Researchers at the University of California, Davis, have demonstrated the long-term home use of a speech neurointerface by a patient with amyotrophic lateral sclerosis (ALS). Over 19 months, Casey Harrell communicated 183,060 sentences—totaling 1,960,163 words—at an average speed of 56 words per minute.

According to a report in Nature Medicine, the authors focused on everyday use of the system rather than laboratory demonstrations, allowing the patient to operate it independently.

Insights from Home Use

Harrell used the neurointerface at home for over 3,800 hours. The system enabled him to communicate with family, friends, colleagues, and doctors, send messages and letters, participate in video calls, browse the internet, and maintain full engagement despite his paralysis.

Source: UC Davis Health.

The patient estimated that the system decoded 92% of his sentences as at least "mostly correct." In formal tests, where words were displayed on a screen, accuracy exceeded 99% with a vocabulary of 125,000 English words, peaking at 99.2%.

At the time of publication, Harrell had used the system for 444 out of 653 days since implantation. After his assistants were allowed to connect and disconnect the equipment without researchers present, the average interaction time increased from 3.7 to 9.5 hours per day.

How the System Works

In 2023, Harrell had four microelectrode arrays implanted in the left precentral gyrus, the brain area associated with speech coordination. The system reads signals from 256 cortical electrodes when the patient attempts to speak.

An algorithm converts neural activity into phoneme probabilities every 80 ms, and then a language model selects the most likely sequence of words from a vocabulary of approximately 125,000 words. The text is displayed on the screen in real-time. Once a phrase is completed, the system can vocalize it using a synthesized voice modeled after Harrell's voice prior to his illness.

To control the computer, the patient also used a cursor decoder. It was previously thought that different brain areas were required for speech and movement, but the team demonstrated that both functions could be achieved through signals from the speech motor cortex.

From Demonstration to Everyday Tool

The team had already shown high accuracy of the neurointerface in laboratory sessions. In 2024, researchers reported that during the first 30-minute training session, the system achieved 99.6% accuracy with a vocabulary of 50 words. On the second day, after an additional 1.4 hours of setup, accuracy was 90.2% with a vocabulary of 125,000 words.

This new work shifts the focus from controlled trials to long-term home use. According to the authors, this is a key step toward practical neurointerfaces for individuals with severe motor impairments. Co-author Sergey Stavisky stated that the 3,800 hours of recorded brain activity during system use represent the largest individual dataset with single-neuron resolution.

The authors emphasized that the study describes a single clinical case. It remains unclear how transferable the results are to other patients, including the implantation sites, types of electrodes, and neurological conditions. The system is still experimental, utilizing wired connections, requiring daily setup by trained assistants, and is only suitable for home use due to its size.

In March, Neuralink CEO Elon Musk announced that their BCI device could restore speech to individuals who have lost it. One patient was able to launch World of Warcraft and play using only his thoughts by the 80th day of using the neuroimplant.

In April, it was reported that Galena Bakwalter, a user of the brain-computer interface, learned to create music using her thoughts.

In spring, media reported that the startup founded by former president of Neuralink would begin human trials of a biohybrid chip.