It turns out that my students do not really believe that sound-change is regular. I didn't figure out why until today. Why didn't they believe in the regularity hypothesis? Because we talked a great deal about analogy, which clearly changes the sounds in words without phonological conditioning. This was indisputable evidence, in the minds of several of my brightest students, that sound change could not be purely regular. The problem is that we typically use the word sound-change to refer to something more confined than “changes in sounds,” rather, it refers to a specific phenomenon (one of many kinds of formal changes that can occur in language). We of a neogrammarian bent actually tend to define this phenomenon as a kind of regular, phonologically conditioned change. This is, as one especially bright student pointed out, completely circular--it renders the statment “all sound-change is regular” vacuous or purely terminological.
It is amazing how long a discussion can continue, each side disagreeing with the other, before both sides realize that they have been using the same term in different ways. I assumed a technical (and non-compositional) interpretation of the compound "sound-change" while my students read this compound compositionally. I thought they were nuts and I can only imagine what they thought of me.