[disgruntledphd2]

And remember that those pictures are themselves the result of lots and lots of clustering and dimension reduction, so are about as useful as cluster analysis of unsupervised data (which is what they are), that is not particularly useful at all.

I mean, the actual problem is that fMRI is expensive, and gives good spatial understanding, but bad temporal understanding (i may be mixing this up, I haven't seriously looked at any brain research in about a decade).

The statistical problems in fMRI are sadly unappreciated, much like the statistical problems in human research more generally.