“Mind-controlled” in context
The phrase “mind-controlled” suggests reading thoughts, which is not what early brain-computer interfaces did. They detected measurable electrical patterns from the brain, usually through the scalp, and learned to match some of those patterns with simple commands. The history here is about how that became possible. Work with electrodes placed in or on the brain came later and is a different story. Nothing here is medical advice; it describes research.
The first human EEG
The basic discovery came from the German psychiatrist Hans Berger. In 1924 he recorded electrical activity from a human head, using metal electrodes and a galvanometer, and he published the results in 1929. He called the recording an electroencephalogram. Berger observed regular rhythms, now called alpha waves, that changed when a person opened their eyes or concentrated. Other scientists doubted the findings at first, but by the mid-1930s they were confirmed and electroencephalography (EEG) became a standard clinical and research tool.
EEG showed that brain activity was measurable from the outside, but the signal is weak and mixed together. Electrodes on the scalp record activity averaged over large numbers of neurons, and the signal passes through skull and tissue, which blur detail. That limit shaped all non-invasive work that followed. Researchers later also tried other ways of measuring brain activity, including magnetic resonance scanning, but EEG stayed popular because it is cheap, portable and fast.
From signals to control
Between the 1940s and 1960s, researchers began to ask whether people could control their own brain rhythms. Experiments in biofeedback found that people could learn, to some degree, to increase or decrease certain rhythms when shown a display of them. A few artists and engineers even used EEG to trigger sounds in music. These were demonstrations, not communication systems, but they supported the notion that the signals could be consciously influenced.
In the late 1960s, the neuroscientist Eberhard Fetz, working with monkeys in Seattle, showed that animals could learn to raise and lower the firing rate of single neurons in the motor cortex when rewarded, as shown by a meter. This was an influential demonstration that brain activity could be voluntarily changed when feedback was provided.
The term “brain-computer interface”
In 1973, the computer scientist Jacques Vidal of the University of California, Los Angeles, published a paper that used the phrase “brain-computer interface” and asked whether EEG signals could be used to control external devices. He and his colleagues set up experiments in which visual evoked potentials, the brain’s responses to flashing stimuli, could be used to steer a cursor through a maze displayed on a screen. The work was funded by a United States defense research agency and was limited by the computers of the era. It is usually cited as the origin of the term and of the field.
The 1980s and 1990s
Progress was slow for many years, then picked up as personal computers made real-time signal processing practical. Several lines of work emerged.
- P300 spellers. In 1988, Lawrence Farwell and Emanuel Donchin described a system in which a person looked at a grid of letters while rows and columns flashed. When the flash included the desired letter, the brain produced a characteristic response about 300 milliseconds later. The computer could detect it and pick the letter. The method is slow, but it gave people a way to spell without movement.
- Mu-rhythm cursor control. Jonathan Wolpaw and colleagues at a New York State research center trained people to change the amplitude of rhythms over the motor cortex, associated with the sensorimotor region, to move a cursor up or down on a screen. By the 1990s, trained users had been able to move a cursor in one and then two dimensions, an approach similar to the idea behind pointing devices like the computer mouse but using brain activity instead of a hand.
- Slow cortical potentials. Niels Birbaumer’s group in Germany trained people to control slow shifts in the EEG. In the late 1990s they reported a “thought translation device” that allowed a person with advanced paralysis to select letters. This was one of the first demonstrations of a brain-computer interface helping a person who could not move to communicate.
During this period, researchers also began to hold meetings and set standards. The first international meeting on brain-computer interface research was held in 1999 in New York.
What had been shown and what had not
By the end of the 1990s, it had been established that people could, with training and feedback, use scalp EEG to select letters or move a cursor on a screen. It was also clear that these systems were slow, that performance varied considerably among individuals, and that the signals could change from day to day. Setting up electrodes required time, and gel-based electrodes were inconvenient outside a laboratory. Communication rates were usually a few letters per minute, far below normal typing.
Brain-computer interfaces at this stage were therefore research tools, mostly for studies with small numbers of participants. They did not control prosthetic limbs in any practical way. That step needed both better signal processing, often now described as a use of artificial intelligence methods, and a more direct way to record from neurons, which led to the implanted-electrode work of the following decades.