The Vestibular System: From Hair Cell to the Map of Place

01Inside the bony labyrinth

The vestibular labyrinth splits self motion across five end organs in each ear

Endolymph runs potassium rich and sodium poor, which is the chemistry of cytoplasm and not of the fluid outside a cell. That inversion puts the ionic driving force on the hair cell's side, so a bent bundle opens a gate and current follows without a pump working against the gradient. The membranous labyrinth holds that fluid around the sense organs of balance and of hearing, which is why one chemistry serves two senses.

Each end organ answers a different question about self motion, and the answers do not substitute for one another. Split that way, each labyrinth hands the brainstem five continuous channels, three reporting rotation and two reporting gravity and linear acceleration.

The three canals of one ear sit close to perpendicular. Three-dimensional multiplanar CT reconstruction of 44 human labyrinths, pooled from 22 subjects, put the angles between them at 90.4 to 94.0 degrees Della Santina 2005. Three tubes at those angles cover every axis a head can turn through, which is why rotation needs no fourth canal.

Everything the labyrinth senses leaves through one cable. Unbiased stereology counted 27,330 plus or minus 2,593 vestibular ganglion neurons per ear in five human specimens with no history of vestibular disease Park 2000. Those bipolar cells carry motion, position and acceleration from five end organs into the vestibular nuclei.

02Findings

What the research shows

03Hair cells and otoconia

Vestibular transduction begins when a hair bundle bends and a tip link opens a channel

Every vestibular signal starts as a bent hair bundle. In the maculae the stereocilia sit embedded in a gelatinous matrix carrying protein bound calcium carbonate crystals called otoconia, dense enough to drag on the bundle whenever the head tilts or accelerates. Each bundle carries one tall kinocilium beside a graded array of stereocilia, and the array moves as a unit when the otolithic membrane is displaced. That drag bends the cilia and turns head position and linear acceleration into a graded change in firing.

Mammalian otoconia are composite calcite crystals. In the guinea pig utricle a network of 20 nanometer beaded filaments covers their surface. The gelatinous matrix beneath is a dense network of 22 nanometer filaments cross-linked by shorter 11 nanometer filaments Lins 2000. The crystals are secured into that scaffold by surface adhesion and by confinement inside a loose interotoconial mesh. In mice, otoconin-90 and otolin assemble the matrix and concentrate calcium in the local environment where crystallization happens. Both proteins are expressed far more strongly in the mouse utricle and saccule than in other inner ear epithelia Yang 2011. The gravity sensor is a protein scaffold that templates its own mineral at one location.

The gate of transduction is a protein tether under tension

Bending a hair bundle pulls on its tip links, and tension in those gating springs opens the transduction channel. Maximal stiffness of about 1200 microN per meter in frog hair bundles occurs in 250 micromolar calcium, close to the concentration measured in frog endolymph. Disrupting the tip links drops that stiffness to roughly 200 microN per meter Marquis 1997. Cutting the tether removes five sixths of the bundle's mechanical stiffness and abolishes transduction along with it.

In the semicircular canals the same gate opens through fluid. Rotating the head leaves the endolymph behind through its own inertia, and the lagging fluid deflects the cupula against the hair bundles set in it. The right horizontal canal is stimulated by turning the head to the right, the anterior canal by flexion, the posterior canal by extension.

Regular and irregular afferents carry different discharge characters

Two afferents in the same nerve can share an average rate and time their spikes differently. In the chinchilla vestibular nerve, regular units outnumber irregular units by roughly three to one and hold the higher resting discharges, with normalized discharge variability spanning 0.020 to 0.60 across the population Baird 1988. Calyx-bearing units are invariably irregular. Hair cell ending type predicts discharge character, and regularity tracks sensitivity to galvanic current.

Both populations fire without any head movement at all. Across 125 identified fibers in the squirrel monkey crista, mean resting discharge was 74 spikes per second Lysakowski 1995. That sample held 10 percent bouton units, 29 percent calyx units and 61 percent dimorphic units. Otolith afferents in the same species rest near 60 spikes per second and shift by 30 to 40 spikes per second for each g of applied force Fernández 1976. Gravity reaches the brain as a deviation from a rate that is already running.

