If therapists were athletes and the model was a javelin, say. And you wanted to work out if one javelin was better than another, and you spent 50 years looking at that question, always thinking it was the javelin, not the athlete, that was getting the results, you'd look pretty silly wouldn't you?
Imagine testing a javelin by randomly assigning athletes, from little athletics level through to Olympic level, without realizing that some were better than others..... getting small statistical differences between this brand of javelin and that one. Then javelin throwers get on the bandwagon and all the really serious ones bought the better javelin, then all of a sudden the better javelin was more obviously better, but we tell ourselves that was just because this new study design was better.
If anyone teaches undergraduates it would be a fun project. Get all the stats for world class javelin players and find out what javelins they use then divvy up the data that way and call yourself a psychotherapy researcher.
Think that's absurd?
Here's my plea. All new studies of psychotherapy effectiveness should not be funded unless they include therapist effectiveness in their study design, and measure this. What do you think?