Middle Knowledge

by Paul Helm

One subtle and ambitious way of attempting to reconcile a 'no risk' view of providence with an account of human action which is indeterministic was provided by the Jesuit theologian, Luís De Molina (1535–1600). His view has recently been revived by Alvin Plantinga and has been given renewed and intensive attention. The idea of middle knowledge can briefly be explained in the following way.

How are we to understand the omniscience of God, the idea that God knows all truths? One helpful way is to pay attention to the sorts of truths that there are for God to know. There are, to begin with, necessary truths; for instance, the laws of logic and arithmetic. Such truths could not be false. Their truth does not depend upon God willing them to be true; he knows them to be true because he is omniscient.

Then there are the myriad truths which are true as a result of God's will. For example, London is the capital of England, and the battle of Hastings was fought in 1066. These, and all such truths, are true in virtue of the fact that God has willed them to be true. Had not God freely willed them, then they would not be true. For this reason his knowledge of them is sometimes said to be free knowledge. They come about as the result of God's free decision. God does not know these truths at a time after willing them, but he knows them in willing them, rather in the way in which we know many of our own actions in doing them.

Besides these two kinds of knowledge, there is also the knowledge that God has of all possibilities which he does not will, but which remain abstract possibilities. For example, Bognor Regis is the capital of England, and the battle of Hastings was fought in 1660. Among such possibilities are conditional propositions; for instance, if Bognor Regis had been capital of England, London would have fewer than one million inhabitants. Again, if John had married Joan they would have had three children. God's knowledge of such possibilities has been called middle knowledge; knowledge midway between God's knowledge of necessary truths, and God's free knowledge.

The Bible gives examples of God's knowledge of such possibilities. Two biblical passages became famous in the discussions: 1 Samuel 23:7–13 and Matthew 11:20–24. What the first passage makes clear is that God knew that if David were to remain in the city of Keilah, then Saul would come for him; and that if Saul were to arrive in Keilah for David, the men of Keilah would surrender David to him. What Jesus affirms in the passage from Matthew is that, if his mighty works had been performed in impenitent Tyre and Sidon, they would have repented.

From these data there can be no doubt of the fact of God's middle knowledge, his knowledge of possibilities which are never brought to pass. God knew what would have happened had David remained in Keilah. But David did not remain in Keilah. Christ knew what would have happened to Tyre and Sidon had his mighty works been performed there. But the works were not performed there.

God, then, in his omniscience, knows propositions which could not be false, and propositions which could be false but which are true. He also knows propositions which could be true, but which are not in fact true, such as these concerning David and Keilah, and concerning Tyre and Sidon and Christ's mighty works. It is from this vast array of possibilities that God in his wisdom and goodness wills the actual world, the world which you and I inhabit.

It ought to be stressed that all this is common ground both to Molina and to his opponents; each side allows for the idea of middle knowledge. What is distinctive of Molina's view is that he claims that among the conditional propositions which God knows are those which indicate what would happen if an individual performed a free (i.e., non-deterministic) action. He knows, that is, myriads of propositions of the form:

(A) In circumstances C, if Jones freely chooses between X and Y, he will choose Y.

And, Molina asserts, it is because God knows all the outcomes of all the possible free choices that people will make, that he is able to create—to actualize—just those possibilities which are necessary for him to fulfil his purposes, and which involve indeterministically free choices. Hence, the free choices of creatures are compatible with God's perfect foreknowledge and his 'no-risk' providential rule of his creation. As William Lane Craig, a defender of middle knowledge, has put it:

Since God knows what any free creature would do in any situation, he can, by creating the appropriate situations, bring it about that creatures will achieve his ends and purposes and that they will do so freely.

Let us see how this works out in detail by taking a trivial but concrete example. It is supposed that among the propositions that God knows are the following conditional propositions:

(A) Only if Jones were placed in circumstances C, and were free to choose between A and B, would he choose A.

(B) Only if Jones were placed in circumstances C*, and were free to choose between A and B, would he choose B.

Let us suppose that God wills that Jones chooses B. In that eventuality, clearly God will actualize circumstances C*.

