Researchers at Mass General Brigham and Brown University have demonstrated an implantable brain-computer interface that restored high-speed typing for two people with paralysis. One participant, a person with advanced ALS who had lost the ability to speak, typed at 110 characters per minute — 22 words per minute — with a word error rate of just 1.6 per cent. The second participant, paralysed by a cervical spinal cord injury but still able to speak, achieved comparable accuracy. The results were published in Nature Neuroscience on March 16.
Both participants used the system from their own homes, not a lab. That distinction matters. Many BCI demonstrations produce impressive numbers in controlled clinical environments that do not translate to real-world conditions. This one did.
How It Works
Microelectrode sensors implanted in the precentral gyrus — the region of the motor cortex responsible for movement — detect electrical activity as the user attempts finger movements. The system maps a standard QWERTY keyboard layout to specific fingers and finger positions: up, down, or curled. Each letter corresponds to a particular finger performing a particular action. An AI model trained to recognise these intended movements decodes the neural signals into characters in real time. A predictive language model then refines the output, correcting errors in much the same way autocorrect works on a smartphone.
Calibration required as few as 30 typed sentences before the system could decode at full speed. Dr. Daniel Rubin, the study’s senior author, compared the daily recalibration process to tuning a musical instrument — necessary each session, but quick. That calibration efficiency determines how quickly a new patient can start using the device productively, and 30 sentences is low enough to be practical in a clinical setting.
How It Compares
The 22 words per minute with 1.6 per cent error rate represents the highest-throughput hand motor BCI typing result published to date. For context, BrainGate’s own previous handwriting-decoding system achieved roughly 18 words per minute (90 characters per minute). A speech-region BCI developed by Edward Chang’s group at UCSF reached 78 words per minute, but with a 25 per cent word error rate — fast, but too inaccurate for reliable independent communication.
The BrainGate system’s advantage is in the balance between speed and accuracy. A 1.6 per cent error rate is on par with able-bodied typing accuracy. For a patient with ALS who has lost both speech and hand function, an unreliable system is worse than a slow one, because every error requires correction and correction requires motor output the patient may not have.
The approach also differs architecturally from handwriting-based BCIs. Francis Willett’s 2021 BrainGate study decoded imagined handwriting — users mentally traced letters — which required a different neural decoding pipeline. The new system decodes finger movements on a familiar keyboard layout, which is intuitive for anyone who has typed before and avoids the abstraction layer of imagined penmanship.
Why Home Use Changes the Equation
Existing assistive communication devices for ALS patients are, as Rubin put it, “slow, error-prone, and difficult to use.” Some patients abandon them entirely. Eye-tracking systems, the current standard, typically achieve 5 to 10 words per minute and degrade further with fatigue and disease progression.
The BrainGate system’s home deployment is significant because it proves the technology can function outside the controlled conditions of a research hospital. Both participants operated the device independently from their residences over extended periods, not as a one-off lab demonstration.
The Research Team and What Comes Next
The study was led by Dr. Daniel Rubin at Mass General Brigham’s Center for Neurotechnology and Neurorecovery, with Dr. Leigh Hochberg of Brown University, who has directed the BrainGate clinical trial since its first human implant in 2004. First author Justin Jude, a postdoctoral researcher at Mass General Brigham, led the technical development. The author list includes Francis Willett and Sergey Stavisky, both central figures in BCI typing research over the past decade.
The team has identified clear next steps: implementing stenographic or personalised keyboard layouts that could push speeds higher, and extending the technology to restore reach-and-grasp movements for upper extremity paralysis. The system is not yet FDA-approved for widespread use — it remains an investigational device within the BrainGate trial — but at 22 words per minute with near-perfect accuracy, it has crossed the threshold from research curiosity to practical communication tool. The gap between what a BCI can offer and what patients actually need to function independently is narrowing fast.