{"paper_id":"4e932e3d-6d5e-4b38-bf42-83f8b1fd20c1","body_text":"A neuroimaging database combining \nmovie-watching, eye-tracking, sensorimotor \nmapping, and cognitive tasks \n \nEgor Levchenko1,2, Hugo Chow-Wing-Bom2, Fred Dick2, Adam Tierney1 & Jeremy I Skipper2  \n1 Birkbeck, University of London, UK \n2 University College London, UK \n \n1 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nAbstract \nWe provide a multimodal naturalistic neuroimaging database (NNDb-3T+), designed \nto support the study of brain function under both naturalistic and controlled \nexperimental conditions. The database includes high-quality 3T fMRI data from 40 \nparticipants acquired during full-length movie-watching and three sensory mapping \ntasks: somatotopy, retinotopy, and tonotopy. Each participant also completed \nsynchronized eye-tracking during movie-watching and retinotopy, physiological \nrecordings, and a battery of behavioral and cognitive assessments. Data were \ncollected across two MRI sessions and a remote testing session, with all data \norganized in a BIDS-compliant format. Technical validation confirms high data \nquality, with minimal head motion, accurate eye-tracker calibration, and robust \ntask-evoked activation patterns. The database provides a unique resource for \ninvestigating individual differences, functional topographies, multimodal integration, \nand naturalistic cognition. All raw and preprocessed data, quality metrics, and \npreprocessing scripts are publicly available to support reproducible research. \n \n \n \nKeywords: human brain, fMRI, naturalistic imaging, multimodal, movie-watching, retinotopy, \ntonotopy, somatotopy, eye-tracker, behaviour, cognition \n \n \n \n \n2 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nBackground & Summary \nOne of the main goals of human neuroscience is to uncover how the brain \noperates in the complex, continuous experiences during everyday life. To achieve \nthis, researchers employ both naturalistic (e.g. watching a movie or listening to a \nnarrative) and task-based paradigms (e.g. n-back task or sensory mapping tasks). \nNaturalistic paradigms tend to elicit higher immersion and attentiveness in \nparticipants (Ki et al., 2016) , which improves ecological validity and also reduces \nhead motion in the scanner (Vanderwal et al., 2019). The naturalistic approach has \nproven effective for studying a range of processes, including the hierarchy of \ntemporal receptive fields (Lerner et al., 2011), event segmentation in memory \n(Baldassano et al., 2018) , default mode network dynamics (Simony et al., 2016) , \nselective attention (Nguyen et al., 2017) , emotions and social cognition (Redcay & \nMoraczewski, 2020), and functional connectivity (Gal et al., 2022). \nOver the past decade, numerous publicly available fMRI datasets have \nembraced naturalistic paradigms to explore how the brain responds to continuous, \nreal-world stimuli. Notable examples include StudyForrest (Hanke et al., 2014) , \nwhich combines 7T fMRI with eye-tracking and physiological recordings during audio \nand audiovisual presentation of the movie “Forrest Gump”; the Sherlock dataset \n(Chen et al., 2017) , which includes free recall during scanning; and the Grand \nBudapest Hotel dataset (Visconti Di Oleggio Castello et al., 2020) , focused on social \ncognition. Other large-scale efforts such as the Narratives  (Nastase et al., 2021) and \nCamCAN (Shafto et al., 2014)  datasets have used spoken stories or short films to \ninvestigate aging, language, and attention. These resources have advanced the field \nby demonstrating that naturalistic stimuli evoke reliable, temporally aligned neural \nresponses across individuals, and can be used to study phenomena like event \nsegmentation, narrative comprehension, and social perception. Despite this \nprogress, most existing datasets emphasize either naturalistic stimulation or \ncontrolled task-based mapping, rarely integrating both within the same cohort. \nFurthermore, multimodal recordings such as eye-tracking and physiological \nmeasures are often missing or only partially available. \nAmong the currently available naturalistic fMRI datasets, the Naturalistic \nNeuroimaging Database (NNDb v1.0) (Aliko et al., 2020) stands out by introducing a \nlarge-scale open dataset in which 86 participants watched one of ten full-length \nfeature films spanning diverse genres, accompanied by extensive behavioral and \ncognitive phenotyping. While NNDb v1.0 was designed to emphasize diversity of \nmovie genres (using ten different films) and a relatively large sample size, the \npresent dataset, Naturalistic Neuroimaging Database  3T+ (NNDb-3T+), provides a \nuniquely rich multimodal resource with several tasks and physiological recordings for \neach participant. The dataset combines high-quality fMRI data from 40 participants \nduring full-length movie-watching (Back to the future), somatotopy, retinotopy, and \ntonotopy tasks, eye-tracking data, physiological recordings, and extensive behavioral \nmeasures. Figure 1 provides an overview of the collected data, preprocessing \ntechniques, and quality control analyses for each task. NNDb-3T+ is, to our \n3 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nknowledge, the first publicly available dataset to combine naturalistic movie viewing, \nthree distinct sensory mapping tasks, synchronized eye-tracking, physiological \nmonitoring, and behavioral profiling within the same participants. We anticipate that \nNNDb-3T+ will support a wide range of future investigations, including individual \ndifferences in naturalistic processing, neural modeling of sensory hierarchies, and \nthe development of novel analytic methods for multimodal neuroimaging. All raw and \npreprocessed data are shared in Brain Imaging Data Structure (BIDS) valid format \n(Gorgolewski et al., 2016) , and quality control metrics are provided. Scripts for \npreprocessing and validation are openly available on GitHub to promote \nreproducibility (https://github.com/levchenkoegor/movieproject2). \n4 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n \nFigure 1. Overview of data collection, preprocessing, and validation pipeline. (1) \nData collection: data were acquired across two MRI sessions and a home \nassessment. At home, participants completed demographic questionnaires, \ncognitive tasks, and self-report measures. Session 1 included acquisition of a \n5-minute high-resolution structural MPRAGE scan and a naturalistic \nmovie-watching task split into three parts, each accompanied by a \nphase-encoding reversed (PErev) scan for distortion correction. Session 2 \ncomprised 2-minute structural imaging and functional localizers: somatotopic \nmapping (2 runs, one with PErev), retinotopic mapping (3 runs, including PErev), \nand tonotopic mapping (2 runs, including PErev). (2) Preprocessing: the pipeline \nwas applied to each task separately. Somatotopy and movie-watching (Back to \nthe Future) tasks followed the same pipeline shown in the current Figure and \nretinotopy and tonotopy tasks followed minimal preprocessing (no alignment, \nmasking and smoothing; see Preprocessing for more details). (3) Validation: Data \nquality and validity were assessed using multiple metrics: amount of head \nmovement for each task, inter-subject correlation during the backtothefuture task, \neye-tracker calibration quality before each run of backtothefuture, and averaged \nactivation maps for somatotopy and retinotopy tasks. \n5 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nMethods \nParticipants \nWe recruited 44 participants using participant pool management software \n(http://www.sona-systems.com/), advertisement on the University campus, and word \nof mouth. Participants were prescreened based on MRI safety criteria (e.g., no metal \nimplants) and self-reported good (or corrected to good) vision and hearing. The final \nsample consisted predominantly of right-handed, native English speakers, between \n18 and 45 years old with no neurological diseases. A few exceptions are noted: one \nleft-handed person with German/English first language and one with Farsi, one \nambidexter, five participants reported prior disorders (Obsessive Compulsive \nDisorder, Autism Spectrum Disorder, Generalised Anxiety Disorder, Depression and \nADHD). Two participants failed to attend and two were excluded due to technical \nproblems during the acquisition (felt uncomfortable inside the scanner). The final \nsample consisted of 40 participants: 21 females, 18–45 years, M = 29.02, SD = 6.20 \nyears (though one participant did not complete the questionnaire). \nThe questionnaire and cognitive tasks collected remotely were approved by \nThe Ethics Committee of the School of Psychological Sciences at Birkbeck \n(Reference number: 2324006). The whole study including the MRI was approved by \nthe ethics committee at the University College London (Reference number: \nfMRI/2023/003). All participants provided written consent to participate in the study \nand share their data. At the end of the study, participants received £67.50 in the form \nof a voucher. \nProcedure \nBefore their visit, participants filled out an MRI safety form to inquire about the \npresence of metal implants, pacemakers, and other potential hazards for the MRI \nexamination. If they were MRI-safe, they filled out a questionnaire about \ndemographic information, language background, musical experience, and knowledge \nof movies. Then the participant completed a set of cognitive tests on the Cognitron \nplatform (https://www.cognitron.co.uk/) and two scanning sessions were scheduled. \nTwo consent forms were signed online: one before the questionnaire and one before \nthe cognitive tests. \nAt the beginning of the first scan day, the participant completed an MRI safety \nform again and signed a paper-based ethics consent form. The participant was \nscreened by an MRI operator once more before entering the scanner room. If the \nparticipant was deemed completely safe to go inside the scanner, we initiated \nSession 1. \nOnce in the scanning room, the participant chose suitable earbud sizes for \nnoise-attenuating headphones and donned a hairnet to prevent hair from getting into \nthe latches of the 30-channel coil. They then put the earbuds in and lay back on the \n6 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nscanner bed, putting the head into the coil. An additional pillow was put under the \nparticipant's legs for comfort and to minimise movements along the bed. The head \nwas fixated in the coil with an in-house developed helmet with inflated pillows \n('MR-MinMo', patent number GB 2205139.5 filed on 07 April 2022). Then the head of \nthe participant was localised and the first-surface mirror was placed on the coil. Once \nthe participant was inside the scanner the light was turned off inside the bore and in \nthe scanning room. Next, the short part of the movie was played to check if they \ncould see the picture and hear the sound clearly. The audio volume was adjusted for \neach participant separately. After that, the quick localizer was run and the field of \nview was adjusted to capture the whole brain. If it was not possible then the operator \nprioritised removing as few slices of the cerebellum as possible. When ready, the \npresentation script was started and eye-tracker calibration and validation procedures \nwere completed. The operator made adjustments to achieve the best validation \nquality for the eye-tracker and then proceeded to the movie-watching task. During \nSession 1, the participant watched the entirety of 'Back To The Future' \n(backtothefuture), divided into three parts (Zemeckis, 1985). Eye-tracker calibration \nwas performed before each part. If they asked to get out of the scanner during the \nbreak between movie parts, the operator assisted but only after encouraging them to \nstay inside the scanner to avoid head displacement. The entire process for Session \n1 took about three hours. \nIn Session 2, we followed the same procedures for preparing participants to \ngo inside the scanner and be scanned as in Session 1. The participant completed \nseveral different tasks inside the scanner during Session 2: somatotopic mapping, \nwhere the participant performed movements inside the