How Balance Works | Vestibular System, Vision, Proprioception and Postural Control

Part VI. Vision: why seeing the world also tells you how your body is moving

54. A moving image has several possible causes

Imagine Alicia sitting beside a large window. A vehicle in the next lane starts moving, and for a moment she feels that her own vehicle has moved instead. The retinal change is real; its cause is uncertain. The same broad movement across the visual field could have been produced by her vehicle, the neighbouring vehicle, a turn of her head or a movement of her eyes. Vision supplies evidence about relative movement. It does not attach a reliable label saying which object supplied the movement.

This is an inverse problem. In the forward direction, a moving eye viewing a particular scene produces a pattern of retinal motion. In the reverse direction, the nervous system must infer the scene and the observer’s movement from that pattern. Several physical situations can produce similar images. A balance explanation that says the eyes simply tell the brain where the body is misses this ambiguity. Eyes provide measurements from which position and movement must be estimated.

The ambiguity is reduced by comparing visual change with other information. A head turn should produce canal activity and a neck movement signal. Eye rotation is accompanied by an ocular motor command. A whole-body acceleration should affect vestibular and contact forces. When these sources disagree, the nervous system must decide whether the disagreement reflects a moving object, a moving support, an unreliable cue or an unexpected bodily movement. No single sensory channel can resolve all four possibilities by itself.

The practical learning question is not whether an image moved but what alternatives remain compatible with that movement. Tricia writes three competing explanations for the vehicle scene. Kai Kai asks what observation would separate them. A stationary pillar beyond both vehicles would provide another visual reference. A view of the road would provide a different motion pattern. Vestibular evidence would contribute during acceleration, but an absence of canal activity would not by itself prove an absence of steady translation. Each additional observation removes some ambiguity without making every other sensor unnecessary.

The mechanism connects directly to the existing How Vision Works article. That article explains visual processing itself. Here the relevant question is how the resulting information enters the estimate of bodily motion. Keeping those questions distinct prevents balance from becoming an incomplete replacement for the vision system.

55. Optic flow is a geometry of movement, not a universal speedometer

When an observer moves forward through a stationary environment while maintaining an approximately fixed viewing direction, many visible features move outward across the retinal field. Features near the direction of travel move relatively little, while features to the sides move more. This structured pattern is optic flow. Backward travel, lateral travel and rotation produce different components of the pattern. The visual system can use these relationships to estimate heading and the movement of the observer relative to surrounding surfaces.

Depth matters. A nearby fence appears to move across the visual field faster than a distant hillside during the same vehicle journey. The observer has not simultaneously acquired two speeds; the image motion differs because the objects are at different distances. Retinal speed therefore cannot be converted into absolute travel speed without assumptions or additional information about distance. An ordinary scene contains many depths, which is useful information but also means there is no single number representing the flow of the entire image.

Rotation introduces a different problem. Turning the head creates visual flow even when the body stays in one place. Eye movements can partly cancel or add to this flow. A complete estimate of self-motion must separate translational and rotational contributions. This is why the vestibulo-ocular reflex and visual motion processing belong together in a balance explanation: stabilising the eyes changes the information that the visual system receives about the moving world.

Consider a paper-only comparison. Tricia draws two corridors with identical shapes but different scales. In one, nearby walls create rapid image expansion during a given forward movement. In the other, the walls are farther away and the expansion is slower. She cannot rank the observer’s physical speed from visual expansion alone. Kai Kai asks her to add a known floor-tile size or a distance estimate. Once scale information is supplied, some of the ambiguity can be reduced. The exercise teaches what a sensory signal contains, rather than claiming the signal contains every variable the observer needs.

The terminology also deserves care. In older research, visual information about one’s own movement was sometimes described as visual proprioception. That does not mean the retina contains muscle-spindle receptors. It describes a function: vision can inform the observer about bodily movement. The same sensory organ can provide information about the outside world and about the observer’s relation to it.

56. A visual surround can influence posture without moving the floor

A classic experimental approach changes the visual environment while the support surface remains stationary. Lee and Aronson’s 1974 study of infants learning to stand showed that movement of the surrounding visual scene could produce bodily sway. The study provides evidence that visual information contributes to postural control; it does not establish a fixed visual contribution for all ages and tasks. The original report is Visual proprioceptive control of standing in human infants.

The reasoning behind this experiment is important. If the floor remains still but the person moves with a changing visual surround, the visual manipulation has entered the postural loop. The response cannot be explained only by a mechanical push from the moving floor. Nevertheless, the observation does not prove that the vestibular system stopped operating. Vestibular and somatosensory signals were still present; they were being combined with a deliberately misleading visual reference.

There are several different outcomes an experimenter might measure. The person might report an impression of movement. Their centre of pressure might shift. Their head might move. Their trunk might sway with a measurable delay. These outcomes are related, but they are not interchangeable. A conscious movement illusion is not a direct recording of ankle torque. Conversely, a postural response can occur without the participant describing a dramatic illusion.

Alicia initially says that vision has tricked the brain. Tricia gives a more useful account: the system interpreted a visual reference according to an assumption that usually works. In ordinary environments, large portions of the room are stationary. When an experiment violates that assumption, a normally useful inference can produce the wrong estimate. The lesson is not that vision is poor. It is that perception and control depend on assumptions about which parts of the world are stable.

This is a thought experiment for readers, not an instruction to move walls, induce dizziness or challenge someone’s standing balance. A diagram of a stationary floor and moving surround is enough to analyse the causal alternatives. The intellectual result is to separate a physical disturbance to the body from a disturbance to information about the body.

57. Visual reliability depends on contrast, distance, motion and the task

Good visual acuity is only one component of useful visual information for balance. A person may read a small stationary letter clearly yet have limited visual information about a poorly lit floor. A distant horizon can provide orientation but little precise information about the next small step. A moving crowd contains many visible edges, but many of those edges are not stable environmental references. The question is whether the visible information constrains the movement estimate needed for the task.

Imagine three versions of the same room. The first has a well-lit floor and stationary furniture. The second has equally sharp central lettering but a dark floor. The third has large moving projections on the walls. An eye chart score could be similar in all three imagined situations, while the visual evidence available for walking and orientation differs. The physical eyes did not change; the scene and the task changed what their signals could usefully mean.

Looking toward the ground can provide information about obstacles and foot placement. Looking ahead can provide information about the route and approaching events. Neither direction is universally optimal. Gaze must be scheduled around the task. A useful explanation therefore avoids replacing one slogan, such as always look forward, with another, such as always look at your feet. What matters is the information required next and whether it is available in time.

Visual correction also has finite delay. The visual system must acquire and process the changing scene, while muscles require time to generate force. Very rapid head movements therefore need fast vestibular contributions to gaze stability rather than waiting for vision to detect and correct all retinal slip. Slower visual feedback remains valuable for orientation and for recalibrating responses over repeated movements.

The main transfer principle is that a sensor has no fixed value independent of the environment. A sharp image of an unstable reference may be less useful than a less detailed image of a reliable one. Kai Kai asks the class to identify the reference, the required variable and the deadline before ranking a visual cue as good or poor.

58. Stable gaze and stable posture solve different errors

The vestibulo-ocular reflex aims to reduce unwanted image movement during head motion. Postural control aims to manage bodily position, momentum and support. These functions interact, but success in one does not certify success in the other. A seated person with a supported trunk can stabilise gaze during a head movement without having to perform the same whole-body correction as a person walking over uneven ground.

Conversely, a person can remain mechanically supported while the visual scene is blurred. A chair may prevent a fall even when the eyes do not compensate well for head motion. The absence of falling in that condition does not demonstrate normal gaze stabilisation. This distinction is useful because the everyday word balance can refer to an inner-ear function, the feeling of steadiness, eye stability, a laboratory sway measure or the ability to recover a step. The meaning needs to be stated.

Target distance adds another distinction. A small head translation changes the viewing direction to a nearby target more than to a distant target. The eye response needed to keep a near object clear therefore differs from the response for a far object. A reflex gain measured in one geometry should not be treated as a geometry-free score. This is an application of elementary perspective, not a reason to dismiss the measurement.

Alicia sketches a head, two target locations and the lines of sight. Tricia marks the larger angular change for the nearer target. Kai Kai asks which feature would be missing from a statement that reports only eye speed divided by head speed. The answer is the task geometry, along with the movement direction and measurement method. A dimensionless ratio can still depend strongly on the conditions under which it was measured.

For the reader, the usable distinction is simple: ask what error the response is reducing. Retinal slip, whole-body sway and foot-placement error are not the same signal. A complete balance system must coordinate them, but an explanation should not merge them before it has shown how the coordination works.

Part VII. Proprioception and touch: how the body describes its own configuration

59. Muscle length is not automatically body orientation

A muscle spindle provides information related to the length and stretch of the muscle in which it sits. To use that information for posture, the nervous system must connect it to joint geometry and then to the position of the body relative to support. That chain works particularly well when the support surface is predictable. It becomes less direct when the foot itself rotates on a moving platform or sinks into a compliant surface.

Consider two drawings with the same ankle angle. In the first, the foot rests on a fixed horizontal floor and the leg leans forward. In the second, the support surface tilts while the leg remains differently oriented relative to gravity. The local ankle relationship can be similar even though the global body orientation is not. An ankle-related sensory signal alone cannot distinguish the two drawings. Vestibular information, vision and other bodily signals help resolve the ambiguity.

This is a coordinate problem. A signal expressed relative to the foot is not yet expressed relative to the room or to gravity. The body must estimate the orientation of the reference segment as well as the joint relationship. A diagram with only the ankle joint misses the part of the inference that turns a local length change into a whole-body state estimate.

Muscle spindles are also active components of a moving system. Their sensitivity is adjusted through gamma motor activity, and muscle contraction changes the geometry around the receptor. The sensory signal is not the output of a passive ruler attached to an unchanging cable. The specialist mechanism is developed in How Muscles Work; its balance role is to provide useful, context-dependent information for reconstructing posture.

Tricia’s corrected sentence is therefore more precise than muscles tell the brain where the body is. Muscle receptors report aspects of local muscle state. The nervous system combines these reports with geometry, support information and other senses to estimate the state of the person. The additional steps are where many apparent contradictions disappear.

