It appears that Dawkins review is behind the NYT firewall, but a science blogger is kind enough to excerpt some parts which he finds devastating.
If mutation, rather than selection, really limited evolutionary change, this should be true for artificial no less than natural selection. Domestic breeding relies upon exactly the same pool of mutational variation as natural selection. Now, if you sought an experimental test of Behe's theory, what would you do? You'd take a wild species, say a wolf that hunts caribou by long pursuit, and apply selection experimentally to see if you could breed, say, a dogged little wolf that chivies rabbits underground: let's call it a Jack Russell terrier. Or how about an adorable, fluffy pet wolf called, for the sake of argument, a Pekingese? Or a heavyset, thick-coated wolf, strong enough to carry a cask of brandy, that thrives in Alpine passes and might be named after one of them, the St. Bernard? Behe has to predict that you'd wait till hell freezes over, but the necessary mutations would not be forthcoming. Your wolves would stubbornly remain unchanged. Dogs are a mathematical impossibility.
Don't evade the point by protesting that dog breeding is a form of intelligent design. It is (kind of), but Behe, having lost the argument over irreducible complexity, is now in his desperation making a completely different claim: that mutations are too rare to permit significant evolutionary change anyway. From Newfies to Yorkies, from Weimaraners to water spaniels, from Dalmatians to dachshunds, as I incredulously close this book I seem to hear mocking barks and deep, baying howls of derision from 500 breeds of dogs -- every one descended from a timber wolf within a time frame so short as to seem, by geological standards, instantaneous.
Wow. I guess Dawkins wins by a knockout.