‘I believe’ in a ranking-theoretic analysis of ‘believe’

Abstract

There is a prima facie tension between two well-known observations about sentences of the form I (don’t) believe p. The first is Moore’s paradox, i.e., the fact that sentences of the form p, but I don’t believe p and ¬p but I believe p sound ‘contradictory’ or ‘incoherent’. The second observation is that I believe often functions as a hedge: A speaker who asserts I believe p often (but not always) conveys that she is not certain that p, or that she does not want to commit entirely to p being true. I argue that (natural explantations of) these two observations are in conflict with respect to the following question: Does saying I believe p commit the speaker to taking p to be true? Moore’s paradox says Yes!, while the fact that I believe functions as a hedge says No! I argue that resolving this tension (along with other desiderata) requires a theory of graded belief. I reject a probabilistic threshold analysis for familiar reasons (lack of closure under conjunction), and show that an alternative based on the ranking theory of Spohn (1988, 1990, 2012) can compositionally deliver the desired results when embedded in a commitment-based theory of declarative force.