Prominent Christian philosopher, apologist, and analytic theologian William Lane Craig is in the process of releasing his magnum opus, a 5-volume Systematic Philosophical Theology (Wiley Blackwell, 2025–2026). Volume 2a (released in 2025) focuses on God’s attributes and includes a 100-page chapter on omniscience. Now aged 76, Craig presumably intends this chapter to be the crowning summary of his nearly 40-year publication record on the topic of divine omniscience and foreknowledge.
In this 4-part blog series I review and critique Craig’s omniscience chapter. The chapter is divided into four main sections—§1 Biblical data, §2 Omniscience, §3 Foreknowledge, and §4 Middle Knowledge—which correspond to the four parts of this blog series. This post is on §3. For my previous post on §1 and §2, see here and here.
Part 3. Foreknowledge (pp. 215–251)
In this section, Craig focuses on two main topics: (1) philosophical arguments for affirming exhaustively definite foreknowledge (EDF) and (2) the compatibility of EDF and human freedom.
But first, Craig introduces the section with an egregious strawman characterization of open theism (which he calls “revisionism”) when he says that on open theism God is “ignorant” in that He “lacks both hypothetical knowledge of how things would go if he were to act in certain ways as well as foreknowledge of how things will in fact go” and so “can neither predict nor plan the course of future events” (p. 215–216). Both claims are completely false, seeing as open theists can and do affirm that God has hypothetical and categorical knowledge of the future. We just don’t think that would bes fully characterize the space of possible futures nor will bes the entire actual future. Moreover, all open theists believe God does extensive, if not exhaustive, contingency planning. Hence (pace Craig) God both predicts and plans for future events and does so while deliberately not ordaining all creaturely events. I could fisk the paragraph further, but let’s dive into the main topic instead.
(1) Philosophical arguments for EDF
(1.a) Conceptual preliminaries
Before looking specifically at Craig’s text, I need to define some terms and introduce some distinctions. These are not definitions or distinctions that Craig makes. Had he made them, his discussion would have been much more accurate.
First, when Craig speaks of “divine foreknowledge” he typically means exhaustive foreknowledge. That is, he’s not thinking of God’s foreknowing this or that future event but “the future” in its entirety. My readers may recall, however, that in my first post in this series I noted that there are two different ways to conceive of “the future”: as a LINE or as a TREE. Open futurists like myself conceive of the future as a branching TREE of possibilities. There are multiple possible futures, multiple possible complete extensions of the actual past and present, and none of them is privileged with respect to actuality. For open futurism there simply is no such thing as a unique actual future (UAF). The future is objectively open-ended and therefore known by God as such. In contrast, settled futurists like Craig conceive of the future as a LINE. So conceived, “the future” refers to the UAF, that supposedly unique and complete linear extension of the actual past and present that, in its entirety, shall eventually come to pass. For God to foreknow this sort of future is for God to have not merely foreknowledge but exhaustively definite foreknowledge (EDF), because it is always wholly definite how reality shall be at every future moment. Mere foreknowledge, as I use the term, is neutral between TREE and LINE. Both open futurists and settled futurists can therefore agree that God has exhaustive foreknowledge of “the future.” EDF, however, is not neutral. It is unambiguously pro-LINE.
