Movement and the Nervous System

Movement is the only source of the body's sense of itself

Catch your heel on a step and your body has already saved you. An arm swings out, the trunk pitches, the other foot lands wider than it meant to. Only afterwards does the alarm reach you, as a jolt of fright about something that is already over. Nobody decided any of it. The correction ran on information the body was generating about itself while it fell.

The English physiologist Charles Sherrington named that stream of information in a paper published in 1907, calling it the proprioceptive system, and the term had arrived in a book he published the year before. A neuroscientist who reread that book a century later found most of the vocabulary students now take for granted was coined in it. The full story of the naming belongs to input quality. One property of the sense decides everything about movement.

Light arrives on its own. Sound arrives on its own. Proprioceptive information has to be made, and the only way to make it is to move. Every other sense can be delivered to a body lying still. This one the body has to generate, and it generates it by moving.

Start from the asymmetry. A nervous system regulating a body it cannot see is working from a model, and the model is only as good as the traffic arriving from muscle, tendon, joint and skin. Move, and tissue slides against tissue, and that traffic runs dense, varied and full of things the system did not expect. Hold still and the supply thins.

Then the harder question. Traffic is not news, and a body can move all day while telling itself nothing it did not already know.

Input quality, prediction, and oscillation carry movement's signature

The case rests on experiments rather than definitions. Receptor recordings, reaching tasks, and instrumented walking mats each isolated one part of what movement does to the body's regulation of itself.

The first is input quality, the fidelity of the body's information about itself. Every regulator depends on its sensors. A body with a thin or distorted report about a region regulates that region badly, however healthy the region is.

The second is prediction. A nervous system does not wait for the world before acting on it. It runs a model of the body and its surroundings, acts on the model, and revises the model where the world disagrees. Revision needs disagreement.

The third is oscillation, the rhythms a system runs and the range those rhythms can move through. Walking is a rhythm generated in the spinal cord. Breathing locks to it. Rhythms in a body do not run side by side; they couple, and coupling is how a change in one place reaches another.

Compose the three and the argument appears. Movement generates information about the body. That information earns its value where it contradicts the model the body was already running. And it is delivered on a rhythm that carries the correction outward, into systems that were never moved.

When a movement problem is not a regulation problem

Degenerative cervical myelopathy is compression of the spinal cord inside the neck, and its early signs are ordinary ones. Hands that fumble buttons and handwriting. An unsteady walk. Numbness. Changes in bladder control. A clinical review in the BMJ reports that the diagnosis takes on average more than two years and around five consultations to reach. That figure comes from a record review of forty-two patients who eventually had surgery. The review warns that delayed treatment leads to poorer outcomes and lifelong disability.

So not every movement problem is a problem of regulation. Compression of a nerve or of the spinal cord, inflammatory disease, infection, metabolic disease, inherited neurological disease, toxic and drug-induced neuropathy, and tumor are real causes with their own real treatments. They are diagnosed and managed by physicians, and delay in finding them costs people function they do not get back.

The rule is simple. New or worsening neurological signs are a reason to see a physician promptly. Weakness, spreading numbness, clumsiness of the hands, a gait that has changed, unexplained falls, or any change in bladder or bowel control all need assessment rather than patience. Nothing on this page is exercise advice, a protocol, or a reason to change anything a clinician has prescribed for you.

Finding the structural cause comes first and is not negotiable. What the model adds runs alongside that search rather than after it. Every disease has a tonal expression, and many are initiated, maintained or amplified by failures of tonal regulation.

That is a different claim from saying every disease reduces to poor tone, which the model does not say. A confirmed compression still arrives in a particular nervous system, which receives it, compensates for it, and either contains or amplifies what it costs. Which is why the same compression runs a different course in two people.

Four receptor classes generate the traffic movement rides on

A nerve is a living wire, a bundle of fibers carrying messages as small electrical pulses. Signals traveling inward toward the spinal cord are called afferent. Signals traveling outward to muscle are efferent. Movement's sensory story rides on the afferent side.

A mechanoreceptor is an ending on such a wire that fires when something mechanical happens to the tissue around it. Stretch, pull, pressure, vibration, a slip of skin. Four classes do most of the work, and each has a different relationship with movement.

Position is computed from populations, not read from any single ending

Not one of these endings knows where your arm is.

Two physiologists, Uwe Proske and Simon Gandevia, established across a review of the proprioceptive senses that position is computed from whole populations of endings rather than read off any one of them. That account is carried in full on the input quality page. They also report that these senses blur with fatigue and decay with age, and that the decay tracks with falling.

