Fixed rest99 SEEG commands matched 195 recorded responses and a mean of 180 responses after local delay swaps in eight scans with a lead window of 500 samples

We tested whether fixed SEEG commands track response time beyond the task cue. We used public OpenNeuro ds004473 version 1.0.2 data from eight patients. The fixed rest99 command trains had 1,153 commands and 816 retained test responses. A command matched a response if it occurred in the prior 500 samples, including both endpoints. There were 195 recorded matches. We held each cue time fixed and swapped response delays within nearby trials. Trials shared the fast or slow condition, response key, and a block of 20 test trials. Across 9,999 joint swaps, the mean match count was 180.12, or 92% of the recorded count. The patient mean recall was 23.76% for recorded times and 21.96% after swaps. The excess was 1.80 percentage points. The 95% patient bootstrap interval was -0.27 to 4.62 points. The primary conditional p value was 0.0231. With blocks of eight trials, it was 0.1646. A post hoc check that kept correctness fixed gave p = 0.0566. The test assumes that response delays can be exchanged within each block and label set. The results show a small timing excess that depends on the control. They do not establish useful online command timing. The package contains the fixed inputs, behavior data, swap assignments, code, figures, and source checks.

stereoelectroencephalography · SEEG · response timing · brain-computer interfaces · permutation test · conditional randomization