Second, Craig speaks interchangeably of “future contingent propositions” and “propositions about future contingents,” but he never carefully defines what a future contingent is or what sorts of propositions he has in mind. This is quite problematic given the centrality of these notions to his arguments. Here’s how I think these concepts should be understood. In the first place, then, a future contingent is a causally contingent event-type, that is, a type of event both the future occurrence and non-occurrence of which are now causally possible in relation to whatever causal constraints (present conditions, laws of nature, etc.) are in effect. By a future contingent proposition (FCP) I mean a proposition about one or more future contingents. More exactly, an FCP says of one or more future contingents that it shall/will/might etc. come to pass. For example, <There will be a sea battle tomorrow> and <There could be a sea battle tomorrow> are both FCPs provided that the occurrence of a sea battle tomorrow is (still) causally contingent. Suppose that’s the case. There is an important further distinction between FCPs that depict the future as determinate in some respect, and ones that depict the future as indeterminate. For example, <There will be a sea battle tomorrow> says that the future is settled with respect to the occurrence of a sea battle tomorrow whereas <There will probably be a sea battle tomorrow> leaves it an open question whether a sea battle occurs tomorrow. I call FCPs like the former settled FCPs (SFCPs) and FCPs like the latter open FCPs (OFCPs). SFCPs are typically expressed in terms of what either “will” or “will not” happen, but can also (though sometimes awkwardly) be expressed in terms of what “does” or “does not” happen in the future. OFCPs are typically expressed in terms of what “might and might not,” “could,” “will probably,” or “will probably not” happen.
When Craig speaks of FCPs it is exclusively SFCPs that he has in mind. That there could even be OFCPs is a thought that seemingly never occurs to him, though perhaps the reason he doesn’t bring them up at all is because he doesn’t see them as relevant for divine foreknowledge. For him expressions of indeterminacy about the future aren’t objective claims about the future but rather subjective claims about the speaker’s state of mind concerning the future. “Could,” “might and might not,” “will probably,” etc. are therefore, for Craig, merely epistemic. They convey a speaker’s uncertainty about a future event and not objective open-endedness or indeterminacy. Because Craig never brings the issue up so as to argue that there are no objectively true OFCPs, his arguments systematically beg the question against open futurism.
(1.b) The argument from omniscience
Craig’s central philosophical argument for EDF is the argument from omniscience (AFO). Streamlining his presentation on p. 216 a bit and applying my distinctions from above, it runs as follows:
- God is omniscient.
- If God is omniscient, then God knows all truths.
- There are true SFCPs.
- Therefore, God knows true SFCPs. (from 1–3)
This argument is clearly valid. Craig takes premise (1) for granted as entailed by perfect being theology (p. 216). Premise (2) follows from his propositional definition of omniscience, namely,
(Oprop) S is omniscient =def. For all propositions, p, if p [is true], then S knows that p and does not believe that ~p. (p. 203)
Craig does qualify premise (2), saying that it is true “modulo heirarchical concerns” (p. 217). See part two of this series for an explanation of why he adds this qualifier. It doesn’t affect his arguments in this section.
Premise (3) is the lynchpin. If there are true SFCPs, and enough of them to specify a unique actual future, then settled futurism follows and any propositionally omniscient God must have EDF. Craig, unfortunately, doesn’t frame his argument in terms of SFCPs but rather FCPs because he question-beggingly takes the two groups to be coextensive, something no open futurist would grant. Because of this his versions of (3) and (4) just say that there are true FCPs and that God knows them, claims that an open futurist can happily accept. (For the open futurist, there are true FCPs, but these are all OFCPs, not SFCPs.) My adjustment of Craig’s argument at this point is therefore necessary to ensure a genuine clash with open futurism, a position Craig clearly wants to refute.
(1.b.1) Denying premise (2)
The first response to the AFO that Craig considers is a denial of premise (2) in favor of a qualified definition of omniscience, one that allows for the possibility of truths that God does not and cannot know:
(Oprop*) S is omniscient =def. For all propositions, p, if p [is true], then S knows that p and does not believe that ~p unless p is not possibly knowable by S. (see pp. 217–218)
Craig complains that this is an unacceptably “cooked” definition of omniscience, one that serves only to protect the label of “omniscience” so that it can be applied to a “cognitively limited” deity (p. 218). This charge is overstated, however, for if as Craig believes there is no difference between all truths and all-truths-knowable-by-God, then Oprop* safely reduces to Oprop.
