Welcome to the pycortex WebGL MRI viewer!

This viewer shows how attention changes the way that information about hundreds of object and action categories is represented across human neocortex. Participants performed three separate tasks while watching natural movies during the experiment: passive viewing, covert visual search for humans, or covert visual search for vehicles. Participants watched the same set of movie trailers for the two visual search tasks. Functional MRI was used to measure brain activity evoked during each task. Computational modelling procedures were used to determine how 935 distinct object and action categories are represented in the brain. Further details on this work can be found in the paper by Çukur, Nishimoto, Huth and Gallant (2013), "Attention during natural vision warps semantic representation across the human brain", Nature Neuroscience 2013.
If you have problems, email us.

Warning: This page uses WebGL, an experimental web technology. It will not work in all browsers or on all platforms. For the best experience we recommend using Google Chrome, maximizing the size of your browser window, and closing other running applications (this viewer takes quite a bit of RAM).

Instructions: On the left is a 3D model of one participant's cortical surface. Use the left mouse button to rotate, right mouse button to zoom, and middle mouse button to pan. The buttons at the bottom change the way the surface is shown. When viewing the 2D flat surface, use the left mouse button to pan.

• Semantic Tuning. The colors on the surface indicate which object and action categories are represented in each part of the brain. These colors match the colors of the categories on the right. The semantic representation during each task can be viewed by clicking the appropriate button on the right panel (e.g., "Attending to Humans"). Click on a location on the cortical surface to see its semantic tuning.
• Category Responses. On the right is a tree showing 935 object and action categories. Mouse over a category to see its name and definition. Click "Category Responses" to select a category from the tree and see how it is represented across the cortical surface. The representation of the selected category during separate tasks can be viewed by clicking the appropriate button (e.g., "Attending to Humans"). To return to the original colorful view click "Semantic Tuning".
• Tuning Shifts. To view the strength of changes in semantic tuning across the cortical surface, click "Tuning Shift Index" (Positive: shifts toward the attended category, Negative: shifts away from the attended category). Click on a location on the cortical surface to see the precise tuning changes between the two search tasks (Task H: attending to humans, Task V: attending to vehicles).
• Model Performance. To view how well our models can predict the responses across the cortical surface, click "Model Performance" (correlation r).

Caveat: All of the category-specific responses shown here are based on blood-oxygen level dependent (BOLD) responses recorded while participants watched 1-2 hours of natural movies. BOLD was measured using fMRI, which is very noisy and which only reveals a small fraction of the information represented in the brain. Please do not try to read too much into the category responses for a single brain area.

Hints: The 3D viewer will run much faster if the ROI labels are disabled. To do this, mouse over the bar at the very top of the brain viewer window and uncheck the box next to "Labels".
While the viewer is showing the inflated or flattened cortex, a double left click and hold will show the same location on the fiducial (folded) 3D surface. Releasing the mouse button will return to the folded/inflated view.

Credits: The pycortex WebGL MRI viewer was written by James Gao. All data are copyright UC Regents and Gallant lab. WordNet is copyright the trustees of Princeton University. This software is copyright UC Regents (see source of this page for details). The software makes use of jQuery, Three.js, and OpenCTM.
The work was supported by grants from the National Eye Institute (EY019684 and EY022454) and from the Center for Science of Information (CSoI), an NSF Science and Technology Center, under grant agreement CCF-0939370.