Data from: Spatial reorganization of object representations in high-level visual cortex distinguishes working memory from perception
Data files
Apr 15, 2026 version files 7.37 GB
-
codes.zip
107.48 KB
-
control.zip
988.65 MB
-
fig1-perception_task_results.ipynb
91.38 KB
-
fig2-wm_task_results.ipynb
125.50 KB
-
fig3-ablation_decoding.ipynb
122.55 KB
-
fig4-attention_task_results.ipynb
50.35 KB
-
fig5-searchlight_RSA.ipynb
3.64 KB
-
fig6-wm_task_results_IPS_FEF.ipynb
49.69 KB
-
figS-cross_decoding.ipynb
170.61 KB
-
fmriprep_brainvoyager.zip
408.09 MB
-
README.md
5.93 KB
-
results.zip
21.58 MB
-
wm_avged.zip
1.29 GB
-
wm.zip
4.66 GB
Abstract
The visual system balances veridical object perception with flexible visual working memory (VWM), both engaging high-level visual regions. Yet how these competing demands shape object-selective cortex (OSC) representations remains unclear. Here, we ask whether VWM inherits the spatial constraints observed in the OSC during perception or reorganizes representations to meet mnemonic demands. Using a matched task design and functional magnetic resonance imaging (fMRI)-based decoding, we systematically compared the localization of object-identity representations during perception and VWM. A robust dissociation emerged: Perception kept object information largely contralateral, whereas VWM produced robust ipsilateral representation even when memorizing bilateral items. These expanded representations exploited over 90 % of ipsilateral OSC vertices and could not be attributed to increased attentional span. Critically, the ipsilateral representations closely mirrored contralateral representations, implicating interhemispheric coordination in VWM. Together, these findings demonstrate that object VWM flexibly recruits distributed high-level visual cortex, with spatial reorganization distinguishing mnemonic flexibility from perceptual fidelity.
Dataset DOI: 10.5061/dryad.3xsj3txts
Code/software
Code and data for Object WM paper
Li, W., Yang, J., & Bao, P. (2025). Spatial Reorganization of Object Representations in High-Level Visual Cortex Distinguishes Working Memory from Perception. bioRxiv. https://doi.org/10.1101/2025.06.29.662186
Core dependencies
Python
numpy==1.20.3
scipy==1.10.1
pandas==1.3.4
matplotlib==3.4.3
seaborn==0.13.2
pingouin==0.5.3
scikit-learn==1.5.0
mne==1.0.2
joblib==1.4.2
tqdm==4.62.3
h5py==3.8.0
MATLAB
BVQXfile (from neuroelf-matlab)
Project structure
codes.zip
Figure codes (.)
Jupyter notebooks for generating all figures in the paper
- fig1-perception_task_results.ipynb - Perception task results
- fig2-wm_task_results.ipynb - WM task results
- fig3-ablation_decoding.ipynb - Ablation decoding analysis
- fig4-attention_task_results.ipynb - Attention task results
- fig5-searchlight_RSA.ipynb - Searchlight RSA analysis
- fig6-wm_task_results_IPS_FEF.ipynb - WM task results for IPS and FEF regions
- figS-cross_decoding.ipynb - Cross-decoding analysis results
Calculation codes (./codes/calculation/)
- decoding/ - Decoding analysis scripts for perception and working memory tasks, related to fig 1, 2, 4, 6
- decoding_ablation/ - Ablation decoding analysis scripts, related to fig 3, 4
- rsa/ - Representational similarity analysis (RSA) scripts, related to fig 1
- searchlight_rsa/ - Searchlight RSA analysis, related to fig 5
Statistics (./codes/stats/)
- stats_table_decoding.py - Statistical analysis functions for decoding results
- stats_table_rsa.py - Statistical analysis functions for RSA results
Utilities (./codes/utils/)
Helper functions for data loading, calculations, plotting, and parallel processing.
Configuration (./codes/config/)
- fmri_config_open.py - Configuration settings for fMRI analysis
Data
Working Memory Task Data (wm.zip, wm_avged.zip)
fMRI data extracted from BrainVoyager files for WM tasks:
- wm1/ - 1-item WM task data
- wm2/ - 2-item WM task data
Tabular-data descriptions for WM label files are provided in:
- wm/wm_task_TABULAR_DATA_DESCRIPTION.md
- wm_avged/wm_task_TABULAR_DATA_DESCRIPTION.md
Control Task Data (control.zip)
fMRI data extracted from BrainVoyager files for perception and attention tasks
Tabular-data descriptions for the control-task label files:
- control/control_task_TABULAR_DATA_DESCRIPTION.md
Searchlight RSA Data (fmriprep_brainvoyager.zip)
Data for searchlight RSA analysis:
- fp_searchlight_rsa/ - Searchlight RSA results
- surface/ - Surface files for visualization
Opening the BrainVoyager searchlight files
The searchlight visualization files were generated in BrainVoyager format and
can be opened in MATLAB using BVQXfile from the open-source NeuroElf
toolbox: https://github.com/neuroelf/neuroelf-matlab.
- .smp - BrainVoyager surface map files containing node-wise searchlight RSA values
- .srf - BrainVoyager surface mesh files
- .vwp - BrainVoyager workspace/viewpoint files for visualization settings
Example for reading the BrainVoyager files with NeuroElf:
addpath(genpath('path_to_neuroelf'));
smp = BVQXfile('fmriprep_brainvoyager\fp_searchlight_rsa\fig5_searchlight_smp_LH.smp');
smp_data = smp.Map(1).SMPData;
srf = BVQXfile('fmriprep_brainvoyager\surface\fsaverage_LH-smoothwm.srf');
vwp = BVQXfile('fmriprep_brainvoyager\surface\fsaverage_LH-smoothwm_INFLATED_gray_177_110.vwp');
For users who do not work with BrainVoyager files directly, we also provide
corresponding MATLAB .mat files containing the node-wise surface-map value
array as the variable smp_data. These files are stored in mat/ subfolders
next to the corresponding .smp files, including:
- fmriprep_brainvoyager/fp_searchlight_rsa/mat/ for the figure-level data
- fmriprep_brainvoyager/fp_searchlight_rsa/sub-XXX/mat/ for subject-level data
These .mat alternatives can be read with open-source Python tools, for example:
from scipy.io import loadmat
mat = loadmat("fmriprep_brainvoyager/fp_searchlight_rsa/mat/fig5_searchlight_smp_LH.mat")
smp_data = mat["smp_data"]
Results (results.zip)
Data needed to replot figures:
- decoding/ - Decoding results
- rsa/ - RSA results
- rdm/ - Representational dissimilarity matrices
- ablation/ - Ablation decoding results
- stats/ - Additional statistical results
Tabular-data descriptions for result tables are provided in:
- results/decoding/results_decoding_TABULAR_DATA_DESCRIPTION.md
- results/ablation/ablation_TABULAR_DATA_DICTIONARY.md
- results/rsa/results_rsa_TABULAR_DATA_DESCRIPTION.md
- results/stats/stats_TABULAR_DATA_DESCRIPTION.md
Unless otherwise noted in a folder-specific result-table description:
- accuracy variables (acc, mean_acc, sem_acc) are SVM decoding accuracy values
- correlation variables (r_val, mean_rval) are Pearson correlation coefficients
- Fisher z variables are Fisher z-transformed correlations
- subject IDs use anonymized labels such as sub-002
Human subjects data
Our participants signed written consent permitting their de‑identified data to be released publicly. Before sharing, we removed all direct identifiers (names, dates of birth, contact information) and replaced subject codes with numeric IDs, ensuring individuals cannot be traced from the dataset.
