Jin Ke, Taylor A. Chamberlain, Hayoung Song, Anna Corriveau, Ziwei Zhang, Taysha Martinez, Laura Sams, Marvin M. Chun, Yuan Chang Leong, Monica D. Rosenberg. (2025). Ongoing thoughts at rest reflect functional brain organization and behavior. bioRxiv
Correspondence to jin.ke@yale.edu and mdrosenberg@uchicago.edu. please feel free to reach out with questions or inquiries.
The Chicago Attention ang Thoughts (CAT) Dataset
60 participants performed up to four 10-minute runs of an annotated rest task as part of a two-session fMRI study that also collected movie-watching, continuous performance task, and post-scan memory and narrative engagement data. Each run of the annotated rest task had 8 trials. In each trial, participants rested for 30 seconds, verbally reported their ongoing thoughts during the preceding resting period for 10 seconds, and rated them on 9 dimensions on a 9-point scale.
- The functional and structural MRI data will be publicly available on OpenNeuro when the paper gets published. Link: https://openneuro.org/datasets/ds006515
- The associated behavioral data are available here: https://github.com/jinke828/rest_thoughts/tree/main/data/beh
- The transcriptions of each individual’s spontaneous speech and its USE-semantic embeddings will not be open sourced due to privacy concerns.
- Preprocessed fMRI and eyetracking data are available upon reasonable request.
Code
We provide a step-by-step, very detailed instructions to run the scripts that replicate the key findings of this paper (i.e., brain decoding of ongoing thoughts): Code Guide; Software's license: Apache License 2.0
The instructions cover code of these following steps:
- Step00: preprocess the fMRI data downloaded from OpenNeuro ./code/a_preprocessing
Code for data analysis after preprocessing can be found here: ./code/b_analysis
- Step01: load the preprocessed fMRI data. step01_load_netts.m
- Step02: calculate FC patterns for each trial. step02_calc_FC.ipynb
- Step03: representational similarity analysis. step03_FC-thoughts_RSA.ipynb
- Step04
- a: connectome-based modeling to predict thought dimensions. step04a_SVR_dimensions.ipynb
- b: plot results from step04a. step04b_plot_SVR.R
- c: predict imagery in the aphantasic twin. step04c_predict_aphantasia.ipynb
- step05: connectome-based modeling to classify thought topics. step05_SVC_topics.ipynb
- step06: generalizing the thought models to Human Connectome Project (HCP). step06_crossdataset_prediction.ipynb
- step07
- a: relating predicted thoughts with behavioral measures suing CCA. step07a_CCA_PC.m
- b: train null thoughts models. step07b_CCA_null-models
- c: compare findings between Ke and Smith. step07c_compare_Ke-Smith.R
Results We share the code output of the main neural findings here: ./results/
- representational similarity analysis (Fig.3a-b). ./results/RSA
- connectome-based modeling (Fig. 3c-d; Fig. S5). ./results/CPM
- model-predicted thoughts in HCP. ./results/predicted
- canonical correlation analysis (Fig.5a-d; Fig. S6-8). ./results/CCA