60. The sole of the foot is a contact-information surface

The skin under the feet experiences pressure, stretch and shear as the body loads the ground. The distribution changes when weight moves toward the heel, forefoot or edge of the foot. These signals help describe the contact between the person and the support. They are not direct measurements of the entire body’s centre of mass, but they constrain how support forces are being produced.

Alicia draws two feet and shades the areas carrying pressure. Tricia asks whether the shaded regions are identical to the base of support. They are not. The current pressure distribution describes where load is acting now. The mechanically available support region describes where contact forces might be generated as the pressure distribution changes. Kai Kai adds a third quantity: centre of pressure, the location of the resultant normal force for the measured contact. These three ideas should not be collapsed into one foot outline.

The skin also provides information about the surface interaction itself. Shear can indicate a tendency to slip. Local deformation can indicate uneven contact. Changing pressure under the toes may accompany a corrective torque or a movement toward push-off. The interpretation depends on the current task: a pressure change during quiet standing means something different from the same change during a deliberate step.

Shoes, insoles and flooring alter this interaction in both mechanical and sensory ways. A softer layer can change pressure distribution while also changing the fidelity of contact information. It is therefore too simple to call every soft surface protective or every firm surface informative. The relevant variables include friction, deformation, fit, support geometry and the person’s task. This article explains the mechanism; it does not prescribe a particular footwear product or claim that one material is best for everyone.

The existing How Skin Works article explains the sensory interface at the tissue level. The balance application is that skin provides a map of interaction with support. The nervous system uses that map together with muscle and vestibular signals, rather than asking the skin to solve the whole orientation problem independently.

61. Light touch can provide information without carrying much body weight

Two mechanisms can explain why contact with a rail changes sway. The hand may transmit enough force to support the body mechanically. Alternatively, small contact forces may provide a stable external reference that improves the estimate of bodily movement. Those mechanisms can occur together, but they are not the same explanation.

Jeka and Lackner’s fingertip-contact experiment distinguished light touch from unrestricted force contact. Under their tested conditions, contact too light to act as substantial bodily support could still reduce postural sway. The result supports an informational contribution to touch; it does not imply that a light fingertip touch can catch a falling body or replace a properly supportive rail. See Fingertip contact influences human postural control.

The mechanical distinction can be understood without performing a balance challenge. Draw a person and a wall. In one drawing, the hand presses hard enough to carry an appreciable external force. In the other, the hand rests lightly on the wall. The second contact can reveal relative motion through changes in skin deformation and arm configuration even when its supporting force is small. Better information allows the leg and trunk muscles to change their own corrective forces.

This does not make touch magical. The reference must be interpreted correctly. A hand resting on a moving surface provides different information from a hand resting on a fixed wall. A hand that slips provides different information again. The nervous system needs an estimate of the reference, not merely the presence of contact.

For teaching, separate information from force on the diagram. Label the sensory arrow from hand to nervous system and the mechanical arrow from wall to body. A rail can participate in both. A force sensor and a sway measurement would help distinguish their contributions experimentally. This is a powerful general lesson: an intervention can change control by improving the estimate of state, not only by adding more power to the correcting muscles.

62. Head-on-trunk and head-in-space are different measurements

The vestibular organs move with the skull. They therefore report the motion of the head rather than the motion of the pelvis, foot or room directly. Neck receptors provide information about the head relative to the trunk. Combining the two makes it possible to distinguish a head turn from a turn of the whole upper body, provided other relevant information and assumptions are available.

Imagine the same head rotation in space in two cases. In the first, Alicia turns only her head to look sideways. In the second, her head and trunk rotate together. The vestibular input related to head rotation could be similar, while the neck relationship differs. A postural response appropriate to one case need not be appropriate to the other. The controller needs to know which bodily configuration produced the sensory change.

Now reverse the comparison. The same neck angle could occur while the entire person is upright or while the trunk is leaning. Head-on-trunk information would be similar, but head-in-space orientation would differ. This is why neither a neck signal nor an inner-ear signal should be treated as a complete body coordinate system.

Predictive information also contributes. A voluntary head turn has an intended motor command and expected sensory consequences. An externally imposed head movement may generate a similar physical rotation but a different mismatch between prediction and feedback. Research on vestibular processing distinguishes these situations, including work reviewed by investigators studying natural self-motion. The distinction does not mean vestibular sensors switch off during voluntary movement; it concerns how central circuits interpret and use their signals.

Kai Kai’s paper exercise uses three labelled reference frames: head, trunk and world. Each arrow must say which frame it relates to which. When a student says the head moved, he asks relative to what. Once that question becomes habitual, many confusing statements about balance, gaze and posture become straightforward comparisons between coordinates.

63. A compliant surface changes both mechanics and information

A foam support does not merely make standing harder by one abstract amount. It deforms under contact, changes the pressure pattern under the foot and alters the relationship between ankle movement and body movement. It may also change the mechanical forces that can be produced at the interface. The person has to solve a modified physical problem while interpreting a modified sensory stream.

That distinction matters when experiments compare firm and compliant surfaces. An increase in sway does not isolate one organ automatically. The surface manipulation changes the usefulness of some somatosensory information, but it also changes the support itself. The vestibular system remains part of the response, as do vision, muscle capacity and the person’s interpretation of the task.

A better causal description specifies the chain. Surface deformation changes contact geometry. Contact geometry changes sensory input and available corrective forces. The nervous system changes sensory weighting and motor output. The body then moves according to the resulting forces. The final sway trace is the combined result of that chain, not a direct reading from one vestibular organ.

For a desk-based exercise, compare two hypothetical support models. Model F has a fixed foot reference and predictable friction. Model C has a foot reference that tilts slightly with pressure. Give both models the same ankle-angle signal. Ask what additional measurements are needed to estimate body orientation. A surface-angle measurement would help directly; visual and vestibular information would provide other constraints. The exercise shows why reweighting is needed without placing a reader on an unstable surface.

This is the transition into multisensory integration. Sensors do not acquire fixed percentages of responsibility at birth. Their contribution must depend on whether the information is currently available, sufficiently reliable and relevant to the variable being controlled. The next part turns that statement into a working model, with clearly marked mathematical examples rather than invented diagnostic scores.

Part VIII. Sensory integration: combining measurements without pretending they are perfect

64. The brain needs an estimate, not a vote among three senses

Vision, vestibular input and somatosensation are often drawn as three arrows converging on a brain. The diagram is useful, but it leaves the difficult operation inside an unexplained box. An image shift, a canal discharge pattern and an ankle-related length signal do not arrive in the same units or reference frame. Before their information can be combined, their relationships to the state being estimated must be understood. Three signals are not three votes on one identical question.

A more useful model begins with a hidden state: head orientation, body configuration, movement velocity and the available support. Each sensory channel provides a different observation of some aspects of that state. The controller combines those observations with a prediction based on recent movement and motor commands. It then selects an action whose expected consequences fit the task. State estimation and action selection are related operations, but one can be uncertain even when the other is mechanically powerful.

Experimental evidence for adaptive sensory contributions comes from Peterka’s study of sensorimotor integration, which manipulated visual surrounds and support surfaces and analysed the resulting sway. The responses could be explained partly through changing sensory weights rather than one permanently fixed mixture. That finding supports a reweighting framework; it does not supply a universal percentage of balance contributed by the eyes, ears or feet.

For a classroom explanation, draw the state first and the sensors second. Ask which variable each sensor constrains, where its reference frame comes from and what could make the signal misleading. Only then draw the corrective command. This order prevents the familiar mistake of treating a list of organs as an explanation of their coordination.

65. A weighted average is a teaching model with explicit assumptions

Suppose two imaginary instruments estimate the same angle in the same coordinate system. Instrument V reports two degrees and instrument S reports four degrees. A simple combined estimate is w times V plus one minus w times S. With equal weights, the estimate is three degrees. With a weight of three quarters on V, the result is two and a half degrees. The arithmetic shows how weighting changes an estimate; it does not establish that the nervous system uses these exact numbers.

Under a particular statistical model, independent unbiased measurements with known noise variances can be combined using weights proportional to inverse variance. A measurement whose random error varies less receives more weight. If one variance is one square degree and the other is four square degrees, inverse-variance weights become four fifths and one fifth. The combined estimate is then closer to the more precise instrument. These are invented quantities for learning the mathematics, not measurements of any reader’s sensory function.

The assumptions matter more than memorising the formula. Both instruments must estimate the same underlying quantity. Their systematic biases must be absent or accounted for. Their errors must be modelled appropriately. Their timestamps must correspond closely enough for the state not to have changed between observations. If the instruments measure different frames or different moments, an apparently sophisticated average can still be wrong.

Alicia initially calls the formula a brain algorithm. Tricia corrects the claim: it is an explanatory model showing why reliability could matter. Kai Kai adds that a model can be useful without mapping one mathematical symbol to one anatomical nucleus. Its scientific value depends on what observations it predicts and where its assumptions fail.

66. Precision and accuracy come apart when a reference is biased

An instrument can repeat almost the same answer every time and still be systematically wrong. A visual surround that moves consistently with an experimental device may provide a clean signal about the surround while providing misleading information about the observer’s orientation. Calling that signal reliable without naming the intended variable confuses repeatability with correctness.

Use a synthetic example. Instrument A repeatedly reports an angle close to five degrees, while the true angle in the constructed dataset is zero. Instrument B scatters around zero with modest random variation. If the model estimates reliability only from trial-to-trial spread, it may trust A too strongly. Averaging more of A’s readings makes its average increasingly precise around the wrong value. Repetition reduces random error; it does not remove an unrecognised fixed bias.

This distinction explains why disagreement between senses can be informative. The disagreement may reveal that the supposed external reference is moving, that a body segment has changed relative to another, or that an expected relationship no longer holds. Automatically averaging every disagreement would discard evidence about the cause of the conflict. Sometimes the useful inference is that two measurements no longer describe the same state variable.

Kai Kai asks for an error budget with two columns: random variation and systematic mismatch. The first can often be reduced by averaging or by using another independent measurement. The second requires checking reference frames, calibration and the physical assumptions connecting signal to state. That distinction transfers to laboratory instruments, school experiments and website claims about balance scores.

67. Correlated cues do not provide as much independent information as they appear to

Two agreeing observations can share the same error source. A video display and a second camera view derived from the same moving display are not independent confirmations of world motion. They may provide different images while inheriting one common reference error. Counting the observations as two separate pieces of independent evidence overstates the information available.

