Egor Levchenko: A neuroimaging database combining movie-watching, eye-tracking, sensorimotor mapping, and cognitive tasks

We provide a multimodal naturalistic neuroimaging database (NNDb-3T+), designed to support the study of brain function under both naturalistic and controlled experimental conditions.

The database includes high-quality 3T fMRI data from 40 participants acquired during full-length movie-watching and three sensory mapping tasks: somatotopy, retinotopy, and tonotopy. Each participant also completed synchronised eye-tracking during movie-watching and retinotopy, physiological recordings, and a battery of behavioural and cognitive assessments. Data were collected across two MRI sessions and a remote testing session, with all data organised in a BIDS-compliant format. Technical validation confirms high data quality, with minimal head motion, accurate eye-tracker calibration, and robust task-evoked activation patterns. The database provides a unique resource for investigating individual differences, functional topographies, multimodal integration, and naturalistic cognition. All raw and preprocessed data, quality metrics, and preprocessing scripts are publicly available to support reproducible research.

Background & Summary

One of the main goals of human neuroscience is to uncover how the brain operates in the complex, continuous experiences during everyday life. To achieve this, researchers employ both naturalistic (e.g. watching a movie or listening to a narrative) and task-based paradigms (e.g. n-back task or sensory mapping tasks). Naturalistic paradigms tend to elicit higher immersion and attentiveness in participants (Ki et al., 2016), which improves ecological validity and also reduces head motion in the scanner (Vanderwal et al., 2019). The naturalistic approach has proven effective for studying a range of processes, including the hierarchy of temporal receptive fields (Lerner et al., 2011), event segmentation in memory (Baldassano et al., 2018), default mode network dynamics (Simony et al., 2016), selective attention (Nguyen et al., 2017), emotions and social cognition (Redcay & Moraczewski, 2020), and functional connectivity (Gal et al., 2022).

Over the past decade, numerous publicly available fMRI datasets have embraced naturalistic paradigms to explore how the brain responds to continuous, real-world stimuli. Notable examples include StudyForrest (Hanke et al., 2014), which combines 7T fMRI with eye-tracking and physiological recordings during audio and audiovisual presentation of the movie “Forrest Gump”; the Sherlock dataset (Chen et al., 2017), which includes free recall during scanning; and the Grand Budapest Hotel dataset (Visconti Di Oleggio Castello et al., 2020), focused on social cognition. Other large-scale efforts, such as the Narratives (Nastase et al., 2021) and CamCAN (Shafto et al., 2014) datasets, have used spoken stories or short films to investigate ageing, language, and attention. These resources have advanced the field by demonstrating that naturalistic stimuli evoke reliable, temporally aligned neural responses across individuals, and can be used to study phenomena like event segmentation, narrative comprehension, and social perception. Despite this progress, most existing datasets emphasise either naturalistic stimulation or controlled task-based mapping, rarely integrating both within the same cohort. Furthermore, multimodal recordings such as eye-tracking and physiological measures are often missing or only partially available.

Among the currently available naturalistic fMRI datasets, the Naturalistic Neuroimaging Database (NNDb v1.0) (Aliko et al., 2020) is notable for a large sample (N=86) and extensive behavioural and cognitive phenotyping for each participant who watched one of ten full-length feature films spanning diverse genres. While NNDb v1.0 was designed to emphasise diversity of movie genres, the present dataset, the Naturalistic Neuroimaging Database 3T+ (NNDb-3T+), provides a uniquely rich multimodal resource with several tasks and physiological recordings for each participant. The dataset combines whole-brain fMRI 3T data (2×2×2mm resolution) from 40 participants during full-length movie-watching (Back to the Future), somatotopy, retinotopy, and tonotopy tasks, eye-tracking data, physiological recordings, and extensive behavioural measures. Figure 1 provides an overview of the collected data, preprocessing techniques, and quality control analyses for each task. NNDb-3T+ is, to our knowledge, the first publicly available dataset to combine naturalistic movie viewing, three distinct sensory mapping tasks, synchronised eye-tracking, physiological monitoring, and behavioural profiling within the same participants. We anticipate that NNDb-3T+ will support a wide range of future investigations, including individual differences in naturalistic processing, neural modelling of sensory hierarchies, and the development of novel analytic methods for multimodal neuroimaging. All raw and preprocessed data are shared in Brain Imaging Data Structure (BIDS) valid format (Gorgolewski et al., 2016), and quality control metrics are provided. Scripts for preprocessing and validation are openly available on GitHub to promote reproducibility (https://github.com/levchenkoegor/movieproject2)