Spatial Disorientation: The Leans · Guide · glidepathscience.org
GuideSpatial Disorientation: The Leans
Contents
This Guide explains the ideas and equations behind every Walkthrough step, Concept Challenge question and Predict round. Part 1 prepares the Intro track: how the inner ear senses rotation, why a steady turn fades from feeling, the after-effect, and the leans. Part 2 prepares the Advanced track: the canal as a torsion pendulum, its frequency response, velocity storage, thresholds, the otoliths and the Coriolis illusion. Each part takes about twelve minutes to read. This sim is for learning, not for flight training: its airplane is a generic drawing and its numbers are typical values.
Part 1 · The essentials
Math you’ll need: algebra, angular velocity, and one exponential, $e^{−t/τ}$, as in a charging capacitor; no calculus is required.
Sensing rotation
Close your eyes in a smoothly turning office chair and you can still tell that you have started to turn. The sense that tells you lives in the inner ear. Each ear holds a small, fluid-filled labyrinth called the vestibular system. Its two otolith organs sense linear acceleration and which way is down; its three semicircular canals sense rotation. The canals are thin rings of fluid, roughly at right angles to one another, so that between them they cover every axis. Shaking your head “no” is a rotation about a vertical axis, a yaw, sensed mainly by the horizontal canal; nodding “yes” (pitch) and tipping an ear toward a shoulder (roll) are sensed mainly by the two vertical canals.
To describe what the canals respond to we need two quantities. The angular velocity $ω$ says how fast something turns, in degrees per second or radians per second; a chair turning at 10 revolutions per minute turns at $ω$ = 60 °/s, since one revolution is 360° and a minute is 60 s (so 1 rpm = 6 °/s). The angular acceleration $α$ says how fast $ω$ changes:
$α = Δω/Δt$
Its unit is degrees per second per second, °/s². The distinction matters, because, as the next sections show, the canals respond to changes in rotation, not to rotation itself.
The fluid lags, and a steady turn fades
Each canal is filled with a fluid called endolymph. Across one end of the ring, in a swelling called the ampulla, stretches a gelatinous flap, the cupula, with sensory hair cells at its base. When the head starts to turn, the canal’s wall turns with it, but the fluid inside, because of its inertia, stays behind, like coffee in a cup you rotate. Relative to the wall, the fluid flows the opposite way and pushes the cupula over. The hair cells turn that bend into a nerve signal: the brain feels a turn.
Now hold the turn steady. Friction with the wall gradually drags the fluid along until it turns with the canal, and the cupula’s own springiness straightens it. The signal fades, even though the head is still turning. The fade is exponential: if the canal reports $c_0$ at the start of a steady turn, then a time $t$ later it reports
$c = c_0e^{−t/τ}$
where $τ$, the canal’s time constant, is about 6 s for a typical person. After one time constant the signal is 37 % of what it was, after three about 5 %. A common slip is to expect a steady turn to be felt for as long as it lasts; in fact, after half a minute of steady turning, the rider feels nothing at all.
The stop, the after-effect and the threshold
Suppose the chair has turned to the right long enough that the rider feels nothing, and then stops. A stop is itself a change of rotation, to the left. The fluid, now turning with the canal, keeps going when the wall stops, and bends the cupula the way a start to the left would. The rider, sitting still, feels a turn to the left. This after-effect starts at about the speed the chair had and fades with the same time constant.
How long is it felt? Below some rate a rotation is not noticed at all. A typical threshold is about 2 °/s. The after-effect is felt until $c_0e^{−t/τ}$ falls to 2 °/s, which happens at
$t = τ\ln(c_0/2)$
Because of the logarithm, doubling the spin rate adds only $τ\ln 2$, about 4 s.
The threshold has a second consequence. If the chair speeds up steadily, with a constant $α$, the cupula settles at a steady bend, and the canal’s signal levels off at about $ατ$ instead of growing with the chair’s speed. If $ατ$ is under 2 °/s, the speed-up is never felt, however long it lasts and however fast the chair ends up turning. With $τ$ = 6 s that is any $α$ under about 0.33 °/s². The usual mistake is to apply the threshold to the chair’s speed; it applies to what the canal reports.
