DECOMEG

Non-invasive brain recordings of people typing from memory, the dataset behind Brain2Qwerty and its companion neuroscience study. The Basque Center on Cognition, Brain and Language (BCBL) recorded 35 healthy native-Spanish-speaking adults, all right-handed skilled touch-typists, as they read a briefly shown Spanish sentence word by word, held it in memory, and then typed it on a custom MR-compatible QWERTY keyboard with no on-screen feedback. Brain activity was captured with magnetoencephalography (Megin/Elekta Neuromag, 306 channels: 102 magnetometers plus 204 planar gradiometers) and electroencephalography (BrainVision actiCAP, 64 channels), both sampled at 1 kHz, alongside behavioral logs of every keystroke, word, and sentence with precise timing. The release contains 231 raw MEG .fif files across 29 recording sessions, 117 EEG BrainVision triplets, and the matching MATLAB behavioral logs, about 262 GB in total covering roughly 5,100 typed sentences in MEG and 4,000 in EEG. It is a rich substrate for motor and language decoding, brain-to-text and brain-computer-interface research, and MEG/EEG methods development. Released by BCBL under CC BY-NC 4.0 (non-commercial); see Lévy et al. (Brain2Qwerty, 2025) and the companion study Zhang et al. (2025, arXiv:2502.07429).

Category: Neuroscience
Provider: Basque Center on Cognition, Brain and Language (BCBL)
License: CC-BY-NC-4.0
DOI: 10.48550/arXiv.2502.07429
Hosted on: Hugging Face, bucket bcbl190626/SpanishBCBL, us
Access: Free and public. No account with the data provider is needed.
Source: https://huggingface.co/datasets/bcbl190626/SpanishBCBL

Data structure and loading code

Fields

`MEG channels`	306 — 102 magnetometers + 204 planar gra
`EEG channels`	64 — BrainVision actiCAP slim @ 1 kHz
`behavioral log (.mat)`	MATLAB struct: stimulus sentence, per-ke
`events`	keystroke, word, and sentence timing mar

Load the data

# Mounted read-only at /mnt/archive. MNE-Python is the tool for this data.
# pip install mne
import mne, glob
fif = sorted(glob.glob('/mnt/archive/MEG/FIF/*/*/Block1.fif'))[0]
raw = mne.io.read_raw_fif(fif, preload=False)   # follows any Block1-1.fif split automatically
print(raw.info)                                  # 306 MEG channels @ 1 kHz
# matching behavioral log (stimulus sentence, keystrokes, timing):
from scipy.io import loadmat
log = loadmat('/mnt/archive/MEG/logs/S1-session1-1.mat', squeeze_me=True, struct_as_record=False)
print([k for k in log if not k.startswith("__")])

Written for a Scigantic notebook, where /mnt/archive is the mounted dataset.

Analyze this dataset without downloading it

On Scigantic, DECOMEG opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

DECOMEG

Fields

Load the data

Analyze this dataset without downloading it

Related Neuroscience datasets