Is Human Freedom Compatible with Divine Foreknowledge?

If God is omniscient and exhaustive knowledge of the future is possible, then God knows (and in fact knew a long time ago) what we will do in the future. But is this compatible with our future actions being free? I address this question by responding to an argument that claims that these things are incompatible. At the heart of this incompatibility argument is the idea that God’s past beliefs about our future actions are “accidentally necessary”—can’t be changed—and that this necessity transfers to our actions. I consider four main responses to the argument: Brian Davies’ claim that God’s past beliefs about our future actions are properly characterised as beliefs about what we will freely do; Alvin Plantinga’s view—stemming from the work of William of Ockham—that God’s past beliefs about our future actions aren’t wholly about the past and so aren’t accidentally necessary; Trenton Merricks’ idea that our having a choice about God’s past beliefs doesn’t require us to be able to change the past; and Alfred Freddoso’s claim that the accidental necessity of God’s beliefs doesn’t transfer to our actions. My response, like that of Merricks, focuses on the idea that God’s past beliefs are dependent on, and to be explained in terms of, our future actions; rather than vice versa. I support this view by appealing to the idea that God’s foreknowledge is simple (an idea I defend against several William Hasker—inspired objections). I argue that neither God’s beliefs, nor their accidental necessity, should be thought of as a threat to our freedom; and that in any event, the accidental necessity of God’s beliefs isn’t something that can coherently be thought to transfer to our actions. I conclude that God’s past beliefs about our future actions are perfectly compatible with our freedom.


Introduction
I'm going to begin by considering the following argument, which dates back at least to the work of St Thomas Aquinas, often used to support the claim that human freedom is incompatible with God's foreknowledge.
Let's suppose that God believed 20 years ago that John will go sailing tomorrow. Since God's knowledge is infallible, it's not possible for him to believe something that's false. So, the fact that God believed 20 years ago that John will go sailing tomorrow entails that John will indeed go sailing tomorrow. This is an important feature of God's belief. Another important feature is that it's unalterable. It's unalterable because it's a past belief-a belief held in the past-and the past is something we can't change. But then it's claimed to follow from these two features of God's belief that John going sailing tomorrow is-like God's belief itself-also unalterable. The idea is that if God's belief is unalterable, and it logically follows from his belief that John will go sailing tomorrow; then John's action is unalterable too. Indeed, not even John can do anything to alter his going sailing tomorrow. But then, the argument goes, his going sailing tomorrow won't be-can't be-something he does freely. In this way, it's claimed that divine foreknowledge is incompatible with John's action being free. But is it? 1 In order to investigate this argument-which I will generally refer to simply as the incompatibility argument-it will be useful to set it out formally in the following way. 2 Let t 1 be any past time (for example, 20 years ago), t 2 be the present, and t 3 be any future time (for example, tomorrow).
(1) Suppose that God believed at t 1 that John will go sailing at t 3 .
(2) The proposition God believed at t 1 that John will go sailing at t 3 is accidentally necessary (that is, fixed or unalterable) at t 2 and at all times thereafter.
(3) If a proposition p is accidentally necessary at some time t and p logically entails q, then q is also accidentally necessary at t.
(4) Since God's knowledge is infallible, that God believed at t 1 that John will go sailing at t 3 logically entails that John will indeed go sailing at t 3 .
(5) The proposition John will go sailing at t 3 is therefore accidentally necessary at t 2 and at all times thereafter.
(6) If the proposition John will go sailing at t 3 is accidentally necessary at t 2 and at all times thereafter, John can't do otherwise than go sailing at t 3 .
(7) If John can't do otherwise when he performs some action, then he doesn't perform it freely.
(8) Conclusion: God believing at t 1 that John will go sailing at t 3 is incompati-1 Although the focus in this paper will be on John's action, the argument outlined of course isn't just a problem for John. After all, the reasoning given could be applied to any human action that God believed will be performed in the future, and so-because God is omniscient and has beliefs (and so knowledge) in relation to every future human action-it seems that it could be applied to any future human action. The argument therefore suggests that divine foreknowledge is incompatible, quite generally, with free human action. In responding to this argument, I'm going to make three main assumptions (many of which are shared by the argument's proponents). The first is an assumption about God's temporal nature. I'm going to assume that God is a temporal being rather than a timeless being, because it seems most natural to think that the problem of incompatibility arises when God is thought of in this way.
After all, the argument appeals to the notion of God's foreknowledge of human action, and this is a knowledge he has which is in time and is temporally prior to the time when John acts. 4 My second assumption relates to the possibility of knowledge of the future. Is such knowledge possible? Aristotle for example, appears to think that it isn't, claiming that it's neither true nor false prior to a given sea battle that a particular side would win it (Aristotle, 1963: Book 9). Now, if knowledge of the future isn't possible because there are no truths about the future to be known, then there is-prior to someone performing an action-no fact of the matter that he or she will act in a particular way and so no possibility of God knowing what will be done. 5 So, if knowledge of our future actions isn't possible, the problem of incompatibility between God's foreknowledge and our future free actions wouldn't arise. However, in what follows I will simply assume that knowledge of the future is possible. 6 3 Although the argument is framed in terms of an alleged incompatibility between God's infallible beliefs and our actions being free, it's interesting to note that there is an argument similar in structure which calls into question the compatibility of God's infallible beliefs and his being able to intervene in the world (for example, to prevent a specific case of natural or moral evil). The similar argument (framed here in terms of a devasting future earthquake) might go something like this: (i) Suppose that God believed at t 1 that there will be a devastating earthquake at t 3 . (ii) The proposition God believed at t 1 that there will be a devastating earthquake at t 3 is accidentally necessary (i.e. fixed or unalterable) at t 2 and at all times thereafter. (iii) If a proposition p is accidentally necessary at some time t and p logically entails q, then q is also accidentally necessary at t. (iv) Since God's knowledge is infallible, that God believed at t 1 that there will be a devastating earthquake at t 3 logically entails that there will indeed be a devastating earthquake at t 3 . (v) The proposition there will be a devastating earthquake at t 3 is therefore accidentally necessary at t 2 and at all times thereafter. (vi) If the proposition there will be a devastating earthquake at t 3 is accidentally necessary at t 2 and at all times thereafter, nobody (not even God) can do anything to prevent it. (vii) Conclusion: God believing at t 1 that there will be a devastating earthquake at t 3 is incompatible with his being able to prevent it-and more generally, God's foreknowledge of any event is incompatible with him being able to intervene to prevent it occurring. Although I won't specifically respond to this argument, I do consider the issue of Divine Providence in a later section: "Simple Foreknowledge: some problems". 4 It's worth noting, however, that this first assumption won't be important if-as some philosophers have claimed-an argument structurally parallel to the argument outlined can be formulated where God is thought of as a timeless being. See, for example, Linda Zagzebski (Zagzebski, 1991: p. 60 That knowledge of the contingent future isn't possible is a view adopted by proponents of open theism. Indeed, open theism is in part a response to the alleged incompatibility between God's foreknowledge and human freedom. Open theists aim to leave aspects of the future "open" and thus unknown to God precisely in order to leave room for our freedom. A position on omniscience similar to that of the open theist is adopted by Richard Swinburne (Swinburne, 1977: pp. 167-178).
