Electroencephalography

Electroencephalography (EEG) is the non-invasive measurement of the brain's electrical activity with electrodes placed on the scalp. The electrodes record voltage potentials that result from current flow in and around neurons. The German psychiatrist Hans Berger made the first recordings of electrical brain signals in humans in 1924 at the University Hospital of Jena.

EEG is the most commonly used brain-sensing method in brain-computer interface (BCI) research, because of its high temporal resolution, low cost, portability and non-invasiveness. Immersive head-mounted displays have been combined with EEG across many fields of research. In VR and AR it is used in two ways: researchers record EEG while people use a headset to study presence, emotion and cybersickness, and BCI systems use EEG signals as an input channel to select objects, navigate or control a virtual body without hand controllers. EEG hardware for VR headsets has been demonstrated or sold by Neurable, Looxid Labs and OpenBCI.

Reviewed 6 October 2026. Checked every citation's bibliographic data against Crossref and each claim against the abstracts, full texts (Rashid 2020, Brake 2025, Nunez and Srinivasan, Tauscher 2019), press articles, the archived Looxid Link page, the Galea brochure and store page, and the Colorado bill page. About review dates.

How it works

EEG records voltage fluctuations caused by the flow of ionic current during synaptic activity in the brain's neurons. According to Paul Nunez and Ramesh Srinivasan, most EEG signals originate in the cerebral cortex, where cortical synaptic action generates electrical signals that change in the 10 to 100 millisecond range. Each scalp electrode records activity at a very large scale, summing potentials generated in tissue that contains something like 10 million to one billion cortical neurons. Because of severe space averaging between the brain's current sources and the scalp electrodes, scalp EEG reflects only large-scale activity. EEG and magnetoencephalography (MEG) are the only widely available technologies fast enough to follow these changes, but their spatial resolution is poor compared with structural imaging methods such as MRI, CT and PET.

The measured signal is the voltage difference between an active electrode and a reference electrode over time. Its amplitude is measured in microvolts and generally falls between -100 and +100 microvolts. Human EEG electrodes are typically 0.4 to 1.0 cm in diameter and are held on the scalp with pastes, caps or nets. In standard clinical practice, 19 recording electrodes are placed uniformly over the scalp according to the International 10-20 System, with one or two reference electrodes (often on the ear lobes) and a ground electrode. Research systems use more channels; the number of electrodes on EEG headsets ranges from 1 to 256. Conventional "wet" electrodes use a conductive gel or paste to improve contact with the skin, which makes setup slow. Dry electrodes that need no gel have been validated, but in BCI use they have achieved, on average, a maximum information rate about 30 percent lower than gel-based electrodes.

Rhythms and evoked potentials

Ongoing EEG is usually described by frequency band. The review by Rashid and colleagues lists the following bands; exact boundaries differ slightly between studies (for example, some define alpha as 8-12 Hz).

Band	Frequency	Associated state (as listed by Rashid et al.)
Delta	0.5-4 Hz	Deep sleep
Theta	4-8 Hz	Drowsiness, light sleep
Alpha	8-13 Hz	Relaxed
Beta	13-30 Hz	Active thinking, alert
Gamma	Above 30 Hz	Hyperactivity

Signals can also be time-locked to events. Averaged evoked potentials are responses to sensory stimuli such as light flashes or tones; event-related potentials (ERPs) are recorded the same way but occur at longer latencies and are more closely tied to the person's internal state. ERP components typically occur less than 500 milliseconds after the stimulus. BCIs use several such control signals. The P300 appears about 300 ms after a stimulus is presented, with a larger peak for less probable stimuli, and P300-based BCIs do not require training. Steady-state visually evoked potentials (SSVEPs) appear when a user looks at a stimulus flickering at a fixed frequency, typically between 6 and 30 Hz, and the brain response carries the same frequency. Motor imagery (imagining a movement) changes sensorimotor rhythms over the motor cortex, which can be detected as event-related desynchronization or synchronization.

Raw EEG contains artifacts that must be removed before analysis. Above about 30 Hz, most of the unprocessed scalp signal reflects muscle activity rather than brain activity.

