There's also an ongoing problem in the analysis of MRI functional connectivity. About 2010, someone had the idea to apply concepts from graph theory to fcMRI. T

There's also an ongoing problem in the analysis of MRI functional connectivity. About 2010, someone had the idea to apply concepts from graph theory to fcMRI. That is fraught with a number of heroic assumptions (fcMRI correlations do not necessarily represent physical connections, and continuous relationships must be binarized with arbitrary thresholds). But most importantly, there are so many different graph theory metrics, and ways to tweak the analysis, that a determined postdoc is basically guaranteed to find something that can reach statistical significance. I suspect something like that has happened here.