What can't functional neuroimaging tell the cognitive psychologist?

In this paper, I critically review the usefulness of functional neuroimaging to the cognitive psychologist. All serious cognitive theories acknowledge that cognition is implemented somewhere in the brain. Finding that the brain "activates" differentially while performing different tasks is therefore gratifying but not surprising. The key problem is that the additional dependent variable that imaging data represents, is often one about which cognitive theories make no necessary predictions. It is, therefore, inappropriate to use such data to choose between such theories. Even supposing that fMRI were able to tell us where a particular cognitive process was performed, that would likely tell us little of relevance about how it was performed. The how-question is the crucial question for theorists investigating the functional architecture of the human mind. The argument is illustrated with particular reference to Henson (2005) and Shallice (2003), who make the opposing case.