List-level HFA varied above chance but did not predict recall in pure-SEEG free recall

We tested whether the subsequent memory effect (SME) in encoding 70–150 Hz power (HFA) reflects slow list-level variation or word-to-word variation within lists. We analyzed archived per-word SEEG features from 74 FR1 patients (primary; OpenNeuro ds004789, snapshot 3.1.0) and 36 catFR1 patients (replication; OpenNeuro ds004809, release not recorded). All included patients had depth contacts only; analyses used session 0, bipolar pairs, and the 0–1.6 s word-encoding window. A preregistered between/within decomposition and subject-level linear probability models showed that 22.0% of FR1 HFA variance and 19.4% of catFR1 HFA variance lay between lists, compared with shuffled-label shares of 7.7% and 8.1%; 71/74 and 36/36 patients, respectively, were above chance. List-level HFA did not predict recall: the FR1 between-list slope was +0.0023 per SD (95% CI −0.0136 to +0.0189; Holm p = 1.00), and the catFR1 slope was −0.0098 (95% CI −0.0379 to +0.0164; p = 0.49). Within-list slopes were small and positive (FR1 +0.0064, p = 0.069; catFR1 +0.0108, p = 0.024), but neither passed Holm correction. The word value averaged all bipolar pairs, which may hide regional effects; list-level variation may reflect non-neural drift, only session 0 was used, and catFR1 lists were semantically categorized. These results do not support using slow list-level HFA alone as a marker of a good encoding state.

SEEG · free recall · subsequent memory effect · high-frequency activity · HFA · list-level variation · within-list effect · memory encoding · OpenNeuro · linear probability model