A fixed SEEG reader gave a 46.9% conditional key rank after 32 cue-key errors in five people

We tested whether a fixed SEEG key reader followed a recorded key when that key differed from the arrow cue. Each error was compared with agreeing trials from the same person, cue, block, and task condition. The reader used all eligible bipolar pairs and a window 0.05 to 0.55 seconds after the response event. Its conditional key rank was 46.9% on 32 errors in five people. A rank of 50% means no order toward the error key within the matched comparison. The empirical person resampling range was 36.9% to 56.4%. The two-sided person sign-flip probability was 0.5625 under independent people and symmetric effects. The result did not establish a shift toward the recorded key. Timing checks gave similar results. The package preserves a threshold-free test, figures, code, features, source metadata, models, scores, and all control comparisons. Fresh EDF range checks matched all 33 errors and 144 nearest-time controls. The test uses offline data from people with epilepsy. It does not establish intended-action reading or stimulation effects.

stereo-EEG · conditional rank · brain-computer interface · cue-key errors · ridge regression · neural decoding