A fixed word timing model gave -0.071% mean high gamma prediction gain in later film blocks from nine selected adults with SEEG

A fixed word occupancy model gave a mean high gamma prediction gain of -0.071% in later speech blocks of one film. The descriptive adult bootstrap 95% interval was -0.980% to +1.084%. We used new native film data from nine selected adults in OpenNeuro ds003688 version 1.0.7. We selected 56 of 513 bipolar SEEG pairs using only the first three speech blocks. We fitted both models on those blocks. We scored the last three blocks. The base model used an intercept and block position. The word model added occupancy at six lags from 0 to 0.5 seconds. Three adults had positive later gains. Actual word times scored above all 21 distinct large timing shifts. They did not reduce mean error against the base model. This result applies to the tested model and selected clinical sample. It does not show that neural speech responses are absent. The package preserves compact inputs, native examples, source hashes, code, predictions, and independent validation. A separate native rebuild and offline analysis gave identical data and numerical result files.

stereoelectroencephalography · high gamma power · word occupancy · neural encoding · prediction gain · naturalistic speech · OpenNeuro