Theta co-fluctuation reduction did not replicate in catFR1 (32 SEEG patients; D = -0.008, Holm p = 0.40)

We tested whether the lower theta (3–8 Hz) power co-fluctuation for later-recalled words reported in FR1 replicated in categorized free recall (catFR1). We analyzed archived catFR1 features for 32 pure-SEEG patients from session 0, using the same code and pre-registered decision rule as the FR1 study. The plan was hashed before the catFR1 statistics were computed. The primary theta result did not replicate: mean D = -0.0080 Fisher z (95% CI -0.0194 to +0.0040; 20/32 negative; Holm p = 0.40), compared with -0.0157 in FR1. HFA (70–150 Hz) did not differ (mean D = -0.0007; 95% CI -0.0115 to +0.0122; Holm p = 0.92). The 17 patients outside the FR1 cohort showed a secondary theta effect (D = -0.0143; Holm p = 0.033); the 15 overlapping patients did not (D = -0.0008; Holm p = 0.94). Recalled words had lower theta power variance, and subject-level theta D correlated with the theta variance ratio (Spearman rho = 0.67). Across the 15 overlapping subjects, FR1 and catFR1 theta effects did not correlate (rho = 0.32, p = 0.25). The result is inconclusive at this sample size: the effect direction matched FR1, but the confidence interval included zero and estimated power to detect the FR1 effect was approximately 0.6. Do not interpret the co-fluctuation result as a change in network coupling without controlling for variance. The data were archived features attributed to OpenNeuro ds004809; the exact dataset release used for those features was not established.

theta power · co-fluctuation · categorized free recall · SEEG · replication · memory · variance