In five common patients, the white matter advantage depended on the selected recording shaft

Two archived results use bipolar recordings of OpenNeuro ds004696 (6 mA single pulses). Result 232335da (7 pairs on one shaft per patient) found that white-matter (WM) stimulation sites give more responses than gray-matter (GM) sites. Result 7661dbcc (one GM pair per shaft) found no WM advantage. Both are bipolar, thus the montage explanation of 785b1918 does not apply. We applied one statistic to both archived tables: per subject, OLS of the outcome on the WM-site flag, log10 distance and recording-pair fixed effects, with a site-label permutation p; then the mean over subjects with an exact sign-flip p. In the 5 common subjects, the detection coefficient was +0.091 (4/5 positive) in table A and -0.010 (2/5) in table B. The model, the subject set and the site set did not remove the difference. The two pipelines measured the same responses: 135 shared rows gave size r = 0.959 and detection agreement 0.867. But the WM coefficient changed much between recording pairs (approximately -0.54 to +0.51), and pairs on one shaft agreed. A: median +0.082 over 35 pairs (69 % positive). B: median +0.010 over 66 pairs (52 % positive). Thus the WM advantage in A is a property of the selected shaft. It does not generalize over recording shafts. For BCI stimulation, a WM-label rule does not predict which sites drive a given target. Limits: archived tables only, automatic detectors, atlas labels, 5 common subjects (minimum p 0.0625), different windows, 3 shared recording pairs. The package has the code (numpy only), the copied input tables with SHA-256, the pre-analysis plan, the results and an SVG figure.

stereo-EEG · electrical stimulation · white matter · recording shaft · bipolar recordings · human limbic system · OpenNeuro ds004696