The inference sits in the next line. A population estimate is only as good as the traffic feeding it, and the traffic is made by moving.

The brain believes the body's receptors even when they lie

Two experiments built the modern account of that traffic. The first is a technique nobody had performed on a conscious human being before the middle of the 1960s.

In a clinical neurophysiology department in Uppsala, the Swedish physiologists Ake Vallbo and Karl-Erik Hagbarth began pushing fine tungsten needles through skin into human nerves, including their own. The technique is called microneurography, and it holds the traffic on a single fiber while the person lies awake and says what they feel. Their full account of what skin receptors report arrived in 1968. Sensation stopped being an inference and became a signal you could record while its owner talked.

The second experiment separated the traffic from the truth. Blindfolded volunteers had a small vibrator pressed against a tendon, which drives the spindle to fire as though its muscle were being stretched. They felt the arm travel into a position it was not in, some of them past the anatomical limit of the joint. That study of vibration-induced illusions of movement is told in full on the input quality page. One line from it decides the argument about movement.

The arm had not moved. The receptor had lied, and the brain believed the receptor over the truth.

So the body's sense of itself is a reconstruction assembled from afferent traffic. It is exactly as good as the traffic, and no better. Which puts the whole question on where the traffic comes from, and on what happens when it stops.

What losing proprioception does, and does not, take away

The cleanest way to see what a signal does is to take it away. The clearest recent attempt went against the people who ran it.

Deafferentation is the loss of the incoming sensory wires while the outgoing motor wires survive. It happens to a handful of people in the world, usually after a disease strips the large sensory fibers.

In 2024 six of them were tested together, one of the largest such groups ever assembled, on a task built to isolate one narrow thing. That thing is the automatic recalibration a motor system performs when vision and movement disagree, the visuomotor loop that runs underneath any decision to correct. The prediction was that these subjects would recalibrate poorly.

They did not. Adaptation was close to normal, and so was the felt sense of where the hand had gone. The authors describe the result as a challenge to their own model, and they say plainly that six people is still six people.

Set it against what the same condition does to a life. In 1982 a London group tested a man whose large sensory fibers had been destroyed by a neuropathy, meaning a disease of the nerves themselves.

In the laboratory he could move his thumb three set distances at three set speeds, produce three set levels of force, and draw shapes in the air with his eyes shut. Outside it he could not hold a cup, fasten a button, or write. There was no automatic correction inside a movement, and no steady contraction lasting more than a second or two without watching it.

A decade later a neurophysiologist named Jonathan Cole measured force and movement perception in a second such man, who had lost his large sensory fibers below the neck as a teenager. With his eyes open he judged a weight nearly as well as anyone. With his eyes shut he could barely tell 200 grams from 400. Both cases are told in full on the input quality page.

Read together, the null is instructive rather than embarrassing. The automatic visuomotor loop is one narrow slice of what movement generates, and losing proprioception does not disable that slice. Everything Cole's patient could not do in his kitchen, he still could not do. Motor output alone is not enough. The command needs the return, and the return is made by moving.

The spinal cord generates walking, and breathing locks to its rhythm

In 1911 everyone believed walking was a chain of reflexes. The foot hits the ground, the sensation triggers the next step, and so on down the line. A young physiologist named Thomas Graham Brown, working in Sherrington's laboratory, cut both the sensory nerves and the descending connections from the brain, isolating the spinal cord. The animal still produced an alternating stepping rhythm.

Graham Brown proposed in his paper on the intrinsic factors in progression that two mutually inhibiting half-centers, one for flexion and one for extension, generate the rhythm themselves. Circuits that produce a rhythm without needing a rhythmic input are now called central pattern generators. Sensory input shapes the rhythm rather than creating it.

Almost nobody believed him. Sherrington himself went back and forth on it. The idea sat unused for nearly five decades, until a Swedish neuroscientist named Anders Lundberg began lecturing on the forgotten work in the late 1950s and writing about it a decade after that. Two historians of the field reconstructed how the half-center idea was lost and recovered. The interneurons Graham Brown had predicted, the small connector cells that sit between an input and an output, were then found.

The human spinal cord reads load itself

Two modern results make it human. In 1997 a group in the neurology department at UCLA, led by Susan Harkema, suspended four people with chronic thoracic spinal cord injuries over a treadmill. Their legs were moved through a stepping pattern by hand.

In two of them no signal from the brain could be detected below the injury at all. The electrical activity in their calf muscles scaled directly with how much body weight was allowed onto the limb, step by step. It did so whether the input coming down from the brain was normal, partial or undetectable.

