Brain Representations of Natural Sound Statistics-fMRI dataset

This dataset contains anonymized functional and anatomical MRI data from participants in a study examining how the human brain encodes the statistical structure of naturalistic sound textures. Participants listened to synthetic sound textures in which time-averaged summary statistics were systematically varied, creating texture sounds with different levels of naturalness, ranging from spectrally matched noise to fully naturalistic exemplars. During MRI scanning, participants performed a perceptual comparison task, judging whether pairs of sounds differed in their level of naturalness. Functional MRI data were acquired on a 3T scanner with a TR of 1.056 seconds and a voxel size of 3 mm³ across five runs. Each run consisted of 52 trials. The dataset includes preprocessed fMRI data, stimulus audio files (.wav), task timing files, and subject metadata. Please refer to the README file and the original paper (https://doi.org/10.1101/2025.06.17.660096) for details on reproducing the results reported in the manuscript and the structure of the dataset. Ethics: All participants gave informed consent for data sharing. Structural images have been defaced to preserve anonymity.