I'm not sure I can add anything to what Bob Imrie and David Ramey have already said, but in fact it seems even easier to "imagine" an improvement when the patient is an animal. In effect, you have interposed another, probably biassed, observer into the system. You don't ask the animal how it feels, you ask the owner. And as the owner is interpreting the animal's behaviour and not feeling the discomfort directly, subjectivity can be fairly marked.
Quack (enthusiastically): Oh, he's ever so much better, isn't he!
Owner (doubtfully): Well, maybe....
Quack (getting carried away): Oh yes, look at that wet nose, those grateful eyes, and he's even wagging his tail!
Owner (convinced): Oh, isn't that wonderful!
Dog: I feel terrible, you freaking morons!
Of course, if you introduce objective measurements into the system this pretty much breaks down. I've seen interesting work on a woo-woo product for arthritis where the blinding broke down in one group, and that was the precise place where a significant positive effect was recorded for the snake-oil by the subjective questionnaire. However, objective force-plate measurements revealed no difference. Quacks, however, aren't very big on objective measurements.
Rolfe.