Hold a newborn upright with the feet touching a surface and the baby steps. Everyone assumed that reflex was discarded and replaced when real walking arrived. In a neuromotor physiology laboratory in Rome, Nadia Dominici and colleagues recorded many leg muscles at once in stepping newborns, toddlers, preschoolers and adults.

From those recordings they pulled out the basic patterns underneath. The newborn's two patterns are still running in the adult, with two more layered on top. The same patterns turned up in rat, cat, macaque and guineafowl.

Breathing locks to the rhythm of gait

And the rhythm reaches outward. In 1983 the biologists Dennis Bramble and David Carrier put jackrabbits on treadmills and watched dogs, horses and people run, recording limb rhythm and breath together. Breathing did not run free of the legs. It locked to them, one breath per stride in trotting and galloping quadrupeds. Humans, freed by two legs from the mechanical constraint, still lock, most often at two strides per breath.

Gait and breath are not two rhythms happening near each other. They are one coupling, which is why a walk reaches so much of a body at once.

Tone, in this model, is the organization of the body's coupled oscillations rather than any one oscillation by itself. Walking is that organization in motion. A rhythm generated in the cord, tuned by load returning from the ground, and locked to the rhythm of breathing.

The brain's body map is redrawn by what you practice

Move up to the cortex, the outer sheet of the brain. It holds a map of the body's movements, and a mirror of that map for the body's skin, where each patch of skin owns its own territory.

In the 1990s neuroscientists including Randolph Nudo and Michael Merzenich charted the movement map in adult squirrel monkeys. They lowered a fine electrode into the cortex at hundreds of separate points and recorded which movement each one produced. Then they trained the animals on a task requiring fine finger work and charted the map again. Finger territory had expanded and wrist and forearm territory had shrunk. A monkey trained on a forearm-turning task showed the reverse. When training stopped, the maps drifted back.

The sensory map behaves the same way. Each neuron up there listens to one patch of skin, and the size of that patch is its receptive field. Owl monkeys trained on a task that repeatedly stimulated one small patch of fingertip skin ended up with far more cortex given to that patch. The receptive fields inside the expanded zone had also shrunk, which means the animal could resolve finer detail there. The map grows and sharpens where use and attention go.

In humans, a German imaging group including Bogdan Draganski and Arne May scanned people who could not juggle. They taught them three-ball juggling for three months and scanned them again. Then they told them to stop and scanned a third time. Gray matter in visual motion areas grew with the training and shrank back after it stopped. Gray matter is the layer where the cells themselves sit, and this is the human result everyone cites.

Where the human imaging evidence is weaker

Two neuroscientists, Cibu Thomas and Chris Baker, then reviewed that entire human literature and concluded that most of it does not survive scrutiny. The changes seen in animals happen at the scale of dendritic spines, far below what an MRI scanner resolves. The human findings are vulnerable to problems of design, statistics and imaging artifact.

The one effect they judged solid was hippocampal volume rising with exercise in older adults. The hippocampus is a seahorse-shaped structure deep in each temporal lobe, central to memory, and one of the few places in an adult brain where new cells are added.

Be precise about what that cuts. It challenges structural MRI claims in humans. It does not touch the map reorganization recorded electrically in monkeys, which is a different instrument measuring a different thing. The animal work stands. The human scanner work is on notice.

Movement without learning leaves the motor map unchanged

Everything so far is consistent with a simple story. Move more, and the system stays better mapped. One experiment kills the simple story.

Nudo's laboratory had already shown that a difficult finger-retrieval task redraws the motor map. So they built the easy version. Same reaching. Same food reward. Same well to reach into, except the well was now wide enough that no skill was required. They counted finger flexions and matched them between the groups, so the amount of movement was equal.

The monkeys on the easy task performed thousands of successful, competent, stereotyped retrievals. They learned nothing, because there was nothing to learn. Their maps did not change at all.

Movement matched. Information not matched. The cortex followed the information.

A rat experiment makes the same cut from the other side. Two neuroscientists, Michael Remple and Jeffrey Kleim, trained animals to grab and snap bundles of dried pasta, thickening the bundles as the animals got stronger. A second group broke a single strand, learning the reach without the load. Both groups ended up with expanded forelimb territory in the motor cortex. Strength added nothing on top of skill.

Two human findings sharpen it further. A pair of researchers had volunteers spend four weeks imagining maximal contractions of a small finger muscle. Strength rose twenty-two percent, against thirty percent for real contractions and under four percent for controls. The twitch evoked by electrically stimulating the nerve was unchanged in every group, so no muscle had grown and whatever changed lived in the nervous system.

