But when we ask whether the hypothesis of a finite universe _must_ be true, the answer is different. It is obvious, on general grounds, that we cannot, from what we know, draw conclusive inferences as to the totality of things. A very slight change in the Newtonian formula for gravitation would prevent masses beyond the limits of the visible universe from having appreciable effects if they existed, and would therefore destroy our reason for supposing that they do not exist. All arguments as to regions which are too distant to be observed depend upon extending to them the laws which hold in our part of the world, and upon assuming that there is not, in these laws, some inaccuracy which is inappreciable for observable distances, but fatal to inferences in which very much greater distances are involved. We cannot, therefore, say that the universe _must_ be finite. We can say that it may be, and we can even say a little more than this. We can say that a finite universe fits in better with the laws that hold in the part we know, and that awkward adjustments of the laws have to be made in order to allow the universe to be infinite. From the point of view of choosing the best framework into which to fit what we know—best, I mean, from a logico-æsthetic point of view—there is no doubt that the hypothesis of a finite universe is preferable. This, I think, is the extent of what can be said in its favor. Let us now see what the two finite universes are like. The difference between them is that in Einstein’s world it is only space that is queer, whereas in De Sitter’s time is queer too. Consequently Einstein’s world is less puzzling, and we will describe it first.