Taxonomic and Thematic Semantic Relationships across the Adult Lifespan

# Taxonomic and Thematic Semantic Relationships across the Adult Lifespan This dataset is associated with the paper: "Beyond dual hubs: Task and aging shape taxonomic and thematic semantic relationships in ATL and TPC" (Kuhnke, Martin et al. 2026; https://doi.org/10.64898/2026.02.19.706838). Structural (T1w, T2w) and functional (task-based and resting-state) MRI data. Tasks: rest (resting state), tax (taxonomic), them (thematic), scram (scrambled). Functional runs were acquired with two echoes (echo-1, echo-2). Fieldmaps (magnitude and phasediff) are provided for distortion correction of the functional data. 85 participants are included: 42 older adults (sub-OA001 to sub-OA042) and 43 younger adults (sub-YA002 to sub-YA044). The `analysis_included` column of participants.tsv marks the 75 participants who contributed to the published analyses. ## Important: event timing and dummy volumes Every functional run begins with **30 dummy volumes (60 s at TR = 2 s)**. These volumes were acquired and **are included** in the `*_bold.nii.gz` files shared here, but were discarded during preprocessing. Onsets in the `events.tsv` files are relative to the **first volume of the raw run**, so the stimulus paradigm starts at t = 60 s. If you model these events against the preprocessed images in `derivatives/spm-preproc/` (which are 30 volumes shorter), subtract 60 s from every onset. > Users of dataset version 1.0.0 please note: the onsets in that version were relative to > the preprocessed series and were therefore 60 s out of register with the raw images. > This was corrected in version 1.1.0. See CHANGES. ## Derivatives Participant-level processed data are provided so that the published analyses can be reproduced from scratch, together with the group maps they lead to. - `derivatives/spm-preproc/` — normalised and smoothed single-subject functional timeseries (SPM12, unified segmentation + DARTEL to MNI, 4.98 x 4.98 x 5.5 mm FWHM smoothing), i.e. the exact images that entered the first-level models, plus per-run confound regressors (realignment parameters, response-time regressor, motion-outlier regressors). Covers the three task runs of 83 participants; the SPM pipeline did not normalise the resting-state runs, so those are shared as raw data only. - `derivatives/spm-firstlevel/` — single-subject first-level GLMs for the 75 analysed participants: `SPM.mat`, parameter estimates, contrast images and t maps, with tables mapping contrast and beta numbers to names. - `derivatives/spm-secondlevel/` — the group-level statistical maps for the semantic > non-semantic contrast (younger adults, older adults, their conjunction, the age-group difference, and its overlap with two published templates). Already thresholded at corrected p < 0.05. See the README in each derivatives directory for the full processing and model specification. The resting-state, functional-connectivity and diffusion analyses reported by this group used separate pipelines (fMRIPrep + CONN, MRtrix3) that are not part of this release. The scripts used to convert the internal analysis outputs into this release are not distributed with the dataset; they are available from the authors on request. ## Anonymization - All structural images (T1w, T2w) were defaced with pydeface. - Identifying scanner/site metadata (institution name and address, device serial number, station name, acquisition time of day, procedure step description) were removed from all JSON sidecars. ## Funding This work was supported by the Max Planck Society. GH was supported by the European Research Council (ERC consolidator grant FLEXBRAIN, ERC-COG-2021-101043747) and by the German Research Foundation (DFG, HA 6314/4-2, HA 6314/10-1). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.