(Interestingly, Swinburne considers the incompatibility argument as applied not only to the question of whether our freedom is consistent with God's foreknowledge of our future actions; but also where libertarian freedom minimally requires an agent having the power to act otherwise. In deciding whether an action foreknown by God is free, the crucial question will be whether or not the agent in question has the ability to ("can") act otherwise than she acts; where this amounts to being able to exercise, in the specific set of actual circumstances she faces, the power to act otherwise. What we really need to know is whether we can be free in this libertarian sense if God has foreknowledge of our actions.
My aim in this paper is to investigate the incompatibility argument to see how we can best respond to it. In the course of doing so, my own response to the issue of incompatibility will emerge. I want to begin by briefly considering premise (1).

God's Belief
Premise (1)-God believed at t 1 that John will go sailing at t 3 -might seem to be relatively uncontroversial, and we might wonder what in any event would be the point of questioning God's belief. After all, we are supposing that God believed 20 years ago that John will go sailing tomorrow and the incompatibility argument proceeds to the conclusion that this is incompatible with him freely going sailing; but if instead, God believed that John will stay indoors tomorrow, wouldn't a structurally parallel argument establish that this would then be incompatible with him freely staying indoors?
to the question of whether God's freedom is consistent with his foreknowledge of his own future actions.) Rejecting the Aristotelian position, Swinburne accepts that claims about the future (and so about our future free actions) can be true or false (now, for example). However, he argues that omniscience should be thought of as amounting not to having knowledge of all truths, but to having knowledge of everything true that it's logically possible to know, and that "…P is omniscient if he knows about everything except those future states and their consequences which are not physically necessitated by anything in the past…" (p. 175) But is the fact that our future free actions aren't physically necessitated by some prior cause, really a good reason to say that it isn't logically possible to know them? Many would find it difficult to accept the view that there are truths about our future free actions that even God can't know. Moreover, this view doesn't seem to cohere well with Swinburne's view that God being "…omniscient only in the attenuated sense would…have resulted from his own choice. In choosing to…give others freedom…he limits his own knowledge of what is to come. He continually limits himself in this way by not curtailing…men's future freedom." (p. 176) It doesn't cohere well with this view because if God deliberately stops himself from forming beliefs about our future actions in order to preserve our freedom, isn't this in order to stop himself from knowing the knowable? If future contingents weren't knowable, then the issue of God deliberately stopping himself from knowing them surely wouldn't arise. A final worry about Swinburne's view might be the effect it would have on a traditional understanding of providence. Although it's beyond the scope of this paper to investigate this issue fully, it's interesting to note that (against this worry) A. A. Howsepian argues that God could "…lack foreknowledge of free human actions, and yet exercise robust providential governance over Creation." (Howsepian, 1993: p. 16 say that what God believed if he instead believed that John will stay indoors tomorrow, is that he will freely stay indoors.) This line of thought is adopted, for example, by Brian Davies: "…if God knows at time 1 that P will freely do X at time 2 then what God knows is that P will freely do X. In other words, if God knows at time 1 that P will freely do X at time 2 then God's knowledge at time 1 is dependent on P freely doing X at time 2. Given that at time 1 God knows that P will freely do X at time 2, it may be unalterable after time 1 that P will freely do X. But the unalterable fact at time 1 just could not be at all if P were not free to do X at time 2. For if P were not free at time 2, then God could not know at time 1 that P would be free at time 2." (Davies, 1982: pp. 88-89) But even if this line of thought seems initially attractive, it does seem to respond to the incompatibility argument by largely ignoring it. After all, it leaves unanswered the question of how-despite the argument-P's action can nevertheless genuinely be thought of as free; and so why it's appropriate to characterise, in the first place, God's knowledge (or belief) as knowledge about what P will freely do. Moreover, although Davies is right in my view to emphasize the dependency of God's knowledge on what the agent in question will do; 7 if God's knowledge depends on John's action, then it will depend on his action quite independently of whether his action is free. Indeed, his knowledge will depend on John's action if his action isn't free just as much as if it is free. What this shows is that even if we do want to appeal to and emphasize the dependency of God's beliefs about our future actions on those actions, we still need to find an adequate response to the incompatibility argument. If we can find a satisfactory response to the incompatibility argument, we might well then say that God's belief isn't fully captured by premise (1)-but the satisfactory response needs to come first. So, in relation to the example concerning John, the question remains: doesn't it follow from the fact that God believed 20 years ago that he will go sailing tomorrow, that he isn't free in respect of his doing so?
For the time being, then, we should perhaps just accept premise (1) and move on to consider the features of God's beliefs in relation to our future actions thought to generate a problem for those of us who want to defend human freedom.

The Accidental Necessity of God's Belief
The claim in premise (2) is that the proposition God believed 20 years ago that John will go sailing tomorrow is accidentally necessary now (and at all future . The idea here is that a proposition reporting a past event is forever afterwards-and, in particular, from now onwards-fixed or unalterable. The claim that past events 8 are accidentally necessary, appeals to the principle of the necessity of the past. This is the idea that there is nothing that can be done to change what's happened in the past-a thought that seems to be closely tied in with the idea that there can be no backward causation (i.e. that effects can't precede their causes).