The simplest mathematical illustration uses duplicated data. Copy one noisy angle record into two columns and average them. The average is identical to the original record; no noise has been removed. Now average two independently generated noisy records of the same fixed angle. Some positive and negative errors cancel. The difference is not the number of columns but the dependence structure of their errors.

Biological sensory signals also share influences. Head movement changes retinal motion and vestibular stimulation together. A moving support can alter ankle signals and plantar pressure simultaneously. Shared causes are often useful because they create coherent evidence of one event, but their common origin must be considered when a model assigns confidence. Agreement does not automatically mean independence.

For a teaching task, give Tricia three labelled traces and identify how each was generated. Ask which pair contributes genuinely new information. This is safer and more revealing than turning multisensory integration into an informal standing challenge. The exercise teaches why a good estimate depends on relationships among observations, not merely on the number of sensors shown in an illustration.

68. Reweighting takes time, so restored information can briefly disturb an adapted controller

A controller adjusted to one environment may respond poorly immediately after the environment changes. This is possible even when the new environment contains more accurate information. The controller’s current settings reflect the previous situation, and those settings may temporarily produce an excessive or poorly timed correction in the new one.

Peterka and Loughlin investigated this transition by changing support-surface relationships during standing. After participants had adapted to altered sensory conditions, restoring the support could produce transient oscillation. Their modelling favoured dynamic sensory reweighting over a fixed sensory combination with only a changed load compensation. The study demonstrates that adaptation state matters; it does not mean accurate sensory information is generally harmful.

An everyday engineering analogy helps. A camera stabiliser tuned for a heavy attachment may overcorrect after the attachment is removed. The hardware has not become defective merely because a setting needs recalibration. The analogy is limited: a biological controller has many interacting processes and is not literally a camera motor. Its value is to separate the physical state of the apparatus from the current state of the controller.

Alicia asks whether the first response after a change should count as the person’s normal performance. Tricia answers that it is normal for that transition, but not necessarily representative of the settled condition. Kai Kai adds three measurement windows: before the change, immediately after it and after adaptation. Without all three, a study can confuse a transition effect with a stable property.

69. Stronger correction is not automatically better when feedback is delayed

Imagine a simplified controller that produces a corrective torque proportional to an estimated lean. If the body leans forward, the controller commands a restoring action. Increasing the proportional gain makes the correction larger for the same estimated error. With no delay and simple mechanics, that may sound unambiguously helpful. With delay, it can create a problem: the command is based on an earlier state that may no longer require the same correction.

A thermostat gives a familiar analogy. If heating continues strongly after the measured temperature has already begun to recover, the room may overshoot. A delayed postural controller has an analogous timing problem, although the body is a far more complicated mechanical system. Position feedback, velocity-related damping, prediction and muscle mechanics must be coordinated. Excessive gain can create oscillation rather than steadiness.

A synthetic simulation can make the distinction explicit. Give a virtual pendulum the same mass, length and initial lean in two runs. Change only the feedback delay or gain. The trajectories can differ even though the actuator’s maximum force remains identical. A poor response therefore need not be explained by weak muscles alone. It may arise from the timing and scaling of the command.

The scientific guardrail is to label the simulation as a model, not a clinical diagnosis. A human sway trace does not reveal one unique controller gain simply by visual inspection. Different combinations of sensory delay, mechanics and noise can produce similar oscillations. The lesson is that correcting a system requires suitable timing as well as sufficient force.

Part IX. Postural strategies: changing torque, body shape and support

70. An ankle strategy changes the ground-force relationship while the body moves approximately as one segment

For small disturbances on a sufficiently firm support, ankle-related muscle activity can change the position of the resultant ground force and the torque acting on the body. In a simplified single-segment description, the trunk and legs move largely together. This is the useful core of the ankle-strategy concept. It does not claim that the knee, hip, neck and arms are literally motionless.

To understand the mechanics, draw the centre of mass above the foot and the centre of pressure under it. Their relative horizontal positions determine the direction of the horizontal acceleration in a simple inverted-pendulum model. Shifting pressure toward one part of the foot is therefore a way of changing the body’s future motion, not merely a visible consequence of the body already having moved there.

The strategy has limits. The pressure resultant cannot be moved arbitrarily far beyond the contact region while the foot maintains ordinary compressive contact. Available friction and muscle torque also constrain the response. A command asking for a corrective force outside those physical limits will fail even if the state estimate is accurate. Control cannot manufacture a larger foot or a less slippery surface by neural intention alone.

The original experimental work of Horak and Nashner on altered support configurations helped establish how postural movement patterns adapt to support geometry. The useful generalisation is that the available mechanical strategy changes with the support. It is not a rule that every disturbance must be classified into one perfectly pure ankle or hip response.

71. Hip and trunk movements alter segmental momentum rather than merely making the body shorter

A multi-segment body can move its upper and lower portions relative to one another. Hip motion changes where segment masses lie and how angular momentum is distributed. This provides possibilities unavailable to a rigid single-link pendulum. It also changes the forces that must be transmitted through the support and the muscles needed to control those forces.

Alicia sketches a rigid upright rod and then a two-link figure with a hinge at the hip. The rod has only one main configuration variable. The two-link figure can bend while keeping a different part of its mass over support. Tricia notices that a single whole-body centre-of-mass location can correspond to several internal configurations. Kai Kai adds that those configurations can have different joint torques, velocities and future movement options.

The extra degrees of freedom are useful but not free. Moving the trunk rapidly requires muscular force and changes angular momentum. A very large trunk motion may be inappropriate when the person is carrying an object or standing near an obstacle. The controller must select among available actions according to geometry, task and time, rather than always using the most visible movement.

Terms such as ankle strategy and hip strategy describe prominent features of a response. Real actions can combine them and transition between them. A careful explanation should therefore identify which joints contributed, when they moved and what mechanical effect they produced. Labelling a trace hip strategy is the beginning of an explanation, not the end.

72. A recovery step is an action that changes the problem

Keeping both feet fixed preserves one support region. Moving a foot creates a different support region. A recovery step therefore changes the boundary conditions rather than merely producing a larger version of the same in-place correction. When the current contact cannot supply a sufficient corrective moment, changing contact can make a previously unavailable recovery possible.

Time is central. The foot has to be unloaded, accelerated, cleared from the ground and placed where it can accept weight. Meanwhile, the body continues moving. A target chosen only from the present centre-of-mass position may be inappropriate by the time the foot lands. This is one reason velocity and movement prediction matter in balance, even when the final action looks as simple as taking one step.

Foot placement is constrained by anatomy and environment. A step cannot pass through an obstacle, land on empty air or instantaneously move an unlimited distance. A mechanically feasible placement may also be unavailable because the other foot is already supporting a demanding action. The set of recoverable states therefore depends on available actions, not just on the original support polygon.

A teacher should not mark every step during a perturbation as failure without defining the task. In a laboratory task requiring feet to stay fixed, a step violates the instruction. In ordinary life, a well-timed step may be a successful recovery. The same behaviour can therefore be a failed task criterion and a successful safety response. Keeping those objectives separate prevents a narrow laboratory score from being mistaken for the entire meaning of good balance.

73. Reaching adds a new contact, but the contact must actually be usable

A reachable handhold can create both an information source and a mechanical support. The arm can help redirect the body once the hand makes secure contact. Before that contact, however, moving the arm changes bodily mass distribution and momentum without receiving the external support force. The reaching phase and the supported phase are mechanically different.

Suppose two imaginary rails occupy the same location in a diagram. One is firmly attached and supports force. The other is a light movable object. A visual plan based only on their appearance may initially be similar, but the available external reaction forces are very different. The controller needs to discover or predict whether the contact can bear the intended load. A reachable object is not automatically a supportive object.

Timing again matters. A handhold that can be reached slowly from quiet standing may not be reached during a rapid disturbance. Distance, arm starting position, grip formation and the need to avoid an obstacle all affect feasibility. A complete model should represent when contact becomes available, not draw the hand on the rail from the beginning of the event.

This distinction also explains why the light-touch evidence discussed earlier should not be turned into a claim about fall arrest. A small sensory reference can reduce sway in a tested condition. Arresting a large moving body requires sufficient force, appropriate geometry and a secure connection. Information improves control; it does not repeal mechanics.

74. Anticipatory postural adjustments prepare for disturbances generated by one’s own movement

A rapid arm movement changes the forces and moments acting through the rest of the body. The disturbance is not entirely unexpected because the nervous system issued the arm command. It can therefore prepare postural activity before or around the onset of the focal movement. This is different from waiting until the body has already swayed and then correcting the error.

Cordo and Nashner studied postural adjustments associated with rapid arm movements. Their results helped show that posture and voluntary action are coordinated, with the postural response depending on task and support conditions. The evidence supports anticipatory organisation; it does not imply that every voluntary movement must be preceded by an identical muscle sequence.

A paper example separates prediction from reaction. In Case A, the model knows that an arm will accelerate forward at a specified moment. It can adjust the supporting forces in advance. In Case B, the same external impulse is imposed unexpectedly. The controller can react only after information about the event becomes available. Equal disturbance magnitude does not imply equal response timing because the available predictive information differs.

Alicia asks whether anticipation is cheating in a balance test. Tricia replies that it depends on what the test claims to measure. A predictable repeated event examines preparation and adaptation as well as recovery. An unpredictable event places more demand on feedback and rapid response. Kai Kai adds that the distinction belongs in the method, not in an after-the-fact story explaining an inconvenient result.

75. The best strategy is the one that meets the task within physical and informational limits

There is no universally superior postural strategy detached from the task. Keeping the feet still can be useful while holding a tray in a constrained space. Stepping can be useful when support must be relocated. Reaching can be useful when a secure handhold is available. Trunk motion can redistribute momentum but may interfere with a precise visual task. The controller is solving a constrained problem, not trying to minimise one number at any cost.

A useful decision description has four parts. First, estimate the current state and its uncertainty. Second, identify which contacts and movements are physically available. Third, predict their consequences over the relevant time interval. Fourth, compare those consequences with the task’s priorities. These are analytical steps for explaining control; they are not a claim that a person consciously narrates them during every step.

