Electrooculography

Electrooculography (EOG) is a method of measuring eye movements and retinal function by recording, with electrodes on the skin around the eyes, the standing electrical potential between the front and the back of the eye. The cornea is electrically positive relative to the back of the eye, so the eye behaves roughly like a dipole. When it rotates, the voltage between a pair of periorbital electrodes changes, and the recorded signal (the electrooculogram) follows the movement. The same standing potential is the basis of a standardized clinical test of the outer retina and retinal pigment epithelium.

EOG is one of the oldest ways to record eye movements. It needs only electrodes and an amplifier, works in darkness and with the eyes closed, and can be sampled quickly, but it is less spatially accurate than camera-based eye tracking and its baseline drifts over time. In virtual reality and augmented reality it appears mainly in research: as a hands-free input channel for games and interfaces, as a fast saccade detector for redirected walking experiments, as a reference signal for measuring the latency of headset eye trackers, and as one of the biosignal sensors built into research faceplates and headsets such as Emteq's Faceteq and OpenBCI's Galea.

Reviewed 13 October 2026. Checked the EOG physiology, ISCEV clinical test, history papers, JINS MEME dates and VR research studies against the cited textbook, standard, papers and press reports. About review dates.

How it works

The standing potential

There is a difference in electrical potential between the anterior and posterior parts of the eye, known as the standing potential of the eye. According to the 2017 standard of the International Society for Clinical Electrophysiology of Vision (ISCEV), it is an indirect measure of the transepithelial potential of the retinal pigment epithelium (RPE), and it changes when the retina adapts to a change in ambient light. Jaakko Malmivuo and Robert Plonsey's textbook Bioelectromagnetism gives the magnitude of this corneoretinal potential as 0.4 to 1.0 mV and notes that it can be detected in total darkness and with the eyes closed.

Electrode placement and signal

For horizontal movements, one electrode is placed at the outer corner (canthus) of each eye. With the eye looking straight ahead the two electrodes are at about the same potential. When the eye turns to the right, the electrode on the right becomes positive relative to the other, and a turn to the left produces the opposite signal. Vertical movements are recorded with electrodes above and below the eye. Malmivuo and Plonsey give typical signal magnitudes of 5 to 20 microvolts per degree of rotation, a typical achievable accuracy of about plus or minus 2 degrees, and a maximum rotation of plus or minus 70 degrees, with linearity becoming progressively worse beyond 30 degrees. The signal is calibrated by having the person look at two fixation points a known angle apart. The ISCEV standard states that a 30 degree saccade typically produces an EOG amplitude of 250 to 1000 microvolts, with essential frequency content from 0 to 30 Hz, and recommends sampling saccades at 1 kHz or more per channel.

Artifacts and limitations

The main drawback for eye tracking is that the corneoretinal potential is not fixed. It varies over the day and is affected by light, fatigue and other factors, so recordings need frequent recalibration; muscle artifacts and the basic nonlinearity of the method add further error. Won-Du Chang's 2019 review of EOG for human-computer interaction lists the usual unwanted components as electromyographic (EMG) activity from jaw clenching, eyebrow movements, smiling or walking, slow baseline drift, high-frequency noise, and crosstalk between the horizontal and vertical channels. Blinks produce a strong deflection above the eyes because the eyelid conducts the positive charge of the cornea toward the forehead. Chang notes that blinks are an artifact for eye-position estimation but a useful, easily distinguished signal for interfaces, for example to select a button. The ISCEV standard recommends AC coupling with a 0.1 Hz high-pass filter in clinical recordings because direct (DC) amplification, though ideal in theory, suffers from baseline drift.

History

Malmivuo and Plonsey date the observation that the cornea is electrically positive relative to the back of the eye to Emil du Bois-Reymond in 1848. Because the potential was not affected by light, it was at first thought of as a resting potential. In the mid-1930s Mowrer, Ruch and Miller published "The corneo-retinal potential difference as the basis of the galvanometric method of recording eye movements" in the American Journal of Physiology; a 2023 review of EOG-based interaction describes their finding, that electrodes placed around the eyes record potential changes caused by eye movements, as the foundation of EOG research. Elwin Marg described the development of the technique in a 1951 paper titled "Development of electro-oculography; standing potential of the eye in registration of eye movement".

