Most developmental biologists would attribute ear-bone evolution to developmental regulatory changes. That is, no new material was/is employed. Instead, the same old material was simply reshaped. And this could occur because of changes in the timing of expression of certain genes. Yet such regulatory/timing schemes seem largely irrelevant to Behe's focus. Proteins do not change because of the time when they are expressed. To change a protein, you need to change the amino acid sequence. To create a molecular IC system, we need to account for the various parts without the help of a developmental program. Thus, unlike the ear-bones, evolution of the cellular systems involve changing the material and coming up with new material.
Behe constrains his IC considerations to "discrete molecular systems" and explains why on pg. 41. I happen to think molecular machines are well-situated for IC considerations. For any machine is a conglomeration of parts. The parts can exist separately in a non-functional state and are then assembled into a system in which function emerges somewhat like a phase transition. And this is why middle ear bones are irrelevant. During embryological development, how are middle ear bones formed? Is there a "bone-synthesizer" that manufactures 200-or-so different human bones separately and then they are all assembled into a functioning skeleton? The answer is NO. If bones were formed like this, then yes, I think their IC state would pose a problem for non-teleological explanations.
Yet molecular machines are built like this. Here, the "parts" correspond to specific gene products (polypeptide chains). The parts are individually brought together by the ribosome and then assembled (often with the help of chaperones) through their complementary conformations. Upon assembly, function emerges (in fact, one way cells use to turn off function is to disassemble the machine partially or completely).
When dealing with molecular machines, defining parts and systems is easy. In fact, IDers don't have to define these, the scientific community has already done it (or in the process of doing it, depending on the machine).
[This message has been edited by Warren, 03-10-2004]