Pretrial SEEG band power did not predict reaction time in 8 subjects (mean held-out r = 0.032)

We tested whether stereo-electroencephalography (SEEG) power in the 0.5 s rest window from 0.75 to 0.25 s before fixation predicted the next cued key-press reaction time in 8 subjects performing the OpenNeuro ds004473 forced two-choice task. Ridge regression used 200–224 bipolar-pair features from 70–150 Hz and 13–30 Hz log power, five contiguous-block folds, and reaction-time residualization on trial index and foreperiod within each training fold. Mean held-out Pearson r was 0.032 (t-based 95% CI −0.025 to 0.088); the exact two-sided sign-flip p value was 0.25, with 6 of 8 subject correlations positive. No subject beat its circular-shift null at p < 0.05. Channel-mean 13–30 Hz power (mean rho = −0.009, p = 0.77), channel-mean 70–150 Hz power (rho = 0.039, p = 0.18), and foreperiod versus log reaction time (rho = −0.018, p = 0.59) showed no group effect. The reported predicted-fast versus predicted-slow reaction-time contrast was −4 ms (p = 0.70). The circular-shift null used fixed alpha = 1e3 without inner cross-validation for speed; per-subject permutation p values compare against the fixed-alpha observed correlation, while the group test uses the nested-alpha result. Only this archived window and these two bands were tested. The task showed no foreperiod effect in these data, so reaction-time variance may be mostly motor noise. There were 132–298 trials per subject. The CI excludes a mean held-out r above 0.09 for this feature set.

stereo-electroencephalography · reaction time · band power · ridge regression · cross-validation · OpenNeuro ds004473