16 Unit [N] — Movement: The Control System in Action
Unit Overview
16.1 From need to action
Every drive we have met so far ends in the same place: a muscle.
Hunger, thirst, the slow drift of body temperature away from its set point, the pull toward a mate — each is a signal that the body’s internal state has wandered, or is about to wander, from where it needs to be. We have spent considerable time on how the brain detects these departures and on the structures, centered on the hypothalamus, that register them as needs. But detection, on its own, defends nothing. The brain can register a falling core temperature or a rising demand for water and still die of either, because a signal changes nothing in the world. To correct the error — to move into the sun, to walk to the stream, to bring the fruit to the mouth — the organism has to act. And to act is to move.
This is why the motor system is not one chapter of the brain among many. It is the output stage of the entire control system we have been building since Unit I. Everything upstream — the drives, the senses, the memories — exists in order to shape what the muscles finally do. The senses tell the organism what the world is like now; memory tells it what the world was like before (where the water was yesterday, which plant made it sick last week); the drives tell it what it needs. Movement is the single channel through which all of that becomes a consequence. An animal that could perceive perfectly and remember perfectly but could not move would be, from evolution’s point of view, already dead.
Seen this way, the boundary between “motor” and “non-motor” parts of the brain begins to dissolve. If we count planning, selecting, learning, inhibiting, and correcting movements as part of the motor system — and we should, because none of them matter except as they bear on action — then a remarkable fraction of the brain turns out to be in the business of movement. We will not try to police a hard border. We will instead start at the muscle and work upward, watching control get richer at each level, and asking at every step what new problem each layer was added to solve.
16.2 A warning about the little man
Before we climb, a caution, in the spirit of the warning about metaphors that opened this book.
Ask most people how voluntary movement works and you will get some version of the following picture. Somewhere in the brain there is a decider — a command center — that settles on a movement and then issues detailed instructions down to the muscles, contracting this one, relaxing that one, like a puppeteer working strings or an operator at a control panel. The picture even seems to be confirmed by anatomy: the strip of cortex most associated with movement carries an orderly map of the body, the famous motor homunculus — literally a “little man” drawn across the cortical surface, toes at the top, face and tongue at the bottom. There he is, it seems: the small person in the head who runs the body.
This picture is the motor system’s version of the mistake we diagnosed in Unit I, where a model borrowed from digital computers tempted us to imagine a central processor issuing commands to passive hardware. It is wrong, and most of this unit is devoted to showing how it is wrong, in two deep and separate ways.
First, a great deal of coordinated, adaptive movement requires no commander at all. As we are about to see, circuits with no brain attached — and, further back still, single cells with no neurons attached — already produce purposeful, well-organized movement. The commander is not doing the work we imagine.
Second, where there genuinely is control from the cortex, it does not take the form of muscle-by-muscle instructions. The cortical map is real, but it is a map of movements, not of muscles, and what looks like a “command” is better understood, as we will argue, as the unfolding of a dynamical system than as a script read out to the periphery. The homunculus is a useful picture of where body parts are represented; it is a misleading picture of how movement is produced.
The little man also has a logical problem worth keeping in mind. If a small person in your head controls your movements, what controls his movements? Presumably a still smaller person inside him — and so on, without end. Any explanation that places a miniature decider at the center has not explained decision or control; it has only hidden them inside a smaller box. Our task is to open the boxes, and we will find, again and again, not a commander but an architecture.
16.3 Movement is older than brains
To see how movement can happen without anyone in charge, it helps to go back further than the brain — further, in fact, than the neuron.
Consider a single bacterium swimming up a gradient toward food. It has no nervous system whatever, yet its behavior is unmistakably organized around its needs. It rotates a corkscrew-shaped flagellum, driven by a tiny rotary motor in its membrane, and by controlling that motor it travels toward higher concentrations of nutrients and away from toxins. It cannot, of course, compute a route. What it does is simpler and rather beautiful: when conditions are improving — when the food is getting stronger from one moment to the next — it keeps swimming straight a little longer; when conditions worsen, it tumbles and reorients at random. Bias the random walk by recent experience and the cell drifts, reliably, toward what it needs. A paramecium does something comparable with the thousands of tiny hairs (cilia) that cover it, beating them to glide toward food and reversing their stroke to back away when it bumps an obstacle.