04Two canals, one plane

The brain reads rotation as a difference between two canals that share a plane

The six semicircular canals form three functional pairings across the two labyrinths. Right and left lateral canals pair together, the left anterior pairs with the right posterior, and the right anterior pairs with the left posterior. Turn the head right and the right lateral canal raises its firing rate while the left lowers its own. What reaches the brainstem is the difference between the two.

Those pairings come close to coplanar without quite reaching it. Across 44 human labyrinths, the anterior canal and its contralateral posterior partner lay 15.3 plus or minus 7.2 degrees apart on CT Della Santina 2005. Magnetic resonance reconstruction in 20 normal subjects put the contralateral offsets at 15.1 degrees between the two horizontal canals and at 21.2 and 21.7 degrees for the two oblique pairs Kim 2015. The same imaging put the three angles inside one ear lower than the CT figures, at 83.7, 82.5 and 88.4 degrees, so the canals sit close to orthogonal by either measure without being exactly so.

Push-pull pairing costs a second canal and returns two things nothing else can provide. Two canals reporting one plane buy redundancy and cancel the noise they share, because fluctuations common to both disappear in the subtraction while true rotation survives it, sharpened. The two directions are not symmetric, which is part of why the pair is needed. In otolith afferents of the barbiturate anesthetized squirrel monkey, the excitatory response of a regular unit runs larger than its inhibitory response Fernández 1976. The arrangement works only because each canal idles high enough to fall.

What happens at the floor and the ceiling of the range

Canal afferents in the macaque, driven by whole body rotations reaching 500 degrees per second, show inhibitory cutoff at the bottom of their range and excitatory saturation at the top Sadeghi 2007. Cutoff is where firing reaches silence and that side can report nothing further. Above it the two sides trade a graded difference. Below it one side has stopped contributing and the comparison rests on its partner alone.

Removing one labyrinth does not retune the survivor. Recordings from the contralesional nerve in three macaque monkeys found sensitivity, phase, inhibitory cutoff and excitatory saturation all unchanged from one to twelve months after unilateral labyrinthectomy Sadeghi 2007. Mean resting discharge did not move either. Steadiness returns after that surgery anyway. The recovery therefore happens in the central comparison and not in the peripheral sensor that survived.

05Holding gaze still

The vestibulo-ocular reflex drives the eyes at the head's own velocity

Measured in 22 healthy volunteers with a motorized head impulse rotator, the reflex ran at a mean gain of 1.08 plus or minus 0.10 Hirvonen 2007. Head velocity reached 100 to 120 degrees per second. The eyes rotate opposite the head, and the two directions matched within 3.7 percent of each other. In 8 of those volunteers gain rose to 1.26 for a target at 15 centimeters, against 1.08 for one at 140 centimeters.

Ascending fibers carry the signal from the vestibular nerve to the vestibular nuclei, the cerebellum and the oculomotor complex, reaching cranial nerves three, four and six and the six muscles they drive. Vestibulospinal reflexes leave the same nuclei to coordinate head and posture. The reflex takes no shortcut from intention. Afferent head sensitivities in the macaque were the same during voluntary and passive head rotations, and neck proprioceptive and efference copy signals left them unchanged before and after labyrinthectomy Sadeghi 2007. The eyes are driven by the head's measured motion and not by the command that produced it.

Vision and the canals share one adjustable store

A velocity storage mechanism holds the rotational signal longer than the canal's own mechanics allow. Its time constant in humans is 13 seconds, falling to 8 seconds under visual suppression, with the direct visual pathway contributing a gain of 0.75 Laurens 2011. The canal reports a sustained turn for a few seconds. The brain keeps reporting it for thirteen, which is how slow steady rotation stays faithful.

Optic flow and canal signal converge before that store. Every unit identified in the vestibular nuclei of the alert monkey responded when only the visual surround rotated around a stationary animal, as well as to true rotation of the animal Waespe 1977. The two inputs do not simply add, since each dominates across its own range of velocities. Optic flow holds gaze through slow sustained rotation and the reflex holds it through fast ones, so the two cover each other's blind range.