The priorities can conflict. Greater co-contraction can increase resistance to a disturbance while increasing energy expenditure. A wider step can increase lateral support while colliding with an obstacle. Looking down can improve foothold information while reducing attention to a moving object ahead. A mechanism-led explanation names the trade-off instead of presenting one posture or one cue as universally best.

This section also provides a useful boundary for educational content. Learning why a strategy works is not the same as selecting a rehabilitation exercise for an individual. The following chapters continue with walking models, research methods and desk-based reasoning. They do not require readers to close their eyes while standing, provoke dizziness, stand on unstable equipment or arrange surprise pushes. The quality of the explanation should come from causal precision, not from making the learning activity physically risky.

Part X. Walking and dynamic balance: controlling a body that deliberately moves beyond one support

76. Walking is not quiet standing repeated on alternating feet

Quiet standing can often be approximated as a body swaying over a relatively fixed base of support. Walking changes the problem. One foot leaves the ground, the available support shrinks, the centre of mass moves forward, and the next foot must arrive at the right place and time. Stability is therefore achieved by coordinating a sequence of deliberately unstable states.

The stance leg produces ground reaction forces that redirect the centre of mass. The swing leg is accelerated forward while avoiding the floor and obstacles. The pelvis and trunk rotate, the arms counter-rotate, and the head is stabilised enough for useful vision. Every step is both locomotion and a prediction about where future support will be available.

Alicia calls walking controlled falling. Kai Kai keeps the phrase but adds a qualification: the system is not passively collapsing. It actively shapes momentum, predicts foot placement and preserves recoverability throughout the step.

77. The centre of mass follows a moving trajectory rather than remaining over one fixed foot

During gait, the centre of mass advances forward and oscillates vertically and laterally. The supporting foot changes, and the body transfers weight from one limb to the other.

At some points the centre of mass lies ahead of the current stance foot. This would be unstable under a static rule if no new support were forthcoming. Walking remains viable because the swing foot is already travelling toward a future contact point.

The important variable is therefore not simply whether centre of mass is inside the current foot boundary. It is whether the current momentum and available actions can create a new support before control is lost.

Tricia upgrades the static support polygon into a time-dependent sequence of support regions.

78. Foot placement is a powerful balance control variable

A step changes the location where the next ground reaction force can act. Placing the foot farther to one side increases the potential lateral moment; placing it farther forward changes braking and support of forward momentum.

Small changes in step width and length can therefore correct developing errors without requiring enormous ankle torque.

The nervous system can alter foot placement from one step to the next using visual, vestibular and proprioceptive estimates.

Kai Kai calls the foot a movable control surface.

79. Step timing matters as much as step location

A mechanically ideal landing point is useless if the foot arrives too late. During a disturbance, the centre of mass continues moving while the swing limb is in flight.

The controller must therefore estimate not only where support should be placed but when contact will occur.

Changing cadence, shortening swing time or making an additional step can recover states that would be impossible with one slow step.

Alicia sees balance as a scheduling problem as well as a geometry problem.

80. Lateral balance during walking relies strongly on step-to-step foot placement

Side-to-side stability is challenging because the base of support becomes narrow during single-limb stance. Ankle inversion-eversion can make corrections, but the available centre-of-pressure range is limited by foot width.

Adjusting the lateral position of the next foot therefore becomes an important strategy. If the body drifts left, a wider or more leftward step can place support under the projected future centre of mass.

Hip abductors also control pelvic motion and influence lateral centre-of-mass acceleration.

Kai Kai connects foot placement and hip torque rather than assigning lateral balance to one mechanism.

81. Forward balance uses both braking and propulsion

After heel or foot contact, the stance limb can generate ground reaction forces that reduce forward momentum. Later in stance, ankle plantar-flexor and hip-related forces contribute to propulsion.

These phases are not simply opposite actions. Their timing shapes how smoothly the centre of mass moves from one support to the next.

Too much braking can waste energy and require more propulsion later; too little can make the next step mechanically difficult to recover.

Tricia sees gait economy and balance sharing the same force-time pattern.

82. Double support is a mechanically special phase because two feet contact the ground

During ordinary walking, brief intervals occur when both feet are in contact with the ground. The base of support then includes both contacts and the region between them.

Force can be transferred from the trailing limb to the leading limb while the centre of mass passes between supports.

Running differs because it contains flight phases with no foot contact rather than double support.

Alicia sees why walking and running are not merely the same motion at different speeds.

83. Running requires balance without continuous ground contact

During running, periods of flight separate stance phases. There is no centre of pressure on the ground while both feet are airborne.

Body orientation and limb configuration must therefore be managed through angular momentum, internal segment movements and prediction until the next contact.

Once the foot lands, large forces act over short times, and rapid muscle responses must control them.

Kai Kai adds intermittent contact to dynamic stability.

84. Arm swing helps manage angular momentum during gait

Leg movements generate angular momentum around the trunk. The arms swing in approximate opposition, helping reduce excessive trunk rotation and lowering the muscular effort needed to stabilise the upper body.

Arm swing also provides additional options for correcting perturbations because changing arm motion redistributes angular momentum.

It is therefore not merely decorative accompaniment to walking.

Tricia restores the upper limbs to a gait diagram often drawn from the hips downward.

85. Head stabilisation supports both vision and vestibular interpretation during walking

The trunk oscillates with each step, yet the head is often stabilised more strongly in space than a rigid body model would predict.

Neck muscles, vestibulocollic reflexes and predictive control reduce excessive head movement.

This helps keep visual information usable and prevents unnecessary vestibular noise from uncontrolled head motion.

Alicia sees the head as a sensor platform whose stability supports the rest of the control loop.

86. Gaze is scheduled ahead of the feet

Walking through clutter requires information about obstacles before the foot reaches them. People often fixate or sample relevant regions several steps ahead rather than staring continuously at the current foot.

The useful visual question depends on the environment: where is the path, how high is the step, where is the landing region, and what objects are moving?

Visual sampling therefore interacts with motor planning over several future steps.

Kai Kai calls gaze a look-ahead sensor for support planning.

87. Uneven terrain increases the value of prediction and rapid reweighting

On irregular ground, the actual surface under the foot may differ from the expected slope, stiffness or friction.

Vision can estimate terrain before contact, while plantar and ankle information update the estimate after contact. Vestibular signals reveal the resulting head and body motion.

The nervous system must combine prediction with rapid correction rather than trusting either one completely.

Tricia sees uncertainty entering the gait loop before every step.

88. A slip and a trip disturb balance through different mechanics

A slip changes the relationship between foot and ground because the support contact moves unexpectedly. A trip obstructs the swing foot and can arrest or redirect the limb while the centre of mass continues moving.

The corrective strategies therefore differ. A slip may require rapid adjustment of support forces and a compensatory step, while a trip can require elevating or repositioning the obstructed foot and creating a new landing point quickly.

The words describe mechanical events, not diagnoses.

Kai Kai asks what changed first: support motion or swing-limb motion.

89. Turning while walking changes the balance problem before the body reaches the new direction

A turn requires redirecting the velocity of the centre of mass and reorienting the feet, pelvis, trunk and head.

Anticipatory head and gaze movements can orient toward the new path before the trunk completes the turn. Foot placement then creates forces that redirect momentum.

A sharper or faster turn demands larger lateral and rotational accelerations.

Alicia sees turning as planned redirection rather than a straight walk bent around a corner.

90. Carrying objects changes balance by changing mass distribution, available arms and sensory priorities

A bag held on one side shifts the centre of mass and changes required joint torques. A large box can obstruct the view of the floor. Carrying a tray can reduce how freely the arms move to manage angular momentum or reach for support.

The same walking path therefore becomes a different control problem when an object is added.

Performance cannot be predicted from the person’s balance system alone without including the task.

Kai Kai restores the carried object to the mechanical model.

91. Dual-task walking changes the allocation of attention, not the laws of mechanics

Talking, calculating or searching visually while walking can change gait timing and variability because neural resources are being shared across tasks.

The ground reaction forces still obey mechanics, but the information available to select and adjust those forces can change.

A dual-task effect therefore does not prove that walking is normally under conscious control; it shows that attention can influence parts of the control process.

Tricia distinguishes automaticity from complete independence of cognition.

92. Gait variability contains both useful flexibility and noise

No two steps are exactly identical. Some variability reflects measurement noise or imperfect control, while some represents useful adjustment to terrain, task and internal state.

A system with zero variability would be unable to adapt to a changing environment. Excessive uncontrolled variability, however, can indicate reduced precision.

Interpreting variability therefore requires asking whether the changes are structured and task-appropriate.

Kai Kai refuses to label every deviation from an average as error.

93. Walking speed changes the balance solution

At higher speeds, momentum is larger and there is less time to process a disturbance before the next step. At very slow speeds, some passive dynamic efficiencies decrease and the person may spend longer in mechanically demanding transitional states.

The most stable or economical strategy therefore depends on speed.

Comparing balance measures across walking speeds without controlling speed can mix task effects with individual effects.

Alicia adds velocity to every gait comparison.

94. Step width and step length solve different components of stability

Step width primarily changes lateral support geometry. Step length influences forward support placement, braking and propulsion.

Increasing one dimension can help one stability problem while increasing energy cost or changing another mechanical demand.

A broad statement that wider steps are safer therefore needs a specific direction, task and population before it becomes meaningful.

Kai Kai replaces generic step size with vector geometry.

95. Dynamic balance is best understood as preserving future options

A state is recoverable when the person still has feasible actions capable of redirecting the centre of mass before mechanical limits are exceeded.

This perspective explains why a person can temporarily lean far beyond the static support boundary and still remain safe if a well-timed step is available.

It also explains why an obstacle can reduce stability without changing strength: it removes some future support options.

Alicia sees balance as maintaining a viable action set rather than merely holding one position.

Part XI. Learning, development, ageing and adaptation: balance is recalibrated throughout life

96. Balance improves through practice because prediction becomes more accurate

Repeated movement gives the nervous system opportunities to compare expected sensory consequences with what actually happened. If the predicted head motion, foot pressure or body trajectory differs from observation, future commands can be adjusted.

This learning can reduce unnecessary co-contraction, improve timing and make sensory weighting more appropriate for the task.

Improvement therefore does not require the vestibular organs themselves to become physically larger or stronger.

Kai Kai separates better sensors from better interpretation and control.

97. Familiar environments allow the controller to use stronger priors

A known staircase, familiar sports court or repeated laboratory task contains predictable geometry and timing.