Now, in claiming that some past event is unalterable, we could mean one of two things. We could mean that there is nothing that can be done (from now onwards) to change the fact that the event occurred in the first place; or we could mean that there is nothing that can be done (from now onwards) to bring about a change in that past event. Thus, in the present discussion, we could mean that there is nothing that John can do (today or tomorrow) to change the fact that God believed 20 years ago that he will go sailing tomorrow; or we could mean that there is nothing John can do (today or tomorrow) to bring it about that God comes to hold the different belief that, for example, tomorrow he will stay indoors.
We might think, considering the second interpretation, that there is indeed nothing John can do to bring it about that God comes to hold a different belief; and we might think this simply because God's knowledge is infallible. After all, because God's knowledge is infallible, we can't make sense of him changing his belief because the implication of such a change is that he was previously in error.
On this second interpretation, then, God's belief would be accidentally necessary simply because his knowledge is infallible, and it wouldn't be at all important that the belief is held in the past. This, then, can't be what's meant when it's said that past events are accidentally necessary.
So-turning to the first interpretation-when we say that God's belief is accidentally necessary, we must mean that there is nothing John can do (today or tomorrow) to change the fact that God believed 20 years ago that he will go sailing tomorrow. To deny that God's belief is accidentally necessary in this sense amounts to claiming that there is something John can do to bring it about that God didn't have the belief that he had. Inspired by the work of William of (Ockham, 1983) (from whom we get the notion of accidental necessity), Ockhamists deny premise (2) and defend the claim that some propositions reporting God's past beliefs aren't accidentally necessary because some of them aren't wholly about the past. 9 This line of thought-that a proposition about the past has to be wholly about 8 In what follows, for convenience I apply the notion of accidental necessity both to events themselves and to the propositions that report them; even though the notion is meant, strictly speaking, to apply only to the latter. 9 Historically, philosophers attempting to explain the notion of the accidentally necessary past have appealed directly to a distinction between soft and hard facts about the past. A good example is in the work of Marilyn McCord Adams (Adams, 1967). However, I avoid this distinction in what follows, primarily because philosophers have struggled to spell it out in a satisfactory way. For example, John Fischer points out that on Adams' own criterion for something being a hard fact, there isn't anything that would count as such (Fischer, 1989 (Plantinga, 1986). In this respect, a contrast can be made between the proposition God believed 20 years ago that Eileen and John will get married in 2005; and our proposition God believed 20 years ago that John will go sailing tomorrow. The first proposition is wholly about the past: the belief is held in the past, and the belief is about a past event. Being wholly about the past, there is nothing that Eileen nor John can now do to bring it about that God didn't have that belief. By contrast, the proposition God believed 20 years ago that John will go sailing tomorrow is only partly about the past: the belief is held in the past, but the belief is about a future event. This might be thought to open up the possibility that John can remain indoors tomorrow, and so can bring it about that God didn't have the belief that he had. Thus, that God's belief isn't wholly about the past might be thought to open up the possibility of claiming that God's belief isn't accidentally necessary, after all.
Plantinga would put this line of thought in terms of the notion of power. In relation to the example being considered, he would say that despite the fact that God believed 20 years ago that John will go sailing tomorrow, John nevertheless has it within his power to stay indoors. Moreover, if he were to exercise this power and stay indoors, it would be false that he goes sailing tomorrow; and the past would therefore have been different. In particular, it wouldn't have been true 20 years ago that God believed that John will go sailing tomorrow. John's power over the past here-his power to stay indoors-is one that Plantinga would think of as a counterfactual power. It's counterfactual because in the actual world John will go sailing tomorrow (Plantinga, 1986: p. 257).
If John has counterfactual power over God's belief, this gives us a sense in which there is something John can do to bring it about that God didn't have the belief that he had. But it's important to realise that this doesn't commit us to think that John has causal power over the past. John doesn't have the power to cause it to be the case that God doesn't believe, 20 years ago, that he will go sailing tomorrow; but the thought is that he perhaps does have the power to do something (namely, to refrain from going sailing tomorrow) such that, if he were to do it, then God would never have believed, 20 years ago, that he will go sailing tomorrow. 10 In response to Plantinga, William Hasker attempts to establish that having counterfactual power over the past entails having causal power over it too. As most philosophers reject the idea that we can have causal power over the past 11 , the success of Hasker's argument here would be a serious problem for Plantinga's position. Hasker's argument appeals to certain "power entailment principles" which he borrows from Thomas Talbott (Talbott, 1986). One such princi- 10 Plantinga makes it quite clear that these two powers aren't the same, claiming-in terms of his own example involving Jones-that "…it was within Jones' power to do something such that if he had done it, then God would not have held a belief that in fact (in the actual world) He did hold. But by no stretch of the imagination does it follow that if Jones had done it, then it would have been true that God did hold a belief He didn't hold." (Plantinga, 1977: pp. 71-72) 11 George Mavrodes (Mavrodes, 1983) seems to be a notable exception. If (a) it is within S's power to bring it about that P is true and (b) it is within S's power to bring it about that P is false and (c) P entails Q and not-P entails not-Q, then it is within S's power to bring it about that Q is true (Talbott, 1986: p. 458;Hasker, 1989: p. 109).
For example, "S" might stand for "John", "P" for "John goes sailing tomorrow", and "Q" for "God believed 20 years ago that John will go sailing tomorrow". Now, if John has causal power over his going sailing tomorrow, then it's clear that he has counterfactual power over God's belief. PEP3 is meant to show that in the light of the logical equivalence between "John goes sailing tomorrow" and "God believed 20 years ago that John will go sailing tomorrow", John's causal power over his going sailing tomorrow (and so his counterfactual power over God's belief) would transfer over to God's belief; and so it's meant to show that he would also have causal power over God's belief.