History

The first known neurophysiologic recordings in animals were made by Richard Caton in 1875; recording from humans took another half century. Hans Berger first recorded electric signals from the human brain through scalp electrodes on 6 July 1924. In a centennial review, past and present officers of the International Federation of Clinical Neurophysiology (IFCN) described this date as marking the beginning of electroencephalography. He identified alpha waves in scalp recordings, and his work was met with skepticism before it was accepted. Berger published his first report, "Über das Elektrenkephalogramm des Menschen", in the Archiv für Psychiatrie und Nervenkrankheiten in 1929. His subjects showed clear changes in the signal when they closed their eyes and when they did mental arithmetic. The scientific community at first doubted that the signals came from brain tissue, but their brain origin had been established by 1934.

In clinical practice EEG is particularly useful for evaluating patients with suspected seizures and epilepsy, and it is also used to monitor the depth of anesthesia during surgery.

In a 2000 paper, Bayliss and Ballard argued that virtual reality could extend the range of possible BCI prototypes and showed that cognitive evoked potentials could still be recorded reliably while subjects moved and acted normally in an immersive virtual environment, with a single-trial accuracy of 85 percent in telling apart the responses to red and yellow traffic lights. A group including Gert Pfurtscheller, Robert Leeb and Mel Slater reported in 2006 that participants in a CAVE projection system could move along a virtual street only by imagining foot movements, without muscular activity.

Use with VR and AR headsets

Practical challenges

Wearing an EEG cap and a headset at the same time causes mechanical problems. In a study presented at IEEE VR 2019, Tauscher and colleagues, mostly at TU Braunschweig, noted that the head straps of an HTC Vive press on the sensors, displacing them and twisting their cables, which degrades data quality. Using a cap with 16 active electrodes, they found that EEG and VR could be combined without modification under certain conditions, that a custom strap leaving room for the electrodes improved signal quality, and that the display latency difference between the headset and a monitor or dome (about 0.05 s) was visible in the brain responses.

Headsets are also a source of electromagnetic noise. A 2021 study by Weber and colleagues tested an Oculus Rift and an HTC Vive Pro with 64-channel EEG. Both headsets consistently introduced artifacts, especially at the 50 Hz mains frequency, at the 90 Hz display refresh rate and at their harmonics, but the frequency range below 50 Hz that matters most in non-invasive EEG research stayed largely unaffected. The authors recommended testing every new headset in advance, since each is likely to have its own electromagnetic footprint.

The headset itself can also be an advantage. At the Game Developers Conference in March 2019, Valve experimental psychologist Mike Ambinder said that measuring brain signals requires "a way to get people to wear a helmet. If only us as game designers had a way of doing that," while showing a slide of an HTC Vive headset. According to Road to VR, he compared EEG data to sitting outside a football stadium and trying to work out what is happening on the field from the intensity of the crowd's reaction; the publication described EEG as one of the noisiest current BCI technologies because it picks up neuronal signals through the skull, scalp and hair.

Hardware

Product	Maker	Headsets	EEG sensing
Awakening demo strap (2017)	Neurable	Modified HTC Vive	Seven dry electrodes from Wearable Sensing in a replacement strap
Looxid Link	Looxid Labs	HTC Vive, HTC Vive Pro (including Vive Pro Eye), Oculus Rift S	Six channels of gold-plated sensors over the prefrontal (forehead) area, 500 samples per second
Galea (announced 2020)	OpenBCI	Varjo Aero, Varjo XR-3	10 channels of dry, active electrodes (2024 brochure)

Neurable unveiled Awakening, which IEEE Spectrum described as the first brain-controlled VR game, at SIGGRAPH 2017 in Los Angeles. It was a collaboration with the Madrid-based VR graphics company estudiofuture. The system replaced the Vive's elastic strap with one carrying seven dry electrodes and read event-related potentials rather than general brainwave patterns, so a player could select objects by focusing on them; CEO Ramses Alcaide said the approach needed "basically no training" compared with motor-imagery methods.

The Looxid Link from Looxid Labs is an add-on mask for existing PC VR headsets. Its sensors read EEG from the prefrontal (forehead) area, and its software reports attention, relaxation and "brain balance" values plus delta, theta, alpha, beta and gamma activity every 100 ms through a Unity API.