Separately, seventeen randomized trials had trained one limb and measured the other. A group including the physiologist Simon Gandevia pooled the thirteen of them that reported enough data. That is a meta-analysis, a study that combines many trials into one estimate. The untrained limb gained around eight percent, roughly a third of the trained gain.

The test follows directly. If the benefit of movement tracked only metabolic dose, and not the novelty or quality of the information the movement generated, volume-matched repetition would reorganize a map as well as skilled practice does. It does not.

Two weeks without movement starts to erase the body map

If a body map is maintained by movement, then stopping should cost something, and the cost should show up quickly.

Three adults in St Louis, two of them members of the research team, wore casts on their dominant arms for two weeks and were scanned for thirty minutes every single day. The scanner records slow spontaneous fluctuations in each region, and two regions whose fluctuations rise and fall together are counted as connected.

Within forty-eight hours the cortex controlling the casted arm had disconnected from the rest of the movement network while staying internally intact. The disconnection deepened across the fortnight and reversed once the casts came off.

Then something nobody had described appeared. Large spontaneous pulses of activity swept through the disused circuit while the cast was on. Nothing the arm was doing could have produced them, because the arm was doing nothing. The authors suggest the pulses may be what holds a disconnected circuit together until its traffic returns. A regulator starved of input appears to manufacture some.

Ordinary injury does the same thing at the level of tissue. A Zurich group scanned ten people whose arms were immobilized after an accident. Within about sixteen days the cortex serving that arm had measurably thinned, and the white matter tract carrying its commands had degraded.

White matter is the insulated cabling that runs between regions. Meanwhile the untrained other hand got better at a motor task. That study, and what a cast does to the room a body has to move in, belong to constraint and slack.

Two weeks. Not two years.

The limits are real. The ten-patient study had no control group, and the participants were injured, so pain and disuse cannot be separated. Cortical thickness measured by MRI is exactly the measure Thomas and Baker warned about. Two of the three casting subjects were investigators on their own study, so blinding was impossible. What the casting study has instead is daily scanning across weeks in each person, which is a rare kind of density.

The body map is not stored. It is maintained, and the maintenance is movement. A limb that stops moving stops reporting, and a system that stops receiving a report keeps regulating that region from a model it can no longer check.

Why the map only updates when the world disagrees

You cannot tickle yourself, and that everyday fact carries the whole mechanism of movement learning. A group at University College London including Daniel Wolpert scanned people during an identical touch, delivered either by their own hand or by somebody else's. The self-produced version came out quieter in the sensory cortex. The prediction page carries that experiment in full. The movement version of it is the one that matters here.

Reach for something after the world has quietly shifted and your next reach corrects. Is the correction driven by the gap between what you predicted seeing and what you saw, or by the motor correction you made in flight? Movement neuroscientists including John Krakauer and Amy Bastian separated the two by comparing shooting movements, which allow no correction once launched, against pointing movements, which do.

Healthy people learned equally well either way. Only the sensory mismatch mattered. People with damage to the cerebellum, the structure at the back of the brain that times and tunes movement, were impaired in both.

Put that beside the easy-well experiment and the argument resolves. Repetition that confirms the model the body is already running generates almost no information, so it changes almost nothing. Movement that contradicts the model is what forces an update.

The map updates where the world disagrees with it. Everywhere else the body is only confirming what it already believed.

This is what the model means when it says specificity is correspondence rather than force. The value of an input is how much of it the system could not already predict, and how well it lands where that system's model has drifted. Size is a separate axis. The easy well delivered thousands of repetitions and changed nothing, because there was nothing in them the monkey did not already know. The general principle sits on the input quality page.

Healthy walking keeps a long memory in its stride times

Something in walking has been measured for thirty years and is still not widely taught. No two of your steps take exactly the same time, and that wobble was treated as noise in the machinery.

In 1995 a group including Jeffrey Hausdorff and the cardiologist Ary Goldberger timed every stride of ten healthy young men walking for nine minutes and asked whether walking is a random walk. It is not. How long one stride takes depends on strides taken hundreds of steps earlier. Shuffle the strides into a random order and the dependence vanishes, which proves the order was carrying it.

That memory gets a single number. Near one, a stride still carries the trace of strides taken minutes ago. Near a half, each stride has forgotten everything before it, which is what a run of coin tosses looks like. The number is a scaling exponent, and the method that produces it is called detrended fluctuation analysis.

Two years later the same group tested the obvious prediction. If the number reflects how well a system is organizing walking, it should fall when the system degrades. Healthy older adults sat lower than young adults. People with Huntington's disease, an inherited illness that destroys the brain's control of movement, sat lower still, and within that group the number tracked how impaired the person was.

