David Perlman, a neuroscience graduate student at the University of Wisconsin, presents a comedic reminder of the dangers of false positives in fMRI data. The researcher conducted an fMRI of a deal salmon’s brain as it was “shown a series of photographs depicting human individuals in social situations with a specified emotional valence.” Moreover, “the salmon was asked to determine what emotion the individual in the photo must have been experiencing.” Curiously enough, several voxels in the dead salmon’s brain activated at statistically significant levels. Mr. Perlman even put together a poster presentation as a finishing touch to this comical yet important reminder about the dangers of false positives in brain imaging technologies. Check the end of this review for links to the original story and the poster presentation.
This Is Not Unique To fMRIs
Keep in mind that false positives exist in ALL research, not just for fMRIs. Many statistical procedures are probability estimates with an established risk of accepting an erroneous result (i.e., false positive, false negative, etc.). As the number of statistical analyses increase, the greater the chance that a statistical error will occur. For example, 5 out of every 100 observations will occur by chance alone with an alpha-rate set to the traditional .05. Thousands of comparisons are made in an fMRI, and a similar number can occur in a QEEG so it easy to understand how a dead salmon’s brain can seemingly perceive stimuli.
Image: Craig Bennett, et al.