In any case, Craig next argues that there can be no divinely unknowable truths by asking what could be the “intrinsic difference” between a truth God can know and a truth God cannot know (p. 218). His assumption seems to be that if there is any such difference it must be one that is intrinsic to the truths in question and not a difference in, say, God’s access to those truths. I doubt this assumption is warranted. Some truths (i.e., analytic ones) can be known simply by considering the propositions in questions (e.g., <A horse is a horse>), whereas others (i.e., synthetic ones) can be known only by acquaintance with something else (e.g., <Water feels wet>). The access requirements in the two cases are different.
Nevertheless, Craig asks, “what more is needed” for a truth to be knowable by God beyond it’s simply being true? Here he makes a huge mistake. He says that truth is an “intrinsic property” of a proposition (p. 218). This is a shocking claim for a philosopher like Craig who purports to affirm a correspondence theory of truth (CTT) (see p. 221). CTT says that truth consists in a relation of correspondence between a proposition (a truthbearer) and something in reality (a truthmaker or “fact”) that grounds or makes that proposition true. According to CTT, therefore, truth is a relational or extrinsic property, not an intrinsic property. The right answer to the “what more is needed” question is simply acquaintance with a truthmaker. God has de dicto or propositional knowledge of all truths because God has exhaustive de re knowledge (i.e., perfect acquaintance) with all truthmakers, i.e., all of reality. Craig can’t give this answer, however, because, as I observe in my previous post in this series, he is largely blind to de re knowledge. The result is that, while paying lip service to CTT, he winds up with what is, in effect, a deflationary conception of truth, one where the “facts” that truths correspond to are not substantively different from the truths themselves. This is why, in his thinking, truth can be an “intrinsic property.” I’ll say more on this below.
Craig imagines that someone who rejects premise (2) might argue that it is “logically impossible” for God to know any FCPs. As he puts it, “The revisionist [i.e., open theist] claims that for any future-tense proposition p it is impossible that God know p and p be contingently true. Therefore, he reasons, if p is contingently true, it is not possible that God knows p” (p. 219). He then proceeds to argue that this is an invalid inference. All that follows, he says, is merely that God “does not know p,” not that God cannot know p. On this last point I agree with him. The inference is invalid. But I disagree with the accuracy of his attributions. Which “revisionists” have made this claim? Craig doesn’t cite anyone. Nor is the claim that “for any future-tense proposition p it is impossible that God know p and p be contingently true” even plausible. One obvious counterexample is <There will probably be a sea battle tomorrow>. This is future-tense proposition. It is, we may suppose, contingently true. And God can know it by virtue of His de re acquaintance with the world as it currently is. If conditions now are strongly tending toward a sea battle tomorrow, and God has no intentions of intervening to prevent such, then it will probably happen. Craig’s criticism of this inference, therefore, is a straw man. There is no compelling reason for an open theist to endorse this particular incompatibility claim.
Moreover, Craig sloppily transitions from knowing truths about future contingents to knowing propositions that are “contingently true.” These are not the same thing. A truth about a future contingent is a truth about a causally contingent event-type, a type of event that, relative to how things are right now, both causally might happen and causally might not happen. For a proposition to be “contingently true,” in contrast, is for it either to have once been possible for it not to have been true or for it still to be possible for it to become untrue. What’s contingent here is the proposition’s truth, not whether an event-type described by the proposition comes to pass in the future. In short, Craig conflates truth-contingency with event-contingency. The event of a proposition’s coming-to-be true can be a future contingency, but its being true cannot be a future contingency unless its becoming untrue is also a future contingency.
(1.b.2) Denying premise (3)
At this point Craig deems premise (2) of the AFO sufficiently established and shifts his focus to premise (3). Open theists unwilling to deny God’s propositional omniscience “have no choice in the end but to deny” premise (3) (p. 219). This means, he says, that “they must deny that any future-tense, contingent statements about events in all but the immediate future” are true (p. 220). This is confusingly stated. It’s not the contingency of statements that’s at issue, but the causal contingency of future event-types and the implications of that for the truth values of SFCPs.
