There are things I wish to say in my heart but do not intend to express verbally yet. Research results have recently emerged indicating that soon, language conducted only in the mind can also be decoded by computers. This will open a pathway for paralyzed patients, who find even attempting to speak with their mouths uncomfortable, to easily convey their thoughts.
A team led by Professor Francis Willett from Stanford University's School of Medicine and Dr. Erin Kunz from the Department of Electrical Engineering noted on the 15th in the international journal Cell that they succeeded in decoding internal language—the thoughts spoken in the mind—using brain-computer interface (BCI) technology with an accuracy of up to 74%.
BCI is a technology that decodes brain signals to control machines or computers. Until now, it has primarily involved reading the neural signals generated when attempting to move arms and legs, or interpreting attempts by paralyzed patients to use their muscles to speak in order to input sentences. However, since even attempts to speak require muscle use, BCI has still been a slow and exhausting process for some patients.
Instead of capturing signals when an individual attempts to speak verbally, the research team experimented to see if they could decode thoughts spoken only in the mind. They implanted microelectrodes in the motor cortex of the brains of four severely paralyzed patients, prompting them to either try to pronounce specific words or to imagine saying those words. The analysis showed that, in both cases, similar brain areas were activated, but as expected, the signal strength while imagining words was relatively weak.
The research team trained artificial intelligence (AI) on the brain signal data obtained in this way. As a result, they were able to decode sentences composed of up to 125,000 words with an accuracy rate of up to 74%. Some participants counted numbers while sending brain signals, and the system was able to capture this as well.
They also tested security features. The setup required participants to think of the phrase "Chitty Chitty Bang Bang" before decoding would begin, and the system recognized this with an accuracy of over 98%.
Dr. Kunz, who led the study, said, "This is the first case of understanding what happens in the brain when someone speaks only internally without voicing it out loud," adding that it could provide a more natural and comfortable means of communication for patients with serious language or motor disabilities.
The research team acknowledges that the technology developed this time has limitations in reading every sentence thought without error, but they believe that applying more sensors and sophisticated algorithms will expand its possibilities. Professor Willett stated, "One day, a time will come when communication can occur just by thinking as naturally as conversing."
References
Cell (2025), DOI: https://doi.org/10.1016/j.cell.2025.06.015