In subject W3, unverified event timing yielded no sentence-locked high-gamma response at 160 SEEG contacts

We tested whether white-matter stereoelectroencephalography (SEEG) contacts carry speech-evoked high-gamma (70–150 Hz) information useful for decoding, using gray-matter contacts as a positive control. The analysis used OpenNeuro ds004993, subject W3, task timit4: 126 sentences and 160 good contacts (77 gray matter, 53 white matter, 30 unknown), with monopolar, common-average, and bipolar references. No contact had a significant response (0.1–1.0 s versus −0.5–0 s; Wilcoxon, Benjamini–Hochberg FDR q < 0.05) with any reference, including 18 superior temporal gyrus or sulcus contacts. Median effect sizes d_z ranged from −0.03 to 0.05; speech-versus-silence decoding AUC ranged from 0.47 to 0.62. Because the gray-matter auditory-cortex positive control also failed, this run cannot determine whether white matter carries speech signals. Event alignment is unverified: the 404.0 s EDF contains 127 embedded annotations at 1060.39–1452.48 s, while events.tsv onsets are 3.07–395.16 s, differing by a constant 1057.315 s. The EDF is therefore a crop of a longer recording, and a wrong crop offset could explain the flat event-locked averages. A vagus nerve stimulator artifact is another possible cause. The offset remains unresolved; the documented next step is to search for it by cross-correlating superior temporal high-gamma with the sentence envelope over lags or obtain it from the dataset authors, then rerun the code unchanged. This report covers one subject and one run; it does not establish that white matter lacks speech signals.

stereoelectroencephalography · high-gamma · speech · white matter · gray matter · event alignment · decoding · TIMIT · OpenNeuro · vagus nerve stimulator