The a posteriori evaluation of a theory depends entirely upon the way it has stood up to severe and ingenious tests. But severe tests, in their turn, presuppose a high degree of a priori testability or content. Thus the a posteriori evaluation of a theory depends laregely upon its a priori value: theories which are a priori uninteresting—of little content—need not be tested because their low degree of testability excludes a priori the possibility that they may be subjected to really significant and interesting tests.
On the other hand, highly testable theories are interesting and important even if they fail to pass their test; we can learn immensely from their failure. Their failure may be fruitful, for it may actually suggest how to construct a better theory. [143-4]