A.J. Ayer on the Probability of GodLanguage, Truth, and Logic (Courier Corporation: April 18, 2012; orig. 1936), p. 115.
What is not so generally recognized is that there can be no way of proving that the existence of a god, such as the God of Christianity, is even probable. Yet this also is easily shown. For if the existence of such a god were probably, then the proposition that he existed would be an empirical hypothesis. And in that case it would be possible to deduce from it, and other empirical hypotheses, certain experiential proposition which were not deducible from those other hypotheses alone. But in fact this is not possible. It is sometimes claimed, indeed, that the existence of a certain sort of regularity in nature constitutes sufficient evidence for the existence of a god. But if the sentence “God exists” entails no more than that certain types of phenomena occur in certain sequences, then to assert the existence of a god will be simply equivalent to asserting that there is the requisite regularity in nature; and no religious man would admit that this was all he intended to assert in asserting the existence of a god. He would say that in talking about God, he was talking about a transcendent being who might be know through certain empirical manifestations, but certainly could not be defined in terms of those manifestation. But in that case the term “god” is a metaphysical term. And if “god” is a metaphysical term, then it cannot be even probable that a god exists. For to say that “God exists” is to make a metaphysical utterance which cannot be either true or false. And by the same criterion, no sentence which purports to describe the nature of a transcendent god can possess any literal significance.
It is important not to confuse this view of religious assertions with the view that is adopted by atheists, or agnostics. For it is characteristic of an agnostic to hold that the existence of a god is a possibility in which there is no good reason whether to believe or disbelieve; and it is characteristic of an atheist to hold that it is at least probable that no god exists. And our view that all utterances about the nature of God are nonsensical, so far from being identical with, or even lending any support to, either of these familiar contentions, is actually incompatible with them. For if the assertion that there is a god is nonsensical, then the atheist’s assertion that there is no god is equally nonsensical, since it is only a significant proposition that can be significantly contradicted. As for the agnostic, although he refrains from saying either that there is or that there is not a god, he does not deny that the question whether a transcendent god exists is a genuine question. He does not deny that the two sentences “There is a transcendent god” and “There is no transcendent god ” express proportions one of which is actually true and the other false. All he says is that we have no means of telling which of them is true, and therefore ought not to commit ourselves to either. But we have seen that the sentences in question do not express propositions at all. And this means that agnosticism also is ruled out.
Thus we offer the theist the same comfort as we gave to the moralist. His assertions cannot possibly be valid, but they cannot be invalid either. As he says nothing at all about the world, he cannot justly be accused of saying anything false, or anything for which he has insufficient grounds. It is only when the theist claims that in asserting the existence of god he is expressing a genuine proposition that we are entitled to disagree with him.