All functional and anatomical data has been formatted in BIDS, for more information visit the Brain Imaging Data Structure documentation site. Some of the files do not follow the main BIDS convention:

  • Anatomical sequences with multiple contrasts are following BEP001.

  • Spinal cord imaging use Body Part tag proposed in BEP025 (bp-cspine) to allow to distinguish them from brain anatomical images acquired with the same contrasts.

Note that BIDS session names have no meaning apart from being data acquired in the same session. The number of runs, the tasks and their order within each session will not match from one participant to another. Note that a few session indices are skipped if the whole session was discarded for various scanning issues.


Six healthy participants (aged 31 to 47 at the time of recruitment in 2018), 3 women (sub-03, sub-04 and sub-06) and 3 men (sub-01, sub-02 and sub-05) consented to participate in the Courtois Neuromod Project for at least 5 years. Three of the participants reported being native francophone speakers (sub-01, sub-02 and sub-04), one as being a native anglophone (sub-06) and two as bilingual native speakers (sub-03 and sub-05). All participants reported the right hand as being their dominant hand and reported being in good general health.

Exclusion criteria included visual or auditory impairments that would prevent participants from seeing and/or hearing stimuli in the scanner and major psychiatric or neurological problems. Standard exclusion criteria for MRI and MEG were also applied. Lastly, given that all stimuli and instructions are presented in English, all participants had to report having an advanced comprehension of the English language for inclusion.


The anatomical dataset includes longitudinal anatomical images of the brain and upper spinal cord at an approximate rate of 4 sessions a year. The primary intended use of this dataset is to monitor the structural stability of the brain of participants for the duration of the study. Many quantitative measures of brain structure can also be derived and included in analyses, such as gray matter morphometry, tractography or measures of myelination.

The MRI sequences are described in more detailed in and Spinal cord anatomical sequences, including pdfs of the Siemens exam cards.

Brain T1w, T2w and DWI were copied from the HCP aging and development protocol for Prisma MRI scanner. Other sequences were selected and optimized by the Courtois NeuroMod team.

All images covering the face were anonymized by zeroing the data in the face, teeth and ears regions with a custom mask warped from the MNI space based on a linear registration of the T1w brain MRI series. This defacing script is available here


This cneuromod dataset is called HCP test-retest (hcptrt), because participants repeated 15 times the functional localizers developed by the Human Connectome Project, for a total of approximately 10 hours of functional data per subject. The protocol consisted of seven tasks, described below (text adapted from the HCP protocol). Before each task, participants were given detailed instructions and examples, as well as a practice run. A session was typically composed either of two repetitions of the HCP localizers, or one resting-state run and one HCP localizer. The e-prime scripts for preparation and presentation of the stimuli can be found in the HCP database. Stimuli and e-prime scripts were provided by the Human Connectome Project, U-Minn Consortium (Principal Investigators: David Van Essen and Kamil Ugurbil; 1U54MH091657) funded by the 16 NIH Institutes and Centers that support the NIH Blueprint for Neuroscience Research, and by the McDonnell Center for Systems Neuroscience at Washington University. Note that in the cneuromod DataLad, functional runs are named func_sub-<participant>_ses-<sess>_task-<task>_run-<run>, where the <participant> tag includes sub-01 to sub-06. For each functional run, a companion file _events.tsv contains the timing and type of events presented to the subject. Session tags <sess> are 001, 002 etc, and the number and composition of sessions vary from subject to subject. The <task> tags are restingstate, gambling, motor, social, wm, emotion, language and relational, as described below. Tasks that were repeated twice have separate <run> tags (01, 02).


The duration of BOLD series are slightly varying across participants and repetitions. If consistent length is required by analysis, series can be trimmed at the end to match duration, task being aligned to the first TR.


gambling duration: approximately 3 minutes. Participants were asked to guess whether a hidden number (represented by a “?” during 1500ms) was above or below 5 (Delgado et al. 2000). They indicated their choice using a button press, and were then shown the actual number. If they guessed correctly they were told they won money (+$1.00, win trial), if they guessed incorrectly they were told they lost money (-$0.50, loss trial), and if the number was exactly 5 they were told that they neither won or lost money ($0, neutral trial). Note that no money was actually given to the participants and, as such, this task may not be an accurate reproduction of the HCP protocol. The conditions were presented in blocks of 8 trials of type reward (6 win trials pseudo randomly interleaved with either 1 neutral and 1 loss trial, 2 neutral trials, or 2 loss trials) or of type punishment (6 loss trials pseudo-randomly interleaved with either 1 neutral and 1 win trial, 2 neutral trials, or 2 win trials). There were four blocks per run (2 reward and 2 punishment), and two runs in total.


motor duration: approximately 3 minutes. This task was adapted from (Buckner et al. 2011; Yeo et al. 2011). Participants were presented a visual cue, and were asked to either tap their left or right fingers (event types left_hand and right_hand, resp.), squeeze their left or right toes (event types left_foot and right_foot, resp.), or move their tongue to map motor area (event type tongue). Each movement lasted 12 seconds, and in total there were 13 blocks, with 2 of tongue movements, 4 of hand movements (2 right_hand and 2 left_hand), and 4 of foot movements (2 right_foot and 2 left_foot), and three 15 second fixation blocks where participants were instructed not to move anything. There were two runs in total, and 13 blocks per run.