So, how might one seek to deny that there are any true SFCPs?
Craig suggests that open theists might try to deny the principle of bivalence (PB). This principle says that every proposition has exactly one of these two truth values: true, false. But, says Craig, the idea that bivalence fails for SFCPs is “difficult plausibly to maintain” (p. 220). He supports this claim by arguing, first, that “there is no good reason” to think PB fails for SFCPs (p. 220); second, that “there are several good reasons” to think PB holds for SFCPs (p. 221); and third, that denial of PB for SFCPs has “absurd consequences” (p. 223). Let’s look at each branch of this argument in turn.
(1.b.2.i) Is there no good reason to deny PB for SFCPs?
According to Craig, the only reason one might think PB fails for SFCPs is because “the future does not exist” and so “there is nothing for future-tense propositions to correspond to or fail to correspond to” (p. 220). Hence, allegedly, SFCPs are neither true nor false. In response Craig thinks those who push this line of argument “misunderstand the concept of truth as correspondence” (p. 221). According to him, CTT “holds merely that a proposition is true if and only if what it states to be the case really is the case” (p. 221). So, in order for an SFCP to be true, “all that is required is that when the moment in question arrives, the present-tense version of the proposition will be true” (p. 221, my emphasis). In general, “[a] future-tense proposition is true if matters turn out as the proposition predicts, and false if matters fail to turn out as the proposition predicts” (p. 221). On Craig’s view then, for <There will be a sea battle tomorrow> to be true today there doesn’t have to exist a future sea battle or any future reality at all. Instead, it must simply be the case when it’s tomorrow that a sea battle is happening. To put the thought more generically, the idea is that <Event E will occur at T2> is true at T1 (and at all previous times) just in case when it’s T2, E occurs. How the world is at all other times is irrelevant.
I think Craig is quite wrong about this. Every tensed proposition locates events in relation to a contextually specified present moment, a now. That’s just what it means for a proposition to be tensed. (A tenseless proposition, in contrast, lacks any explicit or implied reference to a now.) Hence, every tensed proposition explicitly or implicitly makes a claim about how things are now. <Caesar crossed the Rubicon> does not merely say that at some time or other Caesar crosses the Rubicon. It also says that the time when he did that is prior to now. Now is a post-crossing moment. Likewise, <There will be a sea battle tomorrow> does not merely say that tomorrow is a sea-battle day. It also says that now (today) is a pre-sea-battle day. So, if we apply CTT to tensed propositions then we need to consider not merely what happens at the past or future times specified, but also whether things are now such that the past/future was/will be that way. What about present reality makes it such that Caesar crossed the Rubicon? What about today makes it such that a sea battle happens tomorrow? What this means for future-tensed propositions in general terms is that <E will occur at T2> is true at T1 only if at T1 E’s occurrence at T2 is determinately settled. But if E’s occurrence at T2 is (as of T1) a future contingent, then by definition E’s occurrence at T2 is not determinately settled, for whether E occurs at T2 or not remains an open question. Therefore, if E’s occurrence at T2 is a future contingent then <E will occur at T2> is not true at T1.
Now, the above isn’t (by itself) a reason for thinking that PB fails for SFCPs, but it is a reason for thinking that no SFCPs are true. For if any are true, then they implicitly claim that reality is now determinately settled with respect to some future contingent event. But for those events to be future contingents is precisely for reality not to be now determinately settled in that respect. Accordingly, Craig’s first line of attack fails to push back the denial of premise (3).
(1.b.2.ii) Are there good reasons to affirm PB for SFCPs?
Craig’s second line of attack is that there are several good reasons to think PB holds for SFCPs. I’ll look at these in a sec, but it’s worth pointing out initially that these reasons, however cogent, are irrelevant. Once we grant the existence of OFCPs, it becomes possible to affirm PB while also denying that there are any true SFCPs. We can simply say, with Patrick Todd, that SFCPs are all “all false” whenever certain corresponding OFCPs are true.