The leans, and why pilots trust the instruments
A pilot flying in cloud has no horizon to look at. The vertical canals sense roll the way the horizontal canal senses yaw, with the same fading and the same threshold. If the airplane rolls slowly into a bank, at a roll rate $p$ under about 2 °/s, the roll is never felt. The bank builds up as rate times time,
$φ = pt$
so even a gentle 1.5 °/s produces a 30° bank in 20 s.
Could the pilot feel the bank some other way, as a sideways tilt of “down”? In a coordinated turn, no. The airplane’s lift tilts with the wings, and the push of the seat on the pilot, which together with gravity is what the otolith organs feel, points along the pilot’s spine, exactly as in level flight. It is only a little stronger: $1/\cos φ$ times the weight, 1.10 times at 25°. So once any roll sensation has faded, the turn feels like straight and level flight.
Now the pilot notices the instruments and rolls the wings level briskly. The slow roll into the bank was never felt, but the brisk roll out is: to the inner ear the airplane has just gone from level into a bank the other way. The pilot feels banked opposite to the turn they were in, and leans the other way, toward the old bank, to feel upright. This is the leans. In the sim’s model, after a slow roll into a 30° left bank and a roll back at 15 °/s, the pilot feels banked about 23° to the right. Within about 5 s most of that is gone, down to about 7°, because after a brisk roll the canal itself swings back past zero and reports a slower roll the other way; the last few degrees fade over tens of seconds as the otoliths, which say “upright”, slowly win. How long the leans last in a real pilot varies from person to person, and the model’s numbers are illustrative.
The remedy is simple to state and hard to do: trust the instruments. The attitude indicator shows the true bank whatever the inner ear says, and pilots are trained to believe it over every feeling. Rolling until it “feels level” would put the airplane back into a bank, the first step of a descending spiral.
Part 2 · Going further
Math you’ll need: a second-order linear differential equation with constant coefficients, its characteristic roots, and the frequency response of a linear system; rotation vectors and their components.
The canal as a torsion pendulum
In 1931 W. Steinhausen modeled the cupula and endolymph as a torsion pendulum. Let $θ$ be the angle through which the fluid has turned relative to the canal (the cupula’s deflection, measured as an angle of fluid), and $α$ the head’s angular acceleration about the canal’s axis. Then
$Iθ″ + Bθ′ + Kθ = −Iα$
where $I$ is the fluid’s moment of inertia, $B$ the viscous drag of a tube a few tenths of a millimeter across, and $K$ the cupula’s stiffness. The minus sign says that the fluid moves backward relative to the head. In a tube that narrow the drag is enormous, $B^2 ≫ 4IK$: the pendulum is heavily overdamped and never rings. Its characteristic equation $s^2 + (B/I)s + K/I = 0$ has two negative real roots, $−1/τ_1$ and $−1/τ_2$, with $τ_1 + τ_2 = B/K$ and $τ_1τ_2 = I/K$. When the two are far apart,
$τ_2 ≈ B/K$ and $τ_1 ≈ I/B$
Typical values are $τ_1$ ≈ 5 ms, the time the fluid takes to get going, and $τ_2$ ≈ 6 s, the time the cupula takes to creep back. The sim scales the signal so that an ideal step of angular velocity $Ω$ reads $c = Ω(e^{−t/τ_2} − e^{−t/τ_1})$: a rise in a few milliseconds, then the slow fade of Part 1.
