[PRL] circular haskell programs

[Dave Herman]

How do you reason about termination of Haskell programs that use 
circular programming tricks? I'm looking at some examples in a couple of 
papers, and I can't convince myself that the programs should terminate, 
or even make sense -- the result of a function call gets passed as an 
argument to the function, in a way that seems to create an unresolvable 
mutual dependency between the "bottom" of the recursion and the top result.

Dave