scanner; retinotopic mapping, \nwhere different checkerboard patterns were presented while the participant was \ninstructed to fixate on the dot in the middle of the screen and respond every time the \ndot changed colour; and tonotopic mapping, during which a sequence of beeps was \nplayed and the participant was instructed to press a button when they noticed a \ndifference in tone (see ‘Tasks’ for further details). All tasks together took \napproximately 2.5 hours. \nTasks \nSome participants failed to come back for the second day of testing and some \ntasks were not included in the final dataset due to technical reasons. The number of \ntasks completed by each participant is shown in Table 1 below. \nsub id Questionnaires Cognitive tasks Movie-watching Somatotopy Retinotopy Tonotopy \nsub-01 + + + + + + \nsub-02 + + + + + + \nsub-03 + + + + + + \nsub-04 + - + - - - \nsub-05 + + + + + - \n7 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nsub-06 + + + + + - \nsub-07 + - + + + + \nsub-08 - - + - - - \nsub-09 + + + + + + \nsub-10 + + + + + + \nsub-11 + - + + + + \nsub-12 + + + + + + \nsub-13 + - + + + + \nsub-14 + + + + + + \nsub-16 + + + + + + \nsub-17 + + + + + + \nsub-18 + + + + + + \nsub-19 + + + + + + \nsub-20 + + + + + + \nsub-21 + + + + + + \nsub-22 + + + + + + \nsub-23 + + + + + + \nsub-24 + + + + + + \nsub-25 + + + + + + \nsub-26 + + + + + + \nsub-27 + + + + + + \nsub-29 + + + + + + \nsub-30 + + + + + + \nsub-31 + + + + + + \nsub-32 + + + + + + \nsub-33 + + + + + + \nsub-35 + + + + + + \nsub-36 + + + + + + \nsub-37 + + + + + + \nsub-38 + + + + + + \nsub-39 + + + + + + \nsub-40 + + + + + + \nsub-42 + + + + + + \nsub-43 + + + + + + \nsub-44 + + + + + + \nTotal N 39 35 40 38 38 36 \nTable 1. The number of tasks completed by each participant. The ‘+’ means that the \nparticipant fully completed the task, and the ‘-’ means that the data is missing or the \nparticipant didn’t complete the task. \n8 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nHome \nQuestionnaires \nThe entire questionnaire took approximately 30 minutes to complete. The first \nsection consists of questions regarding basic demographics, language proficiency, \nand background in music and movie preferences. The second section included 9 \nvalidated psychological questionnaires. The set of questionnaires was selected to \ncomprehensively assess participants’ mental health, well-being, inner experience, \nself-talk, mindfulness, and awareness. These constructs are fundamental to \nunderstanding cognitive and emotional processes, particularly in relation to individual \ndifferences in subjective experience. The selected measures are widely validated, \nreliable, and efficient, ensuring that they capture a broad spectrum of psychological \nfunctioning. These were implemented on the Qualtrics platform \n(https://qualtrics.ucl.ac.uk), and are described next. \nThe Patient Health Questionnaire (PHQ) is a 9-item tool for diagnosing \ndepression and various other mental health conditions frequently seen in primary \ncare settings (Kroenke et al., 2001) . Each item has a scale from 0 (not at all) to 3 \n(nearly every day). It is notably shorter than many other depression assessments, \nyet it maintains similar levels of sensitivity and specificity. \nThe 7-item scale for General Anxiety Disorder (GAD-7) is a valid and efficient \ninstrument to screen and evaluate the severity of the condition in both clinical and \nresearch settings (Spitzer et al., 2006) . Each item has a scale from 0 (not at all) to 3 \n(nearly every day). GAD is among the most frequently observed anxiety disorders in \nboth general medical practice and the broader population. \n The Warwick Edinburgh Mental Well-Being Scale (WEMWBS) is a widely \nused measure of mental well-being, comprising exclusively positively worded items \n(Tennant et al., 2007). The participants need to evaluate their mental well-being over \nthe last two weeks. The scale has 14 questions with a 5-point Likert scale (none of \nthe time, rarely, some of the time, often, all of the time). \n The Nevada Inner Experience Questionnaire (NIEQ) is used to assess the \nsubjective frequency at which people experience five common phenomenological \ncategories of inner thought (inner speaking, inner seeing, unsymbolized thinking, \nfeelings, and sensory awareness) via a visual analogue scale (Heavey et al., 2019). \nThe NIEQ has 10 items with two types of questions: 'How frequently…?' with a scale \nfrom 0 (never) to 100 (always) and 'Generally speaking, what portion…?' with a scale \nfrom 0 (none) to 100 (all). \n The Self Talk Scale (STS) is used to assess the subjective frequency at which \npeople engage in various modes of self-talk (Brinthaupt et al., 2009) . The modes of \nself-talk assessed by this scale, as delineated by a four-factor structure, relate to \n‘self-regulatory’ elements, including social assessments, self-criticism, \nself-reinforcement and self-management. Each question starts with 'I talk to myself \nwhen…' and the participant needs to evaluate the frequency on a 5-point scale (1 - \nnever, 5 - very often). \n9 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nThe Varieties of Inner Speech Questionnaire - Revised (VISQ-R) is used to \nassess participants’ subjective frequency and phenomenological characteristics of \ntheir experience of inner speech (Alderson-Day et al., 2018) . The characteristics \ncomprise a four-factor model, with factors comprising; dialogical inner speech, \ncondensed inner speech (as compared to spoken aloud), experience of other \npeople’s voices, and self-evaluative inner speech. It consists of 26 items and each \nitem is rated on a scale of 1 (never) to 7 (all the time). \nThe Five Facets of Mindfulness Questionnaire (FFMQ) is a 39-item instrument \nthat uses five polytomous response options to assess five different aspects of \nmindfulness: observing, describing, acting with awareness, non-judging, and \nnon-reactivity to inner experience (Baer et al., 2008). Each item is rated on a scale of \n1 (Never or very rarely true) to 5 (Very often or always true). \n The Multidimensional Assessment of Interoceptive Awareness version II \n(MAIA-II) evaluates eight factors of interoceptive body awareness (noticing, \nnot-distracting, not-worrying, attention regulation, emotional awareness, \nself-regulation, body listening and trusting) (Mehling et al., 2018) . It consists of 37 \nitems and each item is rated on a scale of 0 (never) to 5 (always). \n The White Bear Suppression Inventory (WBSI) is a 15-item questionnaire \nmeasuring thought suppression (Wegner & Zanakos, 1994) . Chronic thought \nsuppression is a variable that is related to obsessive thinking and negative affect \nassociated with depression and anxiety. Each item is rated on a 5-point scale from \nstrongly disagree (1) to strongly agree (5). \nCognitive tasks \nThe battery of cognitive tasks was designed to comprehensively assess a \nrange of cognitive abilities, including memory, executive function, attention, \nreasoning, and creativity. The selection of tasks reflects key domains of cognition \nrelevant to general intelligence, cognitive flexibility, and problem-solving, providing a \nrobust framework for evaluating individual differences in cognitive function. We used \n16 different tasks from the Cognitron platform. A detailed description of each task is \navailable in the original publication and supplementary material by the Cognitron \nteam (Del Giovane et al., 2023) . The participants completed all the tasks remotely. \nThe whole battery took around 45 minutes to complete. A description of each task is \nprovided next in the order presented to the participant. Figures illustrating the trial \nstructure of each task are available in the Supplementary Materials. \nObject Memory Immediate and Delayed . This task measures memory \ncapacity on short and long-term scales. Participants viewed a list of 20 \nblack-and-white objects (for example, stairs, table, ladle etc.), presented once at a \ntime, and were instructed to remember as many as they could. Immediately after \n(Immediate version of the task), participants’ short-term memory was assessed by \npresenting a grid containing one object from the previously presented list and 7 \nsimilar “distractor” objects. Participants had to identify and click the object they \nrecognised from the original list. In total 20 grids were presented (see \n10 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nSupplementary Fig. 1). This task was repeated at the end of the battery of cognitive \ntasks to assess long-term memory (Delayed version of the task). \nWord Memory Immediate and Delayed.  This task is similar to the Object \nMemory Immediate and Delayed but uses words instead of images (see \nSupplementary Fig. 11). \n2D manipulations . This task measures the ability to mentally rotate a grid \nwithin a two-dimensional space. Participants were presented with a target grid \npartially filled with coloured squares, as well as four comparison grids. One of these \ngrids was a rotated transformation of the target grid  (see Supplementary Fig. 2). \nParticipants were instructed to identify the rotated grid as quickly and accurately as \npossible. \nIntra/extra-dimensional set-shifting task (ID/ED). The ID/ED task is a \ncomputerized analogue of the Wisconsin Card Sorting Task  (Grant & Berg, 1948) , \ndesigned to assess cognitive flexibility. Participants were presented with four \nsquares, followed by two objects appearing in two randomly selected squares. They \nwere instructed to identify the underlying rule, which changed after a number of \ncorrect responses, by clicking on the object that matched the current rule (see \nSupplementary Fig. 3). There were two main types of rule changes: (1) an ID rule \nwhere the rule continues to rely on the same dimension (e.g., shape), but the \nspecific features change (e.g., from triangle vs. circle to square vs. star) and (2) an \nED rule where the rule changes to a different dimension altogether (e.g., from \nselecting based on shape to selecting based on line pattern). These shifts require \nincreasing levels of cognitive flexibility, with ED shifts being particularly challenging \nbecause they demand a shift of attention to an entirely new dimension. Performance \non ED trials is therefore considered a strong indicator of flexible thinking and \nattentional control. \nSpatial Span. This task assesses visuospatial working memory capacity. \nParticipants were presented with a 4-by-4 grid  in which a sequence of squares lit up \nand asked to repeat the presented sequence. The sequence began with two lit up \nsquares. Participants were required to repeat the presented sequence by clicking on \nlightened-up squares. After each response, the sequence increased by one square \n(see Supplementary Fig. 4). \nDigit Span. The task measures working memory number storage capacity. \nParticipants were presented with a sequence of digits and asked to recall them by \ntyping on a digital keyboard (see Supplementary Fig. 5). With each correct response, \nthe sequence increased by one digit. \nSwitching Stroop. This task is a modified version of the classical Stroop test \n(Stroop, 1935)  and incorporates a switching condition in addition to the classic \ninterference condition. On each trial, participants were presented with the cue words \n‘Text’ or ‘Ink’ accompanied by two coloured words, ‘RED’ and ‘BLUE’ and a central \ncoloured box - the 'Ink' (red or blue). Based on the given instruction (\"Text\" or \"Ink\"), \nparticipants were required to click on one of the two words. If the rule was 'Ink', \nparticipants selected the word printed in the same ink colour as the central box. If the \n11 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nrule was 'Text', they selected the word whose meaning matched the colour of the \ncentral box (see Supplementary Fig. 6). \nVerbal Reasoning.  