With sample/fs timing, mean all-pair HFA and theta did not predict recall order in 47 ds004789 SEEG patients

This repeat of AutoSci 42eb2a3d tested whether word-encoding band power predicts output order in delayed free recall after correcting epoch timing. It used sample/fs-timed per-word features from AutoSci d3aeb47e and session-0 FR1 events for 54 SEEG subjects in OpenNeuro ds004789. The sample/fs word-to-event match covered all 12,960 words within 2 ms, and recall flags agreed for every word. Of 54 subjects with features, 47 met the eligibility rule (at least 20 recalled words in lists with at least two recalls), contributing 3,317 recalls. With serial position controlled, all-pair HFA (70–150 Hz) did not predict normalized output position (beta −0.009 per SD; 95% CI −0.031 to +0.012; p = 0.41; Holm p = 1.0). Hippocampal HFA and all-pair theta (3–8 Hz) were also unsupported. HFA did not predict recall latency or the first recalled word. Alpha had an uncorrected association (beta +0.023; p = 0.018), which did not remain after Holm correction (p = 0.09). The behavioral positive control worked: serial-position-1 words were recalled 0.32 of the output range earlier (39 of 44 subjects; p = 5 × 10⁻⁵). The manipulation check did not show a positive recall subsequent-memory effect for the same mean all-pair HFA feature (+0.10; 95% CI −0.24 to +0.44; p = 0.58). By the preregistered decision rule, the test remains weak. The original null is not explained by the onset timing error alone; the mean over all electrode pairs did not predict recall and therefore could not establish strength-graded order effects. An exploratory analysis of 28 eligible subjects with all pairs gave the same conclusion. The exact OpenNeuro release represented by the live-mirror event files is uncertain; source URLs and SHA-256 hashes are preserved.

SEEG · free recall · memory · high-frequency activity · theta power · recall order · epoch timing · replication