Because of our language knowledge seems like a unified phenomena. I can know lots of different things, and no matter how different they are from each other they all seem to have something in common, namely that I know them. I claim that, contrary to this intuition, knowledge is actually a sharply divided phenomena, and that it only seems unified because we don’t have a natural way to express and think about this division. The specific division I have in mind is between knowledge of regularities and knowledge of specific facts. Of course they are both broadly knowledge, meaning that to claim to know about either a regularity or a specific fact is to claim that, of the competing beliefs, that belief has the most value in general. Which is a roundabout way of saying that to claim that we know them is to claim that we have good reason to believe that they are true, without invoking the contentious notion of truth explicitly. But beyond that the two are very different, most notably in how we can come to know about them. Of course this is not to claim that they are totally unconnected, but the connections that exist between them only serve to reinforce the idea that we are dealing with two distinct kinds of knowledge.
Let me first describe knowledge of specific facts. As the name implies, to know some specific fact, is to know some claim about a specific thing and its properties and/or relations to other things. Knowing that my room contains 2 chairs and that George W Bush is the current president of America are examples of knowing two specific facts. There are two ways to know a specific fact. One is for the fact to be self-verifying. Self-verifying facts are facts that by our disposition to think they are facts make them facts. Self-verifying facts are restricted almost entirely to being about the content of our experience. For example, the claim that “I think I am seeing red” is a self-verifying fact. The disposition to make that claim while believing it is true implies that at least “I think that I think I am seeing red” is true, which collapses into “I think that I am seeing red”, which confirms that very fact. If a specific fact other than a self-verifying fact is to be knowledge it must be on the basis of a combination of other specific facts and knowledge about some regularity. For example, I know that what I read in the newspaper is usually true (knowledge of a regularity) and I know that the newspaper said that George Bush was re-elected (knowledge of a specific fact), therefore I know that George Bush is president (knowledge of a specific fact). Obviously we can’t know any specific facts of this kind without knowing other specific facts. Which would be an infinite regress if it wasn’t for the existence of self-verifying specific facts.
Knowledge of regularities is a bit different. Knowledge of this sort comes in the form of claims that when some state of affairs holds then some other state of affairs holds (or is likely to hold with some probability). This may be in the form of a causal claim, such that when that state of affairs holds then a certain state of affairs will hold of subsequent events (when I drop teacups they tend to shatter), but it may also simply reveal additional facts about a particular situation (every group has a leader). In contrast to knowledge of specific facts, we don’t come to have knowledge of regularities through a discrete process. In the case of specific facts they either are self-verifying or follow from knowledge of other specific facts and knowledge of certain regularities, and are thus themselves knowledge, or don’t and are thus not knowledge. However, knowing that a regularity holds, and the degree of certainty to which we think it holds, is something that is continuously in flux. To start with we have knowledge of a number of specific facts about various situations. We compare the number of those situations in which the regularity holds to the number in which it doesn’t, and if that percentage is close to how often we claim it holds then we know that regularity holds, to some extent. However we will run into more situations in which the regularity is supposed to hold. If the percentage of those further situations in which the regularity holds remains the percentage that we initially claimed it would hold we become more certain of the regularity; we know to a greater degree that it holds. But those new situations also have the potential to run contrary to our expectations. In this case we will become less certain that the regularity holds, and eventually may come to know the opposite, that the regularity doesn’t hold. Now admittedly our knowledge of specific facts may need to be revised occasionally as well. However, in the case of specific facts stability is expected for the large part, while we expect our knowledge of regularities to be continuously revised in the face of new evidence.
Reflecting on these two kinds of knowledge reveals a circle. Knowledge of specific facts requires knowledge of regularities, and knowledge of regularities requires knowledge of specific facts. Thus it may seem like neither kind of knowledge could be justified. Of course self-verifying facts don’t rely on knowledge of regularities, but the only regularities we could come to know about via such facts would involve only other self-verifying facts, which doesn’t get us anywhere. But consider a hypothesized regularity, namely that our perceptions usually correspond accurately to a real and external world. Given such a regularity everything we wished, both in terms of specific facts and other regularities, could be derived from it. But can we justify that regularity? I think we can. Simply note that the probability that all of our almost uncountable number of perceptions would reveal a consistent world without something guaranteeing their consistency is virtually zero. Given that they are consistent we can thus reasonably conclude that there must be something ensuring their consistency. We can call that thing, whatever it is, the external world. And since our perceptions are consistent with each other they must be correlated with some facts about this external world. This justifies the regularity (without going too far and claiming that the world is exactly as it appears), and from there on out all knowledge has a firm basis (or at least a non-circular basis).