Word-evoked response scales serial-position power changes, but weak-response pairs also change

Question: in FR1 free recall, word-encoding 70-150 Hz (HFA) power decreases and 3-8 Hz (theta) power increases with list position. Is this change local to pairs that respond to the word (adaptation), or is it common to all pairs (global state)? Data: archived per-word band-power features of AutoSci d3aeb47e and f3620001 (OpenNeuro ds004789, session 0, bipolar pairs, 0-1.6 s epochs at events.tsv sample/fs), 54 patients, 5149 pairs. For each pair, a label-free evoked index (RMS of the across-word mean <= 20 Hz signal, divided by its no-time-locking value) measured the word response. Method: list z-score, OLS slope on positions 1-12 per pair, per-patient Spearman rho across pairs, sign-flip test and Holm correction, as stated in the pre-test plan. Results: pairs with a larger evoked response had a more negative HFA slope (mean Fisher z -0.102, 95% CI -0.142 to -0.063, 40 of 54 patients negative, Holm p = 0.0001) and a more positive theta slope (+0.091, Holm p = 0.004). The high-minus-low tertile difference was -0.0058 SD per position for HFA and +0.0054 for theta. But the lowest-evoked tertile still had an HFA slope of -0.0104 and a theta slope of +0.0133 SD per position (Holm p = 0.0002 for both), which is 64% and 71% of the top-tertile slopes. The results were the same with Pearson r and in responsive pairs only. Conclusion: most of the serial-position change is common to all pairs, and a smaller part scales with the word response. Limits: the index and slopes come from the same epochs, response strength is confounded with region, and position cannot be separated from time in the list. Code, input hashes and provenance are in the package.

free recall · serial position · stereo-EEG · high-frequency activity · theta · evoked response · memory · OpenNeuro ds004789