Prediction can therefore narrow the range of plausible future states before each movement begins.

When the environment changes unexpectedly, those priors can become wrong and require rapid revision.

Alicia sees familiarity as information stored from past interaction.

98. Novel movement initially produces larger errors because the internal model is uncertain

When a person learns a new dance step, skating movement or balance task, the relationship between motor command and sensory outcome is not yet well calibrated.

Early movements may therefore be stiff, slow or heavily co-contracted.

Practice allows the system to discover how much torque, foot placement and body motion are required.

Tricia sees awkwardness as part of model identification rather than simply lack of effort.

99. Adaptation can persist temporarily after the environment returns to normal

If repeated exposure teaches the controller to compensate for a changed sensory or mechanical relationship, that compensation may continue briefly when the change is removed.

The resulting after-effect is evidence that the controller itself changed rather than merely reacting moment by moment.

After-effects are common tools in motor-learning research because they reveal hidden recalibration.

Kai Kai calls them fingerprints of adaptation.

100. Development changes both the body being controlled and the controller

Infants and children grow rapidly. Limb lengths, head-to-body proportions, muscle strength and joint geometry change while sensory and neural systems mature.

A movement calibrated last year is therefore being performed by a different mechanical plant this year.

Motor development requires continuous recalibration to changing body dimensions as well as learning about the environment.

Alicia sees balance development as moving-target control.

101. Infants acquire postural control before they acquire adult-like walking

Head control, sitting, standing and walking emerge in stages because each skill requires sufficient muscle capacity, sensory integration and mechanical coordination.

Early postural responses may rely differently on vision and support information than mature adult responses.

Development therefore changes weighting and strategy rather than simply scaling up an adult system.

Tricia avoids treating a child as a smaller adult.

102. Childhood movement builds sensorimotor experience across many environments

Running, climbing, turning, catching and navigating uneven surfaces provide repeated samples of how body commands interact with gravity and support.

The value is not one exercise-specific balance score but the accumulation of varied state-action relationships.

This diversity helps build flexible prediction rather than one narrowly tuned solution.

Kai Kai adds experience distribution to motor development.

103. Ageing changes balance through several systems at once

With age, vestibular hair-cell and nerve populations can decline, visual contrast sensitivity can change, peripheral sensory function can change, muscle power can fall, joints can stiffen and central processing can slow.

No single change explains every age-related balance difference.

Age also accumulates experience and learned compensations, so older adults are not simply younger adults with weaker sensors.

Alicia sees ageing as a multi-system shift with both losses and adaptations.

104. Muscle power matters because some corrections must occur quickly

A person may possess enough maximal strength to support body weight but still struggle with a rapid recovery if force cannot be generated quickly enough.

Rate of force development, joint range and foot-placement speed therefore matter to reactive balance.

The How Muscles Work article owns the underlying contractile and neural mechanisms.

Kai Kai adds time to the concept of strength.

105. Joint mobility affects strategy availability

Limited ankle range can reduce how far centre of pressure can be shifted while the foot remains flat. Limited hip or trunk mobility can constrain alternative strategies.

The nervous system may compensate by stepping earlier or redistributing movement to neighbouring joints.

The How Joints Work article owns the broader movement and tissue mechanics.

Tricia sees mechanical range as part of the controller’s action space.

106. Sensory loss can be partly compensated when other cues remain informative

If one sensory stream becomes less reliable, the nervous system can increase reliance on other available information.

This compensation is not unlimited. Vision cannot replace rapid vestibular information perfectly during every head movement, and vestibular signals cannot identify every obstacle on the floor.

Redundancy therefore increases robustness without creating perfect substitution.

Kai Kai calls compensation task-dependent rather than universal.

107. Practice can improve balance without reducing all sway

A trained person may still show normal exploratory sway during quiet stance while improving recovery speed, step placement or task accuracy.

Reducing one sway metric is therefore not the only sign of better control.

Training effects should be measured against the task that was practised and the mechanism expected to change.

Alicia sees why one laboratory number cannot define all improvement.

108. Transfer from one balance task to another is incomplete

Standing on one support, walking, landing from a jump and stabilising the head during running share principles but require different mechanics and sensory timing.

Improvement in one task can transfer partly through shared capacities, yet task-specific prediction and coordination remain important.

This is why a training effect should not be generalised automatically from one apparatus to everyday balance.

Tricia asks what components are actually shared before assuming transfer.

109. Fatigue changes balance by altering both motor capacity and sensory interpretation

Fatigued muscles generate force differently, motor-unit recruitment changes, proprioceptive signals can shift and attention may decline.

The controller may compensate through wider steps, greater co-contraction or slower movement.

A final sway or gait measure therefore reflects both the fatigue itself and the compensation.

Kai Kai separates disturbance from adaptive response.

110. Sleep and alertness influence balance through central processing

Reduced alertness can slow reaction, alter attention and affect the precision of movement prediction.

The vestibular organs may still transduce head motion normally while the central system uses the information differently.

Balance performance therefore depends on current brain state as well as peripheral sensor integrity.

Alicia adds state of arousal to the control model.

111. Anxiety can change postural strategy without creating a new law of physics

At height or under perceived threat, people often stiffen, reduce movement amplitude and increase co-contraction.

The mechanical environment may be unchanged, but the cost assigned to a potential fall has increased.

Control policy therefore depends partly on perceived consequence and uncertainty.

Clinical anxiety disorders belong to Medicine. The healthy mechanism shows that task value can alter movement strategy.

112. Balance adaptation is strongest when the nervous system can detect a useful error signal

Learning requires a difference between expected and observed outcome that can be attributed to a changeable command or model.

If feedback is too noisy or the environment changes randomly from trial to trial, the controller may not identify a stable relationship worth learning.

Consistent errors are therefore easier to adapt to than completely unpredictable ones.

Kai Kai adds learnability to disturbance design.

113. Overlearning one misleading relationship can create short-term after-effects

A controller that becomes highly tuned to a shifted visual or mechanical environment may initially apply the learned correction when normal conditions return.

This is not failure of learning. It is evidence that the internal model was updated appropriately for the prior environment.

Good adaptation includes the ability to recalibrate again when the world changes.

Tricia sees flexibility as repeated model revision rather than one final perfect setting.

114. The balance system learns across multiple timescales

Fast adaptation can alter a response within minutes. Consolidation across sleep and repeated sessions can stabilise learning over longer intervals. Development and ageing alter the system across years.

A measured change therefore needs a timescale before it can be interpreted.

A transient familiarisation effect, a day-to-day learning effect and a long-term structural adaptation are different phenomena.

Alicia adds clocks to the learning map.

115. The most durable skill is not one perfect posture but the ability to update

Real environments change. Shoes differ, surfaces move, lighting varies, loads are carried and other people create unpredictable motion.

A controller that performs superbly only under one fixed condition is less robust than one that can identify changed conditions and select a new strategy.

Balance therefore depends on adaptability as much as on performance in one steady test.

Kai Kai closes the adaptation section with one question: how quickly can the system learn that the old model no longer fits?

Part XII. The evidence: what balance measurements actually measure

116. A force plate measures external reaction forces, not a direct neural balance signal

A force plate records forces and moments exchanged between the body and the support surface.

From these measurements, centre of pressure can be calculated under defined assumptions.

The plate does not measure vestibular firing, visual confidence or muscle commands directly.

Kai Kai labels force-plate data mechanical output from the whole loop.

117. Centre-of-pressure path is not the same as centre-of-mass path

Centre of pressure is the point of application of the resultant ground reaction force. Centre of mass describes the body’s mass distribution.

During quiet standing, the two are related dynamically, but they are not identical trajectories.

Centre of pressure often moves more rapidly because it is used to accelerate centre of mass.

Tricia separates control action from controlled state.

118. More sway is not automatically worse balance

Sway amplitude depends on stance width, visual conditions, support stiffness, task instructions, age and strategy.

A person can intentionally explore a larger range without losing control, while another can show small sway through stiff co-contraction that may be metabolically costly.

A sway measure becomes interpretable only after the condition and task are defined.

Alicia refuses to rank balance quality from one sway number alone.

119. Romberg-type comparisons manipulate available sensory information but do not isolate one organ perfectly

Comparing standing with eyes open and closed changes visual availability while leaving vestibular and somatosensory systems active.

The difference in performance reflects how the whole system responds to removing vision, not a pure direct test of vestibular function.

Clinical interpretation belongs to trained professionals.

Kai Kai uses the comparison as an example of a perturbation to information rather than an organ-specific meter.

120. Dynamic posturography changes support or visual reference conditions to probe sensory reweighting

Computerised systems can move a visual surround, tilt or translate a platform and measure resulting sway or forces.

The method challenges assumptions connecting visual and somatosensory cues to stable world references.

Observed performance depends on sensory integration, motor capacity and task understanding.

Tricia sees a systems test rather than a single-sensor test.

121. Video head-impulse testing measures eye responses during rapid head rotations

High-speed cameras track eye motion while brief controlled head impulses stimulate semicircular canals.

The relation between head and eye velocity provides information about vestibulo-ocular performance in selected canal planes.

Goggle slippage, calibration and head-impulse quality affect results.

Clinical interpretation belongs to ENT, Neurology and vestibular specialists. The measurement principle is head motion in, eye compensation out.

122. Caloric testing stimulates the horizontal canal system through temperature-driven fluid effects

Warm or cool stimulation near the ear changes fluid density and creates vestibular activation under specific head positions.

The resulting eye movements probe a low-frequency aspect of horizontal canal pathways.

This artificial stimulus is not equivalent to everyday head rotation and should not be treated as a universal measure of vestibular performance.

Kai Kai adds stimulus frequency and mechanism before comparing tests.

123. Rotational-chair testing probes canal responses over controlled frequencies

A motorised chair can rotate a seated person using known velocity and acceleration waveforms while eye movements are measured.

Varying frequency helps characterise vestibulo-ocular dynamics across timescales.

The method measures a defined subsystem in a controlled environment rather than whole-body standing balance.

Alicia sees one instrument isolating a narrower transfer function.

124. Vestibular-evoked myogenic potentials probe otolith-related pathways indirectly

Sound or vibration can stimulate vestibular receptors and evoke short-latency muscle responses measured by surface electrodes.

