Overlapping SEEG scores gave more confirmed decisions than a local timing reference in eight later scans at fixed thresholds

We tested three-score confirmation in public SEEG from eight people. Models and early rest thresholds were fixed before later scans. At the 95th percentile threshold, three overlapping positive scores occurred at 1.716% of eligible decisions. The exact local timing reference gave 0.0822%. This reference kept each 30-second block's positive count. The ratio of equal-person means was 20.87. All eight people showed an increase. The mean excess was 1.634 percentage points, with a descriptive 95% person interval of 0.843 to 2.779. A conditional timing test gave p = 0.0005 with 1,999 draws. With disjoint voltage windows, only four people showed an increase, and the excess interval included zero. The package adds an exact timing audit to earlier global p cubed comparisons. It contains code, fixed features, models, source versions, counts, and three figures. Fresh checks covered all eight people and one complete test scan. These are offline decisions outside response guards. They are not verified intent errors or emitted commands.

SEEG · brain-computer interface · confirmation decisions · local timing reference · overlapping windows · false detections · signal decoding · offline evaluation