Vestibulo-ocular reflex
More actions
The vestibulo-ocular reflex (VOR) is the eye-movement reflex that keeps the image of the world steady on the retina while the head moves. When the head turns, sensors in the inner ear detect the motion and the brainstem drives the eyes in the opposite direction at the same speed, so gaze stays fixed on the target.[1] In humans the reflex responds to a head rotation in less than 10 milliseconds: a search-coil study measured an average latency of 8.6 ms.[2]
The VOR matters for virtual reality (VR) and augmented reality (AR) because a head-mounted display moves with the head. In a 2017 overview of VR headset engineering, Warren Hunt of Oculus Research wrote that the stability requirements of VR headsets are driven by the human perceptual system and described the VOR as one of the fastest reflexes in the body. Because the reflex makes the eyes counter-rotate during a head turn, the view on a head-mounted display slides across the retina and blurs.[3] The reflex is connected to three design problems in VR: motion-to-photon latency, display persistence and judder, and the visual-vestibular conflict linked to simulator sickness.
How it works
The reflex has three parts: a peripheral set of motion sensors in the inner ear, central processing in the brainstem and cerebellum, and a motor output through the eye muscles.[1] The motion sensors belong to the vestibular system. The semicircular canals respond to angular acceleration and so detect head rotation, while the otolith organs (the utricle and saccule) respond to linear acceleration and detect head translation and the head's orientation relative to gravity.[1] Humans have six canals, arranged as three coplanar pairs (anterior, posterior and horizontal) across the two ears. The pairs work in a push-pull arrangement: a rotation that excites the canal on one side inhibits its partner on the other side, and the difference in firing rate between the two encodes the velocity of the movement.[1]
Vestibular signals reach the brainstem through cranial nerve VIII and enter the vestibular nuclei. Second-order neurons then project, through the medial longitudinal fasciculus, to the nuclei of the cranial nerves that control eye movement (III, IV and VI), and a third neuron drives the extraocular muscles of the eye. Hunt gave the reflex's latency as "3 neurons (about 10 ms)".[1][3] Each semicircular canal has excitatory projections to one pair of extraocular muscles in each eye and inhibitory projections to the antagonist pair. The cerebellum, in particular the flocculonodular lobe, receives vestibular input and fine-tunes the reflex.[1]
Latency and gain
Researchers describe VOR performance by its latency and its gain, the ratio of eye velocity to head velocity. A gain of 1.0 means the eyes rotate by exactly the amount needed to cancel the head movement.[1][4] In a 2000 study in the Journal of Neurophysiology, Collewijn and Smeets applied brief rotations of about 1000 degrees per second squared to subjects' heads with a torque helmet and recorded head and eye orientation with magnetic search coils at 1000 samples per second. They found:[2]
| Measure | Result reported by Collewijn and Smeets (2000) |
|---|---|
| Average VOR latency | 8.6 ms (pooled across eyes, directions and subjects) |
| Difference between the eyes | Latency about 1.3 ms shorter for the eye opposite the direction of head rotation, attributed to an extra neuron in the pathway to the other eye |
| Acceleration gain over the first 40 to 50 ms | About 1.1 |
| Instantaneous velocity gain | Built up gradually and converged toward unity |
| Effect of a near target | Gain rose to a higher level than with a distant target |
Because of the latency and an initial passive eye movement in the wrong direction, the authors concluded that gaze stabilization in the first moments of a head movement was never complete.[2] An earlier search-coil study by Collewijn, Martins and Steinman (1983) measured gains during ordinary head movements at 0.33 to 1.33 Hz. With active head movement in the light, the effective gain was mostly between 0.97 and 1.03; in the dark it was mostly between 0.92 and 1.00. During passive movement in the dark it was about 13 percent lower than during active movement.[4]
Adaptation
