Abstract

Belief-revision models offer an ideal of rationality that can serve as a guide to the processes of cognitive inquiry and decision. Lewis, Gardenfors, and other theorists have discovered a problem for natural formulations of belief-revision models. Their so-called triviality proofs seem to demonstrate that only the most trivial of belief-revision models can avoid a conflict between desirable principles of rationality and a natural analysis of conditionals via the Ramsey test. The Ramsey test advocates that a conditional is true and/or acceptable iff the consequent is present in the minimal rational revision of a set of full beliefs to include the antecedent. ;Gardenfors and Levi are followed in the presentation of a belief-revision model, and Rott is followed in the presentation of a version of the triviality proofs. It is argued that the apparent conflict ought to be resolved by recognizing that conditionals ought not to be considered truth-value bearing. ;Some attempts by Rott and Fuhrman to inject explicit epistemic modalities into the representation of belief are argued to be undermotivated, and the representation of such modalities--like the explicit representation of conditionals--is found to be responsible for the production of triviality. The rejection of explicit epistemic modality reinforces the conclusion that conditionals ought not to be considered truth-value bearing. ;In order to illuminate some of the differences between conditionals and truth-value bearing items of belief, some theorems pertaining to the concept of epistemic entrenchment are proved. ;Having removed the potential obstacle of the triviality results, belief-revision analyses are offered for some of the specific grammatical constructions expressive of conditionals. Actually, Dudman is followed in making a distinction between "hypotheticals" and "conditionals". Both types are seen as expressing inferability relations, but the latter also incorporates a sensitivity to time