EEG Dataset for Cortical Tracking of Surprisal in Speech

Unprocessed 64-channel EEG data was recorded from 13 participants listening to English audiobooks in noiseless conditions. The dataset includes the raw EEG signals, the audio stimuli, story transcripts, and derived linguistic features, all time-aligned. It was created to study cortical tracking of surprisal during continuous speech comprehension.

Use Cases

• Modeling cortical tracking of linguistic surprisal based on aligned EEG and feature data.
• Studying neural correlates of continuous speech comprehension using high-density EEG recordings.
• Benchmarking neural decoding algorithms for naturalistic auditory stimuli.
• Analyzing the relationship between acoustic envelopes, linguistic features, and brain activity.

Strengths

• Includes raw, unprocessed EEG data from 13 participants, providing a foundation for varied analysis pipelines.
• Contains multimodal data: 1 kHz EEG, 16 kHz audio, transcripts, and derived linguistic features, all time-aligned.
• Electrode placement followed the standard 10-20 system using a 64-channel cap, ensuring methodological consistency.

Limitations

• Dataset size, row count, and specific column names are not provided in the source metadata.
• The last updated date is unknown, which may affect confidence in the currency of the repository links.
• Relies on external scripts and documentation for full alignment and preprocessing details.

Provenance

Source

Imperial College London, authored by Hugo Weissbart.

Collection Method

EEG was recorded at 1 kHz via an actiCHamp amplifier while participants listened to audiobooks; audio was recorded simultaneously via a StimTrack device for alignment.