Mean encoding HFA and theta power did not predict recalled-word output order in 68 pure-SEEG patients

We tested whether word-evoked high-frequency activity (HFA; 70–150 Hz) or theta power (3–8 Hz) predicted the output order of recalled words in delayed free recall. We used archived band-power features for 83 pure-SEEG patients from OpenNeuro ds004789 (RAM FR1, session 0) and matching BIDS events. The analysis plan preceded all output-order statistics. After matching events and features (83/83 subjects; 0 unmatched words; recall flags agreed for 4,761 words), 68 subjects with 4,354 recalled words met eligibility criteria; 44 had hippocampal pairs. With serial position controlled, all-pair HFA did not predict output position (mean β = +0.0033, 95% bootstrap CI −0.0133 to +0.0195, Holm p = 1.000); hippocampal HFA (44 subjects; β = −0.0060, CI −0.0223 to +0.0115) and theta (β = −0.0015, CI −0.0157 to +0.0126) also did not support predictions. Secondary tests of recall latency, first-recall HFA, alpha, beta, and low gamma were not significant. The behavioral control showed that words at serial position 1 were recalled 0.305 of the output range earlier than words at positions 2–12. The HFA confidence interval bounds the effect for this feature set; it is not a strong test of strength models because these same features did not predict recall. The study used mean power over 0–1.6 s and averaged across all or hippocampal pairs; pair-specific effects may cancel. Findings apply to session 0 and the 83-subject feature set. The exact OpenNeuro release is not established by the recorded URLs and hashes.

free recall · memory · intracranial EEG · SEEG · high-frequency activity · theta power · output order · serial position