In defence of PEP3, Hasker appeals to Talbott's claim that it "…seems not only true but obviously true. Where P and Q are logically equivalent, it could hardly be up to me whether or not P is true unless it were also up to me whether or not Q is true." (Talbott, 1986: p. 458) In reply to Hasker, Thomas Flint argues that "If two propositions are logically equivalent and I have power over the truth of one of them (i.e., its truth is up to me), then it does seem clear that the truth of the other one is within my power as well; what does not seem clear is that I need to have power in the same sense of "power" over the second as over the first. Suppose I have causal power over the truth of one of two logically equivalent propositions; is it not sufficient that I have counterfactual power over the other? Is that not enough for me to say that each of them is such that its truth is up to me?" 12 (Flint 1991: p. 240) I agree with Flint and think that we should reject Hasker's view that having counterfactual power over the past would entail having causal power over it too. 13 But there is, it seems to me, still the question of whether John even has 12 Hasker responds that the "…answer to this should be clear in the light of our previous discussion of 'bringing about.' On the one hand, power to bring about need not be causal power; on the other hand, the counterfactual dependency relation (and therefore also 'counterfactual power') is not 'enough for me to say that each of them is such that its truth is up to me'." (Hasker, 1989: p. 109) Flint quite rightly considers this response to be unsatisfactory (Flint, 1990: pp. 112-114). 13 It should be noted that in support of his claim that counterfactual power over the past implies causal power over the past, Hasker also appeals (Hasker, 1989: pp. 110-111) to a second power entailment principle "PEP4". His argument-an argument he once again gets from Talbott (Talbott, 1986: p. 460)-basically goes as follows: John refraining from going sailing tomorrow entails that God believed he will refrain from doing so. So, if it's not within John's power to bring it about that God believed he will refrain from going sailing tomorrow, then (given this entailment) John can have the power to refrain only if it's already true that God believed that he will refrain. But God didn't believe that John will refrain from going sailing tomorrow. On the contrary, he believed that John will go sailing. So, if John doesn't have the power to bring it about that God believed he will refrain from going sailing tomorrow, then John doesn't have the power to refrain. Therefore, if John does have the power to refrain from going sailing tomorrow, he also has the power to bring it about that God believed he will refrain. But it seems to me that this argument is susceptible to the same kind of objection as the one Flint directs at PEP3. events not wholly about the past aren't part of the accidentally necessary past-appeals to the existence of counterfactual powers over the past; the distinction between events that are, and those that aren't, wholly about the past, doesn't seem to be a sufficient basis for claiming that we have counterfactual powers of this kind.
In any event, we might take the view that all this talk about whether John has counterfactual power over the past isn't really relevant; because what's really needed for God's belief to not be part of the accidentally necessary past is for John to have causal power over the past. If we take this view-and think, along with most philosophers, that we don't have causal powers over the past-then we might well concede that God's belief is accidentally necessary, after all.
However, I want to end this section by considering an argument put forward by Trenton Merricks (Merricks, 2009(Merricks, , 2011) that calls into question the relevance-to the issue of our freedom-of our having causal powers over the past.

Underlying Merricks' argument is a particular view about the dependency of
God's beliefs on human actions-a view that, as we shall see, is an important aspect of my own response to the incompatibility argument. This view dates back at least as far as to Origen who claims that "…it will not be because God knows that an event will occur that it happens; but, because something is going to take place it is known by God before it happens." 14 (Origen, 2002: p. 90 Following Origen, God's belief a thousand years ago that Jones sits at t depends on what Jones does at t-in particular, it depends on his sitting at t. Moreover, we can say that God had that belief because Jones sits at t. But for Jones to have a choice about God's belief at t, it's not enough that God had that belief because Jones sits at t. What's additionally needed is that Jones freely sits at t. As Merricks puts it: "…because some of God's past beliefs depend… on an agent's future free actions, that agent thereby has a choice about God's having had those past beliefs." (Merricks, 2011: pp. 7-8) So, for Jones to have no such 14 In the earlier section on "God's Belief", we saw that Brian Davies appeals to this kind of view. 15 The numbering of this claim (as well as claim (0) below) comes from Merricks (Merricks, 2009: p. 44+51 16 Merricks recognises, however, that attempts to defend (6) which don't beg the question at issue are likely to appeal to claims such as that: (0) The past is appropriately necessary; and necessarily, if the past is appropriately necessary, then no one now has a choice about what the past was like, not even about which propositions were true in the past.
In attempting to defend (0), we might think that the past being appropriately necessary amounts to no one having causal power over (or to no one being able to change) the past. So, let's suppose that Jones doesn't have causal power over, and can't change, the past. Merricks' view is that neither of these things imply that Jones has no choice about what the past was like and in particular about what propositions were true in the past; for Jones having a choice about the past truth of a proposition (and in particular ones relating to God's past beliefs about his future actions) doesn't require him to be able to cause (nor change) the past. (Merricks, 2009: pp. 40-42) So he has a choice about God's past beliefs even if he's unable to do these things. Merricks' thought seems to be that given Origen's insight, all that's required for Jones to have a choice about God's past belief is that he is free; and that no non-question begging reason has been given for thinking that Jones lacks freedom.
But as it stands Merricks' argument doesn't seem to me to be satisfactory. To see what's wrong with it, let's suppose that Jones endorses the incompatibility argument. He might address Merricks as follows: "I accept that I don't have any causal power over (and can't change) the past. You claim that God believed a thousand years ago that I sit at t. Well, even if he had this belief because of what I will do at t, his belief is part of the past and as such I have no causal power over it and can't change it. But because his belief entails that I will sit at t, isn't it also true that I haven't got any causal power over, and can't change, my sitting at t? In other words, doesn't my lack of causal power and inability to change transfer over from God's belief to my future action-so that I'm not free with respect to my sitting at t?
