The new paper (to be published in Perspectives on Psychological Science but available here) is called “Voodoo Correlations in Social Neuroscience,” which gives you a pretty good idea of its argument. Basically, the authors noticed that a lot of papers in social neuroscience that use brain imaging were reporting correlations between brain activity and social/emotional behavior or thoughts that looked too good to be true or, even, mathematically possible (kind of like the years of steady investment returns that Bernie Madoff reported). So the scientists, led by Edward Vul of MIT and Harold Pashler of the University of California, San Diego, picked 54 such studies, many of them published in prominent journals such as Science and Nature, and wrote to the authors, basically asking how they managed to get such impressive correlations.
More than half admitted using a statistical strategy that, write Vul and his colleagues, “grossly inflates correlations, while yielding reassuring-looking scattergrams.” Other statistical snafus, they say, “likely created entirely spurious correlations in some cases,” and they call on social neuroscientists who use fMRI to reanalyze their raw data “to correct the scientific record.”
This could almost be as embarrassing for Social Science as the Alan Sokal hoax in "Social Text"!