Later-recalled words had a flatter encoding aperiodic exponent in pyFR SEEG

In an independent cohort of 20 pyFR SEEG subjects, words recalled later had a flatter aperiodic exponent during 0–1.6 s of encoding. The preregistered primary statistic was the mean across subjects of within-pair Welch t values (recalled minus not recalled): −0.52 (95% bootstrap CI −0.92 to −0.15; two-sided subject sign-flip p = 0.020; 14 of 20 subjects negative). This agrees in direction and size with the RAM FR1 result (t = −0.44). Secondary uncorrected tests found lower OLS exponent (t = −0.58, p = 0.015), lower offset (t = −0.50, p = 0.032), lower raw theta power (t = −0.49, p = 0.041), and higher raw high-frequency activity (HFA; t = +0.40, p = 0.018) for recalled words. Theta and HFA residuals above the spectral fit were not significant (t = −0.21, p = 0.12; t = −0.09, p = 0.41). The raw band effects mostly followed the spectral tilt. Limits include 20 subjects, a whole-implant mean without region selection, no knee term, and no adjustment for word item effects or serial position. The exact OpenNeuro ds004865 release represented by the archived live-mirror files is uncertain.

aperiodic exponent · electrophysiology · free recall · SEEG · pyFR · spectral tilt · memory encoding