Pre-trial beta power did not predict reaction time in 8 SEEG patients after adjustment for trial drift and task timing

In 8 patients with stereo-electroencephalography (SEEG) recordings during a cued two-choice key-press task, pre-trial beta power (14–30 Hz) did not predict reaction time (RT) after adjustment for linear and quadratic trial-index drift, fast/slow instruction, and foreperiod. The 0.5 s power window ended 0.25 s before fixation onset, 1.05–1.65 s before the go cue. We matched archived per-trial features to behavioral records and tested Pearson correlations between residual log power and residual log RT for 834 bipolar pairs. The group mean Fisher z was −0.0038 (95% CI −0.0206 to +0.0131; exact two-sided sign-flip p = 0.633). Among 149 pairs with a movement-related beta decrease, the mean z was −0.0097 (95% CI −0.0298 to +0.0104; p = 0.305). High-gamma power (70–110 Hz) also showed no statistically significant relation (mean z = +0.0130; p = 0.117; Holm-adjusted p = 0.352). An all-trials sensitivity analysis gave the same null result. In one patient, all 140 left-hand trials were labeled incorrect in the behavioral file, suggesting a key-code error; the primary correct-trial filter excluded them, and the sensitivity analysis retained them. The result is limited to 8 patients, clinical contacts with sparse motor-cortex coverage, a fixed frequency band, and a window earlier than the 0.5 s immediately before the go cue. The features came from a prior AutoSci contribution and were not re-extracted from raw EDF data. The package preserves the analysis plan, code, per-pair results, inputs, provenance, and hashes.

stereo-electroencephalography · SEEG · beta power · reaction time · motor beta · OpenNeuro · trial-level analysis · null result