Hippocampal encoding power did not predict later recall in seven SEEG patients, but moderate effects remain possible

We analyzed hippocampal bipolar SEEG recordings from seven participants in the OpenNeuro ds004789 delayed free recall dataset (session 0; 43 pairs). For each pair, we compared per-pair z-scored Morlet log power from 0 to 1.6 s after word onset for later-recalled versus not-recalled words, then averaged Welch t statistics across pairs within each participant. Exact two-sided sign-flip tests across participants found no evidence of subsequent-memory effects in high-frequency activity (70–150 Hz; mean t = +0.04, 95% bootstrap CI −0.71 to +0.74, p = 0.95), theta (3–8 Hz; −0.11, −0.95 to +0.72, p = 0.80), or spectral tilt z(HFA) − z(theta) (+0.08, −0.83 to +1.00, p = 0.89); all three Holm-adjusted p values were 1.0. Five secondary bands also had Holm-adjusted p = 1.0. Participant effects varied in sign. With seven participants, the smallest attainable exact p value was 0.016, so these results do not find or exclude a moderate effect. The analysis used a corrected epoch-mean amplitude check and retained 43 pairs in seven participants, compared with four pairs in three participants in the pilot. It was limited to session 0, atlas labels of either contact, no interictal epileptiform discharge rejection, a 1 mV amplitude rule, recall labels from REC_WORD item names in the same list, no adjustment for serial position or list, and a time-limited, incomplete run.

hippocampus · subsequent memory · SEEG · free recall · high-frequency activity · theta · spectral tilt · memory encoding