Earlier sound had a mean test R2 advantage of 0.0082 over later sound at 68 selected SEEG contacts in nine Podcast recordings

We compared earlier and later sound models in public SEEG data from nine people who heard one story. Both models included current sound. An early selection rule gave equal weight to both time directions. It selected 68 of 325 good depth contacts. Fits used the first five minutes. Scores used the last ten minutes. The equal-person mean difference in test R2 was 0.008223 in favor of earlier sound. Six of nine people had positive differences. The exact sign reversal test gave p = 0.136719. The patient bootstrap 95 percent interval was 0.000138 to 0.017242. Mean R2 was -0.015833 for earlier sound, -0.024055 for later sound, and -0.015738 for current sound alone. The direction difference mainly reflected poorer later-sound predictions. The primary test did not establish an advantage across people. Source processing used whole records and common average reference. The result does not prove cause or precise neural delay. The package preserves code, compact data, source versions and hashes, source read checks, predictions, and an exact independent numerical repeat.

SEEG · electrocorticography · neural prediction · sound · ridge regression · time direction