus of the brainstem for the muscles of the head. Its axon leaves the cord, travels out in a peripheral nerve, and ends on muscle fibres. One alpha motor neuron together with all the muscle fibres it innervates is called a motor unit, and it is the smallest quantum of movement the nervous system can produce: fire the neuron and every fibre in its unit contracts, all of them, all at once. There is no partial firing of a motor unit.

Now consider everything that converges on this one cell. The corticospinal tract from M1. Descending pathways from the brainstem controlling posture and balance. Sensory afferents from the muscle spindle, arriving directly. Inhibitory interneurons carrying the influence of antagonist muscles. Propriospinal interneurons coordinating across segments. The cerebellum and the basal ganglia, by way of the cortex. Every one of these influences, from the loftiest prefrontal intention to the crudest stretch reflex, ends up as excitation or inhibition arriving at the alpha motor neuron, and the neuron adds them all up and either fires or does not.

Charles Sherrington called it the final common path, and the phrase is exact. It is the point at which every neural conversation about movement is resolved into a single answer, and there is no route to a muscle that avoids it.

The gradation of force follows from the same design. A muscle cannot be driven harder by firing its motor neurons harder in any straightforward sense; instead, force is graded in two ways. Recruitment: additional motor units are brought in as more force is required, and they are recruited in a reliable order, small units first and large ones last, so that fine control is available at low forces and raw power is added only when needed. Rate coding: individual motor neurons fire faster, so that successive twitches summate. Between them, these two mechanisms let the same muscle thread a needle and lift a suitcase.

M1: the map that is not really a map

The primary motor cortex occupies the precentral gyrus, immediately in front of the central sulcus and directly across it from the somatosensory strip described on touch and pain. Stimulate a point on it and a movement occurs on the opposite side of the body, and the point stimulated determines which part moves. Move systematically along the strip and the body is traversed in order. Penfield and Rasmussen mapped this in conscious neurosurgical patients and published the result in 1950, and the cartoon figure that came out of it, the motor homunculus, with its swollen hand and mouth and its stunted trunk, is now one of the most reproduced images in all of biology.

The distortion means the same thing it means on the sensory side: cortical territory is allocated by the fineness of control required, not by the size of the body part. The hand and the muscles of speech get an enormous territory because they need exquisitely graded control. The trunk gets very little because it does not.

So far so good. But the homunculus is also, in an important respect, misleading, and a reference that teaches it without saying so is teaching a superseded model as though it were current.

This matters because of the argument at the top of this page. If M1 specified muscles, it would have solved the degrees-of-freedom problem somewhere upstream and would simply be reading out the answer. If instead M1 specifies movements, directions, and goals, then M1 is still working in a language above the level of the muscles, and the final translation into a pattern of muscle activation is being done further down, in the spinal circuitry, where the interneurons and the reflex loops that know about limb position and load are already in place. The cortex, on this view, does not micromanage. It states a movement and lets a lower system, which has better local information, work out the muscle pattern that produces it.

The uncertainty here is genuine and should be named. Exactly what M1 encodes remains contested. Direction, force, muscle activity, joint torque, and end-point position have all been proposed as the variable that best describes M1 firing, and the honest position is that different experiments, using different tasks, favour different answers, and that M1 may well encode a mixture that resists a single clean label. What is no longer defended is the simple homunculus reading.

The corticospinal tract, and where it crosses

The command leaves M1 down the corticospinal tract, the great descending motor pathway, and it is worth following the route in full because its geography is the whole of its clinical significance.

1. Out of the cortex, into the internal capsule

   Axons from M1, along with contributions from the premotor areas and the somatosensory cortex, converge and funnel into the internal capsule, the dense sheet of white matter that threads between the thalamus and the basal ganglia. This is a bottleneck: the fibres for the entire opposite half of the body are compressed here into a narrow band, which is why a small lacunar stroke in the internal capsule can produce a dense hemiplegia out of all proportion to its size.

2. Down through the midbrain

   The tract descends into the cerebral peduncle on the front of the midbrain, still uncrossed, still on the same side as the cortex it came from.

3. Through the pons and into the medullary pyramids

   It breaks into bundles through the pons, then re-gathers on the front of the medulla as a pair of visible ridges, the pyramids, which give the tract its alternative name, the pyramidal tract.