You say that the fact that I have no causal power over and can't change the past doesn't imply that I have no choice about God's past belief that I will sit at t; and you seem to support this view by appealing to the claim (a "reasonable assumption", you say) that I'm free in respect of my sitting at t. But why should I accept your assumption that I'm free? Of course, before I took up Philosophy I used to think that I was free in respect of such things, but the incompatibility 16 This point is made by Merricks against what he calls the Divine Foreknowledge Argument which is an argument having (6) as its first premise (Merricks, 2009: pp. 51-53). See n.24 below. With this question-begging point in hand, Merricks feels entitled to make what he considers to be the reasonable assumption that Jones will freely sit at t. But it's important to realise that this assumption may not be a reasonable one to make when faced with an argument, like the incompatibility argument, which doesn't seem to beg the question. God's past belief; because it implies, in the way set out in the incompatibility argument, that I'm not free in respect of my sitting at t. Unless and until you adequately address the incompatibility argument, you aren't entitled to assume that I'm free with respect to my sitting at t." 17 So, returning to the example involving John going sailing tomorrow, it seems, in the light of these comments, that the main point that Merricks hasn't addressed-but needs to address-is this: if we think that John has no causal power over and can't change God's past belief-if we think that God's belief is accidentally necessary-doesn't the fact that it entails that John will go sailing tomorrow also mean that he has no causal power over, can't change, and so isn't free in respect of, his going sailing tomorrow? In other words, doesn't the accidental necessity of God's past belief transfer to John's action?

The Transfer of Accidental Necessity
Premise (3) of the incompatibility argument is the claim that: If a proposition p is accidentally necessary at some time t and p logically entails q, then q is accidentally necessary at t. This is known as the "transfer of necessity principle". 18 It's the principle that accidental necessity is closed under entailment: that a proposition (perhaps ordinarily thought of as merely contingently true) that is entailed by an accidentally necessary proposition, is itself accidentally necessary. But should we accept this principle and the role that it plays in the incompatibility argument? In particular, can we respond to the argument by claiming that even though God's belief that John will go sailing tomorrow is accidentally necessary (because John doesn't have causal power over it); John does nevertheless have the (causal) power to refrain from going sailing because its accidental necessity doesn't transfer to his action?
Borrowing an example from Alfred Freddoso (Freddoso, 1993: p. 105) 19 , consider God's belief that Peter will refrain from watching the game (on television) tonight. Applying the transfer of necessity principle here, we would say that because God's belief that Peter will refrain from watching the game tonight entails that Peter will indeed refrain from doing so, the accidental necessity attributed to God's belief transfers to Peter's action (where the notion of action being employed here is wide enough to include Peter refraining from doing something). 17 In not directly addressing the incompatibility argument, Merricks' discussion is reminiscent of the argument put forward by Brian Davies, considered in the earlier section on "God's Belief". 18 See (Zagzebski, 1997). The use of the word "transfer" in (3) is potentially misleading. When A transfers something, X, to B, typically A gives up X and as a result B acquires X. However, when we say that accidental necessity transfers from God's belief to John's action, we mean that John's action acquires an accidental necessity (that it perhaps otherwise wouldn't have had), but we don't mean that God's belief thereby loses its accidental necessity. 19 pp. 101-105 are of particular relevance in the discussion that follows. Freddoso's view is that-given that we can reasonably suppose that Peter at all times retains his general causal ability to, inter alia, get up, walk over to his television set and tune it to the appropriate channel-there isn't anything causally preventing him from doing so. In other words: "Peter's exercising the basic powers required to watch the game is fully compossible with all the other causal activity, including God's, that has ever occurred in the past or is anywhere occurring in the present." (Freddoso, 1993: p (Hasker, 1989: p. 143)-though he immediately goes on to admit that he doesn't pretend to know (nor does he feel any obligation to say) what that thing would be. Freddoso's argument is a response to what he calls Hasker's "logical preclusion" argument 21 -an argument which seems to be just an appeal to the transfer of necessity principle. Clearly of the view that accidental necessity doesn't transfer from God's belief to Peter's action, Freddoso claims that: "…even though God's belief about Peter is now part of the fixed past, as long as the causal prerequisites for free action are satisfied, Peter has the power to watch the game; therefore, when he, in fact, refrains from watching it, he does so freely." (p. 103) But is he right?
In assessing whether Peter can exercise his power to watch the game, let's consider whether he does so in at least some possible world relevantly similar to the actual world. Now, if-as Freddoso is willing to accept-God's belief is part of the unalterable, accidentally necessary past, then the possible worlds in question will presumably all be ones in which God believed that Peter will refrain from watching the game. But since God's belief that Peter will refrain from watching the game entails that he will indeed refrain from watching it, in none of these worlds is it the case that Peter will watch the game (for there would otherwise be a world in which God has a false belief). But then how can it be maintained that Peter can exercise his power to watch the game? (And so how can it be maintained that he has this power?) How can it be denied that the accidental necessity of God's belief would transfer to Peter's action? Faced with this line of thought, the mere fact that we admittedly can't spell out precisely what causally prevents Peter from watching the game doesn't seem to be very important.
Freddoso admits that "…there is no possible world in which God believed 20 He notes that this assumption is also made by Hasker (Hasker, 1989: p. 141 Open Journal of Philosophy from eternity that Peter will not watch the game and in which Peter watches the game nonetheless." But he immediately goes on to deny that "…there is any more interesting sense in which it is impossible for Peter to exercise his power to watch the game." (p. 105) Now, we can agree with Freddoso that all possible worlds in which God believed from eternity that Peter won't watch the game are ones in which Peter doesn't watch the game; and we can agree that this doesn't of itself provide us with an interesting sense in which it's impossible for Peter to exercise his power to watch the game. After all, that fact is consistent with there being possible worlds in which Peter will watch the game (these will of course be worlds in which God believed from eternity that he will). But Freddoso seems to be ignoring here the accidental necessity that he accepts is attributable to God's belief that Peter won't watch the game. When we take this accidental necessity into account, we should realise that no possible world in which God didn't have this belief is relevantly similar to the actual world; and this does seem to give us a very interesting sense in which it's impossible for Peter to exercise his power to watch the game. After all, to repeat, all possible worlds in which God believed from eternity that Peter won't watch the game are ones in which Peter doesn't watch the game; and no possible world in which God didn't have this belief is relevantly similar to the actual world (and so no such world should be taken into consideration when assessing a "can" claim). Freddoso claims that he doesn't "…have even a faint idea of how logical preclusion by divine foreknowledge…is supposed to result in an agent's inability to do something." (pp. 104-105) But-echoing the earlier quote from Edwards-in response, we might admit that we don't understand how logical preclusion by divine foreknowledge results in an agent's inability to do something (we might admit that we don't understand the exact causal mechanism) but be nevertheless convinced-for the reasons just given-that it would. So, if we think or concede that God's belief is accidentally necessary, does the transfer of necessity principle force us to think that divine foreknowledge is, after all, incompatible with free human action? Even if we consider Freddoso's argument to be unsatisfactory, I think that we can nevertheless resist this conclusion.
