I'd see classification more as a means of organizing scientific knowledge, and less as scientific knowledge itself. (Like, the question of whether a result belo

I'd see classification more as a means of organizing scientific knowledge, and less as scientific knowledge itself. (Like, the question of whether a result belongs to algebra or number theory isn't math. Those are just labels that we assign in the way we think we makes the body of knowledge most understandable.) A classification is often somewhat arbitrary--a biologist and a furrier might classify rodents differently, and both have the most useful system for their respective work. Though if you specify the basis for your classification, then I think the classification itself can become a testable prediction. You might have a category for "animals with features X and Y", because you've only ever found animals with both. This is implicitly the testable prediction that those two features will occur together. If you found a black-and-white-striped horse-shaped carnivore, then the biologists would surely consider something to have been falsified...