A velocity sensor, and its frequency response
Between the two time constants the drag term balances the inertial push: $Bθ′ ≈ −Iα$, so $θ ≈ −(I/B)ω = −τ_1ω$. The deflection is proportional to the head’s angular velocity: the canal is an integrating accelerometer, and for everyday head movements it acts as a velocity sensor. Driven by a sine of angular velocity at frequency $f$, the signal’s gain is
$|H| = (τ_2 − τ_1)ω/\sqrt{(1 + ω^2τ_1^2)(1 + ω^2τ_2^2)}$
with $ω = 2πf$ here the angular frequency of the motion. The gain is close to 1 between the corner frequencies $1/(2πτ_2)$ ≈ 0.027 Hz and $1/(2πτ_1)$ ≈ 32 Hz. Below the lower corner, where the cupula’s spring wins, it falls like a high-pass filter, and the signal leads the motion. Above the upper corner the fluid’s inertia wins and the cupula barely moves. Below the lower corner, with $x = 2πfτ_2$,
corner $f = 1/(2πτ_2)$; $|H| = x/\sqrt{1 + x^2}$, $x = 2πfτ_2$
Velocity storage
Measured carefully, the felt rotation after a stop outlasts the canal’s own signal: about 16 s rather than 6 s in yaw. The brainstem holds the canal’s signal in a leaky integrator, called velocity storage. In the sim’s version a store $x$ is charged by the canal’s signal $c$, $x′ = (−x + (τ_{vs}/τ_2 − 1)c)/τ_{vs}$, and the felt rate is $y = c + x$. A step of angular velocity then reads $y ≈ Ωe^{−t/τ_{vs}}$, so a stop is felt for
storage: $t = τ_{vs}\ln(y_0/2)$
Thresholds under a steady acceleration
Under a steady $α$ from rest the canal’s signal is $c(t) ≈ ατ_2(1 − e^{−t/τ_2})$. For short times this is $αt$, and the classic observation that a rotation is felt once $αt$ reaches about 2 °/s is Mulder’s product. It is only the short-time limit. Setting $c$ = 2 °/s exactly gives
$t = −τ_2\ln(1 − 2/(ατ_2))$; $α_{max} = 2/(τ_2 − τ_1)$
The second equation is the largest steady acceleration that is never felt, the level at which the signal settles just under the threshold. Mulder’s product fails for weak accelerations: it predicts that any $α$ is felt eventually, which is false.
The otoliths, and why the leans persist
The otolith organs feel the gravito-inertial acceleration, gravity minus the head’s own acceleration, as the push of whatever supports the body. In a level, coordinated turn at bank $φ$ the lift, and with it the seat’s push, tilts with the airplane, and its vertical part balances the weight, so the push is $n$ times the weight along the pilot’s spine:
$n = 1/\cos φ$
This is why the leans persist. The canals report only changes of rotation and fall silent between them; they never report a bank as such. The otoliths say “upright” in level flight, and they said “upright” in the turn too, so they cannot tell the pilot which feeling was wrong. The brain gives way to them only slowly; the sim’s model uses a time constant of 20 s, an illustrative value. (In the model, most of the felt bank goes sooner, in the first few seconds, as the canal’s swing back after the brisk roll is felt; what lingers fades with those 20 s.) The same reasoning drives the graveyard spiral: in an unnoticed descending turn, pulling back adds lift that is tilted with the wings, which tightens the turn and steepens the descent. The only way out is to level the wings by the instruments first.
The Coriolis illusion
Suppose the chair has turned steadily at $ω$ long enough that the canals have settled. The head’s rotation vector is $ω$ along the vertical (pointing down for a turn to the right, by the right-hand rule). Now the rider nods forward through an angle $β$. The rotation vector does not move, but the head’s axes do, and its components along them change suddenly:
$ω_{roll} = ω\sin β$, and the yaw canal’s change $ω\cos β − ω$
The roll canal, at rest a moment ago, reports a sudden roll of $ω\sin β$, and the yaw canal reports a slowing. The rider feels a tumble about an axis nobody turned. This is the cross-coupled, or Coriolis, illusion, and it can be violent: a nod during a prolonged turn is one of the classic ways to disorient a pilot. For a right turn and a forward nod, the downward rotation vector has a positive component along the nose-down forward axis, a roll to the right in this model. The illusion needs a settled turn: the roll canal’s reading is a true component of the rotation, but the yaw canal no longer reports the rest of it. Nod in a turn just begun and the canals together report the true, vertical rotation, and nothing tumbles.
nod: $ω_{roll} = ω\sin β$
Reference
Common mistakes
Part 1
- Expecting a steady turn to be felt as long as it lasts: the canals report changes, and the signal fades with $τ$ ≈ 6 s.
