Let us begin by reviewing our main results to this point. In Chap.  2 it was argued that it is possible to give a probabilistic justification of induction. In Chap.  5 it was argued that a probabilistic argument, resembling an inductive argument, can also be given for the reality of some unobservable entities. The type of inference used in arguments of this sort was called an "Eddington-inference". But, it was also argued that if Eddington-inferences were to furnish us with good reason to believe in the unobservable entities postulated by specific scientific theories, they needed to be supplemented with a means of determining that one theory about the behaviour of observables was more likely to be true than another. It was argued that the notion of the independence of theory from data was able to do this. The viability of that notion for the task at hand was defended in Chap.  6.