There is one major difficulty, however, with this supposition. It is that Jones is supposed to be indeterministically free. Because he is indeterministically free, he has the power, if he is free in a given set of circumstances, to choose any one of a number of alternatives open to him. As the defenders of middle knowledge frequently put it, which of the alternatives Jones chooses is up to him, not up to God.

If this is so, then God cannot know that (A) or (B) is true. And because he cannot know that A is true he cannot actualize (A) as a whole. He can actualize Jones, and he can actualize circumstances C; what he cannot do is actualize Jones's freely choosing A in circumstances C. For whether or not Jones does choose A when placed in circumstances C is up to Jones.

The proponents of middle knowledge present the following seductive picture of God's relation to various conditional possibilities. It is as if God has before his mind's eye innumerable files. Each of these files is consistent and complete. Each one represents a possible segment of the universe. God surveys all the files, and selects those which, together, form that universe which, in his wisdom and goodness, he wills to bring to pass. Among the files are those which contain references to human free actions in certain circumstances. God actualizes those files which refer to circumstances which, if the individuals are placed in them, and act freely, they will choose in accordance with the end which God desires. Thus, it is said, human freedom is preserved, and the 'no-risk' view of providence is also preserved.

The strength of the middle-knowledge view is that it presents the universe, and innumerable other possible universes, as already having run their courses, albeit in conditional form. From the sum total of all these conditions, God selects (he actualizes) some of them in order to actualize one universe. But this is a false picture. The universe cannot, given the strong view of freedom endorsed by the Molinists, have a shadow form; a form of a purely conditional kind which is the mirror-image of how the universe will be when it is actual. For how it will be when it is actual is, at least in part, up to the free actions of the agents who are actualized, once God has decided to actualize that universe.

We should not be seduced by the picture. God could not 'steer' the course of events in this fashion, given that all the while the individuals in the actualized universe have indeterministic freedom. For the circumstances never ensure one determinate freely-chosen outcome; they provide only the conditions for the free choice of one of several outcomes. Hence God cannot 'weakly actualize' certain outcomes; he cannot, that is, use his knowledge of what a free creature would do under certain circumstances to achieve a desired end.

William Lane Craig, in a particularly clear exposition of the position, distinguishes three logical 'moments' in the actualizing of the world. The second of these moments corresponds to God's middle knowledge. Craig claims that

in the second moment, which corresponds to God's middle knowledge, those aspects of the actual world exist which are states of affairs concerning what free creatures would do in any set of circumstances. For example in this second moment the state of affairs 'if Jones were placed in circumstances c, then he would freely do action X' is actual. Of course, neither Jones nor the circumstances yet exist, except as ideas in God's mind. But it is nevertheless the case that if Jones were to be actualized by God and placed in c, then Jones would freely do x. Thus the states of affairs which are expressed by true counterfactuals concerning free decisions by humans are in fact already actual at this second moment. So even though at this moment the actual world in all its fullness does not yet exist, nonetheless certain aspects of it already exist, namely logically necessary states of affairs and states of affairs corresponding to true counterfactuals concerning creaturely freedom.

Thus Craig asserts that prior to God's decision to create a world, all the features of the world he is to create, including all the outcomes of the free decisions of his creatures, are present to his mind. All that God has to do in creating the world is to make such definite actualities (which exist in his mind) exist in fact. But it is just this seductive picture of God's relation to possibility that must be resisted, because, given human indeterministic freedom, it cannot be true. As William Hasker puts it in criticizing the view, 'insofar as an agent is genuinely free, there are no true counterfactuals stating what the agent would definitely do under various possible circumstances.'

The source of this confusion may lie in a failure to distinguish between abstract possibilities and concrete actualities. There are, in the mind of God, sets of abstract possibilities, for example, the idea of a certain possible person. In creating an actual person, however, God creates all his or her physical and psychological powers, and much else besides. Creation is not like opening the door of a cage to free a lion; it is the bringing of the lion into being.

We may also be misled by an analogy to draw the opposite conclusion. We frequently say that we know our friends so well that we definitely know what they would choose in a given set of circumstances. If we know, surely God can know? But this is to forget, not only that God's knowledge is infallible, but also that if our friends are genuinely free, then in any given circumstances, they are free to choose the opposite course of action from the one that we believe, on the basis of past experience, they will in fact choose.

