hen it is stated that the conclusions of knowledge involving inference must be sub- ordinated to knowledge which is had directly and immediately, and must be carried to the latter for proof and verification, we are at once struck by the multitude of theories regarding what is immediately and infallibly known. The diversity and con- tradictions give ground for asuspicion that in no case is the "knowledge" in question as self-evident as it is asserted to be. And there is good theoretical ground for the suspicion. Suppose aman "explains" the eclipse of the moon by saying it is due to the attempt of adragon to devour it. To him the devouring dragon is amore evident fact than is the darkening of the moon. To us the existence of an animal capable of such afeat is the doubtful matter. It will be objected that it is unfair to take such an absurd case as an instance: dragons are not the sort of thing which any philosopher has asserted to be the ob- ject of direct and certain non-inferential knowledge. But the illustration still serves apurpose.