MNE sample dataset (MEG + EEG audiovisual task with a structural MRI)

Description

TODO(confirm) — the author’s catalog does not hold this dataset (§10.1 rule 3); this page carries only the §10.2 facts until a registry stub is added (§10.11 item 8).

Curriculum role: one continuous 277.7-s run (BIDS meg.json RecordingDuration). Checkerboard patterns to the left or right visual field interspersed with tones to the left or right ear, stimulus interval 750 ms, with an occasional central smiley face to which the subject pressed a key with the right index finger (MNE-Python doc/documentation/datasets.rst, §Sample)

Citation

• Paper DOI: 10.1016/j.neuroimage.2014.02.017
• Dataset DOI: 10.18112/openneuro.ds000248.v1.2.4
• Reference: Gramfort, A., Luessi, M., Larson, E., Engemann, D., Strohmeier, D., Brodbeck, C., Parkkonen, L., & Hämäläinen, M. (2014). MNE software for processing MEG and EEG data. NeuroImage, 86, 446-460. doi:10.1016/j.neuroimage.2014.02.017 — the citation the dataset itself asks for (ds000248 dataset_description.json HowToAcknowledge, which also names Gramfort et al. (2013), Frontiers in Neuroscience 7, doi:10.3389/fnins.2013.00267). Dataset: Gramfort, A., & Hämäläinen, M. S. MNE-Sample-Data. OpenNeuro ds000248 v1.2.4. doi:10.18112/openneuro.ds000248.v1.2.4

Download and access

Field	Value
Official source	https://mne.tools/stable/documentation/datasets.html
Access class	open
Size	1,652,774,934 bytes for MNE’s processed archive (MNE-sample-data-processed.tar.gz); 186,216,741 bytes for the OpenNeuro ds000248 v1.2.4 BIDS mirror
Format	fif, bids
BIDS	mirror
Mirrors	MNE-Python’s own download, the one mne.datasets.sample.data_path() uses: OSF node rxvq7, file GUID 86qa2, MNE-sample-data-processed.tar.gz, 1,652,774,934 bytes (mne/datasets/config.py, MNE 1.10.2); OpenNeuro ds000248 v1.2.4 — a BIDS re-packaging by the MNE authors, 22 files, 186,216,741 bytes, sub-01 + sub-emptyroom, with FreeSurfer derivatives; linked from MNE’s own dataset page

License

• Name: none
• Note (verbatim from the directory): CONTESTED at source. Four statements, every one read from primary material on 2026-09-18, none of them agreeing with the others. (1) MNE-Python’s own documentation source (doc/documentation/datasets.rst, §Sample, main branch) states a use restriction and no licence at all: ‘These data are provided solely for the purpose of getting familiar with the MNE software. The data should not be used to evaluate the performance of the MEG or MRI system employed.’ The word ‘license’ occurs exactly once in that whole file and it belongs to the Brainstorm section, not this one. (2) The channel mne.datasets.sample.data_path() actually downloads from carries no licence whatsoever: OSF node rxvq7 returns node_license: None and has no licence relationship (OSF API, 2026-09-18), and a streamed listing of the first 400 entries of MNE-sample-data-processed.tar.gz found no LICENSE, COPYING or NOTICE file and no top-level README — the only README in that prefix is subjects/fsaverage/mri.2mm/README, a technical note about 2 mm resampling. MNE’s fetcher asks for no licence acceptance here, in contrast to Brainstorm and HCP-MMP, which mne/datasets/config.py gates behind an explicit ‘Agree (y/[n])?’ prompt. (3) OpenNeuro ds000248 v1.2.4, the BIDS re-packaging MNE’s own page links to, states ‘License’: ‘CC0’ in dataset_description.json, with DatasetDOI 10.18112/openneuro.ds000248.v1.2.4 and Authors Alexandre Gramfort and Matti S Hämäläinen — and its README reproduces MNE’s acquisition paragraph verbatim while omitting the ‘solely for the purpose of getting familiar’ sentence. (4) The archive bundles the FreeSurfer subject directories subjects/fsaverage/ and subjects/fsaverage_sym/, and MNE’s documentation says ‘The sample dataset is distributed with fsaverage for convenience’ — so part of this dataset is also governed by whatever governs ds-fsaverage, which is the FreeSurfer Software License Agreement. Per §10.7 the most restrictive reading governs: the channel the site would actually use ships no grant of any kind and MNE’s own words restrict use to familiarisation with MNE, so license.name is recorded as ‘none’ and the sole named licence anyone states (OpenNeuro’s CC0) is recorded here and in source.dataset_doi rather than in license.name. snippets therefore derives to ‘no’ twice over (name ‘none’; license_status contested with no license_decision). Only the author can decide this, exactly as §13 item 15 was decided for ds-erpcore; until then nothing derived from this dataset ships.
• Snippets on this site: no — the site ships no snippet or precomputed product from this dataset; lessons link the official access page.

