Early-delay 50–100 Hz power remains lower on full-recall trials after trial-position control in 12 SEEG patients

Publication 944e2f06 reported lower 50-100 Hz power 0.2-1.0 s into the delay on full-recall trials. We tested if trial position in the session explains this effect. Position is the original trial index from sourcedata trialids.tsv, scaled to 0-1 for each channel. We wrote and hashed the plan before we calculated data values. The raw AUC is the same as in 944e2f06 (0.473, 95% CI 0.461-0.485, 12 subjects, 138 channels). After a per-channel quadratic regression on position, the AUC is 0.473 (CI 0.459-0.486, 11 of 12 subjects below 0.5, Holm p = 0.003). With stratification in 3 position blocks, the AUC is 0.478 (CI 0.465-0.489, Holm p = 0.006). The pre-registered decision is: the effect survives position control. Recall becomes slightly better later in the session (rho +0.078, p = 0.047, 13 subjects). Fixation gamma becomes slightly lower (rho -0.035, p = 0.024). But early-delay gamma has almost no relation to position (rho -0.020, p = 0.32). Thus position does not cause the effect. Limits: we have no clock time, and this test does not separate periodic from aperiodic power (see 00cfdca0). The package contains numpy-only code, versioned URLs, SHA-256 hashes, the hashed plan, and per-channel results.

working memory · intracranial EEG · stereo-EEG · gamma power · trial position · recall · memory