At the Dispatches from the Culture Wars blog, Ed Brayton quotes a long section from Behe's cross-examination about this paper about what it actually demonstrates. It has been represented as demonstrating that a particular kind of irreducibly complex system cannot evolve. What it actually shows is something rather different. As Ed puts it:
Yet what does he admit under oath that his own study actually says? It says that IF you assume a population of bacteria on the entire earth that is 7 orders of magnitude less than the number of bacteria in a single ton of soil...and IF you assume that it undergoes only point mutations...and IF you rule out recombination, transposition, insertion/deletion, frame shift mutations and all of the other documented sources of mutation and genetic variation...and IF you assume that none of the intermediate steps would serve any function that might help them be preserved...THEN it would take 20,000 years (or 1/195,000th of the time bacteria have been on the earth) for a new complex trait requiring multiple interacting mutations - the very definition of an irreducibly complex system according to Behe - to develop and be fixed in a population.The full exchange quoted at Dispatches is worth reading, and more commentary can be found at The Panda's Thumb, where John Timmer points out that
In other words, even under the most absurd and other-worldly assumptions to make it as hard as possible, even while ruling out the most powerful sources of genetic variation, an irreducibly complex new trait requiring multiple unselected mutations can evolve within 20,000 years. And if you use more realistic population figures, in considerably less time than that. It sounds to me like this is a heck of an argument against irreducible complexity, not for it.
Based on the math presented there [in Behe & Snoke], it appears that this sort of mutation combination could arise about 10^14 times a year, or something like 100 trillion times a year.