Look at what that second result is. A single inherited gene causes the disease. The flattening of the stride is that disease's tonal expression, and the number is what counts it. The steps had become more random, and random here means less regulated rather than more free.

Rhythms flatten with age and disease

A geriatrician and a cardiologist, Lewis Lipsitz and Ary Goldberger, had stated the general form of the claim in 1992. Healthy physiology is more irregular than aged physiology, and the flattening of a rhythm is what aging looks like at the level of regulation. The oscillation page carries that argument across the body's other rhythms. Stride timing is where a corridor can count it.

Health is the range. A walk that has lost its variety has lost the range before the person has lost the walk.

That is the model's definition of health arriving from a gait laboratory rather than from the model. Health is the breadth and flexibility of the states a system can enter and leave, and disease is the narrowing of that space. Steps are one of the few places where the breadth can be counted in a corridor.

Gait variability is dangerous at both extremes

If health were steadiness, less variable walking would be safer walking.

In one study 503 older adults walked across an instrumented mat, a strip of floor with sensors that time and place every footfall. Their falls over the previous year were recorded. Among the 281 who walked at one meter per second or faster, the answer came out U-shaped. Walkers at either extreme of step width variability carried odds of a fall history about four times those of the middle group. The two extremes were modelled as one category against the moderate middle.

The limits matter. That extreme category was twenty-five people out of the 281, which is why the confidence interval runs from under two to over ten. The design is cross-sectional, meaning everything was measured at a single moment, so it cannot say which came first.

Falls were recalled in an interview rather than recorded as they happened. The pattern held only in the faster walkers, and the other measures of variability showed no relationship at all. It is one result, and it points somewhere useful. The clinical picture around falling itself belongs to balance and coordination.

Two biomechanists, Nicholas Stergiou and Leslie Decker, put the general form of the idea into the literature. Variability in movement is a state to be tuned rather than an error to be minimized. Too little and movement is rigid. Too much and it is noisy. Both extremes leave a system less able to absorb a perturbation, and both show up in disease.

A painful back stops sharing the load

The same shape appears in a painful back. A grid of sixty-five electrodes laid across the low back lets researchers watch which part of a muscle works rather than only how hard it works. Deborah Falla and colleagues, working in a hospital pain clinic, had nineteen people with chronic low back pain and seventeen without perform a repetitive lifting task.

In the healthy backs the activity migrated as the work went on, shifting downward, sharing the load around. In the painful backs it did not move at all. The same region worked harder in the same place, and those backs were more tender to pressure afterwards.

The movement scientists Didier Delignieres and Kjerstin Torre then reassessed a later dataset from the same laboratory, recorded with and without a metronome, using a different time-series model. They confirmed the memory in free walking. They did not confirm the further claim that walking to a beat abolishes it. On their analysis the structure did not disappear; it moved into how the walker drifted relative to the beat. The disagreement is live and it is about method.

Mismatch between movement and its report can hurt without injury

Forty-one healthy volunteers proved it in a rheumatology department in Bath, using nothing more than a mirror.

The volunteers, none with any history of pain, moved their limbs while watching a mirror arranged so that the two sides disagreed. Twenty-seven of them reported something. Numbness, pins and needles, aching, definite pain, a limb that felt heavier, or colder, or not quite theirs. The symptoms peaked at the moment of maximum conflict. Nothing was touched, injured or inflamed.

Mirror therapy, from the phantom limb to the clinic

The clinical use of the same principle began with a neuroscientist named Vilayanur Ramachandran, who worked with people who still felt a limb that was no longer there. He stood a mirror upright in a box so that the reflection of the intact hand appeared where the missing one should be.

Ten patients used it. Six felt the phantom move when the intact hand moved. Five had involuntary painful clenching spasms in the phantom, and in four of those the spasm released, the hand staying open for hours until the next one came on its own. One patient got nothing.

Two conditions carry the clinical literature from here, and they are not the same condition. One is phantom limb pain, felt in a limb that has been amputated. The other is complex regional pain syndrome, in which a limb stays painful, swollen and altered in color and temperature long after a minor injury, out of all proportion to what happened to it. Trials have often pooled them, and the evidence does not sort cleanly between them.

The graded motor imagery record, from success to failed replication

The pain researcher Lorimer Moseley ran a randomized trial of graded motor imagery, a staged program of judging whether pictured hands were left or right, then imagining movements, then mirror work. Fifty-one patients with phantom limb pain or complex regional pain syndrome type 1 took part.

Pain fell about twice as far as in the control group and the gains held at six months. Single center. Run by the developer of the method. The control arm received physical therapy and ongoing medical care rather than a condition matched for attention and contact time.