This task measures participants’ ability to interpret and \nanalyse written material. Participants processed syntactically complex sentences. In \neach trial, a picture of a square and a circle was shown along with the sentence (for \nexample, 'the square is contained by a circle'). Participants needed to answer if the \nsentence was true or false (see Supplementary Fig. 7). \nBeads task. This task measures impulsivity. Participants were instructed to \nwork out which bead colour, out of two colours, is the most prevalent in the jar. \nParticipants could reveal one bead at a time by clicking the button 'Reveal a bead'. \nAt any point, they could choose to guess the dominant colour in the jar, based on the \nbeads revealed so far (see Supplementary Fig. 8). \nAlternative Use Task. This task measures creativity. Participants were asked \nto invent as many alternative uses for a common item as possible (for example, \nnewspaper). After each response, participants were asked if the idea came to their \nmind as an 'Aha' moment. The task was limited to two minutes or 20 alternatives. \nDivergent Association task. This task measures creativity and specifically \ndivergent thinking. Participants were asked to think of 10 words in four minutes that \nwere as different from each other as possible. \nVerbal analogies.  This task assesses verbal reasoning and the ability to \nunderstand and apply logical relationships between word pairs. Participants were \npresented with a statement where the relationship between two pairs of words must \nbe assessed as either ‘true’ or ‘false’. It followed the structure: ‘A is to B as C is to D’. \nParticipants must determine whether the relationship between A and B was indeed \nanalogous to the relationship between C and D (see Supplementary Fig. 9). \nWord Definitions. This task measures the size of the vocabulary and level of \nlanguage comprehension. Participants needed to choose the correct definition of the \nword out of four options (see Supplementary Fig. 10). \nSpotter (digit vigilance) . This task assesses the participant's vigilance. \nParticipants briefly observed a sequence of numbers, obscured by 'noisy' pixels. The \ntask required the participant to spot and click anywhere on the screen as soon as \nthey recognised a zero (‘0’). Occasionally, the participant was asked how motivated \nand tired they felt on a scale from one (not at all) to six (extremely). \nSession 1 \nParticipants watched the movie 'Back To The Future’ (Zemeckis, 1985). The \nduration of the movie was one hour 51 minutes and 14 seconds. They were told to \nremain still inside the scanner and enjoy the movie watching. \nThe movie file was cropped into three runs using the 'ffmpeg' package: \n \nffmpeg -ss 00:00:00 -i back_to_the_future.mp4 -c copy -t 00:33:48 \nback_to_the_future_cut1-34min.mp4 \n \n12 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nffmpeg -ss 00:33:36 -i back_to_the_future.mp4 -c copy -t 00:37:39 \nback_to_the_future_cut2-38min.mp4 \n \nffmpeg -ss 01:11:03 -i back_to_the_future.mp4 -c copy -t 01:00:00 \nback_to_the_future_cut3-40min.mp4 \n \nThe specific time for each cut was chosen to maintain a similar duration between \nruns and to keep the smooth transition between scenes. The 2nd and the 3rd runs \nincluded 12 seconds of the scene from the previous run to provide enough time for \nthe hemodynamic response function (HRF) and psychological functioning to \n(theoretically) recover to a state similar to that in the preceding run. At the beginning \nof each run, there were eight spare TRs received from the scanner to allow the HRF \nto stabilise. Thus, eight, 16 and 16 TRs were removed during the analysis of runs \n1-3, respectively such that the fMRI data matched the length of the full movie, \nwithout overlaps. The length of the resulting runs was 34 minutes, 38 minutes 3 \nseconds and 40 minutes 12 seconds for runs one, two, and three respectively (1360, \n1522 and 1608 TRs). \nThe cropped files maintain the original video size and quality, using all frames \nwith no cropping or other transformations: \n● Video (codec): H.264 (High) \n● Audio (codec, sampling rate, bitrate, channels): AAC (LC), 48.0 kHz, 339 \nkbps, 5.1 \n● Resolution (pixels): 720 x 576 \n● Aspect Ratio: 16:9 \n● Frame rate (fps): 25 \n The movie presentation was implemented using a script executed from \nMATLAB (9.13.0.204977 R2022b) using PsychToolBox (v. 3.0.18) on a Windows PC \n(Windows 11 Pro v22H2, 64-bit operating system) with a GStreamer of version 1.0. \nSession 2 \nSomatotopic mapping \nEach participant underwent a training session before going inside the \nscanner. The participant watched short videos depicting each movement, listened to \nthe instructions from the researcher and practised to perform the movement. The \nresearcher assessed the movement and if all of them were well performed the \nparticipant proceeded further. The participant did one more short training session \ninside the scanner to find comfortable positions for limbs to perform proper \nmovements. The researcher assessed the quality of movements through the camera \ninside the bore. After, the participant underwent two blocked-design runs. \nExperimental conditions consisted of eight movements including left hand, \nright hand, left foot, right foot, left part of the face, right part of the face, left tongue \nand right tongue. The pattern for each movement is outlined in the Table 2 below. \n13 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nExperimental condition (movement) The pattern of the movement \nLeft/right hand Clench and relax the fist \nLeft/right foot Flex and extend the toes \nLeft/right part of the face Pull the corner of the mouth down and \nto the side \nLeft/right tongue \nTouch the last molar with the tip of the \ntongue when the mouth and jaw are \nclosed \nRest No movement, remain still \nTable 2. Somatotopic mapping task. The experimental conditions and pattern for \neach movement. \nIn each condition, the participant maintained eye fixation on the screen where \nthe instructions were presented. The instructions consisted of one line of text with a \nspecific movement (for example, ‘right foot’). The participant kept doing the \nmovement until instructions were changed on the screen. \nThe metronome, with a 30bpm pace, played during each run in the \nbackground. The participants were instructed to keep the pace of each movement \nalong with the metronome clicks. \nThe sequence of runs was counter-balanced across participants. Run one \nand two lasted for seven minutes 47 seconds and seven minutes 49 seconds \nrespectively (excluding eight spare TRs in the beginning). The duration of each \nmovement varied between 15 and 22 seconds and was repeated three times per \nrun. The sequence of movements was pseudorandomised (exact sequences for \neach run are shown in the Supplementary Materials ). The experiment was \nimplemented in PsychoPy (v. 2023.2.3) using standard builder functionality. \nRetinotopic mapping \nTo map how the visual space is systematically represented across cortical \nareas (i.e., retinotopic organisation), we used a population receptive field (pRF) \nmapping task during fMRI scanning (Dumoulin & Wandell, 2008). The stimulus \nconsisted of a black-and-white contrast-reversing checkerboard (2Hz), embedded \nwithin a rotating wedge (20º angle) and expanding/contracting ring. Each run \nincluded six ring cycles (48 seconds each, logarithmic eccentricity scaling) and eight \nwedge cycles (36 seconds each, alternating clockwise/anticlockwise). Stimuli \ncovered a maximum eccentricity of 8.6 from fixation and updated position every one \nsecond/TR. \nBaseline periods (20 seconds) were inserted at the start, midpoint, and end, \nduring which participants fixated on a central dot against a mid-gray background. A \nsmall white fixation dot (0.2° visual angle radius) and a black radial grid were always \n14 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nvisible to support stable fixation. Each run lasted 348 seconds (five minutes 56 \nseconds), and participants completed three identical runs. \nTo maintain engagement, participants performed a simple detection task, \npressing a button whenever the fixation dot changed from white to black. Due to \ntechnical issues, behavioural responses from the first nine participants were not \nrecorded. Finally, eye movements were tracked using an Eyelink 1000 Plus (SR \nResearch, Ottawa, ON), with a 5-point custom calibration performed before each \nrun. The task was implemented in MATLAB using PsychToolBox (v. 3.0.18) on a \nWindows PC (Windows 11 Pro v22H2, 64-bit operating system) with a GStreamer of \nversion 1.0. \nTonotopic mapping \n The full description of the task can be found in the original study on tonotopic \nmapping of the auditory cortex (F. K. Dick et al., 2017) . Participants listened to \nfour-tone motifs and performed a one-back task on infrequent repeats. The full \nfrequency range of the tones (175-5286 Hz) was divided into ten spectrally delimited \nbands, with each band having a 6-semitone range. At any given point in time, tones \nwere selected from only one frequency band; after 10 motifs, the frequency range \nstepped up (in one run) or down (in the other) to the next band. Each run swept \nthrough the full frequency range four times (64 seconds per sweep). In this way, \neach frequency band occurred with consistent timing within a sweep; critically, then, \nvoxels that respond preferentially to that frequency range should also respond at a \nconsistent phase lag (F. Dick et al., 2012; F. K. Dick et al., 2017; Sereno et al., 1995). \nThe participants were instructed to press the button every time they heard the \nsame tone twice in a row (1-back task). The experiment was implemented in \nPsychoPy (v. 2023.2.3) using standard builder functionality. \nData acquisition \nFunctional and anatomical images were acquired on a 3.0T Siemens \nMAGNETOM Prisma with a 30-channel radio-frequency (RF) head coil (Siemens \nHealthcare, Erlangen, Germany) for both sessions and all tasks. \nMRI parameters: backtothefuture task \nWe used multiband echo-planar imaging (TR = 1500 ms, TE = 35.2 ms, 72 \ninterleaved slices, slice thickness 2.0 mm, voxel size 2 mm isotropic, \nanterior-posterior phase encoding direction (A >> P), field of view 212 mm, flip angle \n60 deg, echo spacing 0.56ms, bandwidth 2620 Hz/Px), with a 4x multiband \nacceleration factor. The first, second, and third run had 1360, 1522, and 1608 TRs \nrespectively (including eight spare TRs at the beginning of each run). The phase \nreverse encoding (P >> A) scan was acquired right after each run. \nA 5-min high-resolution T1-weighted MPRAGE anatomical MRI scan followed \nthe functional scans (TR = 2300 ms, TE = 2.98 ms, 208 sagittal slices, slice thickness \n15 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n1.0 mm, voxel size 1 mm isotropic, anterior-posterior phase encoding direction (A >> \nP), field of view 212 mm, flip angle 9 deg, echo spacing 7.1 ms, bandwidth 240 \nHz/Px). \nMRI parameters: somatotopic mapping \nAt the beginning of Session 2, a structural scan was acquired to help position \nthe field of view more accurately for upcoming tasks. A 2-min high-resolution \nT1-weighted MPRAGE anatomical MRI scan was acquired (TR = 1530 ms, TE = 2.98 \nms, 176 sagittal slices, slice thickness 1.0 mm, voxel size 1.0 mm isotropic, \nanterior-posterior phase encoding direction (A >> P), field of view 256 mm, flip angle \n9 deg, echo spacing 7.1 ms, bandwidth 240 Hz/Px). \nWe used multiband echo-planar imaging (TR = 1000 ms, TE = 30.0 ms, 44 \ninterleaved slices, slice thickness 2.0 mm, voxel size 2 mm isotropic, \nanterior-posterior phase encoding direction (A >> P), field of view 212 mm, flip angle \n62 deg, echo spacing 0.7 ms, bandwidth 1814 Hz/Px) with a 4x multiband \nacceleration factor. The first and the second run had 467 and 469 TRs respectively \n(including eight spare TRs at the beginning of each run). The phase reverse \nencoding (P >> A) scan was acquired at the end of the task. \nMRI parameters: retinotopic mapping \nWe used multiband echo-planar imaging (TR = 1000 ms, TE = 35.2 ms, 48 \ninterleaved slices, slice thickness 2.0 mm, voxel size 2 mm isotropic, \nanterior-posterior phase encoding direction (A >> P), field of view 212 mm, flip angle \n60 deg, echo spacing 0.56 ms, bandwidth 2620 Hz/Px) with a 4x multiband \nacceleration factor. All three runs had the same number of 356 TRs (including eight \nspare TRs at the beginning of each run). The phase reverse encoding (P >> A) scan \nwas acquired at the end of the task. \nMRI parameters: tonotopic mapping \nWe used multiband echo-planar imaging with