There are then possibilities—about the outcome of indeterministic free choices—which God does not know completely. Accordingly, he cannot instantiate complete possibilities respecting anyone's free choice. And so, because his middle knowledge of such free choices is necessarily incomplete, he cannot exercise a 'no-risk' providential control over his creation via his middle knowledge.

We began this discussion by referring to God's omniscience. But God's omniscience is limited by what is knowable. If Jones is indeterministically free, then it is not knowable, either to God or to us or to any other observer, what Jones will do when, in a given set of circumstances, he is confronted with a choice.

Does this somehow involve a limit upon God's omniscience? Are we saying that there are truths that God cannot know? Different answers have been given to this question. Swinburne, as we have seen, believes that God freely denies himself the knowledge of what he could know. Others argue that until Jones, as a free agent, has actually made his free decision, there is nothing to know. And because there is nothing to know, there is nothing for God to know. So, far from middle knowledge being a way of reconciling divine omniscience (and foreordination) and human freedom, we must conclude that human freedom limits the scope of divine omniscience.

Such a state of affairs would not preclude God from making an informed guess, an expert guess, about what Jones is likely to do. Perhaps, in circumstances C, it is very probable that Jones will do A. But the knowledge of such probabilities falls well short of that infallibility which is classically predicated of God, and which the proponents of middle knowledge certainly wish to portray him as possessing.


An antinomy?

The appeal to middle knowledge is ambitious. It attempts to preserve both indeterministic freedom, and a 'no-risk' view of providence extending to all particular actions, by demonstrating such compatibility via God's middle knowledge. The approach we shall now consider could be said to attempt to achieve the same result, but by altogether different and more modest means.

It is acknowledged on all hands that the relation between divine action and human actions is incomprehensible—not that it is impossible to understand anything whatsoever about that relation, but that we can never hope fully to understand or to explain how it is possible to preserve both the sovereignty and independence of God, and human responsibility and accountability. The proponents of middle knowledge would no doubt recognize this, as do the proponents of other views to be considered later.

It is possible to argue, however, that the issue of divine sovereignty and human accountability is such a difficult one that it is unwise to expend the effort necessary to gain even a modest understanding of it. Rather, we should simply accept that Scripture teaches both, and leave it at that.

In Evangelism and the Sovereignty of God, J. I. Packer takes what is in effect this position by calling divine sovereignty and human responsibility an antinomy:

The whole point of an antinomy—in theology, at any rate—is that it is not a real contradiction, though it looks like one. It is an apparent incompatibility between two apparent truths. An antinomy exists when a pair of principles stand side by side, seemingly irreconcilable, yet both undeniable. There are cogent reasons for believing each of them; each rests on clear and solid evidence; but it is a mystery to you how they can be squared with each other. You see that each must be true on its own, but you do not see how they can both be true together.

Man is a responsible moral agent, though he is also divinely controlled; man is divinely controlled, though he is also a responsible moral agent. God's sovereignty is a reality, and man's responsibility is a reality too.

The antinomy which we face now is only one of a number that the Bible contains. We may be sure that they all find their reconciliation in the mind and counsel of God, and we may hope that in heaven we shall understand them ourselves. But meanwhile, our wisdom is to maintain with equal emphasis both the apparently conflicting truths in each case, to hold them together in the relation in which the Bible itself sets them, and to recognize that here is a mystery which we cannot expect to solve in this world.

There are a number of things worth noting about these positions. In the first place, in terms of the basic categorization that we are using in this chapter, Packer is taking a 'no-risk' view of providence. This is clear from his statement that man is divinely controlled, and from the fact that he recognizes that there is a problem of an acute kind here. There would be no problem were he prepared to dilute the claim about divine control. It is only because of this control (and because of human responsibility) that there is a problem in the first place. What Packer is saying, in effect, is that given that there is a 'no-risk' view of divine providence, it is impossible to see how this can be reconciled with the view that human beings are responsible for their actions.

