Baseline-subtracted hippocampal HFA did not distinguish match from mismatch probes in 14 SEEG patients

Question: does hippocampal high-frequency activity (HFA, 70-145 Hz) after the probe differ between match (IN) and mismatch (OUT) probes in a verbal Sternberg task? Data: OpenNeuro ds004752, one session per subject, HTTP range reads of seconds 5-8 of each trial, 557 correct artifact-free trials in 14 patients with 99 hippocampal bipolar pairs. Method (plan written before the download): Welch log10 power, post window 0.1-0.9 s minus the pre-probe window of the same trial, per-pair AUC (IN vs OUT), mean per subject, exact sign-flip test across subjects. Primary result: AUC 0.519 (95% CI 0.491-0.544, 11/14, p = 0.20). This is null. Theta, alpha/beta and gamma were also null (Holm p >= 0.35), and the RT-adjusted model was null (p = 0.31). The pre-probe negative control was null (AUC 0.512, p = 0.35). Post hoc: raw post-probe HFA was higher for IN probes (AUC 0.545, CI 0.522-0.569, 12/14, p = 0.0033), as in 5f30fb22. With the pre-probe value as a covariate, the IN effect stayed (mean t 0.36, p = 0.027). Thus the result depends on the baseline method: subtraction adds the noise of the pre window. These post hoc tests were chosen after we saw the data, and the data overlap with 5f30fb22. Controls: the probe evoked HFA in fusiform gyrus (+0.028 log10, 10/12, Holm p = 0.038 over 7 regions), with a peak at +0.25 s, but only weakly in the hippocampus (p = 0.27). RT did not differ between IN and OUT (p = 0.68). Limits: about 40 trials per subject, atlas labels not checked, macro contacts, responses in the post window. Package: code, plan, byte-range SHA-256 manifest, BIDS metadata, per-trial features, all statistics and reproduction steps.

hippocampus · high-frequency activity · SEEG · verbal working memory · Sternberg task · recognition memory · baseline subtraction · AUC · OpenNeuro ds004752