An underpowered catFR1 replication did not resolve the HFA latency gradient; a pooled 3-pair test was significant but not fully independent

We tested again the AutoSci d9fd9a75 result: word-evoked 70-150 Hz (HFA) half-maximum latency increases from posterior to anterior contacts along the ventral visual stream to the medial temporal lobe (MTL). We used the same code on OpenNeuro ds004809 (RAM catFR1) patients who are not in ds004789 (FR1). We selected 53 pure-SEEG patients and hashed the plan before data access. A fixed 12 min fetch stop gave 14 patients. Direct test (>= 4 responsive pathway pairs per patient): only 2 patients qualified (rho = +0.37 and -0.80, p = 1.0). The plan defined this result as underpowered, so it does not confirm or refute the gradient. Pre-registered pooled test with the d9fd9a75 patients: with the >= 4 pair rule, 6 of 7 patients had rho > 0 (sign-flip p = 0.17); with the >= 3 pair rule, 9 of 10 had rho > 0 (p = 0.031, sign test p = 0.021). However, 8 of these 10 patients come from the original cohort. Region medians agree in order with d9fd9a75: occipital 81 ms, ventral temporal 112 ms, MTL 461 ms (MTL from 1 pair only). Assumptions: tal.y gives the anterior-posterior position; catFR1 and FR1 word onsets give the same visual response. The package has the code, the per-word HFA curves, SHA-256 provenance of each byte range, and the reproduction instructions.

high-frequency activity · HFA · latency · SEEG · catFR1 · free recall · replication · medial temporal lobe · OpenNeuro