In 8 ds004473 sEEG subjects, 2.8–10 s after a key press did not change rest-window decoder scores

We tested whether time since a key press changed how move-like a rest window appeared to a movement decoder, using a preregistered plan and archived data from 8 subjects in the OpenNeuro ds004473 sEEG two-choice task. The decoder used bipolar log high-gamma (70–150 Hz) and beta (13–30 Hz) power with shrinkage linear discriminant analysis and 5-fold blocked cross-validation (mean move-versus-rest AUC 0.845). Rest windows began 2.8–10 s after the preceding key press. Across subjects, lag was not associated with the out-of-fold rest-window score (mean Spearman rho −0.013; 5 of 8 subjects had negative rho; exact two-sided sign-flip p = 0.63; exploratory 95% CI −0.078 to 0.049). Holm-corrected secondary tests found no change in pause-trial scores, rest high-gamma or beta power, or false-positive rate. Sensitivity estimates were partial rho −0.023 (p = 0.55) after controlling for trial index and rho −0.011 (p = 0.65) with a standard train/test split. The result applies only to lags of at least 2.8 s: the first seconds after movement, when beta rebound is largest, were not tested. Lag ranges were narrow in 6 of 8 subjects (10th–90th percentile range under 1 s), limiting power to detect small effects. The package preserves the preregistered plan, code, archived inputs and their SHA-256 hashes, results, and figure.

stereoelectroencephalography · brain-computer interface · movement decoding · rest window · high-gamma power · beta rebound · linear discriminant analysis · cross-validation