the same parameters as for the \nsomatotopic mapping task (see above). Both runs had the same number of 264 TRs \n(including eight spare TRs at the beginning of each run). The phase reverse \nencoding (P >> A) scan was acquired at the end of the task. \nEye-tracker apparatus \nEye tracker data was acquired using an MRI-compatible EyeLink 1000 Plus \nlong-range mount (SR Research Ltd., Mississauga, Ontario, Canada). The optic \ncamera head (f=50mm/F1.4 lense) and illuminator (FL-890) were positioned \nhorizontally behind the screen inside the bore on a tray. A front silvered mirror was \nplaced on the anterior MRI coil. The binocular setup was used with a sampling \nfrequency of 1000 Hz and a 9-dot calibration procedure covering the whole screen \n(5-dot calibration for retinotopy task). For accuracy validation, participants had to \n16 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nfixate on the same dots as during calibration (EyeLink software v. 5.15). The \ncalibration and validation procedures were repeated until the best possible accuracy \nwas achieved. \nThe setup inside the bore was the same between days. Stimuli were \npresented in full-screen mode through a mirror-reversing LCD projector (EPSON \nLB-1100U) to a rear-projection screen, with participants viewing through the front \nsilvered mirror attached to the head coil. Participants were positioned 57.5 cm from \nthe screen, which was viewed via a mirror attached to the head coil, and measured \n35.5 cm in width and 26 cm in height. Stimuli were presented in their native \nresolution and subtended 28.9° × 18.3° (29.6x18.5cm) of visual angle. The \neye-tracker camera and illuminator were located inside the bore behind the screen \non a movable platform. The position of the screen, projector, eye-tracker platform, \nand first-surface mirror was controlled between participants and checked to be the \nsame. \nThe eye tracker data was acquired during backtothefuture and retinotopy \ntasks. The MATLAB script started with the calibration and validation of the \neye-tracker, and if the calibration was deemed sufficient, the main experiment began \nafter 8 spare TRs were received from the scanner. The script sent messages (‘MSG’ \nin ASCII eye link data file) indicating the beginning of the event. In the \nbacktothefuture task, at the beginning of the movie part, each pulse and frame was \nsent to the ASCII data file (see Data Records for more details). In both tasks, the \npresentation script was implemented in MATLAB (9.13.0.204977 R2022b) using \nPsychToolBox (v. 3.0.18) on a Windows PC (Windows 11 Pro v22H2, 64-bit \noperating system) with a GStreamer of version 1.0. \nPhysiological recordings \n The pulse oximetry data were acquired using a Siemens wireless peripheral \npulse unit (PPU; Siemens Healthcare GmbH, Erlangen, Germany). The PPU sensor \nwas attached to the index finger of the left index hand. Participants were instructed \nto avoid any left hand movements (including fingers) to prevent movement artefacts. \nThe pulse data was acquired during backtothefuture, retinotopy and tonotopy tasks. \nPreprocessing \nThe raw DICOM data was transformed to a Brain Imaging Data Structure \n(BIDS) valid format using the heudiconv tool (https://github.com/nipy/heudiconv ; v. \n1.3.0) (Gorgolewski et al., 2016) . All anatomical images were defaced using the \npydeface tool (https://github.com/poldracklab/pydeface ; v. 2.0.2). MRI data were \npreprocessed using the AFNI software suite (AFNI_23.0.03 'Commodus') to ensure \ndata quality and prepare it for statistical analysis (Cox, 1996; Cox & Hyde, 1997). \nThe backtothefuture and somatotopy tasks were analysed using the same \npreprocessing pipeline and the tonotopy and retinotopy tasks were preprocessed \n17 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\ndifferently according to the pipeline described in the previous relevant papers \n(Chow-Wing-Bom et al., 2025; Dekker et al., 2019; F. K. Dick et al., 2017). \nAnatomical \nThe anatomical T1-weighted image was skull-stripped and then nonlinearly \naligned to the MNI152 2009 template, which generated a standard-space anatomical \nimage, a nonlinear warp and an affine transformation matrix (using AFNI SSwarper \ncommand). The anatomical surfaces were reconstructed using Freesurfer software \n(recon-all with default parameters, version 7.3.2-20220804-6354275, \nhttp://www.freesurfer.net) (Destrieux et al., 2010; Fischl, 2012) . The resulting \nsurfaces were converted to AFNI friendly format (using Surface Mapper (SUMA) \ntool) and later used to create white matter and ventricle regions of interest to use \nthem as nuisance regressors during preprocessing. These regions were eroded and \nused as noise regressors in the preprocessing of backtothefuture and somatotopy \ntasks. \nFunctional \nFunctional data underwent multiple preprocessing steps to ensure alignment \nand reduce artifacts. First, the initial volumes of each functional run were removed to \neliminate pre-steady-state effects. The first eight TRs were removed from run 1, and \nthe first sixteen TRs were removed from runs 2 and 3 (backtothefuture task). In other \ntasks, the first eight TRs were removed from each run. To correct for geometric \ndistortions caused by susceptibility-induced field inhomogeneities, a \nblip-up/blip-down correction was applied using a reverse-phase encoding field map. \nThe median images of the forward and reverse phase-encoding runs were extracted, \nand their midpoint warps were computed and applied to the functional runs using \n‘3dNwarpApply’. \nMotion correction was performed using a two-pass alignment strategy \n(‘3dvolreg’). First, volumes within each run were aligned to the run-specific \nreference; then, these within-run bases were themselves aligned to a common \nreference volume. This hierarchical strategy allowed all runs to be brought into a \nshared alignment space. To further improve robustness, alignment was performed \nusing a two-pass procedure: an initial low-resolution stage estimated gross head \nmotion, which was then refined at full resolution. All further steps described below \nwere applied only to backtothefuture and somatotopy tasks. The specifics of \nprocessing for the retinotopy task are described in the section below (Retinotopic \nmapping). \nEach volume was aligned to a reference volume determined by the minimum \noutlier fraction across all runs. The aligned functional images were then aligned to \nthe anatomical image (‘align_epi_anat.py ’). Finally, the functional data were \nnonlinearly aligned to MNI standard space. \nA whole-brain mask was generated by computing the union of individual \nrun-based masks. The functional images were then smoothed using a 4 mm \n18 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nfull-width half-maximum (FWHM) Gaussian kernel to improve signal-to-noise ratio \n(‘3dBlurToFWHM’). Next, the data were temporally band-pass filtered (0.01-1 Hz) to \nreduce low-frequency drifts and high-frequency noise. \nBecause the runs of the backtothefuture task were very long the baseline \npolynomial degree was fixed to two. For all other tasks, the degree was computed \nautomatically using AFNI algorithms based on the length of the run. \nRetinotopic mapping \nTo enable surface-based analysis, each participant’s functional data  were first \naligned to their high-resolution anatomical image using the following approach. As \ndescribed previously, the functional volumes used for retinotopic mapping were \nmotion-corrected using a two-pass alignment strategy (3dvolreg), which included \nwithin-run and across-run registration to a common reference volume. Importantly, \nthis motion correction was performed prior to any additional preprocessing (e.g., \nspatial blurring or nuisance regression), so the volumes used here reflect unblurred, \nmotion-corrected data. \nEach participant’s motion-corrected volume was registered to their \nhigh-resolution anatomical image using boundary-based registration (bbregister, \nFreeSurfer). Registration quality was visually inspected and quantified with the cost \nfunction value. All participants with a minimum cost function value above 0.451 were \nre-registered to a single-band reference volume from run one. This applied to \nparticipants 01, 07, 20, 21, 25 and 44. \nFollowing successful registration, volumetric data for each run were then \nresampled onto the surface of both hemispheres using cortical surface models \nreconstructed from each participant’s high-resolution anatomical scan (mri_vol2surf \ncommand from Freesurfer). Sampling was performed at 50% cortical depth, using a \n3 mm FWHM surface-based smoothing kernel to enhance signal-to-noise ratio while \npreserving spatial specificity. This step generated a surface-based time series for \neach hemisphere and run, enabling subsequent retinotopic analysis to be conducted \nin native cortical surface space. \nThe resulting outputs were then converted into a MATLAB-compatible format \nfor further analysis using the samsrf_mgh2srf command from the SamSrf toolbox \n(version 7.13; https://github.com/samsrf). pRF estimates were computed by fitting a \n2D Gaussian model to the data. As a result, three pRF parameters for each vertex \non the cortical surface were computed: the x- and y-coordinates of the pRF center \nand the pRF size (σ). Eccentricity and polar angle maps were computed from the x- \nand y-coordinates. \nTo model population receptive fields, the symmetric bivariate Gaussian model \nwas used, with mean (x, y) representing the preferred retinotopic location, and \nstandard deviation (σ) representing pRF size. To identify the pRF model parameters \n(x, y, σ) that best predict the measured time series, a two-stage fitting procedure was \nemployed. In a coarse fitting step, data were smoothed along the cortical surface \n1 https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg38136.html \n19 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n(Gaussian FWHM kernel of 5 mm) and a grid-search approach was used to identify \nmodel parameters that maximize the Pearson correlation between observed data \nand the pRF model’s predicted time course. Vertices with R 2 > 0.05 were entered as \nstarting value in a fine-fitting step, which used MATLAB’s fminsearch function to \nidentify parameters that minimised the squared residual deviations between the \nmodel and unsmoothed data. Finally, X and Y position estimates were converted to \neccentricity (distance from fixation) and polar angle. After that, each map was \nvisually inspected for quality control before group-averaging (see Technical \nValidation). \nAfter estimating pRF parameters on each participant’s native cortical surface, \nthe resulting maps were projected to a common reference space to enable \ngroup-level analysis. This was achieved by resampling each individual’s maps from \ntheir native cortical surface onto the fsaverage template surface provided by \nFreeSurfer (Native2TemplateMap function in SamSrf v7.13). The resampled maps \nwere then averaged vertex-wise across participants on the fsaverage mesh, resulting \nin group-level retinotopic maps for each parameter: eccentricity, polar angle, and \npRF size. \nData Records \nInformation and anatomical data that could be used to identify participants \nhave been removed from all records. The resulting dataset is available on the \nOpenNeuro.org platform (https://openneuro.org/datasets/ds006642 ). The scripts \nused for data processing are available on the GitHub repository \n(https://github.com/levchenkoegor/movieproject2). \nQuestionnaires \nLocation sourcedata/Questionnaire_participants_values.csv \nFile format comma-separated value (CSV) \n \nParticipants’ responses to a battery of questionnaires (see Questionnaires \nunder Tasks section) in a comma-separated value (CSV) file. Data is structured as \none line per participant with all questions and test items as columns. \nCognitive tasks \nLocation sourcedata/Cognitron_assessment_participants_cleaned.tsv \nFile format Tab-separated values (TSV) \n \nParticipants’ responses to a battery of cognitive tasks (see Cognitive Tasks \nunder Tasks section) in a TSV file. The file contains the sub-id identifier column and \n20 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\ndata with all