One ECoG implant decoded speech and gesture; the papers disagree on how many people
Nature Neuroscience reports a single high-density electrocorticography implant decoded speech and gestures at once in three people with paralysis. New Scientist's same-day story names two, Bravo-1r and Bravo-6, on a simultaneous avatar task. Isolated-movement decoding in three participants and the two-person avatar numbers are not the same claim.
Science··Evening
Three participants in the journal paper
A peer-reviewed study published in Nature Neuroscience on 14 September 2026 shows that a single high-density electrocorticography implant can decode speech and gestures simultaneously in three people with paralysis. Isolated upper-limb and orofacial movements were first decoded reliably in three participants; parallel speech and gesture decoders then let them drive a personalized virtual avatar by attempting speech and gestures together or apart. Models trained on both isolated and simultaneous data performed better across behavioural contexts. A small clinical sample is not a claim of a universal communication prosthesis.[1]
New Scientist names two on the avatar task
Carissa Wong reports that Bravo-1r, paralysed after a stroke, and Bravo-6, who has amyotrophic lateral sclerosis, used an iPhone-sized implant over the sensorimotor cortex. Isolated-then-simultaneous training predicted Bravo-1r's intended gestures and speech at 88 percent and 84 percent, and Bravo-6's at 66 percent and 70 percent, against a 9 percent chance baseline. The team used avatars rather than restoring the patients' own speech and motion. The Nature Neuroscience paper in the same cluster also describes isolated-movement decoding in three participants; the New Scientist account of the simultaneous avatar task names two.[2]
Three is not two
The journal paper's three-person isolated-movement result and New Scientist's two-person avatar accuracies can both be true without making either N the established headcount for every task. Neither report settles 2 or 3 as the headcount for the simultaneous avatar. What both reports share is a single implant, speech and gesture decoded together, and avatars rather than restored anatomy. Sample size remains small either way.[1], [2]