In 1962 G. B. Arden, A. Barrada and J. H. Kelsey described a "new clinical test of retinal function based upon the standing potential of the eye" in the British Journal of Ophthalmology. The test's main measure became known as the Arden ratio. ISCEV first published a standard for clinical electro-oculography in 1993 and revised it in 2011; its 2017 update calls the Arden ratio the light peak to dark trough (LP:DT) ratio.

In the late 2000s, Andreas Bulling, Daniel Roggen and Gerhard Tröster built wearable EOG goggles as an embedded alternative to camera-based trackers for activity recognition and mobile interaction. In a 2011 study in IEEE Transactions on Pattern Analysis and Machine Intelligence, Bulling, Jamie Ward, Hans Gellersen and Gerhard Tröster detected saccades, fixations and blinks from EOG and used 90 eye-movement features to recognize five office activities (copying text, reading, writing notes, watching a video and browsing the web). With person-independent training over eight participants, they reported an average precision of 76.1 percent and recall of 70.5 percent.

Clinical use

In the ISCEV clinical EOG, the patient tracks two red fixation lights 15 degrees left and right of center, making repeated horizontal saccades while electrodes near the inner and outer canthi record the amplitude of each movement, which is proportional to the standing potential. Recordings are made through 15 minutes of darkness, during which the potential falls to a minimum (the dark trough, usually at 10 to 15 minutes), and then 15 minutes of light adaptation, during which it rises to the light peak, usually 7 to 12 minutes after light onset. The LP:DT ratio is affected in some diffuse disorders of the RPE and of the photoreceptor layer, usually in proportion to the severity of rod-mediated electroretinogram (ERG) abnormalities. Notable exceptions are disorders of the bestrophin gene (BEST1): in Best vitelliform macular dystrophy (Best disease), full-field ERGs are usually normal while the LP:DT ratio is abnormal.

EOG is also used to record nystagmus. Malmivuo and Plonsey describe nystagmography as a clinical tool for investigating the vestibular system. Because visual fixation suppresses vestibular eye movements, the examination is usually done in darkness or with the eyes closed, which makes EOG the method of choice.

Comparison with camera-based eye tracking

Stein and colleagues describe video-based eye tracking as today's most used method; Mishra and colleagues note that eye tracking can be built into a VR headset by placing infrared cameras near the lenses. EOG measures an electrical field instead, which leads to a different set of strengths and weaknesses.

Aspect	Electrooculography	Video-based eye tracking
What is measured	Changes in the corneoretinal potential picked up by skin electrodes	Camera images of the eye
Spatial accuracy	About plus or minus 2 degrees typical (Malmivuo and Plonsey); weaker on the vertical axis (Stein et al.)	Advertised accuracy down to 0.15 degrees for a stationary research tracker, per Stein et al.
Speed	Can be sampled at high rates; ISCEV recommends 1 kHz or more for clinical saccade recording	Headset trackers generally sample at around 100 Hz, per Stein et al. (2021)
Eyes closed or darkness	Works	Needs a visible pupil
Main error sources	Baseline drift, diurnal and light-dependent changes in the potential, facial muscle (EMG) artifacts, nonlinearity beyond about 30 degrees	Eyelids and eyelashes covering the pupil
Cost and hardware	Chang's review cites a complete recording system built for less than 100 euros	Infrared cameras placed near the headset lenses

Chang's review also notes that EOG signals are recorded with the eyes closed and in blind people, and cites an achievable precision of up to 1.5 degrees. Tao, Huang and Chen's 2023 review lists further advantages: the signal is less affected by lighting and is not limited by makeup, contact lenses or eyelashes. Its disadvantages are sensitivity to muscle activity and movement noise, and the drying of gel in wet electrodes over time.

Eyewear and wearable EOG

In a 2015 IEEE EMBC paper, Shin'ichiro Kanoh and colleagues described eyewear that measured horizontal and vertical EOG with three metal dry electrodes placed near the nasion and on both sides of the nose, positions that correspond to the bridge and nose pads of a frame, together with a six-axis motion sensor. Chang's review notes that a product with this electrode placement was commercialized by a Japanese company.