4. The decussation

   At the bottom of the medulla, roughly 85 per cent of the fibres cross the midline: the pyramidal decussation. Having crossed, they descend in the lateral column of the spinal cord as the lateral corticospinal tract, and it is these fibres that drive the fine, fractionated movements of the limbs, and above all of the hand. The remaining minority descend uncrossed as the anterior corticospinal tract and mostly serve axial and proximal muscles, crossing at the segmental level or influencing both sides, which is why trunk control is comparatively resistant to a one-sided lesion.

5. Onto the motor neuron

   In the ventral horn, corticospinal fibres end on interneurons and, in humans and other primates especially, directly on the alpha motor neurons themselves. That direct cortical-to-motor-neuron connection is unusually well developed in primates, and it correlates with the capacity for independent finger movements. A monkey can pick up a raisin with a precision grip; the corticomotoneuronal connection is a large part of why.

That last point is worth taking one step further, because it connects to the most useful localising rule in neurology. Cranial nerves leave the brainstem on their own side and never cross. The corticospinal tract crosses at the bottom of the medulla. So a lesion in the brainstem, above the decussation but at the level of a cranial nerve nucleus, will produce a cranial nerve sign on the same side and limb weakness on the opposite side. That combination is called crossed findings, it is close to unique to the brainstem, and the cranial nerve involved names the level. It is derived in full on the brainstem and cerebellum page, and it is a direct corollary of the tract described here.

The corticospinal tract is the most famous descending pathway but not the only one. Several brainstem pathways run in parallel: the reticulospinal tracts, which set postural tone and are heavily involved in gross and automatic movements; the vestibulospinal tracts, which maintain balance against gravity using signals from the inner ear; the tectospinal tract, which orients the head towards a stimulus. These are older, they are largely bilateral, and they matter more than they are usually credited for. Their existence is why some motor function can survive extensive corticospinal damage, and it is also, as the next section will show, why an upper motor neuron lesion produces the strange signs that it does.

Upper and lower: why the signs are opposite

Neurology students are given a table with two columns, one headed "upper motor neuron" and one headed "lower motor neuron", and they memorise it. This is a waste, because the table can be derived from a single question, and once derived it cannot be forgotten. The question is: what has the lesion removed?

The clinical payoff is that this distinction localises. Weakness with brisk reflexes and an upgoing toe means the problem is in the brain or the cord, and further signs then narrow it further. Weakness with dead reflexes, wasting, and fasciculations means the problem is in the motor neuron, the root, the plexus, or the nerve. And a patient who shows both patterns at once, spasticity in some limbs and wasting with fasciculations in others, presents a combination that is almost impossible unless the disease is attacking upper and lower motor neurons simultaneously. That is amyotrophic lateral sclerosis, and the reason the mixed picture is so diagnostically powerful is precisely that the two syndromes are otherwise so cleanly separable.

The two great modulators

The corticospinal tract issues the command, but it does not do so unadvised. Two large subcortical systems shape motor output, and neither of them projects to the spinal cord in any substantial way. Both work by influencing the cortex, through loops that pass up through the thalamus and back down again. Their jobs are different, and stating the difference crisply is the most useful thing this page can do about them, because both are treated at length elsewhere in this library and repeating that treatment here would help nobody.

Set the two contrasts side by side and the division of labour becomes sharp.

The fact that both systems influence movement through the cortex, rather than by sending their own fibres to the spinal cord, is worth pausing on, because it is not obvious and it is easy to draw the diagram wrong. Basal ganglia output goes to the ventral anterior and ventral lateral thalamic nuclei, and those project to the motor and premotor cortex. Cerebellar output, from the deep cerebellar nuclei, goes to the ventral lateral thalamus, and that projects to motor cortex too. Both systems therefore bias the cortex's command before it is issued rather than editing it afterwards, and this is why lesions of either produce disordered movement rather than paralysis. Neither is on the final common path. Neither can prevent a muscle from contracting. They can only change what the cortex asks for.

Together with the corticospinal tract, they constitute the three-way partition that this library uses throughout: the cortex plans and commands, the basal ganglia select and gate, the cerebellum predicts and corrects. Three systems, three failure modes, three quite different clinical pictures: paralysis, akinesia or chorea, and ataxia.

The cord is a controller, not a cable

The last stage of the hierarchy is the one most often traduced. In the standard shorthand the spinal cord is wiring: a cable carrying commands down and sensations up. It is nothing of the kind. It is a controller with a substantial repertoire of its own, and the motor system's design makes no sense unless you understand what the cord contributes.

