Behe essentially claims that irreducibly complex systems cannot be produced directly by gradual evolution ie
(P1) Direct, gradual evolution proceeds only by stepwise addition of parts.
(P2) By definition, an irreducibly complex system lacking a part is nonfunctional.
(C) Therefore, all possible direct gradual evolutionary precursors to an irreducibly complex system must be nonfunctional.
sounds good but the 1st part of the argument is wrong.. evolution can also change or remove parts. In contrast, Behe's irreducible complexity is restricted to only reversing the addition of parts. This is why irreducible complexity cannot tell us anything useful about how a structure did or did not evolve.
eg a stone bridge:
A clear example of this is a simple stone bridge. Consider a crude "precursor bridge" made of three stones. This bridge spans the area needed to be crossed and is thus functional. For step one of the evolution of the bridge, a part is added: a flat stone on top, covering all precursor stones. Whether this improves the functionality of the bridge is irrelevant — it may or may not, the bridge still functions. Now remove the middle stone. Voilá, we have an irreducibly complex bridge, since the last step made the top-stone necessary for the function but only simple gradual steps were used to create this.
There are plenty of biological examples of the stone bridge effect.
this page has in interesting article by behe and a response by miller which uses behe's own pet mousetrap argument against him
http://www.actionbioscience.org/evolution/nhmag.htmlthere are string of essays here
http://www.talkorigins.org/faqs/behe.htmlshowing why Behe is just wrong.