The VOR is not fixed. When the visual consequences of head motion change, the reflex recalibrates. In experiments published in 1976, Gonshor and Melvill Jones had subjects attempt to track a mirror-reversed image of their surroundings during rotation. Sixteen minutes of reversed visual tracking within a 50-minute daily session reduced VOR gain, measured in the dark, by about 25 percent.[5] In a companion study, four adults wore head-mounted dove prisms that reversed vision left to right for 2, 6, 7 and 27 days. Gain fell to a plateau of about 25 percent of normal by the end of the first week, and the changes were specific to the plane of the reversal.[6]
Smaller mismatches are corrected faster. In the 1983 Collewijn study, subjects wore magnifying or reducing spectacles for 40 minutes to 24 hours; the largest required change in eye-movement amplitude was 36 percent, and most of the adaptation of the VOR measured in the dark was complete in about 30 minutes, a rate the authors described as much faster than in earlier experiments that required larger changes.[4] Head-mounted displays can impose similar changes in image scale: in one study of simulator sickness, Draper and colleagues produced magnification and minification in a head-coupled display by varying the geometric field of view angle.[7]
Clinical testing
Damage to the reflex can cause oscillopsia, in which objects appear to jiggle or bounce during head movement because their image no longer stays fixed on the retina.[1] In 1988 Halmagyi and Curthoys described a bedside sign of a lost horizontal canal: when the examiner turns the patient's head rapidly toward the affected side, the patient makes one large or several small corrective "refixation" saccades back to the target. They validated the sign with magnetic search coils in 12 patients who had undergone unilateral vestibular neurectomy.[8] StatPearls lists this Halmagyi-Curthoys (head impulse) test as a way to assess the reflex: when the reflex is damaged on one side, a forceful head turn toward that side makes the eyes follow the head before they come back to the target.[1] In 2009, MacDougall and colleagues reported that a video head impulse test (vHIT) recording at 250 Hz produced VOR gains that did not differ significantly from simultaneous search-coil recordings at 1,000 Hz, and detected both overt and covert corrective saccades as accurately as the coils.[9] Clinicians also test the reflex with caloric irrigation, in which warm or cold water in the ear canal induces nystagmus.[1]
Role in VR and AR displays
Latency
A headset display is fixed to the head, so during a head turn the eyes counter-rotate relative to the screen. Hunt wrote that, because of the VOR, the visual system "expects a stable, crisp image during head rotation even when the screen (attached to the head) is moving at 300 degrees per second". If the rendered view lags behind the head, the virtual world drifts against the stabilized gaze. Hunt gave the example of a 30 ms system latency at a head speed of 100 degrees per second, which makes the world lag by 3 degrees and produces a "swimming" artifact that can be disturbing and lead to motion sickness.[3]
The 3GPP technical report on VR media services over mobile networks (TR 26.918) names the rotational VOR as "the main driver" of the motion-to-photon latency requirement. It states that the reflex acts in about 10 ms, or 7 to 15 ms depending on the source, and treats that as a performance goal. Allowing time to detect user movement and process images, it concludes that a requirement of 20 ms "would seem appropriate", with about 10 ms as a useful objective.[10] See Motion-to-photon latency and Latency for measured values in commercial systems.
Headset runtimes use several methods to cut the delay between head motion and the displayed image, including predictive tracking and timewarp. In a 2015 Oculus developer blog post, Michael Antonov described timewarp as a technique that "warps the rendered image before sending it to the display in order to correct for head motion that occurred after the scene was rendered and thereby reduce the perceived latency".[11]
Persistence and judder
Latency is not the only problem. Even a perfectly timed frame smears if it stays lit while the eye sweeps across it. Hunt defined judder as "the blurring effect caused by the VOR and a static head-mounted display". A conventional display holds each frame for the whole frame period, about 17 ms at 60 Hz or 11 ms at 90 Hz. At a head speed of 300 degrees per second, 11 ms corresponds to 3.3 degrees, and the counter-rotating eye smears each pixel across that angle. VR displays therefore use a "low persistence" mode that lights the panel for only 1 to 2 ms per frame and shows black for the rest, which removes the smear at the cost of a dimmer image and, in some conditions, a strobe effect.[3]
Simulator sickness and VOR change after VR use