Cervical and ocular VEMPs emphasise different pathways and receptor contributions.

The amplitude depends on muscle activation, electrode placement and stimulus conditions as well as vestibular function.

Tricia sees why an electrical waveform requires a full measurement chain before interpretation.

125. Eye tracking measures gaze behaviour, not the intention behind every fixation

Eye trackers estimate where the eyes are directed and how gaze changes over time.

Fixations can reveal which regions were sampled visually, but they do not prove exactly what information the person extracted or consciously attended to.

Head movement and calibration affect the estimate.

Kai Kai separates where the eye pointed from what the brain concluded.

126. Motion capture measures segment motion while depending on the marker-to-bone relationship

Markers placed on skin can estimate body-segment positions, but skin moves relative to underlying bones.

Small joint translations are therefore difficult to infer precisely from skin markers alone.

Whole-body centre-of-mass estimates also depend on assumptions about segment mass distribution.

Alicia adds model assumptions beneath every smooth three-dimensional animation.

127. Inertial measurement units estimate motion from accelerometers and gyroscopes

Wearable IMUs measure angular velocity and specific force, often combined with magnetometer or other information.

Integrating these signals to estimate orientation and position introduces drift unless corrected by constraints or external references.

The device therefore faces a mathematical version of the same state-estimation problem as the nervous system.

Kai Kai likes the comparison because the sensor package cannot escape reference-frame ambiguity either.

128. Plantar-pressure systems map contact distribution rather than whole-body stability

Pressure mats and instrumented insoles measure normal pressure under the feet and can estimate centre-of-pressure trajectories.

They reveal where contact load is distributed but do not measure shear perfectly unless additional sensors are used.

They also cannot tell whether a pressure pattern arose from voluntary strategy, passive anatomy or external disturbance without other evidence.

Tricia treats the foot map as one layer of the control loop.

129. Electromyography shows muscle electrical activity but not joint torque directly

Surface EMG records electrical activity from nearby active muscle fibres.

Amplitude depends on recruitment, firing, electrode geometry and tissue properties.

Torque additionally depends on muscle force, moment arms, contraction mode and other muscles.

The muscle owner explains this distinction in depth.

130. Reaction-time measures mix sensing, decision and movement initiation

Time from an external cue to a visible movement includes stimulus detection, neural processing, motor preparation, nerve conduction and muscle force development.

A slower response therefore cannot be assigned automatically to sensory delay or muscle weakness.

More specific experiments are needed to localise the delay.

Kai Kai treats total reaction time as a chain, not a single process.

131. A standing test and a walking test answer different balance questions

Standing constrains support and places emphasis on in-place control. Walking requires prediction, swing-foot placement and repeated transitions between supports.

A person can perform well in one and less well in the other because the action space differs.

Tests should therefore be selected according to the mechanism or real-world task under investigation.

Alicia sees why one universal balance score is unlikely to capture every domain.

132. Test repetition can improve scores through familiarisation

Once a participant understands the task and anticipates the perturbation, strategy can change.

Repeated testing can therefore measure both the underlying capacity and learning of the test itself.

Randomisation, practice trials or explicit familiarisation periods help separate these effects depending on the study design.

Tricia adds learning to the measurement protocol.

133. A composite score can hide which subsystem changed

Combining several measures into one number can simplify communication but removes detail.

The same composite score can arise from different combinations of sway, stepping, vision use and muscle response.

Composite measures are useful when validated for a clear purpose, but mechanism questions require returning to the component data.

Kai Kai asks what information was compressed before trusting a summary score.

134. Measurement error can be smaller than biological variability and still matter

A force plate may be highly precise while the person’s performance changes naturally from trial to trial.

Conversely, a noisy sensor can obscure a real biological effect.

Good experiments therefore quantify instrument reliability and repeated-measure variability separately.

Alicia keeps technical noise and human variability in different columns.

135. The strongest balance evidence combines mechanics, sensory perturbation and behaviour

A force trace can show what happened mechanically. A controlled sensory perturbation can test which information altered the response. Motion capture can show segment strategy. Eye measurements can show gaze stabilisation. EMG can show muscle timing.

When these independent observations support the same causal chain, confidence increases.

When they disagree, the disagreement often reveals that a missing layer was being assumed.

Kai Kai closes the evidence section with triangulation rather than one preferred machine.

Part XIII. The reasoning laboratory: separate sensation, mechanics, prediction and action

These are fictional learning cases. They are not home balance tests, diagnostic thresholds or instructions to provoke dizziness. The point is to identify which hidden variable could change the same visible outcome.

136. Equal sway can conceal different control effort

The question. Model A and Model B show the same centre-of-pressure excursion during quiet standing. A uses low co-contraction and flexible ankle corrections. B uses strong co-contraction that makes the body mechanically stiffer. Are their control states equivalent?

No. The external sway measure matches while muscle activation, energy cost and joint stiffness differ.

The repair. Similar kinematics do not guarantee similar control effort.

137. Equal muscle strength can produce different recovery when reaction timing differs

The question. Two fictional people can generate the same maximal ankle torque. One begins an appropriate correction quickly after a disturbance; the other begins later. Must recovery be equally successful?

No. A delayed force can arrive after centre-of-mass velocity has increased beyond the range that the same in-place torque can recover.

The repair. Capacity and timing are separate dimensions.

138. Equal vestibular input can produce different posture when support geometry differs

The question. Two models receive identical head-motion signals. A stands with a wide support region; B on a narrow support region. Must body motion match?

No. The available centre-of-pressure range and feasible corrective moments differ.

The repair. Sensory evidence enters a controller whose mechanical action space depends on the environment.

139. Equal visual acuity can coexist with different visual usefulness for balance

The question. Two observers can read the same eye-chart line. A sees a well-lit stable room; B sees a dim floor with moving projections. Must visual contribution to balance be equal?

No. Acuity describes one visual property. Contrast, reference stability, peripheral motion and task relevance differ.

The repair. A sensory system’s usefulness depends on the information available in the scene.

140. Equal ankle angle can represent different body orientation

The question. The ankle joint is at the same angle in two diagrams. In A, the support is level. In B, the support is tilted. Must the trunk be equally oriented relative to gravity?

No. The same joint configuration can exist in different world coordinates.

The repair. Local proprioception needs a reference-frame transformation before it specifies whole-body orientation.

141. Equal retinal motion can arise from eye, head, body or world movement

The question. A vertical edge moves left across the retina at the same rate in two trials. In A, the observer turns the head right. In B, the edge itself moves left while the observer remains still. Must the balance interpretation match?

No. The retinal observation is compatible with several physical causes.

The repair. Combine eye-movement, vestibular and scene information before attributing retinal motion to self-motion.

142. Equal centre-of-pressure speed can accompany different centre-of-mass trajectories

The question. Two force-plate records have similar average centre-of-pressure speed. In A, rapid pressure adjustments keep centre of mass tightly controlled. In B, pressure follows a larger body sway. Are postural states equivalent?

No. Centre of pressure is a control action interacting with body dynamics.

The repair. Interpret centre of pressure together with centre-of-mass or segment motion when the mechanism requires it.

143. Equal step width can provide different stability when momentum differs

The question. Two lateral recovery steps land at the same width. A begins from low sideways velocity; B from high sideways velocity. Must the same foot placement be sufficient?

No. Higher momentum requires a different future support and force history.

The repair. Foot position without centre-of-mass velocity is incomplete.

144. Equal disturbance size can require different strategies depending on prediction

The question. The same forward perturbation occurs twice. In A, its timing is known. In B, it is unexpected. Must muscle timing match?

No. The predictable event permits anticipatory postural adjustment; the unexpected event relies more heavily on feedback.

The repair. Disturbance magnitude and information about the disturbance are different variables.

145. Equal VOR gain can conceal different visual stability for near and far targets

The question. Two simplified trials report the same eye-velocity-to-head-velocity ratio. One uses a distant target and one a near target during head translation. Must retinal stability be equal?

No. Required eye rotation depends on viewing geometry.

The repair. A dimensionless gain still needs target distance and movement type.

146. Equal vestibular firing can have different meaning during voluntary and imposed movement

The question. Similar head acceleration occurs when Alicia turns voluntarily and when an external device moves the head unexpectedly. Must central interpretation be identical?

No. The voluntary movement has an associated motor command and sensory prediction.

The repair. Physical sensory input and predicted sensory consequence can be compared centrally.

147. Equal visual sway response can reflect different underlying cue conflict

The question. Two people sway similarly when a visual surround moves. One is on a firm platform; the other on a compliant platform. Must the visual weight be identical?

No. Somatosensory reliability differs, so the same visible response can emerge from different combinations of cue weighting and mechanics.

The repair. Manipulations interact; one response does not reveal a unique sensory weight.

148. Equal standing performance can conceal different stepping reserve

The question. Two fictional models stand equally well in quiet conditions. A can move the swing foot rapidly; B has limited step speed. Must their recovery from a large disturbance be equal?

No. Quiet standing tests in-place control, while large disturbances may require changing support.

The repair. Measure the action that the real task requires.

149. Equal foot pressure patterns can conceal different upper-body configurations

The question. Two plantar-pressure maps look similar. In A, the trunk is upright. In B, the trunk leans while the hips compensate. Must whole-body posture match?

No. Several segment configurations can produce similar support forces.

The repair. Contact information does not uniquely specify internal body geometry.

150. Equal dizziness reports can have different physiological causes

The question. Two people use the same word dizzy. Must vestibular physiology be the same?

No. The word can refer to spinning, light-headedness, imbalance, visual motion sensitivity or other experiences.

Clinical differentiation belongs to health professionals.

The repair. Symptom labels are not mechanisms.

151. Equal fall outcome can conceal very different recovery attempts

The question. Two model perturbations both end with the body outside the original support. In A, a step was attempted but blocked by an obstacle. In B, no step was initiated. Are failures equivalent?

No. One failure arose from environmental action constraints; the other from a different control sequence.

The repair. Final outcome does not identify the failed layer.

152. Equal gait speed can conceal different stability margins

The question. Two walkers move at the same average speed. A takes regular well-placed steps; B uses variable emergency corrections. Must dynamic stability be equal?

No. Average speed compresses the temporal structure of stepping.

The repair. Inspect step timing, placement, variability and available recovery options.

