The effects of a "bug" in the analysis of functional neuroimages (AFNI) software was greatly exaggerated, a finding that is in defiance of numerous other studies which have found that false positive rates in the analysis of functional magnetic resonance imaging (fMRI) of the brain may negate the findings of countless previous studies. 

Functional magnetic resonance imaging has been used by everyone from social psychologists to neuroscientists to try and correlate behavior to biology, sometimes to ridiculous effect, like that political conservatives in the United States have actual different brains from liberals. While the associations were always suspect - interpreting images in brains is no different than reading tea leaves, pretty subjective - the fact that common software used in analysis was flawed led to concern that thousands of fMRI studies were even more wrong than they were known to be.

Robert Cox, Gang Chen, Daniel Glen, Richard Reynolds, and Paul Taylor, National Institutes of Mental Health, NIH, Bethesda, MD, repeated some of the earlier simulations performed using the AFNI tools to assess spatial smoothness and clustering of false-positive results on brain neuroimaging. The researchers reported some, though not a particularly high rate of false positives, and described new approaches that show promise in controlling false positive rates.

"Vigorous scientific debate is a key for scientific progression," states Christopher Pawela, PhD, Co-Editor-in-Chief of Brain Connectivity. "This manuscript from Dr. Bob Cox and colleagues at the NIH is an important part of that ongoing discourse, especially in light of the many recent media reports calling into question the validity of the fMRI methodology."

Citation: Cox Robert W., Chen Gang, Glen Daniel R., Reynolds Richard C., and Taylor Paul A., 'FMRI Clustering in AFNI: False-Positive Rates Redux', Brain Connectivity. April 2017, 7(3): 152-171.