Whereas Anscombe had found no basis for strict determinism, and argued that indeterminism was necessary if not sufficient for human freedom, von Wright was more of a compatibilist. Like Wittgenstein, von Wright tied a free and open future merely to human ignorance about what is to come.

The truth of Universal Determinism must remain an open question. We can settle it only for fragments of the world. In order to discriminate between ontic and epistemic alternatives (show that some are "merely epistemic") we must admit alternatives for which such discrimination has not yet been made. How this fact is related to the thesis that the concept of cause presupposes that of action.

The idea that human freedom is relative to human ignorance is familiar from the history of thought. It is related to the idea, discussed earlier, of determinism through the foreknowledge of an omniscient being. It was given an aphoristic formulation in modern times by Wittgenstein when he wrote in the Tractatus (5.1362): "The freedom of the will consists in the fact that future actions cannot be known now."
[See Wittgenstein.]

If action is correlative with ignorance of what is going to be, is then not the openness of the future, which may be said to be baked into our very concept of action, an epistemic and not an ontic feature of the world-tree? On this question, hinted at in the first lecture, I shall still have to take a stand before I finish.

Consider a total state of the world and its immediate development. Let us, for the time being, lay aside assumptions about linearity and Universal Determinism.

What is ontically certain about the world is that certain changes will occur, because there are causes of those changes operating. Ontically certain is also that some other changes will not occur, because there are counteracting causes preventing them. For the rest, the development of the world is ontically contingent, i.e. there are alternatives ahead of the given world. If determinism reigns, this "margin of contingency" shrinks to zero.

Similarly, there are two kinds of epistemic certainty about the world. One is the certainty we have that it will change in some features, because we think we know that causes of those changes are operating. The other is our certainty that the world will not change in some other features. This latter certainty may be grounded in knowledge of counteracting causes preventing the changes, but it may also be a "mere certainty" without any further ground. That which is not epistemically certain in either of these two senses is epistemically contingent.

Between the two kinds of certainty and contingency, the ontic and the epistemic, no logical relations hold. The ontic and the epistemic alternatives can be partly overlapping, or they can be inclusive or exclusive of one another. But any attempted description of the factual ontic alternatives would, of course, have to be in the terms of epistemic alternatives. For what we think, rightly or wrongly, that the ontic alternatives are, is reflected in the epistemic alternatives which we admit.

I have argued that if we want to establish, i.e. give our grounds for thinking, that something is ontically certain or is a causal necessity about changes in the world, then this presupposes an epistemic certainty regarding not-changes. This last is the certainty that some things, though not prevented from changing by any cause, will as a matter of fact not change unless we change them. The existence of this peculiar kind of epistemic certainty entails that we consider it an ontic contingency whether the world will or will not change in those features.

Now it may happen that this epistemic certainty becomes, as I shall say, "undermined" — and therewith also the entailed belief in an ontic contingency. This happens when the epistemic certainty is shown to have been "merely epistemic." The change we were certain would not happen would in fact have happened, because of the operation of a cause. That is: this change was ontically certain to occur. But the undermining of the original epistemic certainty and the establishing (as we think) of the new ontic certainty would again presuppose an epistemic certainty of the same kind as the undermined one. And this would entail belief in a new ontic contingency—and so forth for every further case of undermining the epistemic certainty.

Every time an epistemic certainty is undermined, the margin of that which we consider to be ontically contingent will shrink. But the very process of undermining requires that there is some such margin left. And this means that only for fragments of the world can determinism ever become established. It is part of the logic of things here that the validity of the deterministic thesis for the whole world must remain an open question.

To say that to establish the ontic certainty of a change presupposes an epistemic certainty of the peculiar kind which we have described is but another way of saying that establishing causal bonds in nature presupposes action. It is by virtue of these relationships that I say that the concept of cause presupposes the concept of action. Action, however, cannot rightly be said to presuppose the existence of ontic alternatives in nature, i.e. the truth of some form of indeterminism. What action presupposes is only the epistemic certainty which, as long as it is not undermined, entails the belief in the ontic contingency of some changes and thus takes for granted a certain margin of indeterminism in the world.