Abstract

I investigate the extent to which Descartes's sceptical arguments in the Meditations show that, given a traditional view of justification, we cannot have justified beliefs in the external world. On this traditional view, knowledge is essentially justified true belief, and many beliefs are justified only if they are backed by reasons. Consequently, one's justified beliefs are thought to form a system: an inverted pyramid, resting on foundational beliefs, or a 'raft', held together by coherence. ;To illuminate this view, I contrast it with the 'reliabilist' view of knowledge. Within this alternative view, the demand for justification by reasons is largely replaced by a demand that a modal relation holds between a knower's belief and his environment. If this relation holds, then the belief is likely to be true. ;I contend that these views really agree on this point, since a reason for belief, on the traditional view, also involves a relation with this character. Then, starting from some ideas due to Bonjour, I argue that the 'reliabilist' view does not appreciate that epistemic activity is rational activity directed at a goal--the acquisition of true beliefs. Consequently, epistemic agents are normally obligated to have means to this goal, namely, reasons for belief. In making my argument, I develop an original account of epistemic possibility and doubt. ;I then examine Descartes's sceptical arguments. If one accepts the traditional view, I argue, one holds expectations about justification that amount to constraints on acceptable systems of justified belief. One is that a system is acceptable only if the beliefs in it are likely to be true. Another is that a system is acceptable only if it is sensitive to truth: that is, if the world were different, the system could 'shift' towards the truth. But Descartes's Dream and Demon Arguments seem to show that these constraints, together with others, form an inconsistent set. ;I discuss several strategies against these arguments, and show that each rejects or radically revises some of the constraints. I end by proposing a strategy, based on a transcendental argument, that does neither.