With that in mind, here are Craig’s three reasons for thinking that PB holds for SFCPs:
- The same facts that guarantee the truth or falsity of present- and past-tense propositions also guarantee the truth or falsity of future-tense propositions. (p. 222) Thus, for example, if it rains on Tuesday, then just as <It rained on Tuesday> is true on Wednesday (and all subsequent days), so also <It will rain on Tuesday> is true on Monday (and all previous days). The “same fact” (i.e., Tuesday’s rain) that makes <It is now raining> true on Tuesday, also makes past-tensed propositions about Tuesday’s rain true afterwards and future-tensed propositions about it true beforehand.
- If future-tense propositions are not true, then neither are past-tense propositions.” (p. 222) But obviously many past-tense propositions are true, and so many future-tense propositions are as well.
- Every tensed proposition entails a corresponding tenseless proposition, and since “[t]enseless propositions are always true or false” (p. 222), true tenseless propositions are omnitemporally true. And so, if an SFCP’s tenseless correlate is true, and the event in question lies in the future, then the SFCP is also true.
All of these arguments are very weak if not plainly question-begging against open futurism.
Craig’s first argument (i) relies on a thesis that Patrick Todd has aptly dubbed “retro-closure”:
retro-closure (RC): p → HFp (i.e., if p is the case then it has always been the case that p will be the case)
RC is controversial. Indeed, there is only one serious argument for RC. It’s an argument based on the retrospective validation of predictions, and it is not nearly compelling. (For details, see this post.) The weakness of RC becomes especially obvious when we consider OFCPs. Suppose it rains on Tuesday but that, as of the immediately preceding Monday, it is still a future contingency whether it rains on Tuesday. In that case the OFCP <It might and might not rain on Tuesday> is true on Monday. But if on Monday it’s indeterminate whether it rains on Tuesday (might and might not), then it can’t also be true on Monday that it determinately will rain on Tuesday. And so RC is false. It’s a non sequitur. Using F() as a future determinacy operator and M() as a future indeterminacy operator, all that plausibly follows from <p is the case> is HFp OR HMp.
Craig’s second argument (ii) relies on his mistaken, indeed deflationary, understanding of the correspondence theory of truth (CTT). A past-tense proposition does not merely say something about the past. It also says something about the present. Thus, <It rained on Tuesday> does not merely express the tenseless claim <It rains on Tuesday> but also <Tuesday’s rain is prior to now>. The latter is implied directly by the past tense. If we suppose that now = Wednesday and we’re looking for a corresponding fact to make <It rained on Tuesday> true on Wednesday, then it has to be a fact that obtains, at least in part, now (i.e., on Wednesday). Likewise, <It will rain on Tuesday> does not merely express the tenseless claim <It rains on Tuesday> but also <Tuesday’s rain is posterior to now>. The latter is implied directly by the future tense. If we suppose that now = Monday and we’re looking for a corresponding fact to make <It will rain on Tuesday> true on Monday, then it has to be a fact that obtains, at least in part, now (i.e., on Monday). Now let’s add in a metaphysical thesis that Craig endorses, along with most open futurists:
presentism = All that exists, exists now. There are no non-present entities.