Galea is OpenBCI's sensor platform for mixed reality headsets. Road to VR reported its announcement in November 2020 as hardware that attaches to AR and VR headsets and combines EEG with electrooculography (EOG), electromyography (EMG), electrodermal activity (EDA) and photoplethysmography (PPG) sensors. In February 2021 Tobii confirmed it was working with Valve and OpenBCI on developer units that combined Tobii eye tracking with design elements of the Valve Index, then expected for beta partners in early 2022. Speaking about brain-computer interfaces in general rather than Galea specifically, Valve co-founder Gabe Newell had told 1 News that "If you're a software developer in 2022 who doesn't have one of these in your test lab, you're making a silly mistake." In May 2022 OpenBCI announced a partnership with Varjo; Road to VR reported, citing a screenshot obtained by tech analyst Brad Lynch, that the Varjo Aero and Galea kit appeared to be priced at US$22,500, with shipping planned in five batches starting from an earliest estimated date of August 2023. OpenBCI's 2024 brochure lists 10 dry, active EEG channels, two EOG channels, two EXG channels, four facial EMG channels, an ear-clip PPG sensor, a forehead EDA sensor and Varjo eye tracking. The unit is battery powered, sends data over Wi-Fi, and the headset can be removed for use outside VR. In October 2026 OpenBCI's online store still listed Galea, with an option to request a quote.

Research applications

Brain-computer interfaces in virtual environments

VR has long been used as a testbed for EEG-based BCIs. In 2007, Leeb and colleagues trained ten naive subjects in three sessions to navigate freely through a virtual apartment by motor imagery, deciding at each junction where to go. Only three EEG channels were used to turn the imagined movements into navigation commands, and the study found that motivated subjects performed much better than unmotivated ones.

Head-mounted displays have also been used to present BCI stimuli. Käthner, Kübler and Halder compared a P300 speller shown on a 22-inch monitor with two VR headset layouts in 18 healthy participants. Average online spelling accuracy was 94 percent with the monitor and the first headset layout and 96 percent with the second; a person in the locked-in state reached 100 percent accuracy in one session with the headset. For augmented reality, Si-Mohammed, Lécuyer and colleagues found that an EEG headset and a Microsoft HoloLens were compatible and that small head movements could be tolerated. They proposed a design space for SSVEP command menus in optical see-through displays and built a prototype in which a real mobile robot was controlled in AR through the BCI and the HoloLens.

In rehabilitation, the REINVENT platform described by Vourvopoulos and colleagues detects post-stroke EEG signals that indicate an attempt to move and uses them to drive a virtual avatar arm in VR. In a pilot with four chronic stroke patients, the system was used safely over repeated sessions; the authors described their results as preliminary because of the small sample.

Presence and emotion

EEG has been used to look for neural correlates of presence. A 2006 study by Baumgartner and colleagues, which described itself as the first to investigate the neurophysiology of spatial presence, recorded children and adolescents watching virtual roller coaster rides and found activation in parietal brain areas involved in spatial navigation. Kober, Kurzmann and Neuper compared a large-screen single-wall VR system with a three-dimensional view against a desktop system during navigation in a virtual maze. The more immersive system produced stronger presence and a larger parietal decrease in alpha power, while the desktop group showed stronger frontal-parietal connectivity. In a 2021 eLife study, Hofmann and colleagues recorded EEG from 37 adults during an immersive VR experience that included roller coaster rides and confirmed a link between subjective emotional arousal and parieto-occipital alpha power, decoding high- and low-arousal periods with machine learning.

Cybersickness

EEG is studied as an objective measure of cybersickness. In a 2005 study of 61 participants who navigated a virtual environment for 9.5 minutes, Kim and colleagues found that sickness severity correlated positively with EEG delta activity and negatively with beta activity, along with changes in gastric activity, eyeblink rate and heart period. A 2021 test-retest study with 21 sickness-susceptible men, tested one week apart with VR video designed to cause sickness, found that delta, theta and alpha activity over frontal and central areas differed significantly between baseline and VR sickness in both sessions and showed good test-retest reliability.

Games

In his 2019 GDC talk, Ambinder described possible uses of brain data in games, including variable game difficulty, AI responses tailored to the player's state of mind and replacing traditional input methods. When Galea was announced, Road to VR reported OpenBCI's aim of measuring states such as happiness, anxiety, attention span and interest level to personalize immersive experiences.

Neural data and privacy

Because EEG-equipped headsets collect data about the nervous system, they fall under new privacy laws in some jurisdictions. Colorado's HB24-1058, signed on 17 April 2024 and effective 7 August 2024, amended the Colorado Privacy Act to add biological data, including neural data, to its definition of sensitive data. The law defines neural data as information generated by measuring the activity of a person's central or peripheral nervous systems that can be processed by or with the assistance of a device. See Privacy in virtual and augmented reality for the wider topic.

See also

• Brain-computer interface
• OpenBCI
• Electromyography
• Eye tracking
• Presence
• Simulator sickness
• Neuralink