Pause on what this means. Here is behavior in the full sense — movement marshaled in the service of staying alive — with no neurons, no muscles, and certainly no homunculus. The deepest lesson of this unit, and arguably of the book, is already visible in a creature too small to see: behavior organized around survival is the ground state of life. The nervous system is not the origin of purposeful movement. It is a later, spectacular device for doing the same ancient job — defending the body’s needs by acting on the world — faster, farther, and in more complicated circumstances.
When nervous systems do appear, one of their first tricks is to package useful movements so they run almost by themselves. The key invention is the central pattern generator, or CPG: a small circuit of neurons that produces rhythmic, coordinated output on its own, without needing a rhythmic command from above and without a brain to oversee it. Much of the behavior of invertebrate animals runs on such circuits — the swimming of a leech, the wave of muscle that moves food through a crustacean’s gut, the rhythmic crawl of a worm. In several of these animals the entire circuit has been mapped neuron by neuron, and the principle that emerges is the one we will return to constantly: the coordinated pattern is a product of the circuit’s own dynamics. There is no conductor. The wiring, in effect, is the score.
The simplest version of the idea is worth seeing now, because it recurs all the way up into your own spinal cord. Imagine two groups of neurons — call them a flexor group and an extensor group — wired so that each inhibits the other. When one is active, it silences its rival; when it tires or releases, the rival springs free and takes over, silencing the first in turn. The result is a back-and-forth alternation — flex, extend, flex, extend — exactly the rhythm you need to swim, or to step. This reciprocal-inhibition arrangement, often called a half-centre, was proposed for locomotion more than a century ago, and some version of it sits beneath rhythmic movement across an enormous range of animals.
Mutual inhibition alone is not quite enough. Two pools of neurons that simply inhibit each other would, left to themselves, lock into whichever side happened to fire first. A rhythm needs a second ingredient: some way for the active side to fade, or for the silenced side to escape. Several intrinsic properties of neurons supply it. In post-inhibitory rebound, a cell that has been strongly inhibited fires a burst the instant the inhibition lifts, like a spring released from compression. In spike-frequency adaptation, an active cell gradually slows its own firing. Some cells show plateau potentials or endogenous bursting — they oscillate on their own, without any patterned input at all.
Combine mutual inhibition with any of these and you get a self-sustaining alternation: side A bursts and silences B; A adapts while B rebounds; B bursts and silences A; and the cycle continues, with no rhythmic input and no conductor anywhere in the loop.
Notice the shape of this explanation, because it is the shape of nearly every explanation in this book. We are not describing an algorithm the circuit runs. We are describing dynamics its wiring and its membranes produce. As we put it in Unit I, the architecture does not implement the rhythm — it is the rhythm.
16.4 We never gave them up
The most important fact about central pattern generators, for our purposes, is that evolution did not throw them away when brains grew large. It kept them, and tucked them beneath the brain — precisely so that the brain would not have to think about repetitive control.
You are running CPGs right now. Breathing is the clearest case: it is muscular work, rhythmic, and almost entirely automatic. You can seize control of it for a sentence or a swim, but you do not author each breath, and the moment your attention wanders the rhythm carries on without you. Chewing is another — once you begin, the alternation of jaw muscles largely runs itself. Even walking, which feels deliberate, is built on a rhythm generated in the spinal cord. The evidence is old and striking: animals whose spinal cords have been surgically separated from the brain still produce stepping movements, and a cat in this condition, placed on a moving treadmill, will step with proper alternation of its flexor and extensor muscles — despite receiving no signals from the brain at all. The rhythm of locomotion lives in the cord; the brain’s job is not to generate it from scratch but to start it, stop it, and steer it.
