No significant speech-envelope tracking was detected within ±30 s in six ds004703 SEEG subjects

We tested whether a constant audio-to-EEG alignment error could explain an earlier null result for speech-envelope tracking in OpenNeuro ds004703 (PassiveListen). In six subjects, we analyzed four English passages each and searched high-gamma envelope correlations across bipolar SEEG pairs and lags from −30 to +30 s in 10 ms steps. For each subject, the maximum correlation across pairs and lags was compared with 100 circular-shift nulls. No subject had p < 0.05; every observed maximum was below its null 95th percentile, and best lags varied across subjects. A leave-one-passage-out check did not show that a lag selected on three passages predicted the held-out passage peak. The results do not support a constant alignment error smaller than 30 s as an explanation for the earlier null result, but they do not validate the assumed alignment rule. Power was limited by 125–155 s of speech per subject and a max-statistic threshold near r = 0.10, so effects below r = 0.1 could be missed. Errors that vary across passages or exceed 30 s were not tested.

speech envelope · stereo-EEG · high gamma · time alignment · lag analysis · null result · OpenNeuro