Free recall shows a word-locked SEEG memory effect across three datasets, unlike PAL1 cued recall

A cross-task meta-analysis harmonized archived per-patient subsequent memory effect (SME) tables from 62 unique patients in FR1 (n=25), catFR1 (n=16), pyFR (n=12), and PAL1 (n=9). For word-locked z-scored log power, random-effects pooled high-frequency activity (HFA, 70–150 Hz) was +0.043 (p=0.011; I²=0.71; 95% prediction interval −0.096 to +0.183), and pooled theta (3–8 Hz) was −0.078 (p=2.3e−7; I²=0; prediction interval −0.142 to −0.013). HFA cohort means differed (permutation F p=0.008; Holm-adjusted 0.016); PAL1 minus free recall was −0.066 (permutation p=0.023; Holm-adjusted 0.045). Theta was the most stable cross-task component, while PAL1 theta did not differ from free recall in the main analysis (p=0.18). Excluding PAL1, HFA-minus-theta tilt had I²=0 and pooled +0.137 (p=2e−11). Sensitivity analyses strengthened the PAL1 contrast; an exploratory within-patient analysis found higher tilt in FR1 than PAL1 in 6/7 patients (exact sign-flip p=0.031) and in catFR1 than PAL1 in 6/6 (p=0.031). These findings support a word-locked HFA increase and theta decrease in three free-recall datasets, but not a general encoding-success marker across verbal tasks. Task is confounded with epoch length, recall type, and PAL1's all-channel SME summary. FR1/catFR1 conversion from pair-level Welch t to d is approximate, and four cohorts make tau and the prediction interval imprecise. No raw SEEG was reprocessed. The pre-registered plan specified 20,000 permutations; 5,000 were run because of a time limit, giving a smallest attainable p of 2e−4. The exploratory within-patient comparison was added after the plan.

SEEG · subsequent memory effect · free recall · cued recall · high-frequency activity · theta · meta-analysis · memory encoding