Let us consider, first of all, Hume's most sceptical argument (the argument which he merely suggests). This argument suggests that, since our only reason for supposing two facts to be causally connected is that we have found them constantly conjoined in the past, and since it does not follow from the fact that they have been conjoined ever so many times, that they _always_ will be so, therefore we cannot _know_ that they always will be so, and hence cannot know that they are causally connected. But obviously the conclusion does not follow. We must, I think, grant the premiss that, from the fact that two things have been conjoined, no matter how often, it does not strictly _follow_ that they _always_ are conjoined. But it by no means follows from this that we may not _know_ that, as a matter of fact, when two things are conjoined sufficiently often, they are also _always_ conjoined. We may quite well _know_ many things which do not logically follow from anything else which we know. And so, in this case, we may _know_ that two things are causally connected, although this does not logically follow from our past experience, nor yet from anything else that we know. And, as for the contention that our belief in causal connections is merely based on _custom,_ we may, indeed, admit that custom would not be a sufficient _reason_ for concluding the belief to be true. But the mere fact (if it be a fact) that the belief is only caused by custom, is also no sufficient reason for concluding that we can _not_ know it to be true. Custom _may_ produce beliefs, which we do know to be true, even though it be admitted that it does not _necessarily_ produce them.