Inner Speech in Motor Cortex and Implications for Speech Neuroprostheses | Kunz et al.
Appreciation
--
Importance
--
Date Added
6.17.26
TLDR
T15’s 55b decodes attempted speech, mimed speech, listening, and inner speech; T17’s 55b firing rate was too low for useful decoding. In T15, inner speech is a scaled-down copy of attempted speech along the same axes in both 55b and 6v.
2 Cents
Found it very difficult to read, but the scientific findings are extremely interesting once understood.
Tags
- T15’s 55b and inferior 6v are two speech hotspots where inner speech, silent reading, and listening are all robustly represented.
- T15’s 55b decoded attempted vocalized speech at 72.1%, mimed speech at 77.9%, listening at 68.6%, and two seven-way inner-speech conditions above chance.
- In T15’s 55b, inner speech is a scaled-down copy of attempted speech along the same axes, as it is in 6v.
- These region-resolved results are single-word only, so they do not validate or contradict Jude et al.’s finding that 55b decoding improves as language units grow.
- T17’s 55b arrays recorded very little spiking and were dropped for failing to decode anything. Kunz used days 284–287 post-implant, versus Jude’s days 26–165, so drift may have worsened performance.