Why Antirealists Can’t Explain Success

Scientific realism explains the success of novel predictions by assuming that theories are at least partially true accounts of the unobservable reality. Here ‘novel’ means not used essentially in constructing the theory, heterogeneous to previous observations and a priori improbable. I argue that no alternative antirealist explanation is possible: van Fraassen’s evolutionary analogy explains why we don’t have unsuccessful theories, not why we have some successful ones. His empirical adequacy and Fine’s surrealism (i.e., that theories are compatible with all the observable data) do not entail that they predict anything, let alone new phenomena. Hacking’s deflationist explanation, that success is already explained by theories themselves, presupposes the truth of theories. Saying that theories have true novel consequences because they are instrumentally reliable (Fine, Lyons), or predictively similar to the true theory (Stanford) is like saying that they have lots of true consequences, including many novel ones. But this repeats the explanandum rather than explaining it. Moreover, realists have a plausible account of how partially true theories can be found, but antirealists cannot not explain how theories with true novel consequences can be found (except by looking for partially true theories): by inductive extrapolation one can only predict observations homogeneous to the old ones.