Acquisition

Field	Value
Device	Elekta/Neuromag Vectorview — 306 MEG channels (102 magnetometers + 204 planar gradiometers) recorded simultaneously with a 60-channel EEG cap; MRI on a Siemens 1.5 T Sonata with an MPRAGE sequence
Device class	research-cap
Channels	60 (60 EEG + 1 EOG + 306 MEG + 9 trigger channels (sub-01_task-audiovisual_run-01_channels.tsv and _meg.json, OpenNeuro ds000248 v1.2.4, read 2026-09-18))
Sampling rate (Hz)	600.614990234375 (BIDS meg.json SamplingFrequency = 600.614990234375 Hz for the MEG+EEG recording)
Online filters	none recorded (BIDS meg.json SoftwareFilters: ‘n/a’)
Reference	TODO(confirm)
Mains frequency (Hz)	60
Paradigms	audiovisual
Participants	1 (sub-01 plus sub-emptyroom; a single-subject example dataset, not a cohort)
Population	1 participant; the BIDS participants.tsv records age, sex and handedness as n/a. The release adds an empty-room recording (sub-emptyroom)
Clinical groups	none
Sessions	1
Durations	one continuous 277.7-s run (BIDS meg.json RecordingDuration). Checkerboard patterns to the left or right visual field interspersed with tones to the left or right ear, stimulus interval 750 ms, with an occasional central smiley face to which the subject pressed a key with the right index finger (MNE-Python doc/documentation/datasets.rst, §Sample)

Used on this site

L5.4, L5.5

Loader

Fetches only the stated subset. Recorded loader: mne.datasets.sample.data_path() (downloads the whole 1.65 GB archive; the OpenNeuro mirror ds000248 is per-file addressable and much smaller).

# TODO(confirm): loader recorded from the MNE API, not from the catalog.
import mne
data_path = mne.datasets.sample.data_path()  # ~1.5 GB, fetched once
raw = mne.io.read_raw_fif(data_path / 'MEG' / 'sample' / 'sample_audvis_raw.fif')

Caveats worth knowing

• Licence is CONTESTED at source and no asset derived from this dataset ships: MNE’s own documentation grants nothing and restricts use to ‘getting familiar with the MNE software’, MNE’s download channel carries no licence text at all, and only the OpenNeuro BIDS mirror says CC0. §10.7’s most-restrictive rule governs until the author records a license_decision.
• Not in the author’s catalog: the acquisition facts here were read from the dataset’s own BIDS sidecars on OpenNeuro ds000248 v1.2.4 (sub-01_task-audiovisual_run-01_meg.json, _channels.tsv, participants.tsv) and from MNE-Python’s documentation source on 2026-09-18, not from data/catalog (§10.1 rule 3). The author should add a registry stub (§10.11 item 8).
• One subject. Anything this dataset shows about individual-MRI forward modelling is an illustration, not evidence about a population.
• The archive bundles FreeSurfer’s fsaverage and fsaverage_sym subjects, so part of it inherits the ds-fsaverage licence question as well.
• reference: TODO(confirm) — the EEG online reference is not recorded in the BIDS sidecar or in MNE’s dataset documentation.

Related datasets

• ERP CORE (Compendium of Open Resources and Experiments)
• EEG Motor Movement/Imagery Dataset (EEGMMIDB)
• LEMON (MPI-Leipzig Mind-Brain-Body)
• Dortmund Vital Study resting EEG
• SRM Resting-state EEG

# Loader recorded in the catalog: mne.datasets.sample.data_path() (downloads the whole 1.65 GB archive; the OpenNeuro mirror ds000248 is per-file addressable and much smaller)
# Fetch only the subset you need; see the loader's docstring for subject/run arguments.
import mne

files = mne.datasets.sample.data_path() (downloads the whole 1.65 GB archive; the OpenNeuro mirror ds000248 is per-file addressable and much smaller)(...)  # TODO(confirm) subset arguments