responses organised in a JSON format. The table is structured as one \nline per task (16 rows per participant). \nAnatomical MRI \nLocation sub-<ID>/ses-<SES_ID>/anat/sub-<ID>_ses-<SES_ID>_T1w.nii.gz \nSession 001, 002 \nFile format NIfTI, gzip-compressed \nSequence protocol \nsub-<ID>/ses-00<SES_ID>/anat/sub-<ID>_ses-<SES_ID>_T1w.json \n \nThe anatomical scan was acquired at the end of the first session (5-min \nMPRAGE) and at the beginning of Session 2 (2-min MPRAGE). Both raw anatomical \nimages were defaced. They are available as a 3D image file, stored as \nsub-<ID>_ses-<SES_ID>_T1w.nii.gz. The sequence protocol file accompanies \nanatomical images in the same folder, stored in JavaScript Object Notation (JSON) \nfile format. \nFunctional MRI \nLocation \nsub-<ID>/ses-<SES_ID>/func/sub-<ID>_ses-<SES_ID>_task-<TASK_NAME>_run-<\nRUN_ID>_bold.nii.gz \nSession 001, 002 \nTask-name backtothefuture, somatotopy, retinotopy, tonotopy \nRun 001, 002, 003 \nFile format NIfTI, gzip-compressed \nSequence protocol \nsub-<ID>/ses-<SES_ID>/func/sub-<ID>_ses-<SES_ID>_task-<TASK_NAME>_run-<\nRUN_ID>_bold.json \n \nRaw fMRI data is available as individual time series files per each task run. \n \nLocation \nsub-<ID>/ses-<SES_ID>/fmap/sub-<ID>_ses-<SES_ID>_acq-<TASK_NAME>_dir-P\nA_run-<RUN_ID>_epi.nii.gz \nSession 001, 002 \nTask-name func, somatotopy, retinotopy, tonotopy \nRun 001, 002, 003 \nFile format NIfTI, gzip-compressed \nSequence protocol \nsub-<ID>/ses-<SES_ID>/fmap/sub-<ID>_ses-<SES_ID>_acq-<TASK_NAME>_dir-P\nA_run-<RUN_ID>_epi.nii.json \n \n21 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n The phase reverse encoding (P >> A) scan was acquired for each task and is \navailable as an individual time series. The func task-name refers to the \nmovie-watching task (backtothefuture  in all other files). The sequence protocol file \naccompanies functional images in the same folder, stored in a JSON file. \nPhysiological recordings \nLocation \nsourcedata/sub-<ID>/ses-<SES_ID>/func/sub-<ID>_ses-<SES_ID>_run-<RUN_ID>\n_task-<TASK_NAME>_physioPULS.tsv \nSession 001, 002 \nTask-name backtothefuture, retinotopy, tonotopy \nRun 001, 002, 003 \nFile format TSV \nInformation file \nsub-<ID>_ses-<SES_ID>_run-<RUN_ID>_task-<TASK_NAME>_physioInfo.tsv \n \nRaw pulse data (acquired with a pulse oximetry sensor) is available as \nindividual time series files in TSV format. \nEye tracker recordings \nLocation \nsourcedata/sub-<ID>/ses-<SES_ID>/func/sub-<SUB_ID>_ses-<SES_ID>_run-<RU\nN_ID>_task-<TASK_NAME>_eyelinkraw.<FORMAT>.gz \nSession 001, 002 \nTask-name backtothefuture, retinotopy \nRun 001, 002, 003 \nFile format EDF or ASC, gzip-compressed \n \nRaw eye-tracker data is available in two formats: EDF and ASC. Raw \neye-tracker EDF files were converted to ASCII files using the edf2asc tool \n(EDF2ASC version 4.2.1197.0) from the Eyelink Developers Kit. The ASCII files \ncontain messages (‘MSG’) flags to align the EPI and gaze positions data. \n‘MOVIE_START’ indicates the beginning of the movie, ‘FRAMENUMBER_*” \nindicates the number of frame showed on the screen and “PULSE_*” indicates the \nnumber of pulse received from the scanner. The snippet of the ASCII file with \nmessages is provided below: \n \n… \n3976285   980.1   600.7   372.0  1052.0   582.5   508.0 32768.0 ..... \n3976286   980.0   601.7   372.0  1053.3   582.5   509.0 32768.0 ..... \nMSG 3976287 MOVIE_START \n3976287   980.0   601.7   370.0  1053.6   577.6   510.0 32768.0 ..... \n22 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n3976288   984.1   600.8   370.0  1053.6   577.6   513.0 32768.0 ..... \n… \n3976301   981.8   599.4   373.0  1056.6   578.3   514.0 32768.0 ..... \n3976302   981.8   599.4   364.0  1058.3   581.9   516.0 32768.0 ..... \nMSG 3976303 FRAMENUMBER_1 \nMSG 3976303 PULSE_9 \n3976303   985.1   596.7   364.0  1060.9   584.1   517.0 32768.0 ..... \n3976304   983.7   595.9   368.0  1061.3   584.1   516.0 32768.0 ..... \n… \n \nIn the current snippet, there are events in the ‘MSG’ lines. In a similar way, the \nretinotopic mapping eye tracker ASCII files have ‘MSG’ for each volume. \n‘VOLUME_*” indicates the number of pulse received from the scanner. \n \n… \n3623949   867.7   757.3   637.0   882.7   826.4   574.0 32768.0 \n3623950   863.9   756.0   637.0   882.5   826.4   573.0 32768.0 \nMSG 3623951 VOLUME 1 \n3623951   862.3   755.3   636.0   881.8   823.3   572.0 32768.0 \n3623952   862.3   755.3   636.0   881.8   823.3   572.0 32768.0 \n… \n \nIn both cases, the messages (‘MSG’) received from the MATLAB script make it \npossible to align the eye-tracker and EPI data precisely.  \nFreesurfer outputs \nLocation derivatives/freesurfer/sub-<ID> \n \nThe standard Freesurfer outputs after running the recon-all command for each \nparticipant separately. The fsaverage folder is stored along with participants’ files \ntoo. In each folder there is a SUMA folder with all the outputs from the \nSUMA_Make_Spec_FS command. The pRF maps are stored under the retinotopy \nfolder. \nSSwarper outputs \nLocation derivatives/sub-<ID>/SSwarper \n \nThe standard SSwarper outputs after running the SSwarper AFNI command \nfor each participant separately. This procedure both skull-strips the anatomical \nvolume and computes the nonlinear warp to standard space. \n23 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nPreprocessed functional MRI \nLocation \nderivatives/sub-<ID>/<TASK_NAME>/sub-<ID>_task-<TASK_NAME>_run-<RUN_ID\n>_preproc.nii.gz \nTask-name backtothefuture, somatotopy, tonotopy, retinotopy \nRun 001, 002, 003 \n \nThe preprocessed fMRI data after running the AFNI processing script \n(afni_proc.py). The data is available as individual time series files per each task run. \nThe preprocessing script for each task can be found on GitHub repository. \nTechnical Validation \nEvaluation of head motion control \nThe head motion was evaluated using the framewise displacement (FD) \nmetric, which measures instantaneous head motion by comparing the motion \nbetween the current and previous volumes. The FD values were calculated using \nstandard AFNI functionality (3dvolreg function). Figure 4 shows the overall \ndistribution of FD values for each task and run or condition. The distributions (first \ncolumn) are highly skewed toward lower FD values, with the majority concentrated \nbelow 0.2 mm. A dashed red line marks the 95th percentile of the FD values at 0.2, \n0.24, 0.2 and 0.25 mm for backtothefuture, somatotopy, retinotopy and tonotopy \ntasks respectively, indicating that 95% of the data exhibits minimal motion. Violin \nplots (second column) evaluate FD on a run or condition-specific level. Across all \nruns and conditions, FD distributions remain consistent, with median values below \n0.2 mm and most of the values below 0.3 mm (for all tasks). These results \nunderscore that head motion was well-controlled throughout the study, minimising \npotential motion-related artefacts in the data. \n24 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n \nFigure 2. Framewise displacement distributions across all participants for each task and run. \nThe figure shows the distributions for each task separately by row: backtothefuture, \n25 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nsomatotopy, retinotopy, and tonotopy. The first column depicts the distribution of FD across \nall participants and TRs. The second column shows violin plots and overlaid boxplots display \nthe distribution of FD for each run or condition across all participants. Each violin plot \nrepresents the density of FD values, with the boxplot inside showing the interquartile range \n(IQR), median, and whiskers (minimum and maximum values within 1.5×IQR). \nTable 3 shows descriptive statistics for head motion parameters across each \ntask and run. It reports the mean, median, and maximum values for three rotational \n(Roll, Pitch, Yaw; in degrees) and three translational components (Superior-Inferior \n(dS), Left-Right (dL), Posterior-Anterior (dP); in millimetres). Overall, the mean \ndisplacements for both rotational and translational motion remained within ±0.1 \ndegrees and millimetres, respectively. Mean maximum values for all motion \nparameters across all tasks were below 0.5, with the exception of the \nbacktothefuture task, where maximum displacements in pitch and dS approached \n1.0 degree/millimetre across all runs. Despite minor variations between tasks and \nruns, overall motion levels were low, suggesting minimal motion-related artefacts in \nsubsequent analyses. \n  Mean Mean maximum \nTask Run Roll Pitch Yaw dS dL dP Roll Pitch Yaw dS dL dP \nbacktothef\nuture \n1 -0.01 -0.01 0.02 -0.02 -0.08 -0.03 0.25 0.63 0.36 0.6 0.19 0.45 \n2 -0.05 0.09 0.04 -0.05 -0.03 -0.03 0.28 0.92 0.44 0.7 0.25 0.48 \n3 -0.02 0 -0.03 0.02 -0.06 -0.04 0.3 0.84 0.45 0.94 0.28 0.44 \nsomatotopy \n1 -0.02 0.01 -0.01 0.02 -0.01 -0.04 0.17 0.46 0.22 0.32 0.17 0.36 \n2 -0.01 0 0.02 0 -0.02 -0.02 0.18 0.45 0.22 0.38 0.16 0.4 \nretinotopy \n1 -0.01 0.1 -0.01 -0.03 -0.01 -0.08 0.14 0.48 0.17 0.21 0.12 0.28 \n2 0 0.09 0.02 0.03 0.01 -0.08 0.11 0.43 0.17 0.34 0.12 0.23 \n3 -0.02 0.08 0.02 0.04 -0.02 -0.07 0.09 0.48 0.18 0.34 0.1 0.23 \ntonotopy \n1 -0.02 0.09 -0.01 0 0 -0.02 0.12 0.43 0.14 0.26 0.1 0.23 \n2 0.01 0 0.02 0.04 0.02 -0.02 0.15 0.36 0.21 0.44 0.16 0.3 \nTable 3. Descriptive statistics for head motion parameters across each task and run. Each \ncell represents the mean, median, or maximum value of framewise displacement per run. \nTasks include backtothefuture, somatotopy, retinotopy, and tonotopy, with up to three runs \nper task. \nInter-subject correlation during movie-watching \nThe inter-subject correlation (ISC) analysis was undertaken  on all participants \nto show the timing alignment and the quality of functional MRI data during \nbacktothefuture task. ISC is a data-driven method that quantifies shared temporal \n26 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nfluctuations in brain activity across participants (Nastase et al., 2019) . It is commonly \nused to assess synchronization of neural responses during naturalistic paradigms. \nSix participants were excluded from this analysis (see Usage Notes for details). \nThe ISC values were calculated using AFNI tools (3dTcorrelate  and 3dISC). \nPairwise ISCs were computed for every unique pair of participants using Pearson \ncorrelation and Fisher z-transformation. Once all pairwise ISC maps were generated, \na group-level statistical model was estimated using mixed-effects modelling that \ntreated both participants in each pair as random intercepts. To assess statistical \nsignificance, t-statistics were subsequently corrected for multiple comparisons using \nthe False Discovery Rate (FDR) procedure at a threshold of q < 0.001. The ISC \nvalues are shown in Figure 3 below. \n \nFigure 3. Inter-subject correlation map. Voxelwise inter-subject correlation (ISC) during Back \nto the Future viewing, computed across 34 participants after excluding six with alignment \nissues (see Usage Notes). Statistical significance was assessed with voxelwise t-tests and \ncontrolled for multiple comparisons using the False Discovery Rate (FDR) procedure at q < \n0.001 (two-tailed). No additional cluster-size threshold was applied. \nHigh ISC values across early sensory cortices further indicate accurate \ntemporal alignment across participants. This suggests that both stimulus \npresentation and preprocessing preserved the temporal structure of the data, \nensuring that stimulus-locked neural responses were synchronized across \nindividuals. The resulting ISC maps show robust inter-subject synchronization across \nwidespread cortical regions, including primary auditory and visual cortices, as well as \nposterior medial areas (precuneus, posterior cingulate). These regions are \nconsistent with previous findings in the literature and reflect reliable, stimulus-locked \nprocessing during naturalistic movie-watching tasks (Aliko et al., 2020; Hasson et al., \n2004; Lerner et al., 2011). \n27 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nEye-tracker quality control \nEach participant underwent a standard 9-point calibration and validation \nprocedure prior to each run of the backtothefuture task. To quantify calibration \nquality, the average and maximum errors for each eye and each