Recently, John Kay received attention on his essay The Map Is Not The Territory.
He spells out methodological critiques of economic theory in general and rational expectations in particular (interestingly enough 2011 Nobel prize winners Sargent and Sims - received the price for their research having centered around the rational expectation framework).
In the core of Kay's critique: the EMH is an illumination idea, but it is not "reality"
INET forwarded Kay's paper to a handful of economists ...
Guesnerie: The optimism embedded in the EMH has been one of the sources of the recent turmoil.
And I also found Crazy Economics Models in The Physics of Finance a blog emphasizing on positive feed backs that often drive markets away from equilibrium.
It is a complex world and I question myself whether economic models should'nt be calibrated and re-calibrated to real market behavior more frequently? If market participants have rational expectations about the future and the current behavior can only model an approximation, frequent re-calibration (inverting) can help to understand better and predict a little further?