During head movements in a virtual environment, the image on the display and the head acceleration sensed by the inner ear can differ, a visual-vestibular conflict of the kind known to modify VOR gain.[12] Several studies have looked at the effects:
| Study | Setup | Main finding |
|---|---|---|
| Ivanenko et al. (1998), Neuroscience Letters[13] | 25 subjects; 45 minutes of passive 180-degree whole-body rotations paired with only 90-degree rotation of a virtual room | Perceived angle of rotation in darkness fell by 24 percent (plus or minus 16), but VOR gain did not change significantly |
| Di Girolamo et al. (2001), Acta Oto-Laryngologica[12] | 20 healthy subjects; 20 minutes in a virtual environment through a head-mounted display; VOR measured in the dark | VOR gain was reduced in all subjects at the end of exposure; the reduction had disappeared 30 minutes later |
| Draper et al. (2001), Human Factors[7] | 50-minute exposures to a head-coupled virtual interface; image scale factors of 0.5, 1.0 and 2.0, and added delays of 125 and 250 ms over a nominal 48 ms | Simulator sickness reports were significantly higher at 0.5 and 2.0 scale than at 1.0; sickness did not vary with the added time delays |
Di Girolamo and colleagues suggested that because a virtual environment can temporarily change VOR gain, the effect could be used for assessing and rehabilitating vestibular disorders.[12] The 3GPP report cites this line of work and argues that, since adaptation to such conflicts is possible but susceptibility to VR sickness varies between people, VR systems should aim to mimic real-world conditions as closely as possible.[10]
Research using VR headsets
Headsets with built-in eye tracking can record the VOR as well as stimulate it. In a 2023 study in Frontiers in Neurology, Clément Desoche and colleagues performed head impulse tests on 12 healthy subjects wearing an HTC Vive fitted with an SMI eye tracker that sampled at 250 Hz, with head motion captured by an external optical tracking system. In the normal visual condition, VR-based VOR gain differed significantly but slightly, by 3 percent, from a standard vHIT system. The team also altered the virtual scene's response to head motion (halved, frozen or reversed). VOR gain did not differ significantly between these conditions, but covert corrective saccades appeared in about one in four trials under each altered condition.[14]
VR is also used in vestibular rehabilitation. In a 2019 randomized study of older adults with unilateral vestibular hypofunction, with and without mild cognitive impairment, Alessandro Micarelli and colleagues added a home-based head-mounted display program to standard rehabilitation. Within each group VOR gain did not change, but between-group comparisons after treatment showed significantly higher VOR gain in the groups that used the headset.[15] Medical training is another use: a 2023 study in Scientific Reports tested a VR-enhanced mannequin that simulated the head impulse test, and found that training raised novices' sensitivity in detecting catch-up saccades from 69.1 percent to 97.9 percent, above the 87 percent scored by experts.[16]
See also
References
- ↑ 1.00 1.01 1.02 1.03 1.04 1.05 1.06 1.07 1.08 1.09 S. Somisetty, J.M. Das (2023-07-25). "Neuroanatomy, Vestibulo-ocular Reflex". StatPearls, NCBI Bookshelf. StatPearls Publishing. https://www.ncbi.nlm.nih.gov/books/NBK545297/. Retrieved 2026-10-15.
- ↑ 2.0 2.1 2.2 H. Collewijn, J.B. Smeets (2000). "Early components of the human vestibulo-ocular response to head rotation: latency and gain". Journal of Neurophysiology, vol. 84, no. 1, pp. 376-389. https://doi.org/10.1152/jn.2000.84.1.376. Retrieved 2026-10-15.
- ↑ 3.0 3.1 3.2 3.3 Warren Hunt (2017). "Frontiers in Virtual Reality Headsets". Frontiers of Engineering: Reports on Leading-Edge Engineering from the 2016 Symposium. National Academy of Engineering, The National Academies Press. doi:10.17226/23659. https://nap.nationalacademies.org/read/23659/chapter/5. Retrieved 2026-10-15.
- ↑ 4.0 4.1 4.2 H. Collewijn, A.J. Martins, R.M. Steinman (1983). "Compensatory eye movements during active and passive head movements: fast adaptation to changes in visual magnification". The Journal of Physiology, vol. 340, pp. 259-286. https://doi.org/10.1113/jphysiol.1983.sp014762. Retrieved 2026-10-15.
- ↑ A. Gonshor, G. Melvill Jones (1976). "Short-term adaptive changes in the human vestibulo-ocular reflex arc". The Journal of Physiology, vol. 256, no. 2, pp. 361-379. https://doi.org/10.1113/jphysiol.1976.sp011329. Retrieved 2026-10-15.