This should already unsettle the picture of the little man, and it points to a more general truth about the spinal cord: it is not a cable. We are tempted to think of it as a bundle of wires carrying the brain’s orders down and the body’s reports up, but it is itself a sophisticated controller. It contains reflex circuits that protect and stabilize the body on timescales far too short for the brain to be involved — pull your hand from a flame before you consciously feel the heat; tighten a muscle automatically the instant an unexpected load begins to stretch it. Some of these reflexes are genuinely clever: step on a tack and the same spinal circuit that lifts the injured leg simultaneously stiffens the other leg to catch your falling weight. None of this requires a round trip to the brain, and that is the point. Sending every detail up to the cortex and waiting for a reply would make us hopelessly slow and clumsy; so the fast, local, repetitive work is handled locally.
One consequence is worth stating plainly now, because it reshapes how we should think about everything above the cord. Even when the brain does drive a movement, control is not a one-way broadcast from cortex to muscle. The lower circuits talk back — through their own feedback loops and through the steady stream of sensory information from the muscles themselves — so that “command” and “consequence” are continuously compared and adjusted. Movement is better pictured as a loop between higher and lower systems than as an order issued from the top.
We will meet one particularly instructive vertebrate later in this unit: the lamprey. It is not here as an example of a CPG — it has them, but so does the worm, and the worm makes the point about brainless rhythm more starkly. The lamprey earns its place when we reach the basal ganglia, because its brain shows us, with unusual clarity, how a higher structure selects which of these ready-made patterns to release at any moment. Generating a pattern and choosing among patterns are different problems, and the lamprey is our best window onto the second one. For now, hold the thought: we have a body full of automatic, competent movement generators, and a great deal of the brain’s motor machinery exists not to produce movement from nothing but to choose, shape, time, and predict it.
16.5 Posture: the set point you never notice
Two ideas from control theory will follow us through every chapter that remains, and the cleanest place to introduce them is the unglamorous problem of not falling over.
Stand still and you will feel almost nothing happening. In fact your body is in constant motion, making small, ceaseless corrections to keep your center of mass over your feet. Watch someone gesture while they talk — leaning, reaching, shifting their weight — and notice that they never topple, though nothing about gesturing is designed to keep them upright. Their nervous system is silently solving a balance problem the whole time, recruiting muscles in the trunk and legs that the speaker is entirely unaware of.
This is homeostasis in the motor domain. Recall the thermostat from Unit I: a variable (room temperature) is held near a set point by detecting departures and correcting them. Posture works the same way. The defended variable is the body’s orientation and balance against gravity; the corrections are continuous and automatic; and much of the work is done not by the cortex but by older systems in the brainstem that specialize in keeping us upright. Posture is, in the most literal sense, the motor system defending a set point — and it is the precondition for everything else. You cannot reach, walk, lift, or even sit without continuously winning the balance problem first. Before the brain can pursue any goal with the body, it must keep the body from falling down.
16.6 Anticipation: stability through prediction
But pure reaction — wait until you start to tip, then correct — is a poor way to stay upright, for exactly the reason that pure reaction is a poor way to stay alive. Recall the book’s central arc: an animal that waits until it is already dehydrated to seek water has waited too long, and the brain’s great trick is to act before the error occurs, using prediction. We called this allostasis — stability through change, or more loosely, stability through prediction. The motor system runs on the same trick, and here we can see it operate at the scale of milliseconds.
The textbook demonstration is the anticipatory postural adjustment. Stand and hold a load in your hand, then lift it on a signal. Common sense says the lifting muscle — the biceps — should fire first. But if you record the muscles, you find that a postural muscle in the calf begins to fire slightly before the biceps does. Think about why. Flexing your arm to raise a weight will pull your body forward and threaten your balance — but only after the arm starts to move. Your nervous system has forecast that consequence and pre-compensated for it, bracing against a disturbance that has not happened yet. The postural correction is launched in advance, on the basis of a prediction about your own movement.
This is the place to draw the distinction that organizes the rest of the unit: the difference between feedforward and feedback control. Feedforward control acts on a prediction — it pre-compensates for the consequences your own action is about to cause, which is what the calf muscle is doing before you lift. Feedback control acts on a measurement — it corrects for what actually happened, especially for surprises the prediction did not include. If the box turns out to be heavier than you expected, or someone jostles you, no forecast covered that; your proprioceptors report the unexpected error after the fact and the system corrects. A capable mover needs both: feedforward to handle the predictable consequences of its own behavior, feedback to catch everything else.
