How People With Paralysis Play Video Games, From Eye Tracking to Brain-Computer Interfaces

People with paralysis play video games every day. How they do it depends on which movements remain. Someone with limited hand function might use an adaptive controller with oversized buttons. Someone who can move only their eyes might aim and select with an eye tracker. Someone with good breath control might play with a mouth-operated joystick that responds to sips and puffs. At the research frontier, a small number of people with implanted brain-computer interfaces have controlled games directly from signals recorded in the motor cortex.

These methods are not interchangeable. They differ in speed, precision, cost, fatigue and in how much the game itself has to accommodate the player. Understanding those differences explains why a brain implant is not simply a better joystick, and why the most important variable is often the design of the game rather than the input device.

Why gaming matters for people with paralysis

Gaming is sometimes treated as trivial next to medical priorities, but researchers who work with people with paralysis have found otherwise. In a 2025 study published in Nature Medicine, a Stanford-led team noted that people with paralysis report unmet needs for peer support, leisure and sport. For many people without disabilities, social media and online multiplayer games meet those needs. For the participant in that study, flying a virtual quadcopter was his top restorative priority, ahead of other everyday functions.

This matters for how the technology is judged. An input system that can type slowly but cannot keep up with a game fails at something users care about. Games are also a harsh benchmark. They demand speed, continuous control and several simultaneous actions, exactly where assistive input has historically been weakest.

The main ways to play, from least to most invasive

Adaptive controllers

The most widely used option involves no neurotechnology. The Xbox Adaptive Controller, released in 2018, is a flat unit with two large programmable buttons and a row of 3.5 mm ports. Almost any compatible switch, button, joystick or pedal can be plugged into those ports and mapped to a standard controller input. Sony's PlayStation Access controller, released in late 2023, takes a similar approach. It has a reconfigurable circular layout with swappable button caps and stick, and extension ports for external switches.

These controllers do not create new abilities. They move controller functions to whatever body parts a person can use reliably: an elbow, a knee, the chin, a foot. For many people with incomplete spinal cord injury, cerebral palsy or muscular dystrophy, that is enough to play mainstream games with little or no modification.

Mouth- and breath-operated controllers

For people with high cervical spinal cord injury, the head and mouth are often the most reliable remaining sources of movement. Mouth-operated controllers combine a small joystick moved with the lips or chin with sip-and-puff sensors, which detect sucking or blowing through a tube as distinct inputs. Several can be combined to cover most of a standard gamepad. Experienced users can reach impressive speed, but the method takes practice and can be tiring over long sessions.

Eye tracking

Eye trackers use infrared light and cameras to estimate where on the screen a person is looking. They are a standard communication tool for people with ALS and other conditions that leave eye movement intact, and they can also be used for gaming.

The difficulty is well known in the field: the eyes both look and select. A person needs to examine the screen without triggering everything they glance at. The usual solution is dwell selection: an action fires only when the gaze stays on a target for a set time, often several hundred milliseconds to a second. Blinking or pairing with a switch are alternatives. Dwell prevents accidental selections but adds a delay to every action. Eye tracking works well for strategy games, turn-based games, puzzle games and point-and-click adventures. It struggles with fast games that require quick aiming and simultaneous movement.

Switch access and scanning

When a person can reliably produce only one or two distinct actions, such as a head movement, a blink detected by a sensor, or pressure on a single button, they can use switch scanning. The system highlights options one after another, and the person presses the switch when the one they want is highlighted. It is extremely flexible and works with very little residual movement. It is also slow, because each choice takes as long as the scanning cycle needs to reach it.

Non-invasive brain-computer interfaces

Brain-computer interfaces that record from outside the head usually use EEG (electroencephalography), with electrodes on the scalp that detect the summed electrical activity of large populations of neurons. EEG games have been built around three main types of signal: imagined movement (motor imagery), responses to flickering visual stimuli, and the brain's reaction to an expected event.

EEG requires no surgery, but the signal is weak and blurred by passing through the skull. Most EEG gaming systems support only a few commands at relatively slow rates, and they need calibration and practice. Performance varies widely between people, and many users find it hard to operate motor-imagery systems reliably. For a person who can still control an eye tracker, EEG is rarely a faster or more precise alternative.

More advanced non-invasive methods also exist in the laboratory. In 2026, Yale researchers reported a system in which participants controlled a game through real-time fMRI, with brain activity measured inside a scanner. It is useful for studying how the brain learns to control an interface. It is not a home gaming technology, since fMRI requires a hospital-grade scanner.

What implanted brain-computer interfaces have actually shown

Implanted brain-computer interfaces record from electrodes placed in or on the brain. They pick up the activity of individual neurons or small groups of neurons, giving a much richer signal than EEG. Every implanted system described below is experimental. They are used only by research participants in clinical studies, not as a commercial product for gaming or for anything else.

The quadcopter study

The best-documented gaming result comes from the BrainGate2 clinical trial. It was published in Nature Medicine in January 2025 by Matthew Willsey, Jaimie Henderson, Francis Willett and colleagues at Stanford, with collaborators including Brown University and Massachusetts General Hospital. Its key features:

• One participant, a man with tetraplegia from a spinal cord injury, with two microelectrode arrays implanted in the region of the motor cortex that controls the hand.

