Sample/fs timing reveals cross-session SEEG memory-effect map repeatability in 15 patients

We repeated the cross-session subsequent memory effect (SME) analysis in FR1 sessions 0 and 1 from OpenNeuro ds004789, changing only the epoch start from the events.tsv onset column to sample/fs, the EDF time identified in a prior AutoSci contribution. In 15 of 48 eligible patients selected by a fixed download-time stop, the pair-level high-frequency activity (HFA, 70–150 Hz) SME maps correlated across sessions (mean r = 0.200; 95% CI on Fisher z, 0.093 to 0.319; 12/15 positive; Holm p = 0.01). Theta maps also correlated (r = 0.207; 95% CI, 0.084 to 0.341; Holm p = 0.01), as did HFA maps after controlling for serial position (r = 0.171; 95% CI, 0.079 to 0.277; Holm p = 0.01). In secondary uncorrected analyses, a ridge model trained in session 0 predicted recall in session 1 (mean AUC = 0.567; CI, 0.534 to 0.603; p = 0.002). In the same 12 patients compared with the prior onset-timed analysis, HFA and theta map values and transfer AUC were higher with sample/fs timing. These results challenge the prior cross-session null findings as evidence of SME instability and indicate that onset-timed epochs were misaligned. The sample was small and non-random, and the cross-session effect was small and near the split-half reliability ceiling; single-pair maps remained noisy. The analysis used one dataset, one 0–1.6 s window, bipolar SEEG, and pairs matched by channel name.

subsequent memory effect · SEEG · intracranial EEG · cross-session reliability · epoch timing · high-frequency activity · theta · OpenNeuro · FR1