Returning now to the example involving John, we have up till now proceeded on the assumption that it's coherent to think of there being a transfer of accidental necessity from God's belief to John's action; and we've focused just on the more factual question of whether accidental necessity does in fact transfer in this way. But perhaps we can give reasons for thinking that a transfer of this kind would be incoherent-that it doesn't even make sense to think of there being a transfer of accidental necessity from God's belief to John's action.
The transfer of necessity principle, premise (3), claims that if a proposition p is accidentally necessary at some time t and p logically entails q, then q is accidentally necessary at t. But what if p and q are logically equivalent? God's belief that John will go sailing tomorrow (p) entails-given that God's knowledge is infallible-that John will indeed go sailing tomorrow (q); and John going sailing tomorrow entails-given God's omniscience-that God believed that he will. So, If p entails q then it's natural to think in terms of the possibility of any accidental necessity p has being transferred to q. But what if p both entails and is entailed by q? In particular, does the fact that q entails p give us grounds for thinking that any accidental necessity attributable to p couldn't transfer to q?
One good way of addressing this question is to think in terms of the notion of explanation and to ask whether p explains q or q explains p. If p explains q rather than the other way round-if, for example, q is true because p is truethen it seems reasonable to think in terms of the possibility of any accidental necessity attributable to p being transferred to q. But if, on the other hand, q explains p-if, for example, p is true because q is true-then it's perhaps difficult to see how any accidental necessity attributable to p could possibly be transferred to q. We should ask, then, whether God's belief explains John's action, or John's action explains God's belief. Should we, for example, say that John will go sailing tomorrow because God believed that he will; or that God had the belief because John will go sailing? (In terms of the notion of truth, is it true that John will go sailing tomorrow because it's true that God believed that he will; or is it true that God believed John will go sailing tomorrow because it's true that he will?) The question for us, then, is whether we can respond to the incompatibility argument's appeal to the transfer of necessity principle by successfully defending the view that it's John's action that explains God's belief. We saw towards the end of the previous section that this idea-that God has his belief because of the way the world is, rather than vice versa-dates back to Origen, and was an insight central to Merricks' discussion. But we can imagine someone claiming that even though our beliefs are true (when they are) because of how the world is; it doesn't follow that God's beliefs are also true (as they always are) because of how the world is. We perhaps need to say more, then, in defence of the view that God's beliefs are explained by our actions rather than the other way round.
One way to defend the claim that God's belief is explained by John's action is to appeal to the view that God's foreknowledge is appropriately described as simple; for what seems to be a central feature of the simple foreknowledge view is that God believed that John will go sailing at t (for example, tomorrow) because he sees John going sailing at t. 22 We might model this by thinking of God's belief as coming from his somehow seeing the future; that it's his perception 20 22 Origen's insight might be combined with other views about God's foreknowledge. For example, Luis de Molina agrees with Origen in taking God's beliefs about the future to depend on what will happen, rather than the other way around (Molina, 1988: p. 179). One such other view might then be the "middle knowledge" view put forward by Molina and others. Although a full investigation of the merits of Molinism is unfortunately beyond the scope of the present paper, it would be interesting to explore whether the kind of response I make to the incompatibility argument is one that (no doubt with some adaptations) could also be adopted by (at least some) Molinists. Open Journal of Philosophy years ago of what John will do tomorrow-it's his seeing what happens in the future-that gives rise then to his belief. 23 On this view, as Nelson Pike puts it, "…God foreknows things in that he "sees them as if actually placed before him". …God does not predict the future on the basis of his knowledge of presently existing circumstances and causal laws. Like the crystal-ball gazer, God's foreknowledge is not the outcome of a prediction based on evidence. It is the result of a "vision" involving nothing in the way of inference or calculation." (Pike, 1970: pp. 64-65) On the simple foreknowledge view, then, God believed that John will go sailing tomorrow because he sees that it's true that John will go sailing tomorrow (that John will go sailing tomorrow isn't true because God had the belief). What explains God's belief (given his omniscience) is John's action; John's action isn't explained by God's belief. On this view, then, when we make the conditional claim that if God believed John will go sailing tomorrow, then John will go sailing tomorrow, we should be careful to recognise this as merely a reflection of the logical entailment between the two propositions; and shouldn't allow it to mislead us into thinking that John going sailing tomorrow is in any way explained by God's belief.
Let's return now to the question of whether accidental necessity travels from God's belief to John's action. If John's action was explained by God's belief, then we could see how God's belief might plausibly be thought to give something (for example, accidental necessity) to it. But if we instead accept-by appealing, for example, to the simple foreknowledge view-that John's action explains God's belief, could there really be a transfer of accidental necessity from God's belief to John's action, just because of the logical entailment of John's action by God's belief? If-in the way indicated in this brief account of simple foreknowledge-God's belief comes from John's action, how could it give anything to John's action? Returning to the wording of premise (3), what these thoughts indicate is that the mere logical entailment of q by p isn't sufficient to ground the transfer of accidental necessity from p to q; and that whether accidental necessity is indeed transferred depends on the specific content of p and q, and in particular on whether q is true because p is true.
