Dr. Swamidass was kind enough to post a link to the article where Dr. Behe offered a second definition of IC: In Defense of the Irreducibility of the Blood Clotting Cascade | Discovery Institute
It comes in Part VI. As far as I can tell, Behe is not rejecting his first definition of IC. He is offering a second, “evolutionary,” definition in order to focus attention on the number of unselected evolutionary steps that might be needed to evolve a particular IC system. As far as I can tell, Behe is not being self-contradictory. But I will merely quote the section in its entirety, so the reader may decide for herself.
"VI. An Evolutionary Perspective on Irreducible Complexity
In Darwin’s Black Box I defined the concept of irreducible complexity (IC) in the following way.
By irreducibly complex I mean a single system which is composed of several well-matched, interacting parts that contribute to the basic function, and where the removal of any one of the parts causes the system to effectively cease functioning.
(Behe 1996, 39)
While I think that’s a reasonable definition of IC, and it gets across the idea to a general audience, it has some drawbacks. It focuses on already-completed systems, rather than on the process of trying to build a system, as natural selection would have to do. It emphasizes “parts,” but says nothing about the properties of the parts, how complex they are, or how the parts get to be where they are. It speaks of “parts that contribute to the basic function”, but that phrase can, and has, been interpreted in ways other than what I had in mind (for example, talking about whole organs that contribute to complex functions such as “living”), muddying the waters in my view. What’s more, the definition doesn’t allow for “degree” of irreducible complexity; a system either has it or it doesn’t. Yet certainly some IC systems are more complex than others; some seem more forbidding than others.
While thinking of Keith Robison’s scenario, I was struck that irreducible complexity could be better formulated in evolutionary terms by focusing on a proposed pathway, and on whether each step that would be necessary to build a certain system using that pathway was selected or unselected. If a system has to pass through one unselected step on the way to a particular improvement, then in a real evolutionary sense it is encountering irreducibility: two things have to happen (the mutation passing through the unselected step and the mutation that gives a selectable system) before natural selection can kick in again. If it has to pass through three or four unselected steps (like Robison’s scenario), then in an evolutionary sense it is even more irreducibly complex. The focus is off of the “parts” (whose number may stay the same even while the nature of the parts is changing) and re-directed toward “steps.”
Envisioning IC in terms of selected or unselected steps thus puts the focus on the process of trying to build the system. A big advantage, I think, is that it encourages people to pay attention to details; hopefully it would encourage really detailed scenarios by proponents of Darwinism (ones that might be checked experimentally) and discourage just-so stories that leap over many steps without comment. So with those thoughts in mind, I offer the following tentative “evolutionary” definition of irreducible complexity:
An irreducibly complex evolutionary pathway is one that contains one or more unselected steps (that is, one or more necessary-but-unselected mutations). The degree of irreducible complexity is the number of unselected steps in the pathway.
That definition has the advantage of promoting research: to state clear, detailed evolutionary pathways; to measure probabilistic resources; to estimate mutation rates; to determine if a given step is selected or not. It allows for the proposal of any evolutionary scenario a Darwinist (or others) may wish to submit, asking only that it be detailed enough so that relevant parameters might be estimated. If the improbability of the pathway exceeds the available probabilistic resources (roughly the number of organisms over the relevant time in the relevant phylogenetic branch) then Darwinism is deemed an unlikely explanation and intelligent design a likely one."