If presentism is true, then the only corresponding facts that can make anything true on Monday have to obtain wholly on Monday. And, in general, for a proposition to be true at time T, it must correspond to a fact that obtains wholly at time T. From this it follows that present truths about the past and present truths about the future must all be grounded (wholly) in the present. This result leads to a popular objection against presentism, namely, that what exists at present doesn’t seem adequate to ground all the truths about the past that we think there are. Craig’s argument presupposes the validity of this objection as a reason for rejecting, not presentism, but a non-deflationary interpretation of CTT. Thus, he wants to say that truths about the past and future can be grounded in the present “fact” that the events in question did or will exist. This, however, is a cheat, for on Craig’s view there is no substance to these “facts” beyond the truths in question. We’re left with a metaphysically circular and vacuous explanation. Thus, for Craig, <It will rain on Tuesday> is true on Monday because on Monday there exists the “fact” that it will rain on Tuesday. And if you ask what that “fact” consists in over and above the proposition it’s supposed to make true, all he can do is repeat the same thing over and over: it’s the “fact” that it rains on Tuesday. Craig’s approach not only trivializes CTT, it’s also unwarranted. For theists who allow God to be temporal (like all theists should) the grounding objection for truths about the past and future is easy to meet. We can simply say that truths about the past are grounded in God’s present infallible memory of the past. As for truths about the future, we can ground those in the present causal dispositions of things. When it comes to SFCPs, however, no present grounds are sufficient. And so, given presentism and CTT, we should conclude (pace Craig) that no SFCPs are true.
Finally, Craig’s third argument (iii) relies on the dual assumptions of what I call ersatz eternalism (EE) and the omnitemporality of tenseless truth (OTT). Unlike regular eternalism, which posits that concrete past, present, and future things are all equally real, EE just says that true propositions about all past, present, and future things tenselessly exist. So understood, EE is a corollary of a common assumption among analytic metaphysicians that there is a unique albeit abstract “actual world” which contains a complete time-LINE, past, present, and future. (Note the LINE assumption here.) This assumption is rarely defended, and it begs the question against open futurism. The same is true of OTT: often assumed, rarely defended, and question-begging against open futurism. The idea behind OTT is that if we take all true tenseless propositions, then we have a complete, tenseless description of the unique actual world along with its unique actual future (UAF). And since that same actual world (allegedly) obtains at all times, then those true tenseless propositions are true at all times.
This argument is specious, however. When used to argue against open futurism that there are true SFCPs it’s just question-begging. Open futurists don’t believe that there is a unique actual world that contains a complete time-LINE. We reject both ersatz and regular eternalism. And we reject OTT. I give a detailed argument against OTT in this post and an argument against EE (i.e., the modal settledness of the future) in this post. The core problem is that, given open futurism, reality is fundamentally dynamic and open-ended. Hence, there is no timeless or omnitemporal vantage point from which anyone, even God, can survey past, present, and future reality.
(1.b.2.iii) Does denying PB for SFCPs have absurd consequences?
We finally come to Craig’s third line of attack against denying (3) by way of rejecting bivalence (PB): it leads to “absurd consequences.” Craig offers two such consequences. Both are based on truth-functional logic.
First, Craig argues that if PB fails for SFCPs, then both disjuncts of <Either Vance will win in 2028 or Vance will not win in 2028> are neither true nor false. But then it follows, he says, that the entire truth-functional disjunction is neither true nor false, which is “absurd” because “there is no other alternative” (p. 223). This argument fails, however. The reason why is subtle. Basically, Craig conflates a truth-functional interpretation of “or” with an ordinary language interpretation of “or.” On a truth-functional (TF) interpretation, an “or” statement says merely “At least one of these alternatives is true.” On an ordinary language (OL) interpretation it says “These alternatives exhaust the possibilities.” These interpretations are clearly not equivalent. According to TF, for example, if one alternative is true, then the whole disjunction is true whether or not the alternatives exhaust the possibilities. Now, when Craig says “there is no other alternative” he is applying the OL interpretation, but when he argues that the disjuncts being neither true nor false means the whole disjunction is neither true nor false he is applying a TF interpretation. Combining both interpretations at the same time, as Craig implicitly does, is incoherent. If we construe the disjunction in an OL way, then the disjunction as a whole can be true even if none of the disjuncts is true. As long as the alternatives exhaust the possibilities, that’s all that matters. If, however, we construe the disjunction in a TF way, then we have two options. We can understand “neither true nor false” either as a truth value gap (i.e., the absence of a truth value) or as third truth value intermediate between true and false. On the first option we can argue that the disjunction is “false” on the grounds that neither of the alternatives is true. On the second option, we can argue that the disjunction is “neither true nor false” because it’s not clearly either true or false. Whichever option we take, the result is neither surprising or “absurd.” It’s simply a function of how “or” and “neither true nor false” are interpreted.