The stretch reflex. Within a muscle lie muscle spindles: encapsulated receptors, wrapped around specialised muscle fibres, that report how long the muscle is and how fast that length is changing. When a muscle is stretched, its spindles fire. Their sensory axons run into the spinal cord and synapse directly onto the alpha motor neurons of that same muscle, with no interneuron in between. Those motor neurons fire, the muscle contracts, and the stretch is opposed.

That is a two-neuron arc, sensory in and motor out, and it is the only reflex in the body known to be genuinely monosynaptic. It is what a tendon hammer tests: the tap stretches the tendon, the stretch fires the spindles, the spindles fire the motor neurons, the muscle jerks. And it is a feedback controller in the engineering sense, a device that opposes deviation from a set length, which is exactly what is needed to hold a posture against an unpredictable load. Stand holding a tray and let someone drop a book onto it: your arm dips and then, without any decision on your part and faster than any decision could be made, recovers. The spindles detected the stretch and the cord answered it, all within the cord.

The Golgi tendon organ. At the junction between muscle and tendon lies a second receptor, which reports tension rather than length. It is wired, through an inhibitory interneuron, to inhibit its own muscle. It is a protective circuit and also a force regulator: as tension rises, inhibition rises with it, damping the output. Between the two receptors, the cord has a length sensor with a positive feedback loop and a tension sensor with a negative one, which is enough to stabilise both posture and force.

Reciprocal inhibition. The same spindle afferent that excites its own muscle also, by way of an inhibitory interneuron, inhibits the antagonist. Without this, every contraction would be fought by its opponent. The cord ensures that when the biceps is told to contract, the triceps is simultaneously told to relax, and it does so locally, without troubling the brain.

Central pattern generators. The strongest evidence that the cord is a controller is that it can generate a rhythm without being given one. Networks in the spinal cord can produce alternating, rhythmic bursts of activity, appropriate to the step cycle of walking, in the absence of any rhythmic descending input and in the absence of sensory feedback. These are central pattern generators. The brain does not have to specify each step. It supplies a start signal, a speed, and a direction, and the cord supplies the rhythm, with sensory feedback shaping the pattern to the actual terrain.

The picture that emerges is of a brain that does not so much drive the body as configure it. The cord already knows how to hold a posture, how to oppose a stretch, how to relax an antagonist, how to alternate legs. The descending command does not spell these things out. It selects among them, biases them, and turns them on and off. This is the last and cleanest answer to Bernstein's problem: the degrees of freedom are not all resolved by the brain, because a great many of them were already collapsed, long before the command arrived, by the reflex architecture waiting in the cord.

Why movement is a loop, not a command

One thread has been left hanging, and it is the one that ties the page to its opening. Feedback is too slow to control a fast movement, and yet fast movements are controlled, and controlled well. How?

Not by feedback alone. The motor system runs on prediction. It issues a command, and at the same moment it predicts what the sensory consequences of that command ought to be, using the forward model maintained by the cerebellum. The prediction is available immediately, because it is computed rather than measured, and the system can therefore begin correcting on the basis of the predicted state of the limb long before the real state has been reported. Feedback, when it arrives, is used not to steer in real time but to check the prediction and, where it was wrong, to improve it. Feedback trains the model; the model drives the movement.

This is why a movement is properly described as a loop rather than a command, and it is why sensory information is not an optional extra to the motor system but a constituent part of it. That claim can be tested, and the test is stark. There exists a small clinical literature on people who have lost large-fibre proprioceptive input, through rare sensory neuropathies, while retaining normal motor nerves and normal muscle strength. The muscles work. The motor cortex works. The corticospinal tract works. And movement is nevertheless devastated.

Such individuals, in the qualitative descriptions given of them, cannot maintain a posture without watching the limb; the arm drifts away without their knowledge. Movements without visual guidance become wild and unregulated. Walking, standing, and even sitting upright have to be learned again as consciously guided visual tasks, with attention on the body at every moment, and the whole edifice collapses the instant the lights go out. What has been lost is not strength, and not the ability to command. What has been lost is the information the loop was closing on, and without it the loop is open, and an open motor loop cannot control a limb.

Which returns the page to its beginning. The intention to pick up a cup does not become a movement by being transmitted. It becomes a movement by being progressively specified, gated, predicted, executed, measured, corrected, and, over many repetitions, learned. Bernstein's infinity of possible arm trajectories is collapsed to one, not by a single act of choosing, but by a hierarchy in which every level throws away some of the freedom, and by loops that keep checking whether the freedom that was thrown away was the right freedom to lose.

Common misconceptions

Related topics