A fixed envelope model left positive mean shared 1–4 Hz EEG covariance across two ds004408 excerpts under source time alignment
This study tested shared EEG covariance outside a fixed speech envelope model. Nine test listeners were compared with ten separate reference listeners. The data were two ds004408 version 1.0.8 excerpts filtered from 1 to 4 Hz. The mean residual covariance was 0.004391. The conditional 95% listener bootstrap interval was 0.001455 to 0.007182. The residual was 86.9% of total mean shared covariance. Run 01 alone had an interval that included zero. The result assumes the source time alignment and concerns standardized numeric file values. Physical amplitude calibration is unresolved. Shared artifacts and omitted acoustic features remain possible. The result does not identify a neural cause or a language process. The package supplies an exact covariance decomposition, source code, processed data, snapshot hashes, a raw retrieval recipe, and independent checks. An offline repeat matched all 19 numeric artifacts. A fresh raw EEG and audio retrieval reproduced the stored arrays exactly.
EEG · speech envelope · shared covariance · listener bootstrap · signal processing · naturalistic speech
References
6- usesdoi:10.18112/openneuro.ds004408.v1.0.8
This exact OpenNeuro dataset version supplied the EEG and audio excerpts analyzed here.
- doi:10.1016/j.cub.2015.08.030
- doi:10.1016/j.cub.2018.01.080
- doi:10.1016/j.neuroimage.2024.120890
- supportsdoi:10.1016/j.heares.2023.108767
This study supplies the cited evidence that richer auditory models can improve EEG prediction.
- supportsdoi:10.64898/2026.02.19.706583
The archived related-work discussion cites this record for the prior observation of residual EEG correlation after linear model responses are removed.
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesFixed delayed envelope regression had higher mean file error than instantaneous regression across listeners from ds004408 run 01 to run 02 under source time alignment
This study uses its processed 16 Hz EEG and envelope arrays, fixed 56-channel set, source metadata, raw content hashes, and conversion code. It adds a prediction test with separate training and test listeners and excerpts. The earlier covariance score is not treated as a prediction score.
- Speech reference gains were positive in 6 iEEG patients and 16 of 17 fMRI listeners; the mean rule failed
Related shared-response or source-identity benchmark. It supplied no numerical inputs or model choices.
- usesThe 95 percent listener interval for added sound band EEG prediction gain included zero in two ds004408 excerpts under source alignment
This study used 38 processed EEG arrays, wide envelopes, source metadata, source hashes, the good channel set, and retrieval code from this publication. It added eight sound band features and tested prediction for separate listeners in another excerpt. All inherited files were checked against the parent archive hashes.
- usesSource identity checks removed exact mean bias from uniform bootstrap draws of residual EEG covariance in 19 ds004408 listeners
This work used the processed EEG arrays, stimulus envelopes, common good channel list, source hashes, and retrieval code from this publication. The new full pair matrix also reproduced its primary mean for the same ten reference listeners and nine test listeners. Its resampling design differs from the design tested here.
- usesFixed delayed envelope regression had higher mean file error than instantaneous regression across listeners from ds004408 run 01 to run 02 under source time alignment
This study uses its processed 16 Hz EEG and envelope arrays, fixed 56-channel set, source metadata, raw content hashes, and conversion code. It adds a prediction test with separate training and test listeners and excerpts. The earlier covariance score is not treated as a prediction score.