Nikos Logothetis has written an important review about fMRI. "What we can do and what we cannot do with fMRI" -- published in Nature a couple of weeks ago -- is essential reading for anyone using fMRI, anyone playing with fMRI analysis, and anyone thinking about cognitive neuroscience and computational neuroscience.
Nikos has the street cred to say the things he says because he really really does understand what he is doing when he is doing fMRI.
The paper discusses the prospects and problems of understanding the relation between hemodynamics and neuronal activity. But this is not just theoretical analysis -- check out the measurements of the amount of vasculature per unit of cortex compared to the amount of neuronal tissue ... Yikes!
Anyway, in this paper, from one of the premier practitioners of fMRI, come the smackdown. Informative and important, and a good reminder that we should not give up electrophysiology, single-unit recording, or any other techniques, for that matter.
AND: If you want to get a bit more nerdy about this and learn more about how cells generate the stuff we are measuring and interpreting, this news study on the contribution of astrocytes from Mriganka Sur's lab is pretty interesting:
Tuned Responses of Astrocytes and Their Influence on Hemodynamic Signals in the Visual Cortex. James Schummers, Hongbo Yu, and Mriganka Sur. Science 20 June 2008: 1638-1643.