Hippocampal Maintenance Band Power Did Not Change with Load in 14 SEEG Patients; Encoding Power Increased

This independent replication tested whether hippocampal stereo-electroencephalography (SEEG) band power changed with verbal working-memory load (4, 6, or 8 letters) in session 1 of OpenNeuro ds004752. A separate NumPy-only pipeline analyzed 15 patients. For the pre-specified primary test, maintenance-period (3.5–6 s) power was summarized by per-pair Spearman correlations, averaged within subject, and tested with exact sign-flip permutations and Holm correction across five bands. No hippocampal band passed correction in 14 patients. High gamma showed a negative trend (mean rho −0.068, 3/14 positive, p = 0.031, Holm p = 0.15). In exploratory encoding analyses (1.5–3 s), hippocampal theta, alpha, and beta power increased with load (Holm p = 0.014–0.038; 12/14 subjects positive); theta also increased in lateral temporal and other-region pairs. This encoding effect was not specific to the hippocampus and may reflect the greater number of letters on screen at higher loads. Fixation showed no hippocampal load effect (p ≥ 0.48). Accuracy decreased from 0.975 to 0.834 across set sizes, while response time increased with load (mean rho 0.30, p = 0.0002). Limits include session 1 only (about 50 trials per subject), atlas-based rather than visual contact localization, bipolar pairs only, no aperiodic correction, epileptic tissue not excluded, and an exploratory encoding result. EDF hashes matched S3 ETags, but the live-mirror files cannot be assigned with certainty to a version-specific dataset release.

stereo-electroencephalography · working memory · hippocampus · memory load · band power · replication