153. Equal reaction time can arise from different internal delays

The question. Two people begin moving 250 milliseconds after a visual cue. In A, visual detection is fast but muscle force rises slowly. In B, detection is slower but force rises quickly. Are mechanisms equal?

No. Total reaction time is the sum of several processing and mechanical stages.

The repair. Aggregate latency does not localise the delay.

154. Equal head motion can create different sensory conflict when the visual scene differs

The question. Identical head rotation occurs in darkness and in a bright stable room. Must the state estimate be identical?

No. Visual information is absent in one case and available in the other.

The repair. The same vestibular signal is interpreted inside a changing multisensory context.

155. Equal force-plate sway can conceal different body-segment coordination

The question. Two people create similar centre-of-pressure trajectories. A sways mainly at the ankles; B uses substantial hip and trunk motion. Are strategies equivalent?

No. Similar external force patterns can arise from different internal segment coordination.

The repair. Add kinematics when strategy is the question.

156. A failure map separates sensing, estimation, action and environment

This is a healthy-mechanism learning map, not a diagnostic table. Real clinical problems can involve several layers simultaneously.

157. Thirty balance misconceptions that collapse under a mechanism-first model

1. “Balance comes from the inner ear.” Vestibular sensing is essential, but vision, proprioception, skin, mechanics and neural prediction also contribute.
2. “Good balance means no sway.” Quiet standing normally contains continual controlled sway.
3. “The centre of mass must always stay over one foot.” Dynamic balance can move it beyond current support while creating future support.
4. “Centre of pressure and centre of mass are the same point.” One describes resultant support force; the other describes mass distribution.
5. “A larger base of support guarantees success.” Momentum, friction, muscle capacity and strategy still matter.
6. “Vision tells the brain where the body is.” Visual motion is ambiguous and must be interpreted relative to eye, head and world movement.
7. “Closing the eyes tests the vestibular system alone.” Somatosensation and motor control remain active.
8. “The semicircular canals sense absolute orientation.” They respond primarily to rotational dynamics.
9. “The otolith organs distinguish gravity from translation by themselves.” Their signal contains both and requires multisensory interpretation.
10. “Vestibular afferents are silent at rest.” They maintain tonic baseline firing.
11. “The VOR is a fixed reflex.” Its gain can adapt to persistent visual error and task geometry.
12. “Proprioception is one sense located in joints.” Muscle, tendon, skin and joint signals all contribute.
13. “A soft support only makes muscles work harder.” It also changes sensory reference reliability and support mechanics.
14. “Light touch helps because the hand carries body weight.” Very light contact can also provide useful spatial information.
15. “Stronger postural corrections are always better.” Excess gain with delay can create oscillation.
16. “An ankle strategy is the correct response to every small disturbance.” Strategy depends on support, task and available joints.
17. “Taking a step means balance failed.” A step can be the successful strategy that restores support.
18. “Recovery depends only on strength.” Timing, state estimation and action availability matter.
19. “Walking balance is quiet standing on one leg.” Gait includes prediction, swing-foot placement and repeated support transitions.
20. “Running is walking faster.” Running introduces flight phases and different contact timing.
21. “A fall reveals one failed sensor.” Falls are outcomes of a whole system and environment.
22. “A normal scan proves normal balance.” Structural imaging does not measure multisensory control or motor timing.
23. “A force plate directly measures vestibular function.” It measures external forces generated by the entire control loop.
24. “More sway always means worse control.” Strategy, task and exploration alter sway.
25. “Balance training transfers completely to every activity.” Transfer depends on shared mechanics and control demands.
26. “Ageing changes only the inner ear.” Vision, sensation, muscle, joints and central processing also change.
27. “Dizziness means a balance-organ problem.” The word describes several possible experiences and requires clinical context.
28. “A familiar task tests raw balance capacity.” Familiarity contributes prediction and task-specific learning.
29. “One composite balance score explains the mechanism.” Different internal states can produce the same score.
30. “Healthy balance is never being unstable.” Healthy movement often uses temporary instability that remains recoverable.

158. Frequently asked questions about how balance works

What part of the body controls balance?

No single part does. Inner-ear sensors, vision, muscle and skin sensation, spinal circuits, brainstem, cerebellum, cortex and the musculoskeletal system work together.

What does the vestibular system actually measure?

Semicircular canals provide information about head rotation, while utricle and saccule provide information about gravito-inertial acceleration, including gravity-related tilt and linear acceleration.

Why do we sway when standing still?

The body is mechanically unstable, sensory signals contain noise and delay, muscles produce variable force, and breathing and circulation create small disturbances. The controller continually corrects these effects.

What is centre of mass?

It is the weighted average location of the body’s distributed mass. It changes as body segments and carried objects move.

What is centre of pressure?

It is the point at which the resultant ground reaction force acts over a support surface. The nervous system changes it through muscular forces and foot pressure.

Why does closing the eyes make some balance tasks harder?

Removing vision eliminates one source of orientation and motion information. The system must rely more on vestibular and somatosensory evidence.

Why is standing on foam different from standing on a hard floor?

Foam deforms, altering both the mechanical support and the relationship between ankle or plantar signals and body orientation.

Why can touching a wall lightly help?

Small contact changes provide information about relative movement even when the hand is not carrying much body weight.

How do the semicircular canals detect head rotation?

Head rotation creates relative motion between the canal wall and endolymph, deflecting the cupula and changing vestibular hair-cell activity.

How do otolith organs detect gravity?

Dense otoconia load a gelatinous membrane over hair cells. Gravity and linear acceleration shear the membrane and deflect hair bundles.

Can the inner ear tell whether acceleration comes from gravity or movement?

Not from the otolith signal alone. The nervous system combines otolith information with canals, vision, body sensation and prediction.

Why do the eyes move when the head moves?

The vestibulo-ocular reflex rotates the eyes in a compensatory direction to reduce retinal image slip during head movement.

What is sensory reweighting?

It is a modelling framework in which the nervous system changes how strongly it relies on sensory sources according to their current reliability and task relevance.

Is taking a step a sign of poor balance?

Not necessarily. When in-place corrections are insufficient, stepping changes the base of support and can be an effective recovery response.

Why do people use their arms when they nearly lose balance?

Arm movement can redistribute angular momentum, change centre-of-mass position and prepare to create a new external support through reaching or grasping.

How does walking depend on balance?

Walking requires repeated prediction of future support, control of centre-of-mass momentum, swing-foot placement and correction of unexpected changes.

Does stronger muscle always improve balance?

Strength and power expand the available action set, but accurate sensing, timing, prediction and strategy remain necessary.

Does good eyesight guarantee good balance?

No. Visual contribution depends on the scene, movement, contrast, target geometry and integration with vestibular and somatosensory information.

What does a force plate tell us about balance?

It measures ground reaction forces and moments and can calculate centre of pressure. It does not directly measure sensory weighting or vestibular firing.

Can balance improve with practice?

Yes. Practice can improve prediction, sensory interpretation, strategy selection and muscle coordination. Transfer to unpractised tasks depends on how much of the control problem is shared.

Why can balance feel worse when tired?

Fatigue can reduce force speed, alter proprioceptive input, change attention and encourage different movement strategies.

What is dynamic balance?

It is control of the moving body when support, momentum and available actions change through time. Walking and running are examples.

Is dizziness the same as imbalance?

No. Dizziness is an imprecise symptom word that can describe several sensations; imbalance refers more specifically to difficulty controlling posture or movement. Clinical interpretation requires professional assessment.

159. A glossary for balance-system thinking

Base of support: the support region through which external reaction forces can act. Centre of mass: weighted average location of body mass. Centre of pressure: location of the resultant ground reaction force over a support surface. Dynamic balance: control of posture and movement while support and momentum change.

Endolymph: fluid inside the membranous labyrinth of the inner ear. Hair cell: mechanosensory receptor cell whose bundle deflection changes transmitter release. Otolith organ: utricle or saccule, containing weighted sensory membranes that respond to gravito-inertial acceleration. Otoconia: calcium-carbonate crystals adding mass to an otolithic membrane.

Semicircular canal: vestibular sensor structure specialised for rotational dynamics. Cupula: gelatinous structure in a canal ampulla deflected by endolymph motion. Macula: sensory epithelium of utricle or saccule. Striola: curved region around which otolith hair-cell polarity changes.

Vestibulo-ocular reflex: rapid eye response that compensates for head motion. Optic flow: structured retinal motion produced by observer and scene movement. Proprioception: information about body configuration and movement from muscle, tendon, skin and joint-related sensors. Sensory reweighting: adaptive change in reliance on sensory evidence.

Anticipatory postural adjustment: preparatory muscle activity organised before an expected self-generated or predictable disturbance. Reactive postural response: correction triggered after an unexpected disturbance is detected. Postural sway: natural movement during standing. Recovery step: step that relocates the base of support to regain control.

State estimation: inference about hidden body and environmental variables from sensory evidence and prediction. Reference frame: coordinate system relative to which position or movement is described. Gain: ratio between output and input in a defined system. Feedback delay: time between a state change, its measurement and the resulting corrective action.

Inverse dynamics: calculation of net joint moments from motion and external forces. Force plate: instrument measuring ground reaction forces and moments. IMU: inertial measurement unit containing accelerometers and gyroscopes. Posturography: measurement and analysis of posture and sway under controlled conditions.

160. The one-page causal chain: from head motion and support forces to a corrected step

1. The body occupies a state defined by segment positions, velocities and available support.
2. Gravity, movement and external disturbances alter that state.
3. Semicircular canals transduce head rotational dynamics.
4. Otolith organs transduce gravito-inertial acceleration.
5. Vision provides scene motion, orientation and future-path information.
6. Muscle, tendon, joint and skin receptors describe local body configuration and support interaction.
7. The nervous system transforms those signals into compatible reference frames.
8. Prior motor commands provide predictions about expected sensory consequences.
9. Sensory reliability and task context alter how evidence is weighted.
10. The brain estimates current position, velocity, orientation and uncertainty.
11. A control policy selects in-place torque, trunk motion, stepping, reaching or a combination.
12. Motor commands recruit muscles and change joint torque and stiffness.
13. Ground reaction forces move centre of pressure and accelerate centre of mass.
14. If the current support cannot recover the state, a foot or hand creates a new support.
15. The movement produces new vestibular, visual and proprioceptive evidence.
16. The controller compares predicted and observed consequences and updates the next command.
17. Repeated errors drive adaptation across minutes, days and longer learning periods.

