Functional MRI: A Phrenology for the 1990’s? Mark Cohen Page 1 Functional MRI: A Phrenology for the 1990’s? Mark S. Cohen, Ph.D. Director of Magnetic Resonance Imaging UCLA Brain Mapping Division 710 Westwood Plaza Los Angeles, CA 90095 “…In single individuals, ƒMRI (functional MRI) responses have been reported to visual stimuli [1-7], somatosensory/motor activity [5, 8-10], and acoustic stimuli [11].” “…blood flow changes were reported using non-tomographic techniques more than fifteen years ago [12] and through the use of PET [13], but with ƒMRI it becomes possible to interrogate the locus of such activity on single subjects with a high degree of reliability [14].” These are my own words, and though I would stand by them, I may well live to regret writing them. Functional MRI is touted widely, by no means solely by this author, as a method to detect regional brain activation patterns on individuals, and it reasonable to expect that enthusiastic readers might take this rather literally. In fact, however, it is only in the rare case that the validity of single subject work has been tested critically (e.g., in pre- surgical mapping studies, or multiple subject studies with identical outcomes). Taken at face value, the claim for single subject sensitivity, the failure to replicate across individuals, or even across trials for single subjects, have all been interpreted in print as demonstrating “differences in processing strategy” or “learning effects.” Consider a recently presented paper asking, “are the names of living and extinct organisms represented at different loci in the human brain?” (To avoid criticizing the work of an individual author, the experimental details have been altered...). While being imaged in an MR unit, the subjects are presented a series of pictures of obviously living (condition A) and obviously extinct (condition B) animals and asked silently to name them. A set of difference images between conditions A and B is then calculated for each subject. The pictures show a variety of locations that have larger signal in one condition than the other over three repeats of each condition. Our authors state that this is effectively conclusive proof that, in fact, the cerebral representation of these life forms is different. Needless to say, we are then presented with an exhaustive post-hoc review of the linguistic theories that might be cited in support of (or at least consistent with) this result. The data themselves are quite interesting showing widely different areas of “activation” in each subject, demonstrating, “differences in cognitive processing strategy” and a remarkably uniform distribution of active regions within individuals (perhaps a slight predominance of regions on the cortical surface) including both white matter and ventricular areas, and few common areas across individuals. The authors do not find such a result silly, and have been trying for some time to get these data into print, as I am certain they eventually will. The fact that it makes effectively no sense in the context of one hundred years of neuroscience seems to be irrelevant. Just what is our failure here?