run for all \nparticipants were extracted from corresponding eyelink asc files. The table with \naverage and maximum errors per participant, run, and eye is available in the \nSupplementary Materials. \nFollowing SR Research guidelines, calibration accuracy was classified as \nGOOD when the average error was below 1.0° and the maximum error was below \n1.5°, FAIR when the average error ranged between 1.0° and 1.5° or the maximum \nerror between 1.5° and 2.0°, and POOR when the average error exceeded 1.5° or \nthe maximum error exceeded 2.0° \n(https://www.sr-research.com/support/showthread.php?tid=244). Across all \ncalibration entries (N = 234), the mean average error was 0.46° (SD = 0.19°), and \nthe mean maximum error was 1.03° (SD = 0.82°). Based on the classification criteria \nfrom the guidelines, 212 calibrations (90.6%) were labeled as GOOD, 13 (5.6%) as \nFAIR, and 9 (3.8%) as POOR. \nSince we used a binocular setup, we adopted the following approach for \nrun-level quality classification: if at least one eye showed GOOD calibration, the run \nwas considered GOOD. Under this criterion, only 3 runs were labeled POOR and 2 \nas FAIR, while all remaining runs met the GOOD threshold for at least one eye (and \nin most cases, for both eyes). These results indicate a high level of calibration \naccuracy and confirm the overall quality of the eye-tracking data. \nSomatotopic mapping \nTo assess the quality and specificity of the somatotopic mapping, we first \nestimated participant-level beta coefficients using a general linear model \n(implemented in AFNI's 3dDeconvolve function with dmUBLOCK(1) parameter), \nmodeling each body movement separately. Individual results were visually inspected \nto ensure expected activation patterns and data quality. We examined group-level \nactivation maps for three contrasts between different body part movements: Face vs \nFeet, Face vs Hand, and Hand vs Tongue. For each participant, condition contrasts \n(Face vs Feet, Face vs Hand, and Hand vs Tongue) were computed within \n3dDeconvolve, resulting in subject-level contrast beta maps. At the group level, \nthese contrast betas were then entered into a one-sample t-test to assess \nconsistency across participants. As shown in Figure 4, the three contrasts show the \nactivation maps aligned with the canonical somatotopic organisation along the \nprecentral and postcentral gyri. Specifically, face-related activity was localized to the \ninferior-lateral portion of the sensorimotor cortex. Tongue-related activations were \nfound ventrally, with strong effects in lateral sensorimotor regions and inferior \npre/postcentral areas. In contrast, feet-related activity was located dorsomedially, \nwhile hand activations occupied intermediate dorsolateral regions of the central \n28 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nsulcus. The results are consistent with the known motor and somatosensory \nhomunculus, indicating both the spatial specificity and sensitivity of the task and data \nacquisition. \n \nFigure 4. Somatotopic mapping. Group-level statistical activation maps are displayed for \nthree pairwise contrasts: Face vs Feet, Face vs Hand, and Hand vs Tongue, each rendered \non a standard MNI space in axial, coronal and sagittal planes. Colors represent voxel-wise \nt-values, thresholded at t > 5, corresponding to p < 0.00001. Red to white colour indicates \nregions showing significantly greater activation for the first body part in each contrast, while \nblue to dark blue colour indicates greater activation for the second body part. \nRetinotopic mapping \nTo evaluate the quality of the retinotopic mapping, we first estimated pRF \nparameters for each participant separately. Each participant’s eccentricity and polar \nangle maps were visually inspected for anatomical consistency, smoothness of \ntopographic gradients, and the presence of expected retinotopic features. This \nevaluation was qualitative, focusing on whether the maps displayed coherent and \ninterpretable organization across early visual areas. \nParticipants 07, 17, and 29 were excluded from the group-level analysis due \nto poor retinotopic map quality. Although data were collected for these individuals, \n29 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\ntheir maps lacked coherent topographic organization, showing either excessive noise \nor fragmented patterns without the expected polar angle reversals or eccentricity \ngradients. As a result, these maps could not be reliably interpreted and were omitted \nfrom the mean maps across all participants. \n To show the quality of the retinotopy task, the maps were averaged across all \nthe remaining participants to get a grand average map (Fig. 5). Polar angle shows \nthe well-known pattern along the calcarine sulcus and around the occipital cortex. In \nboth hemispheres, polar angle values smoothly transition from the horizontal \nmeridians (coded in green) to upper and lower vertical representations (coded in \nblue and red, respectively) , consistent with known topography in areas V1, V2, and \nV3. Eccentricity maps show a clear central-to-peripheral gradient, with foveal \nrepresentations located at the occipital pole (coded in blue) and peripheral regions \nexhibited more anteriorly (green). The resulting maps show well-organized and \nexpected representations of both polar angle and eccentricity on inflated cortical \nsurfaces of each hemisphere. \n \nFigure 5. Retinotopic mapping. pRF results for the left and right hemispheres displayed on \ninflated cortical surface reconstructions. Columns from left to right show: cortical curvature \n(dark gray, sulci; light gray, gyri) with anatomical landmarks labeled (IPS, intraparietal sulcus; \nPOS, parieto-occipital sulcus; LOS, lateral occipital sulcus; Calc, calcarine sulcus; OTS, \noccipito-temporal sulcus; CoS, collateral sulcus); polar angle maps indicating visual field \nrepresentation (color wheel inset); eccentricity maps showing distance from fixation (scale \ninset, center 0°, periphery 7°); estimated pRF size (in degrees of visual angle, color scale \ninset); and variance explained by the pRF model (in %, color scale inset). Black lines \ndelineate visual area boundaries (V1–V3). Rows correspond to the left hemisphere (top) and \nright hemisphere (bottom). \n \n30 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nUsage notes \nThe NNDb-3T+ dataset provides a uniquely rich multimodal resource for \nstudying the brain under naturalistic conditions and individual differences across \nmultiple domains (somatotopy, retinotopy and tonotopy). The dataset combines \nhigh-quality fMRI data from a full-length movie-watching paradigm and somatotopy, \nretinotopy, and tonotopy tasks, eye-tracking data, physiological recordings, and \nextensive behavioral measures. \nPractical considerations \nSeveral participants were excluded from inter-subject correlation analyses \ndue to interruptions or inconsistencies during the movie task. These participants \nshould be carefully considered for any time alignment-sensitive analysis and \npotentially excluded or adjusted. For many analyses, these participants could still be \nused. \nsub-01: scanning stopped near the end of run 2 which means that the last 152 TRs \nof the run were not acquired. Run 3 was collected with no issues. \nsub-02: incorrect lengths of runs. The movie files were cut incorrectly: run 2 is \nmissing the final 12 seconds, while run 3 includes those 12 seconds again (overlap \nwith run 2). The perfect alignment can’t be guaranteed. \nsub-03: scan stopped 55 seconds (37 TRs) before the end of run 2. Run 3 was \ncollected with no issues. \nsub-05: the anatomical MRI is missing due to technical reasons. \nsub-10: movie stopped ~15 minutes (511 TRs) before the end of run 3 due to \nparticipant scheduling conflict. \nsub-24 and sub-36: paused during run 1 using the implemented pause/resume \nfunctionality (see below). Run 1 is not recommended for time-sensitive analysis; run \n2 and 3 were collected with no issues. \nThe presentation MATLAB script for backtothefuture task included a pause \nfunctionality for situations in which participants squeezed the alert ball, indicating a \nneed for a break during the run, or if the movie needed to be stopped for any reason. \nIf the movie was paused (by pressing ‘P’ on the keyboard), it froze on the last frame \ndisplayed and waited for the ‘R’ key to be pressed to resume. When ‘R’ was pressed, \nthe movie rewound by 12 seconds and paused for eight spare TRs from the scanner. \nThe operator adjusted the protocol to ensure the correct number of TRs for the \nremainder of the movie and resumed scanning when the participant was ready. This \nfunctionality was used with sub-24 and sub-36, both of whom paused during the first \npart of the movie. In the first case, the participant initially needed a bathroom break \nbut decided to finish the first run and use the bathroom during the break between \nruns. In the second case, the participant appeared sleepy, prompting the operator to \nstop the scanning and allow a short break inside the scanner. After five minutes, the \n31 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nscanning resumed. In both cases, the pause resulted in two files for the first run, \nlabelled with the prefixes ‘beforepause’ and ‘afterpause’ to distinguish them. \nThe database includes both raw and preprocessed MRI data, with \npreprocessing pipelines designed to follow best practices for functional neuroimaging \nusing AFNI, Freesurfer, MATLAB and Python. These pipelines are tailored to the \nspecific structure and goals of each task. However, we emphasize that no single \npreprocessing strategy is optimal for all hypotheses or analytical approaches. \nDepending on the research question, users may wish to modify aspects of the \npipeline such as spatial smoothing, alignment methods, filtering, nuisance \nregression, or motion censoring thresholds. We encourage users to consult the \naccompanying GitHub repository (see Code availability ), which provides all \npreprocessing scripts, quality control reports, and logs. Reviewing these materials \ncan help ensure reproducibility, clarify processing decisions, and guide adaptations \nfor custom workflows. \nSeveral components of the NNDb-3T+ dataset are not included in the current \nrelease but will be made available in future versions. Individual-level tonotopy maps \nare currently undergoing preprocessing and validation. These maps will allow \nresearchers to investigate auditory cortical organization and model tonotopic \ngradients across participants. \nLimitations \nDue to copyright restrictions, the full-length movie stimulus used in the \nbacktothefuture task cannot be included in the dataset. However, the copy of the \nmovie can be purchased with their unique Amazon Standard Identification Number \n(ASIN: B000BVK82I) or International/European Article Number (EAN: \n5050582401288). Researchers can perform or apply timecoded annotations using \npublicly available tools or resources. Users with a legally obtained copy of the film \ncan align annotations using the frame-level metadata and scanner trigger messages \nprovided in the dataset. The exact same movie stimulus as in NNDb v1.0 was used, \nand all extracted annotations from that version are equally applicable to the current \ndataset. \nIn recent years, the rise of machine learning and computer vision algorithms \nhas dramatically improved the quality, scalability, and diversity of annotations \navailable for naturalistic stimuli. Tools like Neuroscout (https://neuroscout.org), built \non pipelines such as Pliers, offer frame-level annotations for many popular films, \nincluding Back to the Future, covering features such as dialogue, visual objects, \nfaces, scenes, and even emotional tone. These automated annotations can be easily \nadapted for use with NNDb-3T+. At the same time, it is important to note that while \nsuch annotations are a valuable resource, they are not a prerequisite for meaningful \nanalysis of the database. Many approaches in naturalistic neuroimaging such as \ninter-subject functional connectivity (ISFC) (Ren et al., 2017) , dynamic functional \nconnectivity (Hutchison et al., 2013) , or model-free analyses of temporal reliability \n(Nastase et