- ↑ A. Gonshor, G. Melvill Jones (1976). "Extreme vestibulo-ocular adaptation induced by prolonged optical reversal of vision". The Journal of Physiology, vol. 256, no. 2, pp. 381-414. https://doi.org/10.1113/jphysiol.1976.sp011330. Retrieved 2026-10-15.
- ↑ 7.0 7.1 M.H. Draper, E.S. Viirre, T.A. Furness, V.J. Gawron (2001). "Effects of image scale and system time delay on simulator sickness within head-coupled virtual environments". Human Factors, vol. 43, no. 1, pp. 129-146. https://doi.org/10.1518/001872001775992552. Retrieved 2026-10-15.
- ↑ G.M. Halmagyi, I.S. Curthoys (1988). "A clinical sign of canal paresis". Archives of Neurology, vol. 45, no. 7, pp. 737-739. https://doi.org/10.1001/archneur.1988.00520310043015. Retrieved 2026-10-15.
- ↑ H.G. MacDougall, K.P. Weber, L.A. McGarvie, G.M. Halmagyi, I.S. Curthoys (2009). "The video head impulse test: diagnostic accuracy in peripheral vestibulopathy". Neurology, vol. 73, no. 14, pp. 1134-1141. https://doi.org/10.1212/WNL.0b013e3181bacf85. Retrieved 2026-10-15.
- ↑ 10.0 10.1 "Virtual Reality (VR) media services over 3GPP (3GPP TR 26.918 version 15.2.0 Release 15), section 8.1.2". ETSI TR 126 918 V15.2.0. ETSI. 2018-07. https://www.etsi.org/deliver/etsi_tr/126900_126999/126918/15.02.00_60/tr_126918v150200p.pdf. Retrieved 2026-10-15.
- ↑ Michael Antonov (2015-03-02). "Asynchronous Timewarp Examined". Meta for Developers blog. https://developers.meta.com/vr/blog/asynchronous-timewarp-examined/. Retrieved 2026-10-15.
- ↑ 12.0 12.1 12.2 S. Di Girolamo, P. Picciotti, B. Sergi, W. Di Nardo, G. Paludetti, F. Ottaviani (2001). "Vestibulo-ocular reflex modification after virtual environment exposure". Acta Oto-Laryngologica, vol. 121, no. 2, pp. 211-215. https://doi.org/10.1080/000164801300043541. Retrieved 2026-10-15.
- ↑ Y.P. Ivanenko, I. Viaud-Delmon, I. Siegler, I. Israël, A. Berthoz (1998). "The vestibulo-ocular reflex and angular displacement perception in darkness in humans: adaptation to a virtual environment". Neuroscience Letters, vol. 241, no. 2-3, pp. 167-170. https://doi.org/10.1016/S0304-3940(97)00972-5. Retrieved 2026-10-15.
- ↑ C. Desoche, G. Verdelet, R. Salemme, A. Farnè, D. Pélisson, C. Froment, R. Hermann (2023). "Virtual reality set-up for studying vestibular function during head impulse test". Frontiers in Neurology, vol. 14, 1151515. https://doi.org/10.3389/fneur.2023.1151515. Retrieved 2026-10-15.
- ↑ A. Micarelli, A. Viziano, B. Micarelli, I. Augimeri, M. Alessandrini (2019). "Vestibular rehabilitation in older adults with and without mild cognitive impairment: Effects of virtual reality using a head-mounted display". Archives of Gerontology and Geriatrics, vol. 83, pp. 246-256. https://doi.org/10.1016/j.archger.2019.05.008. Retrieved 2026-10-15.
- ↑ A. Charlery-Adèle, C. Guigou, J. Ryard, M. Chartier, M. Toupet, C. Guillet, F. Mérienne, A. Bozorg Grayeli (2023). "Effects of saccade delay, side of deficit, and training on detection of catch-up saccades during head-impulse test in virtual-reality-enhanced mannequin". Scientific Reports, vol. 13, 2718. https://doi.org/10.1038/s41598-023-29801-8. Retrieved 2026-10-15.