Eight SEEG pairs did not show lower estimated pitch error beyond sound level in periodic speech windows from ten listeners at 400 ms delay

We tested estimated sound pitch in ten public SEEG listeners from OpenNeuro ds004703 v1.1.0. Each reader used eight adjacent electrode pairs and 400 ms delay. Earlier passages supplied all model choices. The test used 38 later excerpts and 1358 admitted periodic sound windows. Adding SEEG to a sound-level reader gave a mean relative error reduction of -1.36%. The descriptive 95% interval was -8.78% to 2.47% when listeners and distinct sounds were resampled. The interval includes zero. Three of ten listeners had lower error. Stricter target quality and removal of clipped excerpts gave the same direction of mean change. The mean within-passage SEEG pitch correlation was near zero. An added sound-level check also showed no mean SEEG gain on these rows. The result does not establish absent pitch information in the brain. It applies to this narrow power reader, an estimated target, and a reused clinical cohort. All 95 native feature rebuilds matched exactly. All 197 numeric result files repeated exactly. The package preserves new pitch targets, source records, code, model states, test outputs, tables, and figures. It gives no stimulation result.

stereoelectroencephalography · pitch estimation · fundamental frequency · speech · neural decoding · brain-computer interface · ridge regression