Abstract

I propose a modest Bayesian reductionism as an alternative to Quine's moderate confirmational holism. Employing only first-order predicate logic with identity and elementary probability theory, I present two models of confirmation for individual hypotheses within blocks of theoretical sentences. Testing these models, I consider: (1) the old evidence problem; (2) the raven paradox; (3) a version of the the thesis of underdetermination which says that the evidence never provides adequate epistemic grounds for deciding between rival theories; and (4) the confirmation of modal conditionals.