The Failure of Blobology: fMRI Misinterpretation, Maleficience and Muddle

[...]basic research discoveries were not far behind, especially concerning dynamics of brain activity (BOLD-blood oxygen level dependency—effectively blood flow through the brain—which is a measure of neural transmission), but then it was clear that something else was missing: how do we make sense of what we are looking at? [...]the solution to this statistical problem was like most every other high dimensional data problem in astronomy, geology, or ocean mapping: just smooth, aggregate, and cluster, consequently reducing the effective number of variables and therefore degrees of freedom (there are other useful frameworks for controlling degrees of freedom as well, including false discovery rates, Euler fields, etc.,). Because functional localization like the “face area” has no actual anatomical anchors in the brain, but was based on a procedure, using another set or all the existing face stimuli in the experiment to, unfortunately, create circular kind of evidence for where the face area was to be found in the fusiform gyrus (“localizer logic”, see Friston et al., 2010 who illustrates the dangers of localizers). Because the procedure could vary across labs, the brain coordinates of the FFA actually depended on the laboratory doing the studies (Berman et al., 2010; reanalysis of their data; LDA showed a significant “lab” predictor, this was also anecdotally confirmed)! Hanson et al., 2004; Hanson and Halchenko, 2008; Hanson and Schmidt, 2011) clearly showed that model-based conditional probability of brain tissue given the “face” stimulus, although might be found in fusiform gyrus, was actually more predictive in a number of other areas, including lingual gyrus, inferior frontal gyrus, inferior parietal lobule, etc.