Randomized 2.36–2.75 s word intervals did not change word-locked SEEG power or recall in 54 FR1 patients

In the RAM FR1 free-recall task, word onsets were separated by randomized intervals (SOAs) of 2.36–2.75 s. Using archived sample/fs-timed bipolar band-power features for 54 pure-SEEG patients (11,738 words, serial positions 2–12), we tested whether SOA predicted 0–1.6 s word-locked high-frequency activity (HFA, 70–150 Hz), theta power (3–8 Hz), or recall. Subject-level models included serial-position and list fixed effects; group effects were tested with sign flips, bootstrap confidence intervals, and Holm correction. Per 100 ms of SOA, HFA changed by −0.0053 SD (95% CI −0.0307 to +0.0207; Holm p = 1.0), theta by −0.0011 SD (−0.0297 to +0.0261; Holm p = 1.0), and recall probability by +0.0069 (95% CI −0.0042 to +0.0185; Holm p = 0.75; n = 53). All three effects were not supported by the pre-registered tests, and their confidence intervals were within the pre-registered negligible bounds (±0.05 SD or ±0.02 probability per 100 ms). Exploratory results found no SOA effect in the other bands or the top 25% most word-responsive pairs. A placebo test using the next interval found a theta result near the threshold (Holm p = 0.053); the interval occurred after the analysis window, and the authors considered this a possible false positive among 20 secondary tests, without testing that interpretation. The SOA range was narrow, so these findings do not exclude adaptation at shorter intervals. Averaging over the full 0–1.6 s window may dilute short early effects; the model was linear in SOA. Participants were patients with drug-resistant epilepsy and clinically placed electrodes.

free recall · stereo-EEG · neural adaptation · theta power · high-frequency activity · word onset · SOA · OpenNeuro ds004789