One reason people think Behe's old argument is new is that he has a new phrase for it: "irreducible complexity." Something is "irreducibly complex" if by removing any one of its components you render it inoperable. The Times accurately describes Behe's inference from this sort of complexity: "If the structure serves no function without all of its parts, Dr. Behe asks, then how could evolution have built it up step by step over the ages?"

Talk about a bad question. Consider software. It often piles up in distinct increments. (Early versions of Windows, for example, were built on top of DOS.) After each incremental addition—Addition A, Addition B, Addition C, Addition D—the software is functional. But if you take the resulting ABCD software and eliminate the A part or the B part or the C part, the remaining package won't function at all.

It may be true that if you took away the final increment—the D—the remaining ABC software would work as it used to. On the other hand, if in the course of putting D on top of C the software engineers did the slightest fiddling with the structure of C, then ABC might well not work. (And natural selection does seem to modify the functionality of pre-existing components in the course of adding new components to them. In fact, it is sometimes the case that the pre-existing components had served a wholly different type of function before being integrated into the new function.)

The software system is thus, by Behe's definition, "irreducibly complex." Yet his inference—that it couldn't have been built in distinct increments, each of which added valuable functionality—is clearly wrong. (For more on the fallacious logic of "irreducible complexity," see Robert Pennock's Tower of Babel.)