In attempting to apply the transfer of necessity principle, the incompatibility argument appeals, in premise (4), to the entailment of John's action by God's belief. But the fact that God has certain beliefs about the world because of how the world is (rather than vice versa) calls this appeal into question. If we think that God's belief depends on the world, what could possibly be the relevance of that logical entailment? 24 23 We shall see in the next section that this is just one of the ways in which we can model God's simple foreknowledge. 24 It's worth at this point returning very briefly to our discussion of Merricks (Merricks, 2009(Merricks, , 2011. Merricks' argument is in part a response to the Divine Foreknowledge Argument (Merricks, 2009: pp. 51-52): (6) Jones has no choice about: God believed that Jones sits at t a thousand years ago. (7) Necessarily, if God believed that Jones sits at t a thousand years ago, then Jones sits at time t. Therefore, (3) Jones has no choice about: Jones's sitting at time t. We have already considered Merricks' objection to (6). He seems-perhaps because he sees this objection as decisive-to completely ignore (7). But I think that as someone endorsing Origen's insight he should have made the point that (7), though true, is irrelevant in that it doesn't (in combination with (6) It seems to me, then, that by appealing to Origen's insight and the view that God's foreknowledge is simple, we can deny that any accidental necessity attributable to God's belief could coherently be thought to transfer to John's action.
But the ultimate success of this response will depend on whether the view that God's foreknowledge is simple is itself acceptable. It's therefore incumbent on me to address some of the objections that this view has faced.

Simple Foreknowledge: Some Problems
William Hasker considers several objections to the view that God's foreknowledge is simple. He begins by outlining the simple foreknowledge view as one according to which God (often thought of as having "… 'direct vision' of the future as if in a crystal ball or a telescope.") "…directly knows the actual future event…" in virtue of the fact that his "…belief about the matter is somehow brought about by the future event itself." His initial objection is that this view "…of course involves something very much like retroactive causation..." (Hasker, 1989: p. 56) The problem for the advocate of simple foreknowledge is that if she is committed to thinking that backward (retroactive) causation is possible, and there are grounds for thinking that the very notion of backward causation is incoherent, then the simple foreknowledge view itself will also be incoherent. A further point, made by Jonathan Edwards, is that even if we think that the appeal to retroactive causation is coherent, it's difficult to see how this could leave room for John's action being free. After all, how could God's already existing belief that John will go sailing tomorrow be the effect of John's future action and yet that future action not occur? (Edwards, 1969: p. 126) But is the advocate of simple foreknowledge committed to thinking that backward causation is possible? It's interesting to note that Hasker is at this stage quite willing to think of the notion of bringing about (in the claim that God's be- Moreover, and relatedly, this simplistic view seems to ignore the fact that my seeing and my believing aren't things that passively happen to me, but are things that I seem to be actively involved in. None of this is to deny that we could express the relation between the book being on my desk and my belief counterfactually, by saying that if it wasn't on my desk, I wouldn't believe it was there. But this is a long way from saying that there is a causal relation between these things. For similar reasons, it seems to me that if we think of God as having a direct vision of the future, we needn't think of this as involving any kind of causation. 25 We have seen that at other points in his discussion, he's keen to emphasise that these two notions are very different. See n. 12 above. have to appeal to a vision 20 years ago? Why can't she instead say that God believed 20 years ago that John will go sailing at t because he sees at t John going sailing then? On this different model, God will see at t John going sailing and because of this, he will believe at t that John is going sailing then. So, God knows-will know-at t that John goes sailing at t. It might be wondered how he then has this knowledge at the earlier time, 20 years ago; but at this point, a proponent of the simple foreknowledge view can (why not?) appeal to features of God's omniscience. An omniscient God doesn't have knowledge at one time and then lose it and so not have it at another, later, time. Nor, importantly, does an omniscient God lack knowledge at one time and gain it at a later time. Rather, the knowledge he has, is knowledge he has at all times. So, since on this alternative model God will know at t-because he sees at t-that John goes sailing at t; he must be thought of as also knowing this at all other times and in particular 20 years ago. On this model, then, God knows 20 years ago that John will go sailing at t even though, as we might say, the evidence for his knowledge isn't before him at that time, 20 years ago. Importantly, on this model God doesn't need (and doesn't have) a vision 20 years ago into the future at t; and so it's no objection that the future time, t, doesn't exist 20 years ago. So, which model of simple foreknowledge should we adopt? Although in what follows I will generally use the language of the first model, I don't think we have to decide between the two. We can simply say that if an advocate of simple foreknowledge finds compelling the objection to the theory of time underlying the first model, she can instead think in terms of the second.
But there is yet another Hasker-inspired objection to the simple foreknowledge view that needs to be considered, one related to the issue of providence (Hasker, 1989: pp. 57-63). Although it's generally reasonably assumed that God's foreknowledge can make an important contribution to his providential action in the world, Hasker argues that the simple foreknowledge view renders God's There are a couple of related worries about how God could use his simple foreknowledge in acting providentially. The first worry is this. How can an event be the (direct or indirect) product of God's act of intervention if he intervenes already knowing that the event will occur-and he knows this because he has seen that it will indeed occur? There is a kind of circle here, and a sense in which it's too late for the event to be the product of his act of intervention. A second worry is that if God knows that an event will occur-and he knows this because he has seen that it will-then how could he make use of this foreknowledge if he decides, for example, to bring it about that it doesn't occur? Here, there's a sense in which it's too late for the event not to occur. As Hasker succinctly puts it: "In the logical order of dependence of events, one might say, by the "time" God knows something will happen, it is "too late" either to bring about its happening or to prevent it from happening." (p. 58) But is this a sufficient basis for saying that divine providence can't be based on simple foreknowledge-that simple foreknowledge would be providentially useless? Hasker's discussion highlights the problems that can occur where there is an attempt at providential control over an event the occurrence of which is foreknown. One and the same event can't be the object both of foreknowledge and of providential control. 26 So, there do seem to be cases where God wouldn't be able to make providential use of his simple foreknowledge. There are, we might say, constraints on its providential use. But is this really enough to justify the claim that simple foreknowledge is in all cases providentially useless?
David Hunt (Hunt, 1993) makes a number of important points on this issue.
