Causal and statistical object representations in the brain (analysis)

Collection of analysis scripts used in "Computational modeling reveals dissociable physics-based and statistical object representations in the human brain during spontaneous visual processing".

📥 Clone the repository and install it as an editable Python package

Download all required data files by running the following command:

git clone <repo-url>
cd <repo-name>
python -m pip install -e .

📥 Download data

Download all required data files by running the following command:

bash get_data.sh

🚀 Run analysis and plotting scripts

Analysis / preprocessing scripts

• preprocess: Scripts for preprocessing raw BOLD signals using fMRIprep.

• run_glmsingle: Run GLMsingle of the cloth run.

• across_run_split_half_beta_reliability.ipynb: Compute model-free odd-even split-half reliability of voxel response patterns across runs.

• leave-one-scene-and-one-run-out-svm: Perform the leave-one-scene-and-one-run-out SVM.

• fig3_analysis: Get neural decoding accuracy per scenario.

• fig4_analysis: Get stiffness similarity matrix for pairs of scenes.

• create_baker_scramble_imgs.py: Generate Baker-scrambled control images used in the scrambling analysis.

• foundation_models: A video regression pipeline using frozen VideoMAE, ViViT, or V-JEPA2 backbones with regression head.

Plotting / Visualization

• plot_roi.ipynb: Plot subject-specific ROI.

• plot_leave-one-scene-and-one-run-out_SVM.ipynb: Generate figures for Fig. 1D.

• plot_svm.ipynb: Generate figures for Fig. 3B-E and Supplementary Fig. 6A-C.

• plot_rdm.ipynb: Generate figures for Fig. 4B-D and Supplementary Fig. 6D-F.

• plot_svm_video_models.ipynb: Generate figures for Fig. 5B and Supplementary Fig. 9A.

• plot_rdm_video_models.ipynb: Generate figures for Fig. 5C and Supplementary Fig. 9B.

• Baker_scramble_analysis.ipynb: Generate figures for Baker’s scrambling test in Supplementary Fig. 8.