BCIwiki (bciwiki.com) — An intracortical brain-computer interface (BCI) let 2 people with tetraplegia type on a two-handed QWERTY keyboard layout using only attempted finger movements, reaching 110 characters per minute, or 22 words per minute, with a word error rate of 1.6%, researchers from Massachusetts General Hospital, Brown University and other institutions reported in Nature Neuroscience on March 17, 2026 (DOI: 10.1038/s41593-026-02218-y).
The team chose keyboard typing as a familiar communication paradigm with a high information rate. Users attempt finger movements, which the system decodes from intracortical recordings; accurate decoding required as few as 30 calibration sentences. A 5-gram language model improved sentence-level decoding.
Both participants are enrolled in an intracortical BCI clinical trial. One has amyotrophic lateral sclerosis (ALS) and the other a spinal cord injury. Typing speed is regulated by the user. The authors say the accuracy resembles able-bodied typing and that throughput exceeds current state-of-the-art hand motor iBCI decoding.
The authors describe the finger-decoding approach as intuitive, familiar and easy to learn for people whose communication is impaired by paralysis. Contributing institutions also include Harvard Medical School, the VA Providence Healthcare System and Stanford University.