Two British specialist centers then delivered the same program to complex regional pain syndrome patients in ordinary practice and audited what happened. Pain did not improve. The paper is titled failure to improve pain, and Moseley is a co-author on it. One center did see meaningful functional gains, and both saw the expected improvement on the left-right judgment task, which is the internal check that patients were doing the training.

In 2022 reviewers for Cochrane, the international body that pools trials for a living, gathered every randomized trial of physical treatment for complex regional pain syndrome. Thirty-four trials, 1,339 people, twenty-seven of the trials at high risk of bias and the rest unclear. Every comparison came out at very low certainty. Thirty-three of the thirty-four enrolled the type 1 form and one enrolled type 2. None of them enrolled phantom limb pain, so that grade covers one of the two conditions.

Very low certainty means the effects could be real and could be nothing. It is not a finding that these approaches fail.

This is the shape of literature that input meeting tone predicts. A fixed protocol delivered to an unstratified sample averages the person for whom the mismatch was the live problem with the person for whom it was not, and reports a mean describing neither. The design that would settle the question has not been run.

Phantom limb pain has two competing map explanations

Underneath the therapy argument sits an unresolved argument about mechanism. Two laboratories disagree in public about it.

When an arm is amputated, the strip of cortex that served the hand loses its input. Neighboring territory, notably the face, spreads into it. In 1995 Herta Flor and colleagues, hunting the brain change behind phantom pain, measured how far that spread had gone in arm amputees and set it against how much pain each person reported. The correlation was extremely tight. The map goes wrong and the wrongness is felt as pain. That became the standard teaching for two decades.

In 2013 an Oxford imaging group led by Tamar Makin found close to the opposite. Losing input did produce degeneration in the deprived cortex. Yet the people with the worst phantom pain were the ones whose former hand area was best preserved, in structure and in function. What was reduced in them was the connection between that area and the rest of the sensorimotor cortex. An intact map that had become isolated.

In 2020 the two of them wrote a review together. They agree the map changes after amputation. They agree the change relates to phantom pain. They do not agree which way the arrow runs, and they say so.

The model is not entitled to pick the convenient side, and it does not need the dispute settled to make its point. Both findings describe a region whose information has stopped arriving. In one reading the organization is invaded by its neighbors. In the other it stays intact and loses its coupling to the systems it should be in step with. Invasion and decoupling are two ways for one organization to come apart, and the variable being lost is the same variable.

What would separate them is measurement over time rather than argument. Follow amputees from before the surgery, record map structure, connectivity and pain together across years, and watch which moves first. Nobody has done it.

The deep-core theory of back pain was revised by its own authors

Raise your arm quickly and your trunk has to brace before the arm moves, or the arm's momentum would tip you.

In 1996 an Australian physiotherapy researcher named Paul Hodges put fine wires into the deep abdominal muscles of fifteen people with low back pain and fifteen without, and timed that bracing. In the healthy backs the deepest muscle, transversus abdominis, fired ahead of the shoulder muscle that started the movement, whichever way the arm went.

In the painful backs the lead was delayed. Back pain became a timing problem in the nervous system rather than a weakness problem in the muscle, and a clinical industry grew on the finding.

Later work asked where the lateness lives. A coil held outside the skull can deliver a magnetic pulse that makes one muscle twitch, and moving the coil around the head maps where that muscle sits on the surface of the brain.

Mapping that way while recording from the deep abdominal muscle, Hodges and colleagues found that center shifted in people with recurrent back pain. How far it had shifted tracked how late the muscle fired. Eleven people per group, measured at a single moment, so the shift could be a consequence of the pain rather than a cause of the delay.

Garry Allison's group, physiotherapy researchers in Perth, then pointed out that the original recordings came from one side of the abdomen only. They recorded both sides at once and repeated the arm-raising experiment. The muscle did not behave like a symmetrical corset. Its two sides differed, and the timing depended on which direction the arm went, which is the opposite of what a general stiffening mechanism would do.

The muscle and the patient do not improve together

Two physical therapy researchers in Alberta, Arnold Wong and Gregory Kawchuk, then asked the question the field had been avoiding. If treatment works by restoring these muscles, the muscle and the patient should improve together. Fifteen studies had measured both. The review found strong evidence that changes in the deep abdominal muscle and changes in the patient do not move together. For multifidus, the small deep muscles stacked along the back of the spine, the evidence came out conflicting, and they say so.

Cochrane then pooled twenty-nine trials and 2,431 patients. On very low to moderate certainty evidence, exercise aimed specifically at deep trunk control beats doing almost nothing by a clinically meaningful margin. Against other exercise there is no clinically important difference at any follow-up point.

