In RAM SEEG free recall, patient-level encoding SME size did not predict recall rate

We tested whether patients with a larger subsequent memory effect (SME)—the difference in brain activity between later recalled and not-recalled words—also recalled more words. The pre-registered analysis used per-word 0–1.6 s log band power from bipolar SEEG pairs in RAM FR1 (54 patients; 51 included) and RAM catFR1 (30 patients; the primary person-disjoint replication set included 14). A patient SME was the mean Cohen’s d over pairs after list and serial-position residualization. Group tilt SME (HFA 70–150 Hz minus theta 3–8 Hz) was positive in FR1 (mean d=+0.022, p=0.009) and person-disjoint catFR1 (+0.047, p=0.004), but SME size did not rank patients by recall rate. In FR1, tilt SME and recall rate had Spearman rho=-0.17 (one-sided p=0.89; bootstrap 95% CI [-0.43, 0.11]); HFA rho=-0.02 and theta rho=+0.26, opposite the predicted sign. In person-disjoint catFR1, tilt rho=+0.24 (p=0.21; n=14). All Holm-adjusted p values were >=0.38. The pooled FR1 and person-disjoint catFR1 tilt rho was -0.08 (n=65, p=0.73). Cross-fitted and partial analyses were consistent with the primary result. Recall-rate split-half reliability was 0.88, while FR1 SME reliability ranged from 0.11 to 0.39; the maximum observable tilt correlation was about 0.59. These results indicate that patient-level SME size is a weak marker of recall ability for these position-residualized features in RAM SEEG patients with epilepsy. The catFR1 person-disjoint sample was small, electrode coverage varied, and clinical state was not controlled. Region-specific SMEs, other time windows, and classifier-based measures were not tested. The features came from earlier AutoSci packages and were not recomputed from raw EEG.

subsequent memory effect · SEEG · free recall · memory · RAM FR1 · catFR1 · reliability