161. A final checklist for any unfamiliar balance question

1. Define the task. Quiet standing, walking, running, turning, reaching or recovery?
2. Define the controlled variable. Gaze, centre of mass, foot placement, joint angle or subjective orientation?
3. Name the support geometry and friction.
4. Include centre-of-mass velocity, not position alone.
5. Separate centre of pressure from centre of mass.
6. For vestibular input, distinguish canals from otoliths.
7. For otolith signals, preserve the tilt-translation ambiguity.
8. For vision, identify which parts of the scene are stable references.
9. For proprioception, specify the local reference frame.
10. Ask whether sensory cues are independent or share the same error source.
11. Separate reliability from systematic bias.
12. Separate prediction from feedback.
13. For recovery, identify which actions remain physically feasible.
14. Include muscle power and response timing when rapid action is required.
15. For walking, include foot-placement timing and momentum.
16. For a force-plate result, do not infer one sensory organ directly.
17. For a composite score, inspect the compressed components.
18. For an adaptation result, distinguish immediate transition from settled behaviour.
19. Test at least one confusable alternative explanation.
20. Keep healthy physiology separate from diagnosis and treatment.

162. Where this article stops

This article owns the healthy whole-balance mechanism: mechanics of support, semicircular canals and otoliths, vestibulo-ocular stabilisation, visual self-motion information, proprioception, skin contact, multisensory state estimation, sensory reweighting, anticipatory and reactive postural strategies, walking, adaptation and measurement logic.

It does not diagnose or treat vertigo, vestibular neuritis, benign positional vertigo, Ménière disease, stroke, neuropathy, visual disease, syncope, falls syndromes, cerebellar disease or rehabilitation problems. Those belong to Medicine, ENT, Neurology, Ophthalmology, Cardiology, Geriatrics, Rehabilitation and other appropriate clinical owners. Veterinary balance disorders remain separately owned.

The eduKateSingapore Vestibular Hair Cell Learning Manual retains the cellular mechanotransduction job. The eduKateSengkang How to Learn the Vestibular System, Balance and Spatial Orientation route retains its learning-and-teaching job. This eduKateSG article remains the broad world-facing synthesis.

163. Further reading and evidence trail

• NIDCD — Balance Disorders: broad overview of vestibular and multisensory balance function.
• OpenStax — Equilibrium: semicircular canals, otolith organs and vestibular pathways.
• Peterka — Sensorimotor Integration in Human Postural Control: experimental and modelling evidence for adaptive sensory weighting.
• Peterka and Loughlin — Dynamic Regulation of Sensorimotor Integration: transition and reweighting evidence.
• Jeka and Lackner — Fingertip Contact Influences Human Postural Control: evidence that light contact can provide useful postural information.
• Lee and Aronson — Visual Proprioceptive Control of Standing in Human Infants: classic moving-visual-surround evidence.
• Horak and Nashner — Central Programming of Postural Movements: altered-support evidence for flexible postural strategies.
• Cordo and Nashner — Properties of Postural Adjustments Associated with Rapid Arm Movements: anticipatory coordination evidence.
• How Vision Works: visual owner.
• How the Brain Works: neural prediction and network owner.
• How Muscles Work: motor-unit, force and adaptation owner.
• How Joints Work: joint mechanics and proprioceptive interface owner.
• How Skin Works: cutaneous mechanosensation owner.

164. The return path: balance is the art of remaining recoverable

Alicia began by standing on one foot and defining success as not falling. Tricia added the inner ear. Kai Kai kept finding missing variables. The inner ear needed vision and proprioception. Sensory evidence needed reference frames. Reference frames needed a model of the body. A model needed predictions and uncertainty. Predictions needed muscles capable of acting in time. Muscle force needed a support surface capable of supplying reaction forces. And when the existing support was no longer enough, balance needed the option to step or reach and thereby change the mechanical problem itself.

This is why the word balance can be misleading when it suggests a static object perfectly centred over a base. Living balance is dynamic. It permits sway, anticipates disturbance, tolerates uncertainty, changes strategy and deliberately moves beyond the current support when the next support can be created safely.

The deepest mechanism is therefore not “the inner ear keeps you upright.” Balance is a closed-loop state-estimation and action system that keeps future recovery possible. Sensors estimate what is happening, prediction estimates what will happen next, muscles and joints change the forces acting on the body, and the environment determines which actions are actually available.

Continue through How Vision Works, How the Brain Works, How Muscles Work, How Joints Work, How Skin Works, or return to the How X Works | eduKateSG library.

Part XIV. Model limits and transfer: what a balance explanation must not pretend to know

165. A single balance model is useful only inside the conditions it represents

The inverted-pendulum model is excellent for understanding small quiet-standing sway, yet it omits knees, hips, arms and stepping. A multi-link model adds segment motion but still simplifies muscle recruitment, soft-tissue deformation and neural processing. A sensory-weighting model can explain why cue reliability matters while leaving the detailed cellular circuitry abstract.

These models are not competitors in which one must defeat the others. They operate at different resolutions. The important discipline is to match the model to the question and to state which effects have been omitted.

Kai Kai adds a model boundary before every conclusion.

166. A useful model explains what would change when one variable changes

Suppose a model says wider support should increase the range over which centre of pressure can move before stepping becomes necessary. That claim can be tested by changing support width while keeping other conditions as controlled as possible.

If a model says visual reliability matters, it should predict different responses when the scene becomes less informative even though the floor remains mechanically unchanged.

A mechanistic explanation therefore earns value by generating discriminating predictions, not by naming many anatomical structures.

Tricia asks what observation would make the explanation less plausible.

167. Balance has no one universal normal value independent of task

Standing with feet apart, tandem stance, walking, running and turning require different support geometries and control strategies. A centre-of-pressure path that is ordinary in one condition may be impossible or irrelevant in another.

Age, body dimensions, footwear, carried objects and environment alter the same measurements further.

A value becomes meaningful only when the protocol, population and purpose are defined.

Alicia sees why reference data must be matched to the measurement conditions rather than borrowed from any convenient study.

168. Better balance can mean different things depending on the goal

For a laboratory standing task, improvement might mean smaller sway under a defined perturbation. For walking, it might mean more accurate foot placement or faster recovery. For a dancer, it might mean maintaining a chosen body orientation while moving rapidly. For an older adult, the clinically relevant goal might involve safe everyday mobility rather than minimising one instrumented variable.

These goals overlap but are not identical.

Mechanism-first writing therefore avoids presenting one metric as a universal definition of improvement.

Kai Kai asks which real-world failure the metric is supposed to prevent or reduce.

169. Prediction and feedback should be treated as partners rather than rival theories

Prediction allows the nervous system to prepare for expected consequences before delayed sensory feedback arrives. Feedback detects mismatches and corrects prediction errors.

A system using prediction alone would drift when the world changed. A system using feedback alone would react too late to many rapid events.

Healthy balance combines both, with their relative importance changing according to task predictability and movement speed.

Tricia replaces the question “Is balance feedforward or feedback?” with “Which part of the event is predictable, and which part still needs correction?”

170. Stability is not the absence of error but the ability to keep error recoverable

Small state-estimation errors are unavoidable. Muscle force fluctuates. Surfaces deform. Visual scenes contain motion. Yet these errors need not cause failure if the controller preserves enough reserve to correct them.

A system can therefore be robust while being imperfect at every instant.

This is the deepest engineering principle inside balance physiology: performance depends on how uncertainty is managed, not on eliminating uncertainty completely.

Alicia sees why small sway, corrective steps and changing sensory weights can all belong to healthy control.

171. The environment belongs inside the organism-level explanation

A slippery floor, dim lighting, a moving platform or an unexpected obstacle changes the balance problem without changing the person’s anatomy.

The nervous system cannot create friction that the floor does not provide or see an obstacle that the scene hides. It can only choose among actions permitted by the physical and informational environment.

Balance is therefore an organism-environment relation rather than a trait stored entirely inside the body.

Kai Kai restores the world to the physiology diagram one last time.

172. The final transfer rule: identify state, evidence, action and constraint

Any unfamiliar balance question can be decomposed into four parts. First, what state of the body must be estimated? Second, what evidence is available from vestibular, visual, proprioceptive or contact signals? Third, what action can the muscles and limbs produce in the available time? Fourth, what environmental constraints limit the success of that action?

If a proposed explanation skips one of these four, it is probably compressing several mechanisms into one label.

This framework transfers beyond balance to robotics, vehicle control, sports movement and any system that must estimate an uncertain state before acting.

Balance becomes understandable when sensation is treated as evidence, movement as controlled action, and stability as the preservation of future options.

Part XV. Final extensions: uncertainty, calibration and the limits of one-number balance scores

173. Sensor disagreement can be useful because it exposes a hidden change in the world

If every sensory disagreement were treated as noise and immediately averaged away, the nervous system could miss important events. A moving visual surround, a slipping foot or an unexpected head movement creates precisely the kind of mismatch that reveals that the previous model no longer fits.

Disagreement therefore has diagnostic value inside the control system. It can trigger reweighting, a change in strategy or a search for a new reference. The useful question is not merely which sensor is right, but which physical explanation makes the conflicting signals mutually consistent.

Alicia sees conflict becoming evidence rather than failure.

174. Calibration is continuous because the body and environment both change

Eye-to-head geometry changes with viewing distance. Muscle strength and fatigue change through a day. Shoes change contact mechanics. A carried bag changes mass distribution. Growth and ageing change body proportions over years. The controller therefore cannot rely on one permanent calibration learned once in childhood.

Repeated movement provides opportunities to compare prediction with outcome and update the internal relationships among sensory input, motor command and mechanical consequence.

Kai Kai calls calibration maintenance for a moving biological machine.

175. A single balance score is a compressed summary, not the mechanism itself

One number may be useful for screening, tracking or research when it has been validated for that purpose. Yet the same score can arise from different combinations of sensory weighting, muscle power, joint mobility, strategy, attention and environment.

A mechanism question therefore requires unpacking the score into the measurements and assumptions from which it was built. If two people improve by the same amount, one may have changed foot placement while the other changed visual reliance or reaction timing.

The broad lesson extends beyond balance: a summary metric is most useful when the reader remembers what information was compressed to create it.