In 2011 Hodges rewrote his own theory with Kylie Tucker. Pain redistributes activity within and between muscles rather than switching fixed ones off and on. It protects the region in the short run and narrows the range of ways a person can move in the long run. The constraint and slack page carries that revision in full. The field revised itself in public, and the revision landed on range.

Note what that null does and does not touch. The model does not rank doorways, so a result showing one brand of exercise matching another is not a result against correspondence, which operates inside whichever door is used. And these trials sorted people by diagnosis rather than by a tone measure taken first. So they cannot tell a world where any door works from a world where matched inputs work and mismatched ones dilute the mean.

There is a third reading and it is the one this literature keeps stumbling into. The loudest tissue is often a compensation splinting a distortion that began somewhere else. A program aimed at the painful trunk is aimed at the site. Which is one reason it performs like a program aimed anywhere else, and why the muscle and the patient were never going to move together.

Pooled exercise trials report means that describe nobody

Cochrane has reported on exercise for chronic low back pain. Moderate-certainty evidence that it probably beats no treatment, usual care or placebo for pain. For function the effect is small enough to miss the threshold the reviewers themselves set for a difference a patient would notice. Against other conservative treatments the effects are small and not clinically important once everything is pooled.

The Cochrane overview of exercise for chronic pain in general reports quality of evidence that is low, effects that are small to moderate, and results inconsistent across reviews. Adverse events were few.

The sharpest null concerns the brain. An umbrella review, which pools meta-analyses rather than trials, gathered twenty-four of them on exercise and cognition in healthy people. The raw benefit was small. It shrank once active control groups and baseline differences were handled. Publication bias is the tendency of disappointing results to go unpublished. After correcting for it, the benefit was indistinguishable from zero.

Individual responses, and the statistical objection

Then the finding that gives those averages their shape. Claude Bouchard, a geneticist who studies why bodies answer training differently, pooled six supervised exercise studies covering 1,687 people. He first measured how much each marker naturally wobbles between visits, then counted only changes clearly larger than that wobble.

Adverse moves ran from about eight in a hundred for fasting insulin to about thirteen in a hundred for HDL cholesterol. Roughly seven in a hundred moved the wrong way on two markers or more. Same program. Same dose. Opposite direction.

Two exercise scientists who work on research methods, Greg Atkinson and Alan Batterham, took that kind of claim apart.

Measure anyone twice and the people who scored highest the first time will average lower the second, because an unusually high first reading was partly luck. That is regression to the mean. Sort people by their first reading and the arithmetic manufactures responders and non-responders out of nothing.

They built a simulated dataset in which every single person responded by exactly the same amount, then plotted it the way exercise studies usually plot their data. The familiar picture appeared. Dramatic responders. Clear non-responders. Apparently large individual differences. All of it came from ordinary within-person variation plus that arithmetic. Real individual differences can only be established by comparing the spread of change in a treated arm against the spread in a control arm, and most studies have never done it.

The objection is correct. What it establishes is that the usual evidence for responder variation is not evidence. It does not establish that the variation is absent, and it names precisely the design that would settle the question. Stratify people in advance on a measure taken before the outcome is known, then compare the treated spread against a matched control arm.

That is the same design this model specifies for its own central prediction. It sorts people in advance on a tone measure recorded before any outcome is known. It also carries a sham arm, a matched dummy input, precisely because sorting people by their starting side produces some convergence from regression to the mean alone.

Their demand and the model's prediction are one demand. The trial has not been run. Until it is, a mean of this shape stays compatible with two things at once. A real effect being averaged away, and no effect at all.

A timing cue and active movement each reorganize walking

Two results isolate single variables the argument has been leaning on.

The first isolates rhythm. Twenty people relearning to walk after a stroke were split into two groups doing the same six-week program, except that one group had a steady auditory beat to step to, gradually sped up.

Tested at the end with no beat present, the rhythm group had gained more walking speed and a longer stride, and the electrical activity in the calf muscle had become markedly less variable. A Colorado group ran it with ten people per group at one center, without blinding, and early stroke recovery happens anyway.

The input carried no force and no position information. It carried timing, delivered to a different sense entirely, and the walking reorganized around it.

The second isolates the act. A systematic review screened 3,297 studies down to seventy that trained the sense of body position and measured both that sense and motor performance afterwards. Training the sense improves it, and motor performance improves along with it, across healthy and clinical groups. Then the conclusion adds, almost in passing, the sentence that matters most here. The regimens requiring active movement of the trainee tended to be the most successful. Being moved did less than moving.