The same store sets what a turn feels like

The time constants of the reflex and of perceived rotational velocity co-varied at r equal to 0.95 in yaw and 0.93 in pitch Bertolini 2012. That held in healthy controls and in patients with chronic degeneration of the vestibulo-cerebellum alike. When the nodulus and uvula degenerate, the reflex and the sensation shorten together.

06Neck, head and body

The vestibular system cannot report body motion without a signal from the neck

The vestibular labyrinth measures head motion and nothing else. No sensor anywhere in the body measures the motion of the body itself. Turn the head on a still trunk and the labyrinth signals a rotation the body never made. The brain recovers body motion by combining the labyrinth signal with a report of where the head currently sits on the neck.

That subtraction has been recorded cell by cell. In the cerebellar interpositus nucleus of alert squirrel monkeys, 44 of 67 neurons were vestibular sensitive, and 29 of those also answered neck proprioceptive input Luan 2013. The neck stimulus was head-on-trunk rotation about the C1 to C2 axis and lateral flexion about the C6 to T3 axis. In all 29 convergent neurons the two inputs cancelled each other during combined stimulation, in orthograde and pronograde posture alike.

The product of that arithmetic appears one nucleus away. In monkeys, about half of rostral fastigial neurons encode the motion of the body in space Brooks 2009. The other half encode motion of the head in space, in the manner of vestibular nucleus cells. Bimodal cells receiving both vestibular and neck input carry the body estimate. Unimodal cells receiving vestibular input alone carry the head estimate. The body signal is computed rather than sensed.

Neck and canal signals meet again on single cortical neurons

Almost every vestibular neuron tested in the parietoinsular vestibular cortex of the macaque also answered visual and somatosensory stimulation, which classified them as polymodal vestibular units Grüsser 1990b. Of the vestibular units held long enough to test, 62 received visual stimulation and 79 received somatosensory stimulation, and with very few exceptions all of them responded to both. Pressure or movement applied to the neck and shoulder drove those same cells, so canal input and neck input cancel or facilitate one another on single cortical neurons.

Nothing in the labyrinth separates a body turn from a head turn. Rotate the trunk under a stationary head and the canals stay quiet while the neck alone reports the movement. That stimulus was the most effective somatosensory drive in the macaque recordings, and cortical vestibular units answered it with a gain in the same range as their response to vestibular stimulation itself Grüsser 1990b. The distinction between the two events is made downstream, by cells holding both terms at once.

The Unified Model of Tone works on the neck side of that subtraction, the only term in it reachable from outside the skull. Because the brain's answer for body motion is labyrinth minus neck, a change in what the neck reports changes the answer with nothing at all happening to the labyrinth. A cervical segment holding an altered mechanical state feeds an altered neck term into the same computation Luan recorded in the interpositus, and the muscle spindles of the deep neck are what write that term. Chiropractic care loads those spindles along the C1 to C2 rotation axis and the C6 to T3 flexion axis that Luan used as his neck stimulus.

The model predicts that returning the neck term to correspondence with where the head actually sits changes the output of the subtraction. The change appears in gaze stability and postural sway, not at the point of contact. Balance and eye control are the readout because they are what the subtraction is used for.

07From reflex to place

The vestibular signal reaches thalamus, cortex and the brain's map of place

The vestibular signal ascends through the thalamus to cortex, where it becomes the felt sense of being upright. Thalamic connections with vestibular cortex and the reticular formation carry arousal and the awareness of where the body sits. The same relay separates self motion from motion of the world.

The vestibular nuclei turn the signal into eye and postural reflexes inside a few milliseconds. Thalamus and cortex turn it into the felt sense of being upright. The hippocampal formation turns it into position inside a room the body has been in before. The first of those answers is measured in milliseconds and the last in percent of hippocampal volume.

The principal cortical target is the parietoinsular vestibular cortex. About two thirds of the neurons recorded there answered vestibular stimuli, and the rest responded mainly to somatosensory stimulation of the neck and shoulder Grüsser 1990a. The survey yielded 152 vestibular units across a region spanning 6 to 8 millimeters anteroposteriorly in four awake Java monkeys. Of 64 units tested by rotation in more than one plane, 54 answered rotation in more than one of roll, yaw and pitch. This cortex holds a three dimensional account of turning.

