Here, therefore, we have a very different view suggested, as to the limits of human knowledge. And even this is not all. There is yet a third view, inconsistent with both of these, which Hume suggests in some parts of his work.
So far as we have yet seen, he has not in any way contradicted his original supposition that we can know _some_ matters of fact, which we have never ourselves observed. In the second theory, which I have just stated, he does not call in question the view that I can know all such matters of fact as I know to be causally connected with facts which I have observed, nor the view that I can know some facts to be thus causally connected. All that he has done is to question whether I can know any _external_ fact to be causally connected with anything which I observe; he would still allow that I may be able to know that future states of my own, or past states, which I have forgotten, are causally connected with those which I now observe or remember; and that I may know therefore, in some cases, what I shall experience in the future, or have experienced in the past but have now forgotten. But in some parts of his work he does seem to question whether any man can know even as much as this: he seems to question whether we can ever know any fact whatever to be causally connected with any other fact. For, after laying it down, as we saw above, that we cannot know any fact, A, to be causally connected with another, B, unless we have experienced in the past a constant conjunction between facts like A and facts like B, he goes on to ask what foundation we have for the conclusion that A and B _are_ causally connected, even when we _have_ in the past experienced a constant conjunction between them. He points out that from the fact that A has been constantly conjoined with B in the past, it does not follow that it ever will be so again. It does not follow, therefore, that the two really are causally connected in the sense that, when the one occurs, the other _always_ will occur also. And he concludes, for this and other reasons, that _no argument_ can assure us that, because they have been constantly conjoined in the past, therefore they really are causally connected. What, then, he asks, is the foundation for such an inference? _Custom,_ he concludes, is the only foundation. It is nothing but custom which induces us to believe that, because two facts have been constantly conjoined on many occasions, therefore they will be so on _all_ occasions. We have, therefore, no better foundation than custom for any conclusion whatever as to facts which we have not observed. And can we be said really to _know_ any fact, for which we have no better foundation than this? Hume himself, it must be observed, never says that we can't. But he has been constantly interpreted as if the conclusion that we can't really know any one fact to be causally connected with any other, did follow from this doctrine of his. And there is, I think, certainly much excuse for this interpretation in the tone in which he speaks. He does seem to suggest that a belief which is _merely_ founded on custom, can scarcely be one which we _know_ to be true. And, indeed, he owns himself that, when he considers that this is our only foundation for any such belief, he is sometimes tempted to doubt whether we do know any fact whatever, except those which we directly observe. He does, therefore, at least suggest the view that every man's knowledge is entirely confined to those facts, which he is directly observing at the moment, or which he has observed in the past, and now remembers.