The structure of perceptual justification

Abstract

When does a perceptual experience as of my hands provide justification for me to believe that I have hands? One initially plausible positive requirement is that I must have reasons to believe something else: that my experience is veridical. Also plausible is the negative requirement that I must not have reasons to believe that my experience is in this case non-veridical. Dogmatists about perception reject the positive requirement and embrace the negative one, holding that perceptual justification is immediate — it doesn’t rest upon any other justification that I possess — but it is also defeasible, and in particular it is underminable. Whereas Dogmatism is a theory in the epistemology of perception, Bayesianism is a theory of coherence for partial beliefs and of how coherence it to be maintained in the face of new evidence. Just as it is incoherent to believe both I have hands and ¬(I have hands), it is incoherent to be highly confident in both of those propositions. Bayesianism offers a formal account of this latter type of coherence. Importantly, though Bayesianism is a theory of coherence, it is not a coherence theory of the sort defended by Davidson and BonJour: it is not an attempt to explain all facts about justification in terms of facts about coherence. Hence the Bayesian’s claim that partial beliefs are subject to norms of coherence is at least prima facie consistent with the Dogmatist’s claim that some beliefs are immediately justified by experience. I develop and defend Bayesian Dogmatism. I begin by responding to an argument advanced by Cohen, Hawthorne, Schiffer, and White, which purports to show that Bayesians cannot model episodes of perceptual learning in which the proposition learned is both immediately justified and defeasible. I go on to respond to arguments from Weisberg, which purport to show that Bayesianism is inconsistent with underminable perceptual learning. Finally, I defend the superiority of Jeffrey Conditionalization to Holistic Conditionalization as a rule of updating on propositions learned from experience.