Discover datasets

This dataset comprises simultaneous EEG-fMRI recordings from 22 participants (ages 23-51) during naturalistic viewing, collected at the Nathan Kline Institute. The 64-channel EEG data (61 cortical channels, 2 EOG, 1 ECG) was acquired concurrently with fMRI, supplemented by eye-tracking and respiratory recordings. The dataset includes demographic information and behavioral assessments (sleep quality, caffeine intake) to investigate correlations between electrical brain activity and hemodynamic fluctuations during naturalistic stimulation.

Participants

22

Channels

61 (10-10)

Citations

40

Size

256 GB

Version

vv1.0.0

Updated

Jul 10, 2026

Brain Treebank is a large-scale intracranial EEG dataset comprising 43 hours of iEEG recordings from 10 epilepsy patients watching naturalistic Hollywood movies, with 1,688 electrodes sampled at 2048 Hz. The dataset includes time-aligned linguistic annotations with word-level transcripts and Universal Dependencies syntax trees, providing a unique resource for studying neural language processing during naturalistic stimulation.

Participants

10

Size

508 GB

Version

vv1.0.0

Updated

Jul 10, 2026

VocalMind is a stereotactic EEG (sEEG) dataset comprising intracranial recordings from one participant performing vocalized, mimed, and imagined speech tasks in Mandarin Chinese, a tonal language. The dataset supports research on speech decoding, brain-computer interfaces, and neural mechanisms of overt and covert speech production. Raw recordings sampled at 1000 Hz were converted to BIDS iEEG format with event markers indicating stimulus onsets for each speech modality.

Participants

1

Citations

14

Size

1.89 GB

Version

vv1.0.0

Updated

Jul 10, 2026

Participants

2

Size

1.87 GB

Version

vv1.0.0

Updated

Jul 10, 2026

Participants

85

Size

126 GB

Version

vv1.0.0

Updated

Jul 10, 2026

THINGS-EEG2 is a large-scale EEG dataset comprising recordings from 10 subjects viewing 16,540 distinct training images and 200 test images presented via rapid serial visual presentation at 5 Hz. The dataset includes 63-channel EEG data sampled at 1000 Hz across 4 sessions per subject, with approximately 32,540 training trials and 16,000 test trials, designed to support computational modeling of human visual object recognition. Stimuli are drawn from the THINGS database, and the dataset includes resting-state recordings and behavioral annotations for each trial.

Participants

10

Channels

63 (10-10)

Size

198 GB

Version

vv1.1.0

Updated

Jul 10, 2026

A large-scale multi-laboratory replication study (N=356 participants across 9 UK institutions) investigating probabilistic prediction in language comprehension using event-related potentials. Participants read sentences word-by-word while EEG was recorded, with critical manipulations of indefinite articles (a/an) preceding expected or unexpected nouns. The study challenges strong prediction accounts by demonstrating reliable N400 effects on target nouns but not on preceding articles, contrary to the original DeLong et al. (2005) findings.

Participants

356

Channels

63 (10-10)

Citations

97

Size

102 GB

Version

vv1.1.0

Updated

Jul 10, 2026

A polysomnographic dataset comprising 1,983 overnight sleep recordings from Massachusetts General Hospital, including 994 training subjects with expert annotations and 989 test subjects. The dataset was created for the PhysioNet/Computing in Cardiology Challenge 2018 and contains 13 channels of physiological signals (EEG, EOG, EMG, respiratory, SpO2, ECG) sampled at 200 Hz, with annotations for sleep stages, respiratory events, and arousal types in the training set.

Participants

1983

Channels

6 (other)

Citations

92

Size

401 GB

Version

vv1.1.0

Updated

Jul 10, 2026

197 whole-night PSG, 100 subjects, expert 30s sleep staging (W/N1/N2/N3/REM) in events.tsv. EEG Fpz-Cz/Pz-Oz + EOG + EMG at 100Hz. PhysioNet DOI:10.13026/C2X676

Participants

100

Channels

7 (other)

Citations

90

Size

8.12 GB

Version

vv1.0.0

Updated

Jul 10, 2026

2417 clinical EEG recordings, 19-ch 10-20 montage, normal/abnormal labels. EDF format. Zenodo DOI:10.5281/zenodo.10909103

Participants

2417

Channels

21 (10-20)

Size

13.8 GB

Version

vv1.0.0

Updated

Jul 10, 2026