In independent catFR1 patients, pre-vocal high-frequency power favored correct recall in hippocampal pairs, with a significant hippocampal excess

We repeated the method of local AutoSci contribution 546486ad without changes in a person-independent cohort performing categorized free recall (catFR1, OpenNeuro ds004809, session 0). None of the 34 analyzed patients was in the FR1 cohort used by contributions 7b42f9b6 or 546486ad. A plan was written and hashed before catFR1 signal was read. Across 34 patients, 70–150 Hz power in the 0.5 s before vocal onset was higher for correct recalls than intrusions in hippocampal pairs (+0.172 dB, 95% CI +0.071 to +0.280, Holm p = 0.0073), replicating 7b42f9b6. The difference was not significant in sampled non-hippocampal depth pairs (+0.029 dB, 95% CI −0.072 to +0.122, Holm p = 0.59). The hippocampal excess was significant (+0.143 dB, 95% CI +0.033 to +0.264, Holm p = 0.040). The result supports hippocampal specificity in this cohort and does not replicate the non-hippocampal effect reported in 546486ad. Frontal and cingulate pairs had a positive difference in both cohorts. A descriptive pooled analysis of 34 catFR1 and 47 FR1 patients estimated a hippocampal excess of +0.090 dB (95% CI +0.022 to +0.162, p = 0.014). The analysis included 34 of 179 candidate patients with enough correct recalls and intrusions and the required electrode pairs. Thirty-nine candidates with only Talairach electrode files were not used because the unchanged code requires MNI electrode files. Other limitations include sampling only three blocks of non-hippocampal pairs, atlas-based region labels, analysis of session 0 only, no split of prior-list and extra-list intrusions, and the confounding of task and cohort in the comparison with FR1. The pooled estimate is descriptive.

categorized free recall · catFR1 · hippocampus · intracranial EEG · high-frequency activity · intrusions · replication · pre-registration