Skip to main content
. 2021 Mar 2;10:e63591. doi: 10.7554/eLife.63591

Figure 2. IPL parcellation extracted from neural activity responses across attention, semantics, and social cognition.

Neural activity estimates from all three task experiments were pooled to achieve a data-driven segregation of the inferior parietal lobe, separately in each hemisphere. (a) Cytoarchitectonic boundaries defined the contours of our region of interest (ROI) (Zilles and Amunts, 2010). (b) The derived ROI was submitted to automatic parcellation into subregions to capture neural activity profiles of the three domains in each hemisphere. L-ant: left anterior subregion. L-post: left posterior subregion. R-ant: right anterior subregion. R-post: right posterior subregion. (c) Mapping of large-scale brain networks (Yeo et al., 2011) to the ROI. Posterior subregions are predominantly populated by the default mode network (DMN), while the anterior regions mainly host the ventral attention network (VAN). DAN: dorsal attention network. FPN: fronto-parietal network. SMN: somatomotor network. (d) Similarity matrix of voxel-wise neural activity estimates of single trials, reordered according to (c). The similarity structure reproduces the parcellation results.

Figure 2.

Figure 2—figure supplement 1. The final subregion solution shows a high degree of symmetry between hemispheres.

Figure 2—figure supplement 1.

Evidence supporting the hemispheric symmetry of the final two-subregion solution in the region of interest parcellation by clustering algorithms. The illustration shows the proportion of subregion-voxels falling into the homologue by flipping the x-axis, and the proportion of subregion-voxels falling into other areas.
Figure 2—figure supplement 2. Task-evoked neural responses in the IPL region with explicit modeling of motor responses.

Figure 2—figure supplement 2.

Task-dependent BOLD responses from the model that explicitly accounts for general motor responses across tasks (GLMcond+RT) resemble results from the original model (GLMcond, Figure 1). Colors indicate unthresholded T-values. Warm colors: higher GLM beta estimates for the target conditions. Cold colors: higher GLM beta estimates for the control condition.
Figure 2—figure supplement 3. Explicitly modeling motor responses yields a similar clustering solution.

Figure 2—figure supplement 3.

The illustration shows the cluster-wise similarity of the parcellation solution based on the model that captured preparatory motor responses across tasks (GLMcond+RT) compared to the basic model (GLMcond).