Firstly, he recognises that it's very easy to think that because God has complete foreknowledge, any attempt he makes at providential control will inevitably be over an event the occurrence of which is foreknown. He emphasises, however, that the problem being addressed depends on how foreknowledge is used, not on how much is possessed. So, the problem doesn't arise just because God has complete foreknowledge. Rather, the problem arises where, for example, there is an attempt at providential control over an event the occurrence of which is foreknown, where that foreknowledge is used in the providential decision making. Secondly, there's nothing that requires God in acting providentially to attempt the providential use of an item of simple foreknowledge in such a way that the problem arises. As he puts it: "The problem requires that foreknowledge be used in quite specific ways, but we have failed to turn up any feature of God's providential situation that would force Him to comply with this requirement." (Hunt, 1993: p. 407) Thirdly-recognising that the danger is of using simple foreknowledge in such a way that circles of dependence are generated-the completeness of God's simple foreknowledge can be thought of as giving him opportunities for providential action not available to those whose foreknowledge is 26 Hasker considers that because God knows any event's "complete causal antecedents", similar problems arise where the event over which there is an attempt at providential control isn't the event foreknown, but is instead part of the causal stream leading to the foreknown event (p. 61). items that won't generate circles of this kind. Fourthly, even though God's providential control is traditionally thought of as extending to all events in the universe, if we really are free we shouldn't think of this control as extending to our actions. We therefore shouldn't attempt to defend a traditional conception of providence but should instead focus on the issue of whether God can be said to exercise maximal providential control where this refers to "…the strongest control compatible with there being free agents other than God." (p. 406) Fifthly, and lastly, "…since maximal providential control does not encompass all foreknown events (in particular, it excludes events which are exercises of free will), this leaves open the possibility that foreknowledge of these uncontrolled events might be utilized in the production of maximal providential control in such a way that circularity is avoided…" (p. 406) These thoughts lead Hunt to tentatively endorse the thesis that it's possible for God's complete foreknowledge to contribute to maximal providential control. We might summarise this discussion by saying that Hasker shows us what God needs to avoid doing if his simple foreknowledge is to be of providential use; and Hunt indicates to us just how much providential control God, if he avoids these things, can nevertheless have. It's clear that we have no reason to accept Hasker's claim that simple foreknowledge would be providentially useless.
In the previous section, I claimed to have shown-by appealing to Origen's insight and to the view that God's foreknowledge is simple-that any accidental necessity attributable to God's belief couldn't coherently be thought to transfer to John's action; and I ended that section by recognising that the ultimate success of this response depended on the acceptability of the view that God's foreknowledge is simple. On the basis of the discussion in the current section, I think that we can now say that the view that God's foreknowledge is simple is a defensible view. Putting these two sections together, then, we can say that an advocate of simple foreknowledge is entitled to respond to the incompatibility argument by claiming that accidental necessity doesn't-couldn't coherently be thought to-transfer from God's belief to John's action.
We have now what seems to me to be a satisfactory response to the incompatibility argument. But there is a sense in which in accepting that God believed 20 years ago that John will go sailing tomorrow and that this belief is accidentally necessary-and instead focusing on (indeed, arguing against) the possibility of transfer-this response doesn't go far enough. If we endorse Origen's insight and the view that God's foreknowledge is simple, should we really just be accepting these things? Shouldn't we instead also be questioning the claims that God believed 20 years ago that John will go sailing tomorrow and that this belief is accidentally necessary?

God's Belief and Its Accidental Necessity, Revisited
Consider the premise of the incompatibility argument that God believed 20 The false impression that God's belief is independent of John's action might lead us to also see the accidental necessity of his belief as something independent of his action and so as something that similarly needs to be overcome in order to establish his freedom. Again, we might then also feel obliged to say how John can be free to stay indoors tomorrow despite the fact that God's belief that he will go sailing is accidentally necessary. But we can imagine John continuing: "Once God has formed his belief (whatever it is), many philosophers defending the incompatibility argument think that the appropriate question to then ask is whether there's anything I can do to change it (for example, to bring it about that God didn't have the belief in the first place). Answering this question, they think that because God's belief is past, I can't change it and that it's therefore part of the accidentally necessary past. But this question of whether there's anything I can do to change God's belief is misguided; and by appealing in their answer to the fact that God's belief is past, they are appealing to a feature of his belief that's irrelevant to the issue of its accidental necessity. Let me explain.
Of course, I don't think I can change God's belief (in the above or in any other sense). But this isn't because I lack any power in any sense relevant to my freedom, and isn't because God's belief is (part of the accidentally necessary) past. I'm not denying that there's a sense in which God's belief is accidentally necessary. But I think that its accidental necessity-the fact that I can't change it-is a direct consequence of the fact that it tracks my decision, and has nothing to do with the fact that it's a belief held in the past. After all, God's belief held 20 years in the future that I will go sailing tomorrow also tracks my decision; and as such is just as unchangeable as the same belief held 20 years in the past. If God's belief tracks my decision, then I can't change it; and this is the case whether his belief is held in the past, present or future.
Finally, on the issue of counterfactual power, I recognise that some philoso-29 This isn't to say that God's belief is indeterminate. God either believed that John will go sailing tomorrow, or he believed that John will stay indoors tomorrow. He had one of these beliefs. But though the incompatibility argument makes the assumption that we know which belief he had; we don't know which one he had and won't know which one, until John makes his choice tomorrow.

Conclusion
To summarise, we shouldn't allow the use of God's belief as a first premise to give us the false impression that it exists independently of John's action. Recognising that God doesn't have a belief about John's action which is independent of what John decides to do, we should be suspicious of the incompatibility argument's use of God's belief as a first premise. In particular, we shouldn't feel obliged to ask how, despite God's belief, John can nevertheless be free. That God's belief depends on John's action does give us a sense in which God's belief is accidentally necessary, but to be accidentally necessary in this sense shouldn't be thought of as a problem but as merely a reflection of this dependency-and in any event we shouldn't think of the accidental necessity attributable to God's belief as something that could coherently be thought to transfer to John's action. So, neither God's belief nor its accidental necessity, should be seen as a barrier to John's freedom. If we think that God's beliefs about our future actions depend on our actions in the way envisaged by Origen, and adopt a simple foreknowledge view of how his beliefs come about; then we can say that his beliefs are, despite the incompatibility argument, perfectly compatible with our freedom.