Language processing

language duration: approximately 4 minutes. Participants were presented with two types of events. During story events, participants listened to an auditory story (5-9 sentences, about 20 seconds), followed by a two-alternative forced-choice question. During math events, they listened to a math problem (addition and subtraction only, varies in length), and were instructed to push a button to select the first or the second answer as being correct. The task was adaptive so that for every correct answer the level of difficulty increased. The math task was designed this way to maintain the same level of difficulty between participants. There were 2 runs, each with 4 story and 4 math blocks, interleaved.

Social cognition

social duration: approximately 3 minutes. Participants were presented with short video clips (20 seconds) of objects (squares, circles, triangles) that either interacted in some way (event type mental), or moved randomly on the screen (event type random) (Castelli et al. 2000; Wheatley et al. 2007). Following each clip, participants were asked to judge whether the objects had a “Mental interaction” (an interaction that appeared as if the shapes were taking into account each other’s feelings and thoughts), whether the were “Not Sure”, or if there was “No interaction”. Button presses were used to record their responses. In each of the two runs, participants viewed 5 mental videos and 5 random videos, and had 5 fixation blocks of 15 seconds each.

Relational processing

relational duration: approximately 3 minutes. Participants were shown 6 different shapes filled with 1 of 6 different textures (Smith et al. 2007). There were two conditions: relations processing (event type relational), and control matching condition (event type control). In the relational events, 2 pairs of objects were presented on the screen, with one pair at the top of the screen, and the other pair at the bottom. Participants were instructed to decide what dimension differed in the top pair (shape or texture), and then decide if the bottom pair differed, or not, on the same dimension (i.e. if the top pair differed in shape, did the bottom pair also differ in shape). Their answers were recorded by one of two button presses: “a” differ on same dimension; “b” don’t differ on same dimension. In the control events, participants were shown two objects at the top of the screen, and one object at the bottom of the screen, with a word in the middle of the screen (either “shape” or “texture”).They were told to decide whether the bottom object matched either of the top two objects on that dimension (i.e., if the word is “shape”, did the bottom object have the same shape as either of the top two objects). Participants responded “yes” or “no” using the button box. For the relational condition, the stimuli were presented for 3500 ms, with a 500 ms ITI, and there were four trials per block. In the controlcondition, stimuli were presented for 2800 ms, with a 400 ms ITI, and there were 5 trials per block. In total there were two runs, each with three relational blocks, three control blocks and three 16-second fixation blocks.

Emotion processing

emotion duration: approximately 2 minutes. Participants were shown triads of faces (event type face) or shapes (event type shape), and were asked to decide which of the shapes at the bottom of the screen matches the target face/ shape at the top of the screen (adapted from Smith et al. 2007). Faces had either an angry or fearful expression. Faces, and shapes were presented in three blocks of 6 trials (3 face and 3 shape), with each trial lasting 2 seconds, followed by a 1 second inter-stimulus interval. Each block was preceded by a 3000 ms task cue (“shape” or “face”), so that each block was 21 seconds long, including the cue. In total there were two runs, three face blocks and three shape blocks, with 8 seconds of fixation at the end of each run.

Working memory

wm duration: approximately 5 minutes. There were two subtasks: a category specific representation, and a working memory task. Participants were presented with blocks of either places, tools, faces, and body parts. Within each run, all 4 types of stimuli were presented in block, with each block being labelled as a 2-back task (participants needed to indicate if they saw the same image two images back), or a version of a 0-back task (participants were shown a target at the start of the trial and they needed to indicate if the image that they were seeing matched the target). There were thus 8 different event types <stim>_<back>, where <stim> was one of place, tools, face or body, and <back> was one of 0back or 2back. Each image was presented for 2 seconds, followed by a 500 ms ITI. Stimuli were presented for 2 seconds, followed by a 500 ms inter-task interval. Each of the 2 runs included 8 event types with 10 trials per type, as well as 4 fixations blocks (15 secs).

Resting state

restingstate duration: 15 minutes. In every other session, one resting-state fMRI run was acquired, giving 5 runs per participant. Participants were asked to have their eye open, be looking at fixation cross in the middle of the screen and be instructed to not fall asleep. A total of five resting-state fMRI runs were acquired per subject.


This dataset includes about 10 hours of functional data for all 6 participants. The python & psychopy scripts for preparation and presentation of the clips can be found in src/tasks/ of the following github repository. Session tags <sess> were 001, 002 etc, and the number and composition of sessions varied from subject to subject. The <task> tags used in DataLad corresponded to each movie (bourne, wolf, life, figures) and a numerical index of the segments shown as each movie was cut into roughly ten minutes segments presented in separate run. Exact cutting points were manually selected to not interrupt the narrative flow. Fade out to a black screen was added at the end of each clip, and with a few seconds overlap between the end of a clip and the beginning of the next clip. The movie segments can be found under movie10/stimuli/<movie>/<movie>_seg<seg>.mkv, and the functional runs are named func_sub-<participant>_ses-<sess>_task-<movie><seg>, where the <participant> tag ranges from sub-01 to sub-06. A companion file _events.tsv contains the timing and type of conditions presented to the subject.