Hold onto the feedforward half, because it raises a demanding question that will drive the deepest chapter of this unit. To pre-compensate for the consequences of a movement before the movement produces any sensory feedback, the brain must somehow know in advance what those consequences will be. It must, in effect, carry an internal model of the body — a model that can take a planned movement as input and predict the forces and sensations it will produce. Building, running, and continuously correcting such a model is, we will argue, much of what the cerebellum is for. Posture and anticipation are not two more facts to memorize; they are the set-point and the prediction from Unit I, reappearing in flesh and muscle, and together they explain why the elaborate machinery above the spinal cord had to evolve at all.
There is a small, eerie demonstration that the motor system runs its own predictive model — one that can diverge from what you consciously perceive. Take two objects of identical weight but different size and lift each in turn. You will be certain the smaller one is heavier. This is the size–weight illusion, and it is remarkably stubborn: it does not fade no matter how many times you lift the objects.
Now measure the forces your hand actually applies. On the first lift you over-prepare for the larger object, expecting it to be heavy — but within a few lifts your grip and lifting forces settle to the same values for both objects. Your motor system has quietly worked out that they weigh the same, even as you continue to feel the smaller one as heavier. Two estimates of the world’s weight now coexist in one head: a perceptual estimate that is wrong and won’t budge, and a motor estimate that is right and silently self-correcting.
This is a first glimpse of something we develop fully when we reach the cerebellum. The motor system maintains an internal model of the body and its loads, predicts the forces a movement will require, and updates that prediction from error — and this model can run independently of, and more accurately than, conscious perception. Prediction, once again, is doing the real work.
16.7 The architecture we will climb
We can now lay out the architecture this unit will ascend, and map it onto the chapters to come. The motor system is often described as a hierarchy, and it is — but we should be careful with the word. In keeping with everything above, “hierarchy” here does not mean a boss handing orders to subordinates. It means layers of control added over evolutionary time, each new layer offloading routine work onto the layers below and reshaping what they do, rather than replacing them. The old machinery keeps running underneath; the new machinery learns to use it.
At the base sits the layer we have already begun to explore: the spinal cord and brainstem, with their reflex circuits and pattern generators, and the lower motor neurons that actually contract muscle — the final common pathway, in Sherrington’s enduring phrase, through which every command, from whatever source, must ultimately pass. Whatever the rest of the brain decides, it can only move the body by way of these cells.
Above this sits the cortical layer — the upper motor neurons of the motor and premotor cortices — which plan, select, initiate, and direct voluntary movement. Here the most useful organizing distinction is not the old textbook split between “upper” and “lower” but a functional and anatomical one between proximal and distal control: between the medial systems that govern the trunk and the muscles near the body’s core, the machinery of posture and locomotion, and the lateral systems that govern the skilled, independent movements of the extremities — the hand that writes and grasps. The cortex also turns out to be, as already promised, organized around movements rather than muscles, a claim we will defend with some of the most interesting experiments in neuroscience. And it is here that the motor system reveals a feature we have met before: it is almost entirely crossed, each hemisphere governing the opposite side of the body, which is why a stroke on one side of the brain weakens the other side of the body. This first cortical chapter is where the proximal/distal distinction, the motor cortices, and the disorders of cortical motor control belong.
Finally, two great structures sit off to the side of this main pathway. Neither talks to muscles directly; both work by shaping the cortex’s output, through loops that pass mainly through the thalamus — and although they are often introduced as a pair, they solve different problems, and we will take them one at a time.
The cerebellum we take first, because its story is the more self-contained. It is the brain’s system for prediction and calibration — the structure that keeps movements smooth, accurate, and well-timed, and that learns from movement error. The leading idea, which we will weigh rather than simply assert, is that it builds an internal model of the body capable of forecasting the consequences of a movement before the senses could report them — making it the machinery behind the anticipation we have just met. How literally to read that “internal model” language turns out to be one of the liveliest disputes in the unit, and the chapter is built around taking the idea seriously while marking exactly how far the evidence carries it.
