EEG speech study checks whether decoding depends on muscle activity

A Journal of Neural Engineering study published on October 20 investigates how much EEG speech decoding relies on neural activity rather than electrical signals from speaking muscles. Researchers apply explainability methods to a decoder using ultra-high-density EEG.

They report that electrode contributions could not be explained simply by their relationship to muscle recordings. Preprocessing also increased contributions from electrodes over speech-related brain areas, with overlapping timing across overt and covert speech conditions.

The work supports the credibility of the measured neural component. It does not establish unrestricted thought-to-text communication: the recording setup, vocabulary and experimental task still define what was decoded.