al., 2020)  do not require any annotations. Thus, the availability of \n32 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nannotations enhances the versatility of the dataset, but the scope of potential \nanalyses extends well beyond annotation-based methods. \nAlthough high-quality annotations can be generated or sourced for Back to the \nFuture, the fact that the database includes only one movie presents an inherent \nlimitation. Even with frame-level labels for visual, auditory, or linguistic features, the \nnumber of unique events, scenes, or semantic contexts remains finite. This may limit \nthe statistical power and generalizability of certain modeling approaches. Similarly, \nattempts to study rare or specific event types (e.g. non-verbal social interactions \nwithout dialogue, ambiguous language use, animal sounds) may be constrained by \ninsufficient samples. \nCode availability \nAll the code is available on the GitHub repository: \nhttps://github.com/levchenkoegor/movieproject2/. \nAcknowledgements \nThis work was supported by the London Interdisciplinary Doctoral Programme, \ntraining grant from the Biotechnology and Biological Sciences Research Council \nBB/T008709/1. We would like to thank Letitia Schneider, Raha Razin, Oliver \nJosephs, Joerg Magerkurth, Winnie Yeh and other staff at the Birkbeck-UCL Centre \nfor Neuroimaging (BUCNI) for their support. We thank Agata Czarnecka and the \nwhole Cognitron team for providing the platform for behaviour data collection. JIS \nsupported by the Wellcome Leap. \nCompeting interests \nThe authors declare no competing interests. \n33 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nSupplementary materials \nFigures illustrating the trial structure of cognitive tasks \n \n \nFigure 1. Schematic representation of the Object Memory Immediate and Delayed \ntask. Panel a) shows the list of objects presented to the participant in the first part of \nthe task. Panel b) shows the grids where participants needed to find and click on the \nobject which was previously presented in the list. \n \n \nFigure 2. One trial for a two-dimensional manipulation task. The target grid is shown \nat the top, with four possible options displayed below. Participants were asked to \nidentify which of the four grids represented a rotated transformation of the target grid. \nIn the example shown, the correct answer is the second grid from the left, which \ncorresponds to a 180-degree rotation of the target grid. \n34 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n \nFigure 3. Example trial from the Intra/Extra-Dimensional Set-Shifting (ID/ED) task. \nEach trial consists of three consecutive screens: a blank screen briefly appears \nbefore stimulus onset; two compound stimuli are presented, requiring the participant \nto choose one; and feedback follows the participant's response, indicating whether it \nwas correct (e.g., a red cross for an incorrect response). \n \n \nFigure 4. Trial of the Spatial Span task. Panel a) shows the presented sequence of \nsquares lightening up (2 in the current example). Panel b) shows the empty grid and \nthe correct response in this trial. The feedback was received after each trial. \n35 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n \nFigure 5. Trial of the Digital Span task. Panel a) shows a presented sequence of \ndigits (2 in the current example). Panel  b) shows the digital keyboard and the correct \nresponse in this trial. The feedback was received after each trial. \n \n \n \nFigure 6. Two trials with two different conditions (Text or Ink) of the Switching Stroop \ntask. Trial one (on the left side of the figure) instructed participants to click on the \n'RED' box (written in blue colour). In the second trial (on the right side of the figure), \nthe correct response was the 'BLUE' box (written in red colour). \n \n36 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n \nFigure 7. Two trials of the Verbal Reasoning task. The correct response in both trials \nis True. \n \n \nFigure 8. A trial for the Beads task. In this example, the participant revealed three \nbeads and decided to pick green as the dominant colour in the jar. Feedback was \nshown after each trial. \n \n \nFigure 9. A trial in the Verbal Analogies task. \n \n37 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n \nFigure 10. One trial in the Word Definitions task. \n \n \nFigure 11. Schematic representation of Word Memory Immediate and Delayed . \nPanel a) shows the list of words presented to the participant in the first part of the \ntask. Panel b) shows the recall where participants responded whether the word was \npreviously presented in the list or not. \nThe sequence and duration of movements for the somatotopic \nmapping task \nMovement Duration (seconds) \nright hand 18 \nright foot 17 \nleft tongue 15 \nleft hand 15 \nleft foot 18 \n38 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nright part of the face 17 \nrest 20 \nright tongue 18 \nleft hand 18 \nright part of the face 15 \nright foot 18 \nright tongue 17 \nleft part of the face 17 \nrest 18 \nleft hand 17 \nright hand 16 \nleft tongue 18 \nright part of the face 18 \nright foot 15 \nleft foot 15 \nrest 22 \nleft tongue 15 \nleft part of the face 15 \nright hand 16 \nleft part of the face 18 \nleft foot 17 \nright tongue 16 \nTable 1. Sequence of trials for run 1 of the somatotopy task. \n \nMovement Duration (seconds) \nright part of the face 16 \nleft hand 17 \nleft tongue 17 \nleft part of the face 16 \nright foot 18 \n39 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nleft hand 16 \nrest 22 \nright tongue 18 \nright hand 18 \nleft foot 16 \nleft part of the face 18 \nleft tongue 17 \nright hand 15 \nleft hand 17 \nrest 18 \nright foot 15 \nright tongue 15 \nright part of the face 16 \nleft foot 18 \nright hand 17 \nrest 20 \nright foot 17 \nleft tongue 17 \nleft foot 16 \nright part of the face 18 \nleft part of the face 16 \nright tongue 17 \nTable 2. Sequence of trials for run 1 of the somatotopy task. \nEye-tracker calibration quality of each participant, eye and run \nfor backtothefuture task \nSubject Run Eye Average error (deg) Maximum error (deg) Quality \nsub-01 \n1 \nleft 0.31 0.59 GOOD \nright 0.46 0.8 GOOD \n2 \nleft 0.54 1.6 FAIR \nright 0.62 1.77 FAIR \n40 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n3 \nleft 0.43 1.15 GOOD \nright 0.29 0.62 GOOD \nsub-02 \n1 \nleft 0.5 0.77 GOOD \nright 0.34 1.22 GOOD \n2 \nleft 0.53 1.34 GOOD \nright 0.56 0.91 GOOD \n3 \nleft 0.36 0.72 GOOD \nright 0.32 0.7 GOOD \nsub-03 \n1 \nleft 0.34 0.68 GOOD \nright 0.25 0.7 GOOD \n2 \nleft 0.47 0.68 GOOD \nright 0.45 0.74 GOOD \n3 \nleft 0.29 0.73 GOOD \nright 0.3 0.61 GOOD \nsub-04 \n1 \nleft 0.6 0.92 GOOD \nright 0.56 0.83 GOOD \n2 \nleft 0.27 0.89 GOOD \nright 0.48 0.86 GOOD \n3 \nleft 0.42 1.34 GOOD \nright 0.25 0.58 GOOD \nsub-05 \n1 \nleft 0.78 1.12 GOOD \nright 0.33 0.63 GOOD \n2 \nleft 0.59 1.44 GOOD \nright 0.41 0.65 GOOD \n3 \nleft 0.71 1.23 GOOD \nright 0.55 0.89 GOOD \nsub-06 \n1 \nleft 0.92 3.4 POOR \nright 0.45 0.77 GOOD \n2 \nleft 0.52 1.62 FAIR \nright 0.47 0.79 GOOD \n3 \nleft 0.54 1.8 FAIR \nright 0.33 0.76 GOOD \nsub-07 \n1 \nleft 0.56 0.84 GOOD \nright 0.49 1.51 FAIR \n2 \nleft 0.35 0.9 GOOD \nright 0.25 0.65 GOOD \n3 left 0.43 0.67 GOOD \n41 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nright 0.32 0.8 GOOD \nsub-08 \n1 \nleft 0.76 1.06 GOOD \nright 0.77 1.39 GOOD \n2 \nleft 0.41 0.77 GOOD \nright 0.51 0.77 GOOD \n3 \nleft 0.65 1 GOOD \nright 0.44 0.76 GOOD \nsub-09 \n1 \nleft 0.33 0.47 GOOD \nright 0.35 0.63 GOOD \n2 \nleft 0.57 0.83 GOOD \nright 0.62 0.86 GOOD \n3 \nleft 0.48 0.78 GOOD \nright 0.49 0.68 GOOD \nsub-10 \n1 \nleft 0.39 0.86 GOOD \nright 0.33 0.98 GOOD \n2 \nleft 0.31 0.76 GOOD \nright 0.56 1.94 FAIR \n3 \nleft 0.27 0.71 GOOD \nright 0.4 0.76 GOOD \nsub-11 \n1 \nleft 0.37 0.94 GOOD \nright 0.27 0.93 GOOD \n2 \nleft 0.34 0.7 GOOD \nright 0.45 0.83 GOOD \n3 \nleft 0.25 0.6 GOOD \nright 0.39 0.72 GOOD \nsub-12 \n1 \nleft 1.26 3.31 POOR \nright 1.29 5.94 POOR \n2 \nleft 0.94 1.9 FAIR \nright 0.79 1.06 GOOD \n3 \nleft 0.58 1.28 GOOD \nright 0.48 1.28 GOOD \nsub-13 \n1 \nleft 0.37 0.98 GOOD \nright 0.58 1.5 FAIR \n2 \nleft 1.31 5.79 POOR \nright 1.41 5.69 POOR \n3 \nleft 0.47 0.94 GOOD \nright 0.36 0.7 GOOD \n42 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nsub-14 \n1 \nleft 0.77 3.15 POOR \nright 0.64 2.37 POOR \n2 \nleft 0.66 0.88 GOOD \nright 0.4 0.83 GOOD \n3 \nleft 0.49 0.68 GOOD \nright 0.39 0.76 GOOD \nsub-16 \n1 \nleft 0.26 0.65 GOOD \nright 0.32 0.59 GOOD \n2 \nleft 0.35 0.67 GOOD \nright 0.37 0.61 GOOD \n3 \nleft 0.31 0.86 GOOD \nright 0.28 0.66 GOOD \nsub-17 \n1 \nleft 0.42 0.57 GOOD \nright 0.28 0.55 GOOD \n2 \nleft 0.37 0.61 GOOD \nright 0.25 0.42 GOOD \n3 \nleft 0.38 0.61 GOOD \nright 0.43 0.62 GOOD \nsub-18 \n1 \nleft 0.51 0.89 GOOD \nright 0.7 1.04 GOOD \n2 \nleft 0.32 0.63 GOOD \nright 0.38 0.56 GOOD \n3 \nleft 0.35 0.91 GOOD \nright 0.27 0.39 GOOD \nsub-19 \n1 \nleft 0.43 0.96 GOOD \nright 0.34 1.03 GOOD \n2 \nleft 0.61 1.33 GOOD \nright 0.41 0.81 GOOD \n3 \nleft 0.52 1.59 FAIR \nright 0.4 0.86 GOOD \nsub-20 \n1 \nleft 0.56 1.01 GOOD \nright 0.67 1.16 GOOD \n2 \nleft 0.64 1.41 GOOD \nright 0.44 0.95 GOOD \n3 \nleft 0.41 1.17 GOOD \nright 0.48 1 GOOD \nsub-21 1 left 0.34 0.82 GOOD \n43 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nright 0.32 0.55 GOOD \n2 \nleft 0.22 0.53 GOOD \nright 0.29 0.76 GOOD \n3 \nleft 0.27 0.68 GOOD \nright 0.36 0.64 GOOD \nsub-22 \n1 \nleft 0.29 0.74 GOOD \nright 0.35 0.59 GOOD \n2 \nleft 0.61 0.82 GOOD \nright 0.37 0.91 GOOD \n3 \nleft 0.47 1.11 GOOD \nright 0.37 0.61 GOOD \nsub-23 \n1 \nleft 0.24 0.62 GOOD \nright 0.51 0.8 GOOD \n2 \nleft 0.22 0.61 GOOD \nright 0.36 0.73 GOOD \n3 \nleft 0.24 0.54 GOOD \nright 0.34 0.92 GOOD \nsub-24 \n1 \nleft 0.58 0.89 GOOD \nright 0.29 0.82 GOOD \n2 \nleft 0.44 0.92 GOOD \nright 0.25 0.56 GOOD \n3 \nleft 0.86 7.64 POOR \nright 0.32 0.63 GOOD \nsub-25 \n1 \nleft 0.37 0.82 GOOD \nright 0.58 0.78 GOOD \n2 \nleft 0.5 1.01 GOOD \nright 0.78 1.31 GOOD \n3 \nleft 0.33 0.76 GOOD \nright 0.51 0.86 GOOD \nsub-26 \n1 \nleft 0.37 0.61 GOOD \nright 0.41 0.74 GOOD \n2 \nleft 0.3 0.46 GOOD \nright 0.3 0.81 GOOD \n3 \nleft 0.35 0.45 GOOD \nright 0.29 0.46 GOOD \nsub-27 1 \nleft 0.41 1.13 GOOD \nright 0.51 0.82 GOOD \n44 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n2 \nleft 0.35 0.64 GOOD \nright 0.47 0.73 GOOD \n3 \nleft 0.38 0.8 GOOD \nright 0.34 0.94 GOOD \nsub-30 \n1 \nleft 0.6 1.12 GOOD \nright 0.63 1.21 GOOD \n2 \nleft 0.47 0.95 GOOD \nright 0.44 0.87 GOOD \n3 \nleft 0.73 1.36 GOOD \nright 0.57 1.06 GOOD \nsub-31 \n1 \nleft 0.79 2.31 POOR \nright 0.69 1.74 FAIR \n2 \nleft 0.38 0.54 GOOD \nright 0.55 1.4 GOOD \n3 \nleft 0.33 1.02 GOOD \nright 0.64 1.95 FAIR \nsub-32 \n1 \nleft 0.33 0.77 GOOD \nright 0.23 0.69 GOOD \n2 \nleft 0.48 0.81 GOOD \nright 0.55 0.76 GOOD \n3 \nleft 0.45 0.77 GOOD \nright 0.7 0.87 GOOD \nsub-33 \n1 \nleft 0.58 1.22 GOOD \nright 0.45 1.31 GOOD \n2 \nleft 0.41 0.58 GOOD \nright 0.42 1.2 GOOD \n3 \nleft 0.33 1.07 GOOD \nright 0.43 0.75 GOOD \nsub-35 \n1 \nleft 0.31 0.84 GOOD \nright 0.26 0.68 GOOD \n2 \nleft 0.74 0.92 GOOD \nright 0.62 1.01 GOOD \n3 \nleft 0.66 0.82 GOOD \nright 0.54 1.25 GOOD \nsub-36 \n1 \nleft 0.28 0.51 GOOD \nright 0.49 0.64 GOOD \n2 left 0.54 1.02 GOOD \n45 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nright 0.47 0.9 GOOD \n3 \nleft 0.37 0.88 GOOD \nright 0.55 0.85 GOOD \nsub-37 \n1 \nleft 0.47 0.71 GOOD \nright 0.55 0.93 GOOD \n2 \nleft 0.71 1.3 GOOD \nright 0.79 1.08 GOOD \n3 \nleft 0.46 0.61 GOOD \nright 0.37 0.58 GOOD \nsub-38 \n1 \nleft 0.53 1.04 GOOD \nright 0.49 1.06 GOOD \n2 \nleft 0.44 0.63 GOOD \nright 0.36 1.34 GOOD \n3 \nleft 0.41 1.24 GOOD \nright 0.7 0.92 GOOD \nsub-39 \n1 \nleft 0.38 0.49 GOOD \nright 0.44 0.94 GOOD \n2 \nleft 0.24 0.52 GOOD \nright 0.24 0.57 GOOD \n3 \nleft 0.2 0.43 GOOD \nright 0.39 1.08 GOOD \nsub-40 \n1 \nleft 0.58 1.35 GOOD \nright 0.58 1.14 GOOD \n2 \nleft 0.56 0.94 GOOD \nright 0.54 0.89 GOOD \n3 \nleft 0.31 1.01 GOOD \nright 0.66 1.04 GOOD \nsub-42 \n1 \nleft 0.33 0.52 GOOD \nright 0.26 0.48 GOOD \n2 \nleft 0.45 0.71 GOOD \nright 0.26 0.4 GOOD \n3 \nleft 0.3 0.48 GOOD \nright 0.23 0.72 GOOD \nsub-43 \n1 \nleft 0.65 0.81 GOOD \nright 0.37 1.03 GOOD \n2 \nleft 0.28 0.65 GOOD \nright 0.36 0.83 GOOD \n46 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\n3 \nleft 0.62 1.11 GOOD \nright 0.56 1.5 FAIR \nsub-44 \n1 \nleft 0.57 0.85 GOOD \nright 0.53 1.75 FAIR \n2 \nleft 0.55 1.44 GOOD \nright 0.29 0.76 GOOD \n3 \nleft 0.52 0.88 GOOD \nright 0.45 0.89 GOOD \nMean   0.46 1.03  \nTable 3. Average and maximum errors in degrees per participant, run, and eye for the \nbacktothefuture task.  \n47 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nReferences \nAlderson-Day, B., Mitrenga, K., Wilkinson, S., McCarthy-Jones, S., & Fernyhough, C. (2018). \nThe varieties of inner speech questionnaire – Revised (VISQ-R): Replicating and \nrefining links between inner speech and psychopathology. Consciousness and \nCognition, 65, 48–58. https://doi.org/10.1016/j.concog.2018.07.001  \nAliko, S., Huang, J., Gheorghiu, F., Meliss, S., & Skipper, J. I. (2020). A naturalistic \nneuroimaging database for understanding the brain using ecological stimuli. Scientific \nData, 7(1), 347. https://doi.org/10.1038/s41597-020-00680-2  \nBaer, R. A., Smith, G. T., Lykins, E., Button, D., Krietemeyer, J., Sauer, S., Walsh, E., \nDuggan, D., & Williams, J. M. G. (2008). Construct Validity of the Five Facet \nMindfulness Questionnaire in Meditating and Nonmeditating Samples. Assessment, \n15(3), 329–342. https://doi.org/10.1177/1073191107313003  \nBaldassano, C., Hasson, U., & Norman, K. A. (2018). Representation of Real-World Event \nSchemas during Narrative Perception. The Journal of Neuroscience, 38(45), \n9689–9699. https://doi.org/10.1523/JNEUROSCI.0251-18.2018  \nBrinthaupt, T. M., Hein, M. B., & Kramer, T. E. (2009). The Self-Talk Scale: Development, \nFactor Analysis, and Validation. Journal of Personality Assessment, 91(1), 82–92. \nhttps://doi.org/10.1080/00223890802484498  \nChen, J., Leong, Y. C., Honey, C. J., Yong, C. H., Norman, K. A., & Hasson, U. (2017). \nShared memories reveal shared structure in neural activity across individuals. Nature \nNeuroscience, 20(1), 115–125. https://doi.org/10.1038/nn.4450  \nChow-Wing-Bom, H. T., Lisi, M., Benson, N. C., Lygo-Frett, F., Yu-Wai-Man, P., Dick, F., \nMaimon-Mor, R. O., & Dekker, T. M. (2025). Mapping Visual Contrast Sensitivity and \nVision Loss Across the Visual Field with Model-Based fMRI. \nhttps://doi.org/10.7554/eLife.105930.1  \nCox, R. W. (1996). AFNI: Software for Analysis and Visualization of Functional Magnetic \nResonance Neuroimages. Computers and Biomedical Research, 29(3), 162–173. \n48 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nhttps://doi.org/10.1006/cbmr.1996.0014  \nCox, R. W., & Hyde, J. S. (1997). Software tools for analysis and visualization of fMRI data. \nNMR in Biomedicine, 10(4–5), 171–178. \nhttps://doi.org/10.1002/(SICI)1099-1492(199706/08)10:4/5<171::AID-NBM453>3.0.C\nO;2-L  \nDekker, T. M., Schwarzkopf, D. S., De Haas, B., Nardini, M., & Sereno, M. I. (2019). \nPopulation receptive field tuning properties of visual cortex during childhood. \nDevelopmental Cognitive Neuroscience, 37, 100614. \nhttps://doi.org/10.1016/j.dcn.2019.01.001  \nDel Giovane, M., Trender, W. R., Bălăeţ, M., Mallas, E.-J., Jolly, A. E., Bourke, N. J., \nZimmermann, K., Graham, N. S. N., Lai, H., Losty, E. J. F., Oiarbide, G. A., Hellyer, P. \nJ., Faiman, I., Daniels, S. J. C., Batey, P., Harrison, M., Giunchiglia, V., Kolanko, M. \nA., David, M. C. B., … Hampshire, A. (2023). Computerised cognitive assessment in \npatients with traumatic brain injury: An observational study of feasibility and \nsensitivity relative to established clinical scales. eClinicalMedicine, 59, 101980. \nhttps://doi.org/10.1016/j.eclinm.2023.101980  \nDestrieux, C., Fischl, B., Dale, A., & Halgren, E. (2010). Automatic parcellation of human \ncortical gyri and sulci using standard anatomical nomenclature. NeuroImage, 53(1), \n1–15. https://doi.org/10.1016/j.neuroimage.2010.06.010  \nDick, F. K., Lehet, M. I., Callaghan, M. F., Keller, T. A., Sereno, M. I., & Holt, L. L. (2017). \nExtensive Tonotopic Mapping across Auditory Cortex Is Recapitulated by Spectrally \nDirected Attention and Systematically Related to Cortical Myeloarchitecture. The \nJournal of Neuroscience, 37(50), 12187–12201. \nhttps://doi.org/10.1523/JNEUROSCI.1436-17.2017  \nDick, F., Taylor Tierney, A., Lutti, A., Josephs, O., Sereno, M. I., & Weiskopf, N. (2012). In \nVivo Functional and Myeloarchitectonic Mapping of Human Primary Auditory Areas. \nThe Journal of Neuroscience, 32(46), 16095–16105. \nhttps://doi.org/10.1523/JNEUROSCI.1712-12.2012  \n49 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nFischl, B. (2012). FreeSurfer. NeuroImage, 62(2), 774–781. \nhttps://doi.org/10.1016/j.neuroimage.2012.01.021  \nGal, S., Coldham, Y., Tik, N., Bernstein-Eliav, M., & Tavor, I. (2022). Act natural: Functional \nconnectivity from naturalistic stimuli fMRI outperforms resting-state in predicting brain \nactivity. NeuroImage, 258, 119359. https://doi.org/10.1016/j.neuroimage.2022.119359  \nGorgolewski, K. J., Auer, T., Calhoun, V. D., Craddock, R. C., Das, S., Duff, E. P., Flandin, \nG., Ghosh, S. S., Glatard, T., Halchenko, Y. O., Handwerker, D. A., Hanke, M., \nKeator, D., Li, X., Michael, Z., Maumet, C., Nichols, B. N., Nichols, T. E., Pellman, J., \n… Poldrack, R. A. (2016). The brain imaging data structure, a format for organizing \nand describing outputs of neuroimaging experiments. Scientific Data, 3(1), 160044. \nhttps://doi.org/10.1038/sdata.2016.44  \nGrant, D. A., & Berg, E. (1948). A behavioral analysis of degree of reinforcement and ease \nof shifting to new responses in a Weigl-type card-sorting problem. Journal of \nExperimental Psychology, 38(4), 404–411. https://doi.org/10.1037/h0059831  \nHanke, M., Baumgartner, F. J., Ibe, P., Kaule, F. R., Pollmann, S., Speck, O., Zinke, W., & \nStadler, J. (2014). A high-resolution 7-Tesla fMRI dataset from complex natural \nstimulation with an audio movie. Scientific Data, 1(1), 140003. \nhttps://doi.org/10.1038/sdata.2014.3  \nHasson, U., Nir, Y., Levy, I., Fuhrmann, G., & Malach, R. (2004). Intersubject \nSynchronization of Cortical Activity During Natural Vision. Science, 303(5664), \n1634–1640. https://doi.org/10.1126/science.1089506  \nHeavey, C. L., Moynihan, S. A., Brouwers, V. P., Lapping-Carr, L., Krumm, A. E., Kelsey, J. \nM., Turner, D. K., & Hurlburt, R. T. (2019). Measuring the Frequency of \nInner-Experience Characteristics by Self-Report: The Nevada Inner Experience \nQuestionnaire. Frontiers in Psychology, 9, 2615. \nhttps://doi.org/10.3389/fpsyg.2018.02615  \nHutchison, R. M., Womelsdorf, T., Allen, E. A., Bandettini, P. A., Calhoun, V. D., Corbetta, M., \nDella Penna, S., Duyn, J. H., Glover, G. H., Gonzalez-Castillo, J., Handwerker, D. A., \n50 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nKeilholz, S., Kiviniemi, V., Leopold, D. A., de Pasquale, F., Sporns, O., Walter, M., & \nChang, C. (2013). Dynamic functional connectivity: Promise, issues, and \ninterpretations. NeuroImage, 80, 360–378. \nhttps://doi.org/10.1016/j.neuroimage.2013.05.079  \nKi, J. J., Kelly, S. P., & Parra, L. C. (2016). Attention Strongly Modulates Reliability of Neural \nResponses to Naturalistic Narrative Stimuli. The Journal of Neuroscience, 36(10), \n3092–3101. https://doi.org/10.1523/JNEUROSCI.2942-15.2016  \nKroenke, K., Spitzer, R. L., & Williams, J. B. W. (2001). The PHQ-9: Validity of a brief \ndepression severity measure. Journal of General Internal Medicine, 16(9), 606–613. \nhttps://doi.org/10.1046/j.1525-1497.2001.016009606.x  \nLerner, Y., Honey, C. J., Silbert, L. J., & Hasson, U. (2011). Topographic Mapping of a \nHierarchy of Temporal Receptive Windows Using a Narrated Story. The Journal of \nNeuroscience, 31(8), 2906–2915. https://doi.org/10.1523/JNEUROSCI.3684-10.2011  \nMehling, W. E., Acree, M., Stewart, A., Silas, J., & Jones, A. (2018). The Multidimensional \nAssessment of Interoceptive Awareness, Version 2 (MAIA-2). PLOS ONE, 13(12), \ne0208034. https://doi.org/10.1371/journal.pone.0208034  \nNastase, S. A., Gazzola, V., Hasson, U., & Keysers, C. (2019). Measuring shared responses \nacross subjects using intersubject correlation. Social Cognitive and Affective \nNeuroscience, 14(6), 667–685. https://doi.org/10.1093/scan/nsz037  \nNastase, S. A., Liu, Y.-F., Hillman, H., Norman, K. A., & Hasson, U. (2020). Leveraging \nshared connectivity to aggregate heterogeneous datasets into a common response \nspace. NeuroImage, 217, 116865. https://doi.org/10.1016/j.neuroimage.2020.116865  \nNastase, S. A., Liu, Y.-F., Hillman, H., Zadbood, A., Hasenfratz, L., Keshavarzian, N., Chen, \nJ., Honey, C. J., Yeshurun, Y., Regev, M., Nguyen, M., Chang, C. H. C., Baldassano, \nC., Lositsky, O., Simony, E., Chow, M. A., Leong, Y. C., Brooks, P. P., Micciche, E., … \nHasson, U. (2021). The “Narratives” fMRI dataset for evaluating models of \nnaturalistic language comprehension. Scientific Data, 8(1), 250. \nhttps://doi.org/10.1038/s41597-021-01033-3  \n51 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nNguyen, V. T., Sonkusare, S., Stadler, J., Hu, X., Breakspear, M., & Guo, C. C. (2017). \nDistinct Cerebellar Contributions to Cognitive-Perceptual Dynamics During Natural \nViewing. Cerebral Cortex, 27(12), 5652–5662. https://doi.org/10.1093/cercor/bhw334  \nRedcay, E., & Moraczewski, D. (2020). Social cognition in context: A naturalistic imaging \napproach. NeuroImage, 216, 116392. \nhttps://doi.org/10.1016/j.neuroimage.2019.116392  \nRen, Y., Nguyen, V. T., Guo, L., & Guo, C. C. (2017). Inter-subject Functional Correlation \nReveal a Hierarchical Organization of Extrinsic and Intrinsic Systems in the Brain. \nScientific Reports, 7(1), 10876. https://doi.org/10.1038/s41598-017-11324-8  \nSereno, M. I., Dale, A. M., Reppas, J. B., Kwong, K. K., Belliveau, J. W., Brady, T. J., Rosen, \nB. R., & Tootell, R. B. H. (1995). Borders of Multiple Visual Areas in Humans \nRevealed by Functional Magnetic Resonance Imaging. Science, 268(5212), \n889–893. https://doi.org/10.1126/science.7754376  \nShafto, M. A., Tyler, L. K., Dixon, M., Taylor, J. R., Rowe, J. B., Cusack, R., Calder, A. J., \nMarslen-Wilson, W. D., Duncan, J., Dalgleish, T., Henson, R. N., Brayne, C., \nMatthews, F. E., & Cam-CAN. (2014). The Cambridge Centre for Ageing and \nNeuroscience (Cam-CAN) study protocol: A cross-sectional, lifespan, \nmultidisciplinary examination of healthy cognitive ageing. BMC Neurology, 14, 204. \nhttps://doi.org/10.1186/s12883-014-0204-1  \nSimony, E., Honey, C. J., Chen, J., Lositsky, O., Yeshurun, Y., Wiesel, A., & Hasson, U. \n(2016). Dynamic reconfiguration of the default mode network during narrative \ncomprehension. Nature Communications, 7(1), 12141. \nhttps://doi.org/10.1038/ncomms12141  \nSpitzer, R. L., Kroenke, K., Williams, J. B. W., & Löwe, B. (2006). A Brief Measure for \nAssessing Generalized Anxiety Disorder: The GAD-7. Archives of Internal Medicine, \n166(10), 1092. https://doi.org/10.1001/archinte.166.10.1092  \nStroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of \nExperimental Psychology, 18(6), 643–662. https://doi.org/10.1037/h0054651  \n52 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint \n\nTennant, R., Hiller, L., Fishwick, R., Platt, S., Joseph, S., Weich, S., Parkinson, J., Secker, J., \n& Stewart-Brown, S. (2007). The Warwick-Edinburgh Mental Well-being Scale \n(WEMWBS): Development and UK validation. Health and Quality of Life Outcomes, \n5(1), 63. https://doi.org/10.1186/1477-7525-5-63  \nVanderwal, T., Eilbott, J., & Castellanos, F. X. (2019). Movies in the magnet: Naturalistic \nparadigms in developmental functional neuroimaging. Developmental Cognitive \nNeuroscience, 36, 100600. https://doi.org/10.1016/j.dcn.2018.10.004  \nVisconti Di Oleggio Castello, M., Chauhan, V., Jiahui, G., & Gobbini, M. I. (2020). An fMRI \ndataset in response to “The Grand Budapest Hotel”, a socially-rich, naturalistic \nmovie. Scientific Data, 7(1), 383. https://doi.org/10.1038/s41597-020-00735-4  \nWegner, D. M., & Zanakos, S. (1994). Chronic Thought Suppression. Journal of Personality, \n62(4), 615–640. https://doi.org/10.1111/j.1467-6494.1994.tb00311.x  \nZemeckis, R. (Director). (1985). Back to the Future [Video recording].  \n53 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted September 25, 2025. ; https://doi.org/10.1101/2025.09.25.678556doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}