THINGS-data

A multimodal, large-scale collection for studying natural object representations in human brain and behavior.

All data are sampled from responses to the same stimulus set: the THINGS database of 1,854 object concepts and 26,107 natural images. THINGS-data links fMRI, MEG, and behavioral similarity judgments measured with these stimuli.

Get the data Read the paper

1,854

Object concepts

26k+

Naturalistic object images

4.7M

Behavioral similarity judgments

fMRI · MEG

Neuroimaging responses resolved in space and time

All data are openly available via Figshare, OpenNeuro, and OSF, with ready-to-use derivatives and analysis code.

Data components

Every component uses the same THINGS stimulus set (object concepts and images). Data are openly available: preprocessed fMRI and MEG derivatives via the Figshare collection, raw fMRI and MEG in BIDS format on OpenNeuro, and behavioral similarities and embeddings on OSF.

Behavior

Triplet odd-one-out similarity judgments (~4.7M) over THINGS images, plus an semantic embedding model capturing interpreable dimensions underlying perceived objects similarity.

Behavioral data on OSF

fMRI

High-resolution functional magnetic resonance imaging (fMRI) with dense sampling per participant during natural object viewing. Three participants, each saw > 8,000 images of 720 objects.

How to use Preprocessed fMRI responses (Figshare) Raw fMRI data (OpenNeuro) Browse all derivatives (Figshare collection)

MEG

Time-resolved magnetoencephalographic (MEG) recordings during object viewing.

Four participants, twelve sessions each
Preprocessed sensor-level time courses and eyetracking

Raw MEG on OpenNeuro (ds004212) Preprocessed MEG data (Figshare)

How THINGS-data has been used

A few concrete example use cases from Teichmann, Contier, and colleagues.

1. Replicating seminal findings and integrating modalities across brain and behavior.

The main THINGS-data paper (Hebart, Contier, Teichmann et al., 2023) shows how to (i) replicate classic findings on the representation of object animacy and size in the human brain, (ii) link behavioral embedding to cortical response, and (iii) integrate object responses resolved in time and space.

2. Time-resolved object properties in MEG

Using THINGS MEG in “Dynamic representation of multidimensional object properties in the human brain” to model the temporal dynamics of neural signals reflecting human perception of objects.

3. Behavior-derived object dimensions in visual cortex

Using the behavioral and fMRI parts of THINGS-data, this paper “Distributed representations of behavior-derived object dimensions in the human visual system” mapped interpretable behavioral dimensions onto cortical response patterns, revealing human understanding of objects is distributed across the visual system.

How to cite THINGS-data

If you use THINGS-data in your work, please cite the main paper:

Hebart MN, Contier O, Teichmann L, Rockter AH, Zheng CY, Kidder A, Corriveau A, Vaziri-Pashkam M, Baker CI (2023). THINGS-data, a multimodal collection of large-scale datasets for investigating object representations in human brain and behavior. eLife 12:e82580. https://doi.org/10.7554/eLife.82580.

Show BibTeX

@article{THINGSdata,
  title   = {THINGS-data, a multimodal collection of large-scale datasets for investigating object representations in human brain and behavior},
  author  = {Hebart, Martin N and Contier, Oliver and Teichmann, Lina and Rockter, Adam H and Zheng, Charles Y and Kidder, Alexis and Corriveau, Anna and Vaziri-Pashkam, Maryam and Baker, Chris I},
  journal = {eLife},
  volume  = {12},
  pages   = {e82580},
  year    = {2023},
  doi     = {10.7554/eLife.82580},
  url     = {https://doi.org/10.7554/eLife.82580}
}

Analysis code in this repository

This GitHub repository contains the analysis code used for the neuroimaging results in the THINGS-data paper. It focuses on fMRI and MEG; raw data live on OpenNeuro, and derivatives on Figshare. Source code: github.com/vicco-group/THINGS-data

MRI

The MRI/ folder contains Python modules, scripts, and notebooks for working with THINGS-fMRI:

ThingsMRIdataset and related utilities to navigate BIDS-formatted data.
ThingsmriLoader for loading single-trial betas, voxel metadata, and masks.
GLM and encoding model workflows (e.g. glm.py, animacy_size.ipynb).

See MRI/README.md and example notebooks in MRI/notebooks/ for usage.

MEG & eyetracking

The MEG/ folder provides a stepwise pipeline for preprocessing, quality control, validation analyses, and decoding/RSA using MEG and eyetracking data.

Preprocessing of BIDS MEG data and head position checks
Validation analyses for animacy/size, MEG–fMRI combination, and noise ceilings
Pairwise decoding and representational similarity workflows

See MEG/README_MEG.txt for a description of the pipeline and dependencies.