- Expecting a stop to feel like nothing: it is felt as a turn the other way.
- Applying the 2 °/s threshold to the chair’s speed rather than to the canal’s signal: a steady $α$ under about 0.33 °/s² is never felt.
- Thinking the seat pushes the pilot sideways in a turn: in a coordinated turn the push is along the spine.
- Thinking the leans tilt the pilot toward the real bank: they lean toward the bank they never felt, after feeling the roll back to level.
- Trusting the feeling over the attitude indicator.
Part 2
- Taking $I/B$ for the long time constant: it is the short one; the long one is about $B/K$.
- Using Mulder’s product $αt$ = 2 °/s for weak or long accelerations, where the signal has already leveled off.
- Using the canal’s $τ_2$ for the felt yaw when velocity storage is on: the felt decay is $τ_{vs}$.
The equation sheet
| Part 1 | |
|---|---|
| rotation | $α = Δω/Δt$; 1 rpm = 6 °/s |
| a steady turn | $c = c_0e^{−t/τ}$, $τ ≈ 6$ s |
| the after-effect | felt for $t = τ\ln(c_0/2)$ (threshold 2 °/s) |
| steady speed-up | steady $α$: $c → ατ$ |
| a slow roll | bank $φ = pt$ |
| Part 2 | |
|---|---|
| the canal | $Iθ″ + Bθ′ + Kθ = −Iα$; $τ_2 ≈ B/K$, $τ_1 ≈ I/B$ |
| frequency | corner $f = 1/(2πτ_2)$; $|H| = x/\sqrt{1 + x^2}$, $x = 2πfτ_2$ |
| storage | storage: $t = τ_{vs}\ln(y_0/2)$ |
| thresholds | $t = −τ_2\ln(1 − 2/(ατ_2))$; $α_{max} = 2/(τ_2 − τ_1)$ |
| a coordinated turn | $n = 1/\cos φ$ |
| a nod | nod: $ω_{roll} = ω\sin β$ |
Symbols: $ω$ angular velocity; $α$ angular acceleration; $c$ the canal’s signal and $c_0$ its value at the start of a fade; $τ$ (or $τ_2$) the canal’s long time constant and $τ_1$ its short one; $τ_{vs}$ the velocity-storage time constant; $y_0$ the felt rate at the start of a fade; $θ$ the fluid’s turn relative to the canal; $I$, $B$, $K$ its inertia, drag and stiffness; $f$ a frequency and $x = 2πfτ_2$; $|H|$ the gain; $p$ a roll rate; $φ$ the bank; $n$ the load factor; $β$ a nod’s angle. Typical values: $τ_2$ = 6 s, $τ_1$ = 5 ms, $τ_{vs}$ = 16 s, threshold 2 °/s.
Sources and further reading
- Federal Aviation Administration, Pilot’s Handbook of Aeronautical Knowledge (FAA-H-8083-25), the chapter on aeromedical factors: spatial disorientation and its illusions, in a pilot’s words.
- OpenStax, Anatomy and Physiology 2e, Chapter 14, “The Somatic Nervous System,” section 14.1 “Sensory Perception” (equilibrium), openstax.org; OpenStax, University Physics Volume 1, for angular velocity and acceleration.
- This sim’s Field Manual (H) for the model, its typical values and its limits; Accelerating Reference Frames for apparent gravity; Circular Motion & Banked Turns and G-Force & G-LOC Centrifuge for the load factor.
Spatial Disorientation: The Leans v1.1 · © 2026 Dr. David Meier · CC BY-NC-SA 4.0 · glidepathscience.org