Why go inside the head
Scalp EEG, the earlier approach, records a blurred mix of activity. Electrodes placed on or in the brain can record much more detail, including the firing of individual nerve cells. The tradeoff is surgery. Everything here is research, performed in studies with small numbers of volunteers, and it is described as history, not as an available treatment.
Animal groundwork
By the 1990s, neuroscientists had shown that neurons in the motor cortex fire in patterns related to the direction of arm movements. Microelectrode arrays, small grids of tiny needles, were developed to record from dozens or hundreds of neurons at once. In 2000, a group led by Miguel Nicolelis and John Chapin reported in a major journal that monkeys with implanted arrays could control a robotic arm, using neural activity recorded in their brains. Other labs showed monkeys using brain activity to move a cursor and later to feed themselves with a robotic arm. These studies provided the basis for human trials.
First human implants
In the late 1990s, the neurologist Philip Kennedy implanted a “neurotrophic electrode”, a glass cone containing wires and growth factors, in the motor cortex of a man with locked-in syndrome. The patient, known publicly as Johnny Ray, was reported to have learned to move a cursor to select letters, though communication was slow. It was one of the first reports of an implanted brain-computer interface in a person.
A larger step came in the mid-2000s. A team at Brown University and collaborators, who later formed the BrainGate research group, implanted a small silicon array of about a hundred electrodes into the motor cortex of a man with tetraplegia, Matthew Nagle. Their 2006 report described how he was able to move a cursor, open emails, and operate a simple robotic hand using neural activity. The array was connected to equipment through a socket in the skull, so the person had to be linked by cable.
Robotic arms
In 2012, the BrainGate team reported that two people with long-standing paralysis had used a robotic arm to reach for and grasp objects, with one participant bringing a bottle of coffee to her mouth. A group at the University of Pittsburgh then reported on a woman with tetraplegia who used implanted arrays to control a robot arm with several dimensions of movement, including hand shaping, and to carry out tasks such as picking up objects. These were widely reported in the press, but they were carried out in laboratories with technical staff and with specialist equipment, and the arm was not an everyday device.
Researchers also explored feedback. By stimulating the sensory cortex with small currents, studies reported that participants could feel touch-like sensations in the hand, which could help with grasping. The work is early and based on a small number of participants.
Communication and decoding
In the 2010s and 2020s, attention moved to communication for people who cannot speak or move. Experiments reported that a participant who imagined writing letters by hand could have that activity decoded into text, at speeds that were far faster than earlier cursor-based typing, though still below natural handwriting. Other work, using electrodes on the brain’s surface or arrays in speech areas, decoded attempted speech into text or synthesized voices for people who had lost the ability to talk. The decoding relies on machine learning and artificial intelligence methods trained on each participant’s data over weeks or months.
Open questions
Many basic questions remain unanswered.
- Durability. Electrodes can lose signal over months or years because of scar tissue and movement. Some participants have kept useful signals for years, while others have not.
- Safety. Any brain surgery carries risks such as infection or bleeding. Long-term safety data for people are limited.
- Wireless operation. Early systems needed a cable. Wireless and fully implanted systems are in development, but few people have used them.
- Daily use. Decoders usually need frequent recalibration, and performance varies with fatigue, medication and mood.
- Generalization. Most studies are small and involve a few volunteers. Results may not apply to all people or conditions.
- Ethics. Questions about consent, privacy of brain data, who pays, and what happens to participants when a study ends are widely discussed.
Several companies have since entered the field, and some have begun early human trials. Because claims in this area move quickly, it is wise to read results in peer-reviewed reports and to note how many participants were involved.
Where this sits among devices
Implanted brain interfaces are one of several approaches to restoring function. Some use signals from the muscles, as in myoelectric arms. Some use nerve signals, and others use implantable devices that stimulate tissue, like the cochlear implant and retinal implants. The brain-based route has the highest potential bandwidth and the highest barrier, and for now it remains a research effort whose future usefulness is uncertain.