1. Wittgenstein's remarks on knowledge present a fragmented and inconclusive picture. It would, I think, be a mistake to try to extract from them anything that could properly be called a conception of knowledge. However, there are two themes that Wittgenstein returns to again and again, not only in OC but throughout the later philosophy: first, that the idea that ‘I know’ describes a state of affairs which ‘guarantees’ what is known is a wrong picture; and second, that there are circumstances in which doubt is excluded and questions of justification do not arise, and in which the use of ‘I know’ is equivalent to ‘There is no such thing as doubt in this case’ (OC §58). At least some of these remarks focus on the question of what is sometimes called ‘non-inferential knowledge’, that is, on how we are to understand the status of judgements that we make ‘straight off, without any doubt interposing itself ‘ (OC §524). It is this issue, and Wittgenstein's treatment of it, that I want to focus on in this chapter.

2. The idea that some knowledge is non-inferential is prima facie compelling. What is non-inferentially known provides known premises for arguments that justify knowledge based on inference. In this way, non-inferentially acquired knowledge provides both the starting point and an ultimate court of appeal within our overall system of empirical knowledge. However, for those who are moved by these reflections to hold that at least some knowledge is non-inferential, there is a question of what the authority of the relevant knowledge claims derives from. On the one hand, the non-inferential nature of the knowledge seems to preclude the idea of a rational base for the relevant claims, if ‘rational base’ is understood as essentially involving the idea of arguments in which the relevant claims figure as conclusions. On the other hand, the concept of knowledge is internally linked with the idea of being able to justify what one claims to know, where justifying involves giving grounds that are epistemically more secure than the proposition that is the content of the claim.