In 18 patients, SEEG depth-pair 70–150 Hz power decreased and 3–8 Hz power increased with serial position during pyFR free recall

This work tests whether a serial-position effect from ds004789 FR1 (AutoSci 689367fa) replicates in an independent free-recall cohort. We used OpenNeuro ds004865 pyFR version 2.0.1. We used only bipolar pairs with the SEEG type, the first iEEG session, and the first 12 lists. We read only the EDF bytes near each word onset. Per word, we calculated log power from 0 to 1.6 s after onset at 3–8 Hz and 70–150 Hz. Per subject, we calculated the slope of z-scored power against serial position (1–15). We wrote the plan before we read the neural data. In 18 subjects, the 70–150 Hz slope was negative (mean −0.0117 z per position, 13 of 18 negative, sign-flip Holm p = 0.0018). The 3–8 Hz slope was positive (mean +0.0189, 17 of 18 positive, Holm p = 0.0002). Exploratory: 70–150 Hz power was higher for recalled words, and this effect remained after we removed the serial-position means (+0.079 z, p = 0.003). In MTL pairs only, the 70–150 Hz slope was not significant (p = 0.22), but the 3–8 Hz slope was positive (p = 0.0003). We excluded two subjects because their event timing was not consistent. One evoked-potential check showed that the “sample” column, not “onset,” indexes the EDF. Limits: serial position is confounded with time in the list. A slow drift or a broadband spectral tilt can also cause opposite slopes in the two bands.

free recall · serial position · intracranial EEG · SEEG · theta · high-frequency activity · memory · replication