Pegleg,

This information is enormously helpful. Our research group took a scale that had been used extensively in substance abuse research, but which was tremendously unreliable and changed it dramatically into a computer administered measure that has excellent psychometric characteristics. It is now very widely used.

Perhaps something similar could be done in this area. I need to think about it. But your links are a great start. When your measurement tool has more variability than your target outcome, how can you possibly be able to determine whether a given therapy is working?