Truth and the Pragmatics of Explanatory Success

Bruce Glymour
Kansas State University

It is a plausible requirement on any account of scientific explanation that the correctness of otherwise successful explanations be epistemically accessible.

Since Hempel and Oppenheim in 1948, philosophers of science have taken it to be the case that fully successful explanations are correct in the sense that they contain only literally true sentences.

In this paper I argue that if correctness requires literal truth, then, on standard models, probabilistic statistical explanations are inaccessible. I further argue that we should revise our standard of correctness rather than reject the requirement of accessibility, and I suggest a class of alternative standards of correctness. Finally, I argue that the choice of standard in the suggested class is determined only by constraints of merely conventional force, and hence the notion of a "fully successful" explanation is essentially pragmatic.