Introduction

In the past decade, there has been a growing interest in the complex topology of human functional brain networks. Resting state (RS) and task-based (TB) fMRI provide complementary information regarding the human connectome, while meta- analysis provides an opportunity to assess convergence of task- based co-activation across a wide range of cognitive domains. Graph analysis offers a powerful means for characterizing brain networks; however, it is not yet clear what the optimal parameterization is for application to functional neuroimaging data. Specifically, node selection plays a critical role; however, this analysis stage has not been fully examined. Here, we investigated the effects of three node parcellation schemes on their respective RS and BrainMap TB graphs. We predicted that brain graphs based on functionally derived nodes are more informative than those using anatomically derived nodes.