Abstract

After critically discussing the traditional account of knowledge and alternatives to it, this thesis offers an account of empirical knowledge and rational belief then applies the proposal to various issues about skepticism and epistemic logic. In the first part, I consider the causal theory, defeasibilism, the conclusive reasons analysis and Robert Nozick's tracking theory. In the second part I propose an account of empirical knowledge. S noninferentially knows that h if, and only if, S's belief that h is caused by a sequence of events each member of which is both a reliable and a conclusive indicator that h, and some member of which is an infallible indicator that h. An event E is a reliable indicator that h if, and only if, in the E-worlds close to the actual world, it is usually the case that E occurs only if h holds. E is a conclusive indicator that h if, and only if, in S's circumstances C, E would not occur unless h held. And E is an infallible indicator that h if, and only if, E would not occur unless h held. S inferentially knows that h if, and only if, S's belief that e is caused by a sequence of events each member of which is both a reliable and a conclusive indicator that h, and some member of which is an infallible indicator that h, and S's inference of h on the basis of e is rational. In the last part of the thesis I defend, against Nozick's important criticisms, the Principle of the Closure of Knowledge Under Entailment. At stake is the skeptic's argument that we cannot know that we are not in a skeptic's scenario sk, such as that we are disembodied brains misled by direct stimulation to believe what we do; so we cannot know anything incompatible with sk because if we did know that e we would, assuming closure, know that not-sk. Nozick agrees that we do not know not-sk but retains our knowledge that e by rejecting closure. I defend closure and explain the sense in which we can always know the consequences of what we know : given that it is via sequence F that we know our belief that e, the consequences of e, including not-sk, hold in all possible F-worlds except those that are far from the actual world.