According to Packer, the difficulty about reconciling divine sovereignty and human responsibility is due to our ignorance. It is important to see that he is not arguing that the difficulty arises because of some basic contradictoriness about the nature of things. He is not arguing that the claims that God is sovereign and that men and women are responsible are logically contradictory—like the claims that Smith is to the left of Robinson and Robinson to the left of Smith. On the contrary, he is emphasizing that the sovereignty of God and the responsibility of men and women must be consistent, since both are true. The point is, though, that from our present vantage point we cannot see how they can be consistent.

There are different kinds of ignorance. I may know your name but not your telephone number. This ignorance of mine is, however, fairly easily remedied, for I can look up your number in the directory. Even if your number is exdirectory I can obtain it from you. We might call this contingent ignorance.

Such contingent ignorance can be contrasted with necessary ignorance. We do not know what a shrimp's view of a barnacle is, or what it is like to be a bat. This is because we are not shrimps or bats, and are necessarily not shrimps or bats. In a similar way, it is impossible for us to know what it is like to be God, and to have his timeless vision of his creation. There are hosts of such matters of which we are necessarily ignorant.

Yet Packer is not arguing that among the things of which we are necessarily ignorant is the recognition of the consistency of God's sovereign control of human beings and their personal responsibility. For he expresses the hope that, in heaven, Christians will understand every antinomy presented to them in the Bible, including the antinomy of divine sovereignty and human responsibility. Yet Packer seems to be saying that our ignorance, though not absolutely necessary, is necessary in this life. This is either because there are data relevant to the reconciliation of the antinomy which are withheld from us, or because our faculties of understanding are constricted in some way in this life (or for both of these reasons).

What are we to make of Packer's proposal that, in effect, Christians should confess both divine 'no-risk' providence and human responsibility, admit their ignorance, and leave the matter there?

The proposal does have several strengths. It sacrifices nothing in terms of the positions that a Christian might wish to hold on this issue. There are no theoretical constraints or considerations ruling out any particular position. In the construction of Christian doctrine there is always the danger of rationalism, of imposing some a priori principle upon the data from which the doctrine is to be constructed or restricting a doctrine to what we can understand about it. Clearly enough, Packer's proposal avoids such dangers. By the same token—and this is the second of its strengths—Packer can take into account the full range of the scriptural data in an uninhibited fashion, reserving any difficulties that the consideration of such data raises until, in heaven, we possess the necessary faculties, or additional data, or both.

These are considerable gains. Yet there are a number of disadvantages to be set against them. As we have seen, Packer affirms that certain biblical doctrines constitute antinomies. But by what means is it established that there is an antinomy? Packer says that 'it is forced upon us by the facts themselves'. But how do we establish the difference between facts which are very hard to reconcile, and facts which are antinomic in character? Clearly it is not sufficient to say that in the one case the facts do not force an antinomic conclusion, whereas in the other case they do. Packer has reasoned his way to the conclusion that certain data constitute an antinomy. But the process of his reasoning is unclear.

Whatever the details of that reasoning, in order to be justified it will have to have included a general statement to the effect that all reasonable steps have been taken to reconcile the two claims about divine sovereignty and human responsibility, and have failed. The number of failures of reconciliation must be sufficiently large to generate the conclusion, on inductive grounds, that it is reasonable to believe that the two claims are antinomic in character. But has such an enquiry ever been undertaken?

The second difficulty is perhaps the more serious. Packer distinguishes between an antinomy and a self-contradiction or logical incoherence. An antinomy is an apparent inconsistency, not a real one. But it is an apparent inconsistency which, in this life, we shall be for ever unable to clear up. There is no way of effecting a reconciliation, and we know that there is no such way. But then, in these circumstances, what is the difference between an apparent inconsistency and a real one? How do we know that what is called an antinomy might not turn out to be a real inconsistency?

Packer would presumably reply to the question by an appeal to Scripture. He might argue that since both divine sovereignty and human responsibility are taught in Scripture, and are divinely revealed truths, they must be consistent, since truth is one. Nothing that is divinely revealed can be inconsistent with anything else that is divinely revealed, however much it may seem to us to be inconsistent.