Beyond the insula, zone seven of the parietal cortex responds to optic flow and governs spatial orientation of the head in space. The ventral and medial intraparietal areas and the medial superior temporal area answer optic flow the same way. Electrical microstimulation of dorsal MST biased where monkeys judged themselves to be heading when the cue was optic flow, and left vestibular heading judgments unaffected Gu 2012. Inactivating the same area raised thresholds for visual and vestibular headings alike, and the animals still combined the two cues near optimally. Those inactivation effects were substantially stronger for optic flow, and the combined cue deficit landed between the two single cue deficits.

The brain's compass runs on labyrinth input

Head direction cells in the rat anterior thalamic nucleus lose their directional firing entirely after vestibular lesions, and the collapse follows the time course of the vestibular loss Stackman 1997. The angular head velocity influence on thalamic firing disappears with it. A subset of cells in those lesioned rats fired intermittent bursts unrelated to heading, a pattern never seen in a vestibular intact rat. Caloric stimulation in humans activates the hippocampal formation along with retrosplenial cortex and the subiculum Vitte 1996, the stations that carry heading and place. Cold water in one ear reaches the brain's map of where the body is.

Losing both labyrinths shrinks that machinery. Ten patients with acquired chronic bilateral vestibular loss showed a selective 16.9 percent decrease in hippocampal volume against controls, with navigation deficits matching the atrophy and general memory spared Brandt 2005. The deficits appear while the patients are sitting still. Labyrinth input maintains the spatial map as a standing condition.

The labyrinth tracks the rest of the body's physiology. Among 5,086 US adults aged 40 and over tested with a modified Romberg standing test, diabetes raised the odds of failing that test by 70.0 percent Agrawal 2009. Systemic metabolic state reaches the balance system without any lesion appearing in the temporal bone.

Tone inside its healthy range keeps the weighting free to move, and that freedom is what steadiness costs. It is why balance returns after one labyrinth is destroyed while the surviving nerve never retunes: the brainstem cannot change what the sensor reports, so it changes what the report is worth. Tone drifted outside that range holds an old weighting in place after the evidence for it has gone. The neck keeps its full weight in the subtraction after cervical position sense has stopped matching the labyrinth, and the person reports unsteadiness with a normal scan. The labyrinth keeps its weight after inhibitory cutoff has taken one side to silence, and the comparison then rests on a partner the brain is still treating as half of a pair. Lose both labyrinths and the weighting has nothing left to redistribute, which is where the 16.9 percent hippocampal loss and the navigation failure appear.

08Tone

How this system expresses tone

The vestibular system states tone in plain units. Every fiber holds a standing rate with the head still, and every value the brain takes from the labyrinth is a comparison between two reports, never a reading from one.

Every value in this system is a rate, an angle or a time constant, and the rest of tone is legible in those units. Gain: the vestibulo-ocular reflex runs at 1.08 for a target at 140 centimeters and rescales to 1.26 for one at 15, so the same head turn buys a different eye movement. Oscillation: the slow drift and quick reset of nystagmus is the reflex arc writing motion onto the eyes, named by the direction of its quick phase. Prediction: velocity storage holds a turn for 13 seconds after the canal's own mechanics have decayed, so the brain keeps reporting a rotation the sensor has stopped reporting. Time course: an afferent answers a head turn in milliseconds while central compensation after one labyrinth is lost takes weeks, and the peripheral sensor was still unchanged at 12 months. Load: 27,330 ganglion neurons per ear hold a resting discharge through every hour of the day, and that baseline is paid for whether or not the head moves. Constraint: the three canals of one ear are fixed close to orthogonal, and each partner pair sits 15.3 degrees off coplanar. One labyrinth reports only the planes its geometry allows, so the two ears have to be compared to cover the rest.

09Across the library

How this page relates to the rest of the library

10Frequently asked

Questions about this topic

11The sources

References

Sources: primary literature, linked inline.