Correlations and Multiple Comparisons in Functional Imaging: A Statistical Perspective (Commentary on Vul et al., 2009)

Vul, Harris, Winkielman, and Pashler (2009), this issue) claim in their article that the correlations reported in fMRI studies are commonly overstated because researchers tend to report only the highest correlations or only those correlations that exceed some threshold. Their article has in a short time given rise to a spirited debate about key statistical issues at the heart of most functional neuroimaging studies. The debate provides a useful opportunity to discuss core statistical issues in neuroimaging and ultimately provides a chance for the field to grow and move forward. This commentary approaches the debate from a fundamentally statistical perspective. We begin by summarizing several of the key points under discussion, followed by our own commentary on these issues from a statistical point of view. We conclude our discussion by contemplating whether it may be time to move beyond the correlation and multiple comparisons framework that is causing so much confusion and instead represent all relevant research questions as parameters in one coherent multilevel model.