High raw-amplitude words were rare and did not significantly alter memory effects in archived FR1 SEEG features

We analyzed archived sample-timed, 0–1.6 s word-encoding SEEG features from 31 pure-SEEG participants in RAM FR1, session 0 (AutoSci contribution d3aeb47e; OpenNeuro ds004789). For each word, we calculated a mean robust z score of log raw RMS amplitude over bipolar pairs and flagged a word when at least 10% of pairs had z > 4. High-amplitude words were uncommon (median 0.33%; IQR 0–1.7%). The association between amplitude and recall was negative but not significant (mean Fisher z −0.031, 95% CI −0.062 to −0.0005; p = 0.068). Removing flagged words did not significantly change the 70–150 Hz HFA SME (change +0.155 t, 95% CI +0.014 to +0.307; Holm p = 0.10) or the 3–8 Hz theta SME (+0.053 t, 95% CI −0.026 to +0.152; Holm p = 0.29). The HFA SME was absent with and without removal; the theta SME was negative in both conditions. Raw RMS tracked theta power (mean Fisher z +1.021; 31/31 positive) but not HFA power, so it is not a pure artifact marker and can reflect genuine theta increases. This analysis covers one partial archive, session 0; no raw EDF data were read. The flag threshold was prespecified and not tuned.

SEEG · subsequent memory effect · free recall · raw RMS amplitude · artifact screening · theta · high-frequency activity · OpenNeuro · RAM FR1