That product was JINS MEME. The Japanese eyewear company JIN Co., Ltd. announced it in May 2014 as eyewear with a three-point EOG sensor reading from the nose pads and between the eyebrows, able to measure line-of-sight shifts in eight directions and blinks in real time. The JINS MEME ES model, with the three-point EOG sensor and a six-axis sensor, went on sale in Japan on 5 November 2015 at 39,000 yen before tax; the cheaper MT sports model left out the EOG sensor. JINS ended general consumer sales of JINS MEME on 27 March 2024, after which it was sold only to businesses; existing users were promised app support until the end of August 2025.

Applications in VR and AR

Input and interaction

Chang's review describes gaming in VR, where conventional input devices are not easy to use, as an application field for EOG, and several research groups have tested eye movements recorded by EOG as a hands-free input with a head-mounted display; in such interfaces saccades are often used for movement commands and blinks for selection.

• Devender Kumar and Amit Sharma (ICACCI 2016) built a VR game, "VRrailSurfer", controlled by EOG blink detection and gaze calibration. They reported average efficiencies of 96 percent for online blink detection and 80 percent for gaze calibration, and an average real-time game control accuracy of 78 percent across ten subjects.
• Junichi Shimizu and George Chernyshov (UbiComp 2016 adjunct) added the EOG module from disassembled commercial smart glasses to a Google Cardboard viewer, presented a first signal-level evaluation, and outlined concepts for VR interaction techniques based on relative eye movements.
• Jing Xiao, Jun Qu and Yuanqing Li (IEEE Access, 2019) used a single-channel EOG signal to drive a VR interface in which buttons flash in random order and the user blinks in time with the target. They built a music-on-demand system on it and tested it online with ten healthy subjects and five patients with spinal cord injuries.

Saccade detection for redirected walking

During a saccade the visual system suppresses perception, so small changes to a scene made during the eye movement can go unnoticed. To exploit this in VR, the scene change must be time-locked to saccade onset. Benjamin Bolte and Markus Lappe (IEEE TVCG, 2015) combined an EOG amplifier with a Sensics zSight HMD for this purpose; they wrote that fast video-based trackers were too bulky to use with an HMD and that smaller ones were not fast enough, whereas EOG amplifiers are small with high temporal resolution. Their setup placed an electrode at the outer canthus of each eye, and they adapted two eye-movement classification algorithms for online saccade detection. Participants could not detect translations of about 0.5 m along the line of gaze or rotations of about 5 degrees in the transverse plane during 15 degree saccades. The authors proposed the approach as a way to improve redirected walking and change blindness techniques.

Reference for eye-tracker latency

EOG can also serve as a timing reference. Niklas Stein and colleagues (i-Perception, 2021) recorded EOG at the same time as the built-in eye trackers of the FOVE 0, the Varjo VR-1 and the HTC Vive Pro Eye, while participants made 60 saccades between targets 20 degrees apart. Measured against the EOG, tracker delays ranged from 15 ms to 52 ms and end-to-end latencies from 45 ms to 81 ms. The authors wrote that the Fove-0 appeared to be the fastest device and the best suited to gaze-contingent rendering. EOG itself showed no delay relative to an EyeLink 1000 research tracker, whose advertised latency is below 2 ms, from which they concluded that the EOG delay is similarly low.

Biosignal faceplates and headsets

In 2016 Road to VR reported on Faceteq, a faceplate system from Emteq that used EMG, EOG, heart rate and other sensors to register eye, forehead and cheek muscle movements underneath a VR headset such as the Oculus Rift. OpenBCI's Galea combines EEG, EOG, EMG, electrodermal activity and PPG sensors; in the Varjo Aero version, the facepad carries two EOG channels, and eye tracking comes from the Varjo Aero itself. Road to VR reported the Varjo Aero version in May 2022.

Vision therapy

Saswat Mishra, Woon-Hong Yeo and colleagues (Science Advances, 2020) built a wireless system of soft, skin-conformal EOG sensors made by aerosol jet printing and paired it with a Samsung Gear VR headset running vergence training programs, aimed at portable therapy for disorders such as convergence insufficiency. They argued that camera-based tracking inside a headset depends on clear images of the pupil, which eyelashes and eyelids can block. In their comparison of electrode placements, averaged over three subjects, the best configuration classified vergence movements with 91 percent accuracy, and the full system was demonstrated with 14 human subjects.

See also

• Eye tracking
• Saccade
• Electromyography
• Electroencephalography
• Brain-computer interface
• Redirected walking
• OpenBCI