• A decoder that turned his neural activity into continuous, independent movement of three finger groups: the thumb, the index and middle fingers together, and the ring and little fingers together. The thumb moved in two directions, giving four degrees of freedom in total.

• In target-acquisition tests he reached an average of 76 targets per minute, about 1.6 seconds per target.

• The decoded finger positions were then mapped to the controls of a virtual quadcopter, which he flew through obstacle courses.

The study matters because it achieved simultaneous, continuous control of several dimensions, which gaming requires and which earlier interfaces struggled to provide. The authors reported that the participant expressed a sense of enablement, recreation and social connection.

The limits matter too. This was one person, in a research setting, supported by a scientific team, using a system with wired connections through the skull. It shows what is possible, not what is available.

Company demonstrations

Neuralink has shared videos of participants in its PRIME feasibility study playing computer games, including chess and strategy titles, using its wireless implant to control a cursor. These are real demonstrations by implanted participants, but they come mainly from company announcements and livestreams rather than peer-reviewed publications. They should be weighed accordingly: they indicate capability, but the performance data have not been independently analysed and published in the same way as the Stanford results.

What remains unsolved

Several obstacles separate these results from everyday use.

• Surgery and risk. Implants require neurosurgery, with the associated risks of infection and complications. Weighing that risk makes sense for people with severe paralysis, but it makes brain implants unsuitable as a general gaming technology.

• Decoder drift. The relationship between neural signals and intended movements changes over days and weeks, so decoders need recalibration. Reducing that burden is an active area of research.

• Durability. How long implanted electrodes keep recording high-quality signals over years is still being studied in long-term follow-up.

• Access. No implanted brain-computer interface is currently approved for sale for any purpose. Access is limited to trial participants.

The overlooked factor, game design

The input device is only half of the problem. The other half is the game's own design, and here the gap between research and daily life is widest.

Every assistive input method trades speed for reliability. Dwell selection adds a delay to each action. Switch scanning adds a longer one. Even the best brain-computer interfaces need time to register an intended action reliably. Games that assume a player can press a button within a fraction of a second, hold several inputs at once, or react to sudden events exclude these players, no matter how good the hardware is.

The same principle applies beyond games to other interactive content. The Web Content Accessibility Guidelines, the standard many regulators refer to for digital accessibility, include a requirement that users be able to turn off, adjust or extend time limits. The requirement exists because timed interactions block people who operate a computer slowly. Some interactive formats are fundamentally built around timing, and cannot offer that flexibility without changing what they are. Among the fast browser games common across digital entertainment, many resolve an entire round in one timed decision. The crash-style instant games listed on the Crazytower site are one example, where the outcome depends on when the player acts. For someone who needs a second or more of dwell time to register a single selection, a format that turns on split-second timing is effectively inaccessible, however capable the input device. Real-money games also raise separate questions that no accessibility feature solves, including age restrictions, laws that vary by country and the risk of gambling harm. From the neurotechnology perspective, the point is narrower: a game's timing model determines who can play it.

Game studios have started to take this into account. Widely used industry resources such as the Game Accessibility Guidelines recommend full remapping of controls, options to hold or toggle inputs, adjustable game speed, generous timing windows, and the option to skip or simplify quick reaction sequences. Several major titles now include extensive accessibility menus. The result is that a slow input method can still reach the ending of many modern games. That would have been rare a decade ago.

Comparing the options

For readers weighing options, the trade-offs look roughly like this:

MethodWhat it needs from the userStrengthsMain limitsAdaptive controllerReliable movement in any body partWorks with mainstream games, affordable, flexibleRequires some voluntary movementMouth/breath controlHead, lip and breath controlCovers many controller functionsFatiguing, takes practiceEye trackingControlled eye movementPrecise aiming in slower gamesDelay from dwell selection, trouble with fast actionSwitch scanningOne or two reliable actionsWorks with minimal movementVery slowEEG-based BCINo movement neededNon-invasiveFew commands, variable reliability, calibrationImplanted BCINo movement neededRichest signal, multidimensional controlSurgery, experimental, trial access only

In practice, many people combine methods. They might aim with an eye tracker and use a switch to fire, or use an adaptive controller with a mouth-operated joystick. Occupational therapists and specialist charities often help find the combination that uses a person's most reliable movements while keeping fatigue manageable.

What comes next

The near future of accessible gaming will be shaped mostly by gradual improvements: better eye trackers, more configurable controllers, and games designed from the start with adjustable timing and remappable inputs. Those changes reach many people quickly and carry no medical risk.

Brain-computer interfaces are on a slower and more uncertain path. Clinical trials of several implanted systems are growing from single participants to dozens. Games are likely to remain one of the standard ways researchers test and demonstrate multidimensional control, because they are demanding, measurable and motivating. Whether any implanted system becomes an approved medical device, and whether gaming would be among its supported uses, depends on data not yet collected, especially on long-term safety and reliability.

For now, the realistic answer to how people with paralysis play video games is a toolbox rather than a single technology. Most of the tools are available today. Brain-computer interfaces are an important research addition, not a replacement.

---

This article is for general information about assistive technology and neurotechnology research. It is not medical or clinical advice. People interested in assistive input or in joining a brain-computer interface trial should speak with an occupational therapist, rehabilitation specialist or the relevant clinical research team.