Scientists have vocalized the "inner voice" of people who cannot speak

Scientists at Stanford University have been able to decipher the brain activity associated with a silent "inner voice" in the heads of people with severe paralysis caused by amyotrophic lateral sclerosis (ALS) with up to 74% accuracy, Gizmodo reports.

The authors of the study, published in the journal Cell, believe that their work could potentially help people who cannot speak communicate using brain-computer interfaces (BCIs).

Previously, BCIs were used to decode speech attempts, where people who could not speak physically tried to speak out loud, using the muscles associated with speech. However, this method was tiring for people with limited muscle control. In the new study, BCIs are used to directly interpret the "inner voice."

"This is the first time we’ve managed to understand what brain activity looks like when you just think about speaking. For people with severe speech and motor impairments, BCIs capable of decoding inner speech could help them communicate much more easily and more naturally," says lead author Erin Kuntz.

The study involved four participants who had microelectrodes implanted in their motor cortex, the area of the brain responsible for controlling voluntary movements, including speech. The researchers were able to record activity in the motor cortex, and found that attempting and imagining speech activate similar, but not identical, patterns of brain activity.

The researchers then used a large language model to interpret these signals, decoding sentences from a vocabulary of up to 125,000 words with 74% accuracy. In some cases, the system even picked up spontaneous internal thoughts, such as numbers that participants silently counted during the task.

To prevent people using the new technology from vocalizing their every inner thought, the researchers added a system to start and stop the BCI using a special password. To start or stop the vocalization of the "inner voice," individuals had to think the phrase "Chitty chitty bang bang." The system successfully recognized it 98% of the time.

"The future of BCIs is bright. This work gives real hope that speech BCIs can one day restore communication that is as fluent, natural, and comfortable as conversational speech," says Frank Willett, another author of the study.

In general, in recent years, the topic of neurocomputer interfaces has become increasingly popular among both researchers and investors. On the eve, there were reports of OpenAI and Sam Altman's desire to invest in Merge Labs, which is developing brain implants. So far, the undisputed leader in this direction is Elon Musk's Neuralink, which previously received permission to test implants in Europe.

Other, less prominent players include Paradromics, which is developing the Connexus Brain-Computer Interface, which could help people with severe motor impairments speak through a computer. Valve founder Gabe Newell invested in Starfish Technologies, a startup working on implants to treat neurological disorders, this year.

In addition, China also completed the first clinical trials of a technology that allows brain signals to be used to control an external device in June 2025. China became the second country after the United States to reach this milestone.