Abstract

The aim of this essay is to argue for a new version of ‘inference-to-the-best-explanation’ scientific realism, which I characterize as Best Theory Realism or ‘BTR’. On BTR, the realist needs only to embrace a commitment to the truth or approximate truth of the best theories in a field, those which are unique in satisfying the highest standards of empirical success in a mature field with many successful but falsified predecessors. I argue that taking our best theories to be true is justified because it provides the best explanation of the predictive success of their predecessors and their own special success. Against standard and especially structural realism, I argue against the claim that the best explanations of the success of theories is provided by identifying their true components, such as structural relations between unobservable, which are preserved across theory change. In particular, I criticize Ladyman's and Carrier’s structural account of the success of phlogiston theory, and Worrall's well-known structural account of the success of Fresnel’s theory of light. I argue that these accounts tacitly assume the truth of our best theories, which in any case provides a better explanation of these theories’ success than the structural account. Structural realism is now defended as the only version of realism that is able to surmount the pessimistic meta-induction and the general problem that successful theories involve ontological claims concerning unobservable entities that are abandoned and falsified in theory-change. I argue that Best Theory Realism can overcome the pessimistic meta-induction and this general problem posed by theory-change. Our best theories possess a characteristic which sharply distinguishes them from their successful but false predecessors. Furthermore ‘inference-to-the-best-explanation’ confirmation can establish the truth of our best theories and thus trumps the pessimistic inductive reasoning which is supposed to show that even our best theories are most likely false in their claims concerning unobservable entities and processes.