The participants watched the following movies (cogatlas):

  • <task> name bourne: The Bourne supremacy. Duration ~100 minutes.

  • <task> name wolf: The wolf of wall street. Duration ~170 minutes.

  • <task> name figures: Hidden figures. Duration ~120 minutes. This movie was presented twice, for a total duration of ~240 minutes.

  • <task> name life: Life Disc one of four: “Challenges of life, reptiles and amphibian mammals”. DVD set was narrated by David Attenborough. Duration, and lasted ~50 minutes. This movie was presented twice, for a total duration of ~100 minutes.

It should be noted that although three of the participants are not native anglophones, all participants watched the movies in English. The three native francophone participants are fluent in English and report regularly watching movies in English.


The duration of BOLD series are slightly varying across participants and repetitions. If consistent length is required for analysis, series can be trimmed at the end to match duration, movie segments being aligned to the first TR.


There are instances of re-scanned segments (due to scan QC fail), these re-scans will be in separate sessions. These should be handled or excluded in analysis requiring continuity of the presentation of the story.


This dataset contains functional data acquired while showing participants episodes of the Friends TV show in English. It includes seasons 1-6 for all subjects, except sub-04 who only completed seasons 1-4 (and a few segments of season 5). Each episode is cut in two segments (a/b) to allow more flexible scanning and give participants opportunities for breaks. There is a small overlap between the segments to allow participants to catch up with the storyline. The task BIDS entity identifies the season, episode and segments (a/b) as such task-s<eason>e<pisode>[ab].


A mistake happened when ripping the first season, causing s01e01 and s01e06 to be swapped in name and order of presentation. Files were renamed afterward to match external data such as annotations. However the order of presentation remains, slightly disrupting the storyline presented to the participant.


This dataset contains a single session per participant (N=5) when they read chapter 9 from chapter 9 of Harry Potter and the Sorcerer’s Stone. The text was presented word by word, at a 2Hz pace (each word presented for .5s). This chapter was split over 7 runs of approximate equal length. The stimuli used in this dataset are taken from the experiment reported by Wehbe et al. (2014) for which a separate fMRI dataset (N=9) has been collected and shared.


This is a pure behavioral dataset collected while participants trained at home on the videogame Shinobi III The Return of the Ninja Master. A subset of 3 levels of the game was selected for their similarity in terms of core gameplay although some mechanics were specific to each level. These levels were the same than those used in the shinobi dataset.

The participants were first introduced to the game by playing the Level-1 at least once. This level introduces the basic game mechanics (moving from left to right, avoid or kill enemies) at an abordable difficulty level for most players. Then, the participants were able to freely choose the level on which they played as well as the training sessions frequency. No training regimen was imposed to the participants making that dataset highly heterogeneous. The dataset consists of sessions of gameplay as collections of .bk2 files recorded by the gym-retro API.

This dataset can be used to analyze learning or individual game-play styles, and can be investigated in conjunction with the fMRI dataset.


This dataset contains about 10h of gameplay on the videogame Shinobi III The Return of the Ninja Master, for N=4 participants (sub-01, sub-02, sub-04 and sub-06). Participants used a custom-built fully fiber-optic MRI controller, designed by the team and described in Harel et al. (2022). In each run, participants played 3 levels in cycles and always in the same order. These levels were selected in the game to have fairly homogeneous core game mechanics (see the Sega documentation for more details on game structure):

  • Level-1 corresponded to round 1 of the original game, “Zeed’s Resurrection”. It included one mini-boss and one boss fight.

  • Level-4 corresponded to the beginning of round 4 of the original game, “Destruction”. It included no mini-boss or boss fight.

  • Level-5 corresponded to the beginning of round 5 of the original game, “Electric demon”. It included one mini-boss fight and no boss fight.

Participants moved to the next level if they successfully completed a level, or lost three lives. A new level was then initiated unless 10 minutes had elapsed from the start of the run, at which point the run ended. The duration of each run is thus variable to a degree, with a minimum of ten minutes. Due to the fixed order in the cycle, Level-1 was repeated more often than Level-4 and Level-5.

In this dataset and the related documentation, we use the term run to designate a single functional sequence acquisition (per the usual in neuroimaging). The term repetition is used to designate the play of a single level (from start to either completion or the loss of three lives). As such, each run contains around 3 to 5 repetitions.

For each functional run, a companion file _events.tsv contains the timing and duration of each repetition played, as well as a _annotated_events.tsv file that additionally contains richer annotations, including button presses, handcrafted annotations (Kills, Health losses), and frame-wise RAM values. Additional documentation on the available annotations can be found here.

The companion .bk2 files can be found in the <participant>/<sess>/gamelogs folder.


Due to a programming error a certain number of game recording files were lost during acquisition, these repetitions are still listed in the events file but their stim_file field is left blank. Choice is left to the user whether to exclude the corresponding fMRI volumes or not for their analysis.