I think one of the main problems with fMRI right now, as you’ve alluded to, is that people rush off to try out this new and exciting method before seriously considering whether or not the same question could be answered with a carefully designed (and much less expensive) behavioral study. It’s these studies, which employ the method without paying deference to its unique capabilities, that discredit the field.

In fact, these methodological issues of fMRI completely overshadow the significant lingering questions about some of the data preprocessing that goes on “in the black box.” How many cog neuroscientists can claim to understand the nature of that extensive preprocessing before you get a voxelized image? I have heard that at least one person who was involved in developing the math inside those “black boxes” is now on a crusade to make people aware of his math’s limitations - presumably because people have been making claims that rely on possibly artifactual data. (A friend of mine pursued her PhD in the lab in question, and actually dropped out 3 years in, supposedly because of the stress caused by being a “bad cop” 24/7).

I’ll end with a relevant quote from Dave Weich:

“At whatever level of science you are working, the level just beneath you seems a little more scientific. Biologists kind of envy biochemists who kind of envy chemists who kind of envy physicists. Everybody wants to wear a slightly whiter coat.”

A provocative article in Seed by Yale’s Paul Bloom, Seduced by the Flickering Lights of the Brain, suggests that scientists are getting carried away with their reliance on fMRI studies.
…This is more than just phrenology. But it is not so …

VTW, I agree completely. Hindsight is always 20/20, but when the first (now ‘light bright’) studies came out, I think most people were pretty stunned to see the degree to which cognitive operations could be localized. And as you say, we probably still don’t have enough of them; there’s such a strong push to do something new with every study that it’s hard to get a good sense of what the basics are in a lot of cases.

I’m especially concerned with the criticism, both from within and outside the field, that classic brain mapping studies are recieving.
The derogatory term “light bright imaging” comes to mind. While those types of studies do lack sophistication, in a science so young as this I believe they are essential. After all, pretty much every cog neuro experiment these days is building on the brain mapping studies of the mid 90s, back when it was enough to just get people to wiggle their fingers in the scanner. Take the Superior Temporal Sulcus for instance, we wouldn’t know that it was involved in the perception of biological motion if someone hadn’t done the very trite study having people passively view point light displays and such. To think that after 8 years we’ve come so far that we no longer need to do simple brain mapping studies, is absurd. If anything I think we need to redo them as a lot of the early fMRI work had power issues (SNR problems, outdated registration techniques, small number of subjects, etc.).

I use fMRI in my own research and I still think it’s overrated and abused as a method (both in its use and in the interpretation of resulting studies.) But, like you said in your defense of neuroimaging, it offers different information from other methods. For that reason, I will continue to use it in my research!