Good theory is testable

Much of what passes for theory in studies  of animal behavior and sociobiology is semantic maneuvering to obtain maximum congruence of classifications. This process is useful but better described as concept formation. Real theory is postulational-deductive. To formulate it, we must first identify the parameters, then we define the relations between them as precisely as we can, and finally we construct models in order to relentlessly extend and to test the postulates. Good theory is either quantitative or at least cleanly qualitative in the sense that it produces easily recognized inequalities. Its results are often nonobvious or even counterintuitive. The important thing is that they exceed the capacity of unaided intuition. Good theory produces results that attract our attention as scientists and stimulate us to match them with phenomena not easily classified by previous schemes. Above all, good theory is testable. It's results can be translated into hypotheses subject to falsification by appropriate experiments and field studies. 

-E. O. Wilson, Sociobiology