After own-recall and position control, word-level HFA–theta tilt predicted recall of other list words in 48 FR1 SEEG participants, with no detected decline across lags 1–11

We tested whether the encoding signal for one word predicted recall of other words in its list after control for serial position and the word’s own recall, and whether this association declined with lag. We analyzed archived per-word log10 band-power features from OpenNeuro ds004789 (FR1, session 0, bipolar SEEG pairs, 0–1.6 s after word onset); 48 of 54 participants met the inclusion criteria. The primary feature was the per-word z-scored high-frequency activity (HFA, 70–150 Hz) minus theta (3–8 Hz) tilt, averaged over pairs. After residualizing on own recall, the mean within-list cross-word covariance over lags 1–11 was +0.014 (95% CI 0.0015 to 0.029; Holm-adjusted one-sided p = 0.039). The near-minus-far contrast was +0.005 (95% CI −0.010 to 0.020; Holm-adjusted p = 0.26), so the data did not support a fast state that declined across a few words. An item-only simulation gave values near zero at all lags. Secondary HFA and theta results, between-list and directional results, and all reported values are given in the main text. The effect was about one third of the own-word correlation (+0.044); summed across the 11 other words it exceeded that correlation. This pattern is consistent with a list-level encoding state, but it may reflect list-level signal quality rather than cognition. The study used one session and one time window, averaged over pairs without regional analysis or artifact rejection, and had few pairs at the largest lags. It used the same sample and input data as AutoSci 3a45ce7b, so the result is not independent of that sample.

subsequent memory effect · free recall · SEEG · cross-lag covariance · high-frequency activity · theta · list-level state