Before turning to the body of the paper, we note that the word “network” in the MRI literature has an unfortunate ambiguity in the broader world of network studies. Across many disciplines (including graph theory), “network” explicitly indicates a collection of items with pairwise relationships. This sense is sometimes employed in the MRI literature, but “network” may also refer to groups of voxels or regions of interest (ROIs) that co[de]activate in PET or fMRI data (e.g., the dorsal attention network (Corbetta and Shulman, 2002; Corbetta et al., 1995), or the default mode network). It can also denote the so-called “resting state networks”, which may be defined in a variety of manners, often with component analyses (e.g. (Damoiseaux et al., 2006)) or seed correlation maps (e.g. (Biswal et al., 1995; Fox et al., 2005)). Seed correlation maps are formed by correlating the timecourse of a seed ROI (a voxel or group of voxels) with all voxels in the brain, revealing the spatial locations where rs-fcMRI activity is similar to the seed’s. For example, seeds in visual cortex tend to result in
