Via the Online Papers in Philosophy blog, I discovered a nice paper by William Bechtel titled, "The Epistemology of Evidence in Cognitive Neuroscience." In the paper, he discusses in some detail, and with great examples, the ways in which neuroscientists have used three different methodologies: lesion/deficit studies, single-cell recordings, and neuroimagining (PET and fMRI in particular). The first two methods have been used for decades in neuroscience proper, and with a great deal of success. Single-cell recording studies, for instance, have painted an incredibly thorough picture of the visual cortex, especially its early parts (and most especially V1, about which we probably know more than we do about any other brain area). Still, both lesion and single-cell recording methods have their limits (it's unlikely that either will tell us much about higher-level cognitive processes anytime soon), and this has led cognitive neuroscientists to rely heavily on neuroimagining techniques. In my last post, I made some pretty snide remarks about neuroimagining studies (heck, the whole post was a snide remark about neuroimagining studies). I thought it might be appropriate, then, to link to a paper which does a good job of discussing the problems with neuroimagining techniques, and in particular, the problems with the subtraction method that is most commonly used in cognitive neuroscience.
If that's not enough for you, you might also check out Sartori and Umilta's excellent paper on the subtraction method, and alternative methods, titled, "How to Avoid the Fallacies of Cognitive Subtraction in Brain Imaging." It was published in Brain and Language in 2001 (vol. 74, pp. 191-212), but I can't seem to find an online version that doesn't require a subscription. If anyone knows of one, please let me know.
And finally, while you're reading Bechtel, you might also check out his very interesting paper on the symbolic-connectionist debate, which was also linked on the Online Papers in Philosophy blog. That paper is here.