Read through the model, those two results meet in the middle. What decides whether an input does anything is whether the system registers it and reorganizes around it. An input the body has to compute against is one reliable way to be registered.

So is a signal arriving in a sense the body was not using for the job. Registration belongs to the receiver, which is why the same beat passing through a different tone is a different biological event. The input does not carry the outcome. The system that meets it does.

Leoni Winter and Jurgen Konczak, who led the review of position-sense training, also say plainly that nobody knows how long the gains last. They add that there is no solid evidence yet about what changes in the nervous system to produce them.

Movement is how tone stays current

The fields these findings come from rarely cite each other. Receptor populations that only make sense together. A cord that generates rhythm without a brain. Maps redrawn by skill and faded by two weeks in a cast. A forecast that subtracts the expected. Stride timing with a memory hundreds of steps long. Variability that is dangerous at both ends. A back that stopped sharing the load.

The Unified Model of Tone reads them as one thing.

Tone is the integrated organization of the body's interacting state transitions, at every scale, carried by coupled oscillation. Movement is how that organization is refreshed. Tissue moves against tissue, receptor populations fire, the report comes back, and the model the body was running is corrected wherever the report contradicts it. The rhythm carries the correction outward, which is why gait and breath do so much at once.

Read that way, the three dimensions compose rather than sit side by side. Movement generates the information. The information earns its value where it disagrees with the prediction. And the rhythm delivers the result into systems that were never moved. Remove any one and the account collapses. No movement, no information. No surprise, no update. No rhythm, no reach past the limb that moved.

The body's model of itself is maintained rather than stored. A body that stops moving stops correcting it.

Restoring against masking, in movement

The distinction between restoring and masking is generous by design. A painkiller lowers a signal. An anti-inflammatory lowers a chemical driver. A brace lowers a mechanical demand. A walking aid gives a person their day back. Surgery changes a constraint that nothing else could change. All of that has a real and honorable place, and people who need it should have it.

Relief also does more than make a day bearable. Pain that costs a person sleep and work is itself a load on regulation, and lifting that load can be what gives a system back the reserve it needs to reorganize at all. On this account relief is a participant in restoration rather than an alternative to it.

The line the model draws runs by aim rather than by who delivers the input. Masking quiets an output while the organization underneath goes on operating as it was. Restoring reorganizes the regulator, so the output settles because the system found its middle again. Every doorway contains both, and the same input can be given either way. The auditory beat in the last section was delivered to a person by a machine, and it reorganized how they walked.

What the model aims at is the second achievement. Widening the range of states a system can enter and leave, and improving the fidelity of the information it holds about itself. Those two targets can look identical on a 12-week outcome scale and behave differently across years.

What a restoring input does that a masking one cannot

One trial separates the two. A directional input moves a value one way regardless of where the person started. A restored regulator moves it toward the middle from either side. Movement offers an unusually clean test, because the variability findings are already U-shaped.

So the design is specified in advance. Take two groups of walkers, one whose stride variability sits above the healthy window and one whose sits below it. Fix the input and the site beforehand, from a tone measure recorded before any outcome is known, so the site is the one the measure picks rather than the one that hurts. The input can be as plain as a stepping beat, or position-sense training the person has to move themselves to receive.

Deliver it to half of each group. Give the other half a sham matched for contact, time and attention. The model predicts both groups move toward the middle, further than the sham arms move, and predicts the spread of the treated cohort narrows around it. A uniform shift in one direction marks an input that pushed the output and carried the group on the wrong side further from the middle. Convergence no greater than the sham means the input did nothing the sham did not.

The dose question runs in the same corridor. Match two programs for energy cost and vary only how much a person has to learn to complete them. If stride variability moves the same distance under both, then novelty added nothing on top of work.

What a walk can measure today

Finally the measure, which is the part a clinician can act on today. This model holds that tone shows itself in three places. In the variability of a signal rather than its average. In the coupling between two rhythms rather than either alone.

And in how a system answers a challenge and returns from it. Stride timing carries all three. So does the ratio of breaths to strides, the error in a person's sense of joint position, and how quickly a gait pattern settles after somebody nudges it.

Most of tone needs a laboratory to read. Movement can be read from a walk, and it is not the distance that carries the reading.

How movement relates to the rest of the library

Movement runs through more of this library than any other subject, because it is how the body generates the information every other page's regulation depends on. The machinery that produces a movement in the first place, meaning the forward model, efference copy, the paired predictors and controllers, and the descending systems that carry each part of a command, is taught on the neurology lesson on internal models and the motor hierarchy. This page is what movement generates and what reads it.

Frequently asked