Template white matter at stimulated depth pairs was not associated with recall change in the primary FR2 analysis

We tested whether template white matter (WM) probability at a stimulated bipolar depth pair predicted the change in free recall during 50 Hz encoding stimulation. The preregistered-in-advance primary analysis used 37 depth pairs from 29 patients in OpenNeuro ds005489 (FR2). Exposure was mean ICBM152 2009a WM probability within a 3 mm sphere centered at the MNI midpoint of the pair; outcome was recall for stim-on minus stim-off words in the same stimulation lists. The association was small and uncertain (Spearman rho = 0.18, subject-bootstrap 95% CI -0.17 to 0.47; two-sided permutation p = 0.29). Secondary analyses found no clear association: log-odds meta-regression p = 0.78, median-split difference +3.1 points p = 0.14, and adjusted WM coefficient p = 0.94. In ds005491, rho = -0.00 (n = 11, p = 0.99); pooled across both datasets rho = 0.16 (n = 48, p = 0.27). One of nine sensitivity tests, contact-level WM, had rho = 0.38 and uncorrected p = 0.022; it did not pass the Bonferroni threshold of 0.0056 and is hypothesis-generating only. The data do not establish that site WM predicts the stimulation effect. Small sample size, noisy per-pair recall estimates, use of population rather than patient tissue maps, and clinical site selection limit inference; a small association is not excluded. Analysis code, source manifests, input tables, results, and a figure are included. The ICBM152 maps are available from the Nilearn source URL documented in the README and are linked externally because their bundled source distribution is under BSD terms.

free recall · intracranial stimulation · SEEG · white matter · memory · FR2 · OpenNeuro