What ever other problems we can find with Thomas Kuhn's Structure of Scientific Revolutions, hist discussion of the importance of technique and instrument in the advancement of science is invaluable.  Every introduction of a new technique appears to be followed by a period in which nobody does anything other than simply play with the technique - difficult questions are abandoned in favor of 'cheap and easy' experiments that tend to follow the 'let's turn the machine on and see what will happen' model of experimentation.

I've long been worried about neuroimaging for exactly this reason. For many years there, it seemed like the only work in cognitive science that could get funding required and fMRI machine. Moreover, the only stuff reported in the mainstream press was about the neural correlates of gender differences, terrorist plots, of affections for the president. I have a file of these in my office somewhere. I'll post them someday.

All of this may be coming to an end. A paper in Nature calls into question the assumption underlying all neuroimaging studies that bloodflow correlates with neuronal activity. I first heard of the story from:
Deric Bownds' MindBlog: Brain imaging can reflect expected, rather than actual, nerve activity

The full paper is here: Sirotin & Das, (2009) "Anticipatory haemodynamic signals in sensory cortex not predicted by local neuronal activity" Nature 457, 475-479 (22 January 2009).

The abstract is here. You'll need access to Nature through your library to get the full text.