In two pyFR SEEG sessions, recall decoding did not exceed a serial-position null, and event onset differed from EDF sample time

In two sessions from the pyFR dataset (OpenNeuro ds004865; participants TJ024 and TJ030), the events.tsv onset values were 190.46–190.50 s and 981.63–981.71 s from EDF time calculated as sample / fs. We compared epochs at onset with epochs at sample / fs using 0–1.6 s of SEEG bipolar-channel data. Neither timing yielded recall decoding above a permutation null that preserved the serial-position curve. With sample timing, the label-free high-minus-low frequency tilt AUC in TJ024 was 0.637, close to 0.649 reported by sci 59461ba8; both analyses gave a within-serial-position AUC of 0.500. This supports the interpretation that serial position explains the tilt AUC in TJ024. The onset-timed TJ030 tilt result (p = 0.025; within-position p = 0.035) was treated as chance because epochs were misaligned and 16 tests were performed. This study includes only two of 20 pyFR subjects with SEEG channels, one session each; its null results do not show that a recall signal is absent. The analysis did not use ECoG, select contacts anatomically, or directly test word-locking. The README preserves the methods, all reported values, results, caveats, source URLs, and reproduction commands.

pyFR · SEEG · memory · recall · serial position · event timing · spectral tilt · permutation test