The basal ganglia we take second, because their problem opens onto the one the unit ends with. They are the brain’s system for selection and gating — for releasing the movement you want while suppressing the many competing movements you do not. This is the problem the lamprey will illuminate for us, and the problem that fails in two famous and opposite ways: in Parkinson’s disease, where wanted movements become hard to release, and in Huntington’s disease, where unwanted movements become hard to suppress. But selection raises a question it cannot answer on its own. To release the movement you want is to presuppose that some actions are worth more than others — and nothing so far says where that worth comes from. So the basal ganglia hand us, in their final turn, the problem of value: how the brain learns what an action is worth, how dopamine carries that lesson, and why wanting something and liking it turn out to be separate machinery. The cerebellum, the basal ganglia, and this problem of value each receive a chapter of their own — three chapters, not two, because choosing what to do and learning what it is worth are not the same act.
This architecture even gives us a diagnostic logic that runs through the whole unit: because the system is layered, the kind of deficit a patient shows points to the level of the damage. Destroy the final pathway in the cord and you get paralysis — no plan, however intact, can reach the muscle. Damage the cortex and you can get apraxia — the strength and the muscles are fine, but the organization of skilled action falls apart. Damage the cerebellum and you get ataxia — movement that is strong but clumsy, poorly aimed, poorly timed. Damage the basal ganglia and you get the disorders of release and suppression, Parkinson’s and Huntington’s. And when we reach value, we will meet a failure of a different kind altogether — addiction, in which the machinery that learns what is worth wanting is captured and turned against the organism, a disorder not of moving but of wanting. We will use this logic repeatedly: the failures of the system are among our best evidence for how the intact system is organized.
16.8 Where we are going
From a bacterium biasing its random walk toward food to a pianist’s hand finding a chord, movement is the brain doing its oldest and most basic job: turning the body’s needs into action on the world. The chapters ahead begin at the muscle and work outward through the control system — first the spinal cord and the cortical systems for proximal and distal control; then the cerebellum’s work of predicting and refining movement; then the basal ganglia’s work of choosing among the actions on offer; and finally, because choosing presupposes worth, the dopamine system that learns what an action is worth and energizes its pursuit — the point at which the motor system passes the body’s needs on to valuation, and at which this unit hands its question to the next. At each step the same two questions recur, the questions this book has asked from the beginning: what problem of control did this layer evolve to solve, and what does its architecture do, that we are tempted, and should resist the temptation, to describe as a little man issuing commands?
We begin at the bottom, where movement actually meets muscle.
It is worth separating, at the outset, the parts of this story that are settled from the parts that remain at the frontier — a distinction we will revisit at the end of each chapter.
What is well established. The spinal cord contains reflex circuits and locomotor pattern generators that produce coordinated movement without the forebrain. Lower motor neurons are the final common pathway to muscle, so that damage to them produces paralysis. The corticospinal system is overwhelmingly crossed, so that one hemisphere governs the opposite side of the body. Balance is defended continuously, and postural muscles are recruited in anticipation of voluntary movements, not merely in reaction to them. The basal ganglia and the cerebellum shape movement indirectly, through the thalamus and brainstem rather than by contacting muscle, and their diseases produce recognizable and distinct disorders.
What remains contested or unsettled. Exactly what activity in primary motor cortex fundamentally encodes — particular muscles, abstract directions of movement, or the evolving state of a dynamical system that generates muscle activity — is a live debate we take up directly when we reach the motor cortex. How literally to read the language of “internal models” and “forward models” for the cerebellum is not fully resolved. How much genuine ipsilateral motor control the human brain possesses, and whether it can be recruited to aid recovery after stroke, is an open question with real consequences for rehabilitation. And whether so-called mirror neurons underpin our understanding of others’ actions, or are instead a consequence of that understanding, is far from settled. Keeping the established core and the open frontier distinct is part of reading this unit well.