The problem with such a reply is that it is too permissive. During the history of Christianity there is scarcely a limit to the nonsense that has been believed because it is allegedly biblical in character. No doubt the doctrines of divine providence and human responsibility are centrally biblical in a way that other themes are not. Nevertheless, in the light of what has happened in history, it may be thought that some attempt ought to be made to show their consistency.

In fact, even in Packer's brief treatment of the issue, and despite what he says about simply accepting the antinomy, there are traces of an attempt to provide some of this reasoning. Thus it is interesting, and surely significant, that Packer does not speak in terms of human freedom but of human responsibility. Yet there is nothing in his appeal to the idea of an antinomy that would prevent him arguing that divine sovereignty and human indeterministic freedom constitute an antinomy. He contents himself with referring to responsibility only, and says nothing about the conditions for human responsibility. So, although appealing to an antinomy could be a licence for accepting nonsense (or, to be more polite, for accepting statements of a strongly counterintuitive nature), Packer restricts his use of it.

It ought to be emphasized that these critical comments are not offered as decisive against Packer's approach. Its attractions remain, and readers must make up their own minds as to which approach to take. Nevertheless, because of the theoretical unsatisfactoriness of appealing to an antinomy, an attempt will be made to take the discussion a stage further.



We have looked at several ways in which a 'no-risk' view of providence might be established (notably by appealing either to middle knowledge or by invoking the idea of an antinomy), and have found difficulties with each. In the case of middle knowledge, the difficulty is with the indeterministic view of freedom its proponents adopt. In the case of the idea of an antinomy, the difficulty lies with its theoretically permissive stance.

In the history of debate about human freedom and responsibility, two contrasting views of freedom have been identified: indeterministic freedom (sometimes referred to as the liberty of indifference), and deterministic freedom (sometimes referred to as the liberty of spontaneity). As we have noted, the attractiveness of the middle-knowledge approach lies in its assumption that human freedom requires the liberty of indifference. By contrast, the view taken in what follows favours deterministic freedom. According to this view, people perform free acts when they do what they want to do, not when they have the power of self-causation, or some other version of indeterminism. That is, they are not constrained or compelled in their actions, but what they do flows unimpededly from their wants, desires, preferences, goals and the like.

The great advantage of such a view of human freedom is that, being compatible with determinism, it is also compatible with a full view of divine omniscience and omnipotence, and thus with a 'no-risk' theory of providence. Possible disadvantages lie in the areas of human responsibility and the problem of evil. We shall attempt to handle these difficulties in later chapters. It is perhaps sufficient at this stage to indicate the assumption upon which the subsequent discussion rests.

Needless to say, the policy of taking up one position, sticking to it, and following through its consequences, is not intended to constrain the reader. Authors naturally hope that whoever reads their books will end up holding views corresponding to those set out in them. But whether or not my readers conclude by holding a 'risk-free' view of divine providence, and the compatibilist view of human freedom and determinism which I believe most naturally coheres with it, the important thing is that readers make up their own minds in the light of the evidence presented to them.

Each of the views just discussed (middle knowledge, the antinomic view, and compatibilism), is consistent with the 'no-risk' view of divine providence. Each position has its own strengths and weaknesses which have repercussions in other areas of Christian theology. In following this study as it traces some of the repercussions of compatibilism, readers will no doubt wish to keep the other views in mind as alternatives. It is hoped that the discussion of this chapter will make it easier for them, if they find themselves disagreeing with the main thread of argument, to make the necessary adjustments to it, and thus, by the end, to have developed a view of divine providence of their own which they believe to be Christian, coherent, and defensible.


Excerpt: Helm, P. (1993). The Providence of God. (G. Bray, Ed.) (pp. 55–68). Downers Grove, IL: InterVarsity Press.


By Topic


By Scripture

Old Testament









1 Samuel

2 Samuel

1 Kings

2 Kings

1 Chronicles

2 Chronicles








Song of Solomon


















New Testament







1 Corinthians

2 Corinthians





1 Thessalonians

2 Thessalonians

1 Timothy

2 Timothy





1 Peter

2 Peter

1 John

2 John

3 John



By Author

Latest Links