Hippocampal SEEG power showed no detectable load effect during verbal working-memory maintenance in 14 patients

We asked whether spectral power during the 3-s maintenance period of a verbal Sternberg working-memory task varied with set size. We analyzed sessions 1 and 2 from 15 epilepsy patients in OpenNeuro ds004752 v1.0.1, using 1,254 correct, artifact-free trials and 749 adjacent-contact bipolar SEEG pairs. For each pair, we measured the Spearman association between set size (4, 6, or 8 letters) and log power in theta, alpha, beta, gamma, and high-gamma bands, plus the 30–150 Hz aperiodic exponent. We averaged pair associations within subjects and used exact two-sided sign-flip tests with Holm correction across six features per region set. In the primary hippocampal analysis (99 pairs in 14 patients), no feature varied significantly with load (mean rho −0.031 to +0.053; all uncorrected p ≥ 0.11; all Holm p ≥ 0.69). No secondary region set passed Holm correction. Reaction time increased with load (mean subject rho +0.25, p = 1.2 × 10⁻⁴), while accuracy fell from 97% at set size 4 to 85% at set size 8. A nominal hippocampal high-gamma decrease in an earlier 11-patient pilot (p = 0.037, uncorrected) was absent in the full sample (p = 0.44). The findings indicate small or absent load effects on maintenance-period bipolar LFP power in this dataset; they do not test single-unit firing or phase synchrony. Only two sessions per patient and one 3-s window were analyzed, with no baseline normalization or time-frequency resolution. Atlas labels and a one-contact regional rule may mix white-matter contacts into regions; epileptic-zone tissue was not excluded, and per-pair correlations may be noisy with few trials.

SEEG · verbal working memory · hippocampus · spectral power · memory load · intracranial EEG · Sternberg task · aperiodic exponent