Audiovisual Learning MEG Dataset

OpenNeuro/NEMAR Dataset:ds002598 #Files:396 Dataset size:136.7 GB #MEG Channels:306 #EOG Channels:1 #Trigger Channels:11

BIDS Version: 1.2.0 HED Version: Version: 1.0.2

On Brain life: True Published Date: 2020-02-28 10:36:16 Tasks: AVLearn

Available Modalities


#Sessions: 2 #Scans/session: 8 #Participants: 30 Ages (yrs): 19 - 36 License: CC0

Dataset DOI: 10.18112/openneuro.ds002598.v1.0.2

Uploaded by Weiyong Xu on 2020-02-25 19:07:51

Weiyong Xu, Orsolya Beatrix Kolozsvari, Robert Oostenveld , Jarmo Arvid Hämäläinen

We would like to thank Suvi Karjalainen, Aino Sorsa and Ainomaija Laitinen for their help in data collection. This work has been supported by the European Union projects ChildBrain (Marie Curie Innovative Training Networks, no. 641652), Predictable (Marie Curie Innovative Training Networks, no. 641858) and the Academy of Finland (MultiLeTe #292466). The authors declare no conflicts of interest.

How to Acknowledge
Please cite: Weiyong, X., Orsolya, B. K., Robert, O., & Jarmo, A. H. Rapid Changes in Brain Activity During Learning of Grapheme-Phoneme Associations in Adults. NeuroImage, 117058. Audiovisual Learning MEG Dataset.

  • ChildBrain (Marie Curie Innovative Training Networks, no. 641652)
  • Predictable (Marie Curie Innovative Training Networks, no. 641858)
  • Academy of Finland (MultiLeTe #292466)
  • References and Links
  • Ethics Approvals


    ========= Dataset description: Magnetoencephalography (MEG) dataset on grapheme-phoneme learning in 2 consecutive days

    This MEG dataset was prepared in the Brain Imaging Data Structure (MEG-BIDS, Niso et al. 2018) format using MNE-BIDS (Appelhoff et al. 2019).

    In total 36 people were scheduled to participate in the study. The data from 6 participants are not included in this dataset; 2 were excluded due to a low learning accuracy during the whole training sessions on Day 1, and for the other 4 participants MEG data were not measured because of cancellation.


    Day 1: 12 training and testing blocks Day 2: 6 training and testing blocks

    The auditory and visual stimuli were 12 Georgian letters and 12 Finnish phonemes in 2 sets:

    Set 1

    a, ä, e, t, s, k

    ჸ, ჵ, ჹ, უ, დ, ჱ

    Set 2

    o, ö, i, p, v, d

    ც, ჴ, ნ, ფ, ღ, წ

    Set 1 was used as Learnable (LB) and Set 2 was used as Control (CT) for participants with odd ID numbers. Set 2 was used as Learnable and Set 1 was used as Control for participants with even ID numbers.

    The following triggers are included in the raw .fif files and are also in the “trigger” column of the events files:

    LB/A/Training: 101,102,103,104, ...

    LB/V/Training: 110,120,130,140, ...

    LB/AVC/Training: 111,122,133,144, ...

    LB/AVI/Training: 112,113,114,121,123,124,131,132,134,141,142,143, ...

    LB/AVC/Testing: 311,322,333,344, ...

    LB/AVI/Testing: 312,313,314,321,323,324,331,332,334,341,342,343, ...

    CT/A/Training: 201,202,203,204, ...

    CT/V/Training: 210,220,230,240, ...

    CT/AV/Training: 212,213,214,221,223,224,231,232,234,241,242,243, ...

    CT/AV/Testing: 412,413,414,421,423,424,431,432,434,441,442,443, ...

    LB/YES: 511

    LB/NO: 510

    CT/UNKNOWN: 520

    Test trial: 600

    Correct button press: 610

    Wrong button press: 630


    Three anatomical landmarks were used to define the MEG head coordinate system: Nasion, LPA, and RPA.

    The position of the HPI coils and the head shape (>100 points evenly distributed over the scalp) were digitized using the Polhemus Isotrak digital tracker system (Polhemus, Colchester, VT, United States).

    MEG was recorded using the Elekta Neuromag TRIUX system (Elekta AB, Stockholm, Sweden) at the Centre for Interdisciplinary Brain Research, University of Jyväskylä.

    Data were acquired from 306 MEG channels and 2 EOG channels with a sampling rate of 1000 Hz, an online band-pass filter of 0.1-330 Hz, and a 68° upright gantry position.

    Maxfilter version 3.0.17 was used for movement compensation using temporal signal-space separation (tSSS).

    Bad MEG channels were identified manually and were interpolated by Maxfilter.


    Individual structural MRIs were not acquired. For source reconstruction, it is recommended to use a template, for example the “fsaverage“ brain from Freesurfer (, and to scale the template head model and source space to the shape and size of the individual participants (as obtained from the head shape points).