Most serial-position change in RAM FR1 SEEG band power preceded word onset

We tested if the serial-position change of encoding power in free recall is a pre-word state or a change of the word response. We used archived per-word band power from 2ea9970e (OpenNeuro ds004789, FR1 session 0, sample/fs epochs, bipolar depth pairs, hippocampus and ventral temporal cortex). For each subject, we fit the slope of power on serial position with list fixed effects. We did this for the pre-word window (-0.75 to 0 s), the post-word window (0 to 1.6 s) and their difference. We used a sign-flip test and Holm correction over 12 tests. The plan was hashed before analysis. Post-word slopes were significant for VTC HFA (-0.012 dB per position, Holm p 0.014), VTC theta (+0.042, p 0.002) and HIP theta (+0.033, p 0.03). Pre-word slopes had the same sign and were 65-89% of the post-word slopes (VTC theta Holm p 0.012). No post-minus-pre slope was significant (primary Holm p > 0.7). This null result did not change without the first word or with recall covariates (Holm p > 0.4). Thus, most of the serial-position effect is a slow state change within the list, not a change of the word response. A null difference does not prove a zero evoked slope; the CIs allow a small evoked part.

free recall · serial position · intracranial EEG · SEEG · high-frequency activity · theta power · hippocampus · ventral temporal cortex · word encoding