Second, Craig argues that if PB fails for SFCPs, then both conjuncts of <Vance will win in 2028 and Vance will not win in 2028> are neither true nor false and, therefore, the entire “and” proposition becomes neither true nor false. This is absurd, he says, for the conjunction is self-contradictory (p. 223) and so necessarily false. The main problem with this argument is that the conjunction as a whole is not an SFCP. There is no causally possible future in which the conjunction comes out true. So, even if we deny bivalence for SFCPs, we aren’t forced to reason truth-functionally and say that the conjunction is neither true nor false. We can say instead that it is just plain false precisely because the conjuncts can’t be co-realized.
(1.b.2.iv) What about all-falsism?
Craig now considers one final strategy for rejecting premise (3). According to logician Arthur Prior, instead of denying bivalence for SFCPs one can instead hold that all SFCPs are false. Craig makes like this is a bizarre suggestion, but the idea is actually quite straightforward. Using our future determinacy operator F() and our future indeterminacy operator M(), the idea is that the negation of F(p) is not F(~p) but simply ~F(p). This works if F(p) and F(~p) do not exhaust the possibilities. If, for example, we follow Patrick Todd in supposing that F(p) means “In all available futures, p” and F(~p) means “In all available futures, ~p,” then there is clear room for another possibility, namely, that p obtains in some available futures and not in others. That is precisely the case with future contingents. If we use M(p) to represent that case, then it is clear that the negation of F(p) is not F(~p) but F(~p) OR M(p). That is, ~F(p) = F(~p) ∨ M(p). This result is devastating to Craig’s argument. He rambles on for a few pages trying to poke holes in Prior’s logic, but it’s clear he doesn’t comprehend what’s going on because he has no category in his mind for OFCPs, which is what M(p) represents. Thus, he thinks Prior’s writing the negation of <Vance will win in 2028> as <It is not the case that Vance will win in 2028> instead of <Vance will not win in 2028> is just an ambiguous “reinterpretation” of “the normal understanding of negation” (p. 225). On the contrary, Prior is simply following the standard practice of negating a proposition by putting a not in front of it. For example, to negate p we write ~p. When we add in a tense operator like F() we naturally get a scope ambiguity. That is, we have to consider whether ~F(p) is equivalent to F(~p) or not. Does it matter whether the negation goes in front of the operator or inside the operator? Prior thinks it matters, and as soon as we countenance OFCPs it’s easy to see why. Craig doesn’t think it matters because OFCPs aren’t on his cognitive radar. And they aren’t on his radar because he simply assumes a LINE conception of the future. Because of this, Craig’s objections to Prior ultimately beg the question against open futurism.
Side note: Even if we set OFCPs aside, it can be argued that ~F(p) is not equivalent to F(~p) because they have different truth conditions. Let F(~p) be <Vance will not win in 2028>. This says that it will be the case that <Vance does not win> is true in 2028. Thus, for it to be true it has to be the case when the 2028 election comes around that Vance is not the winner. Now let ~F(p) be <It is not the case that Vance will win in 2028>. This simply denies that it will be the case that <Vance does not win> is true in 2028. That, however, could be true even if the 2028 election never comes around. Suppose, for example, that God has already decided to end creation in 2027. It seems logically possible then that <It is not the case that Vance will win in 2028> can be true even if <Vance will not win in 2028> is false. Hence, ~F(p) is not equivalent to F(~p).
(2) Divine foreknowledge and future contingents
Craig finally comes to the foreknowledge and future contingency problem. This is a long subsection—roughly 25 pages. He spends more than half of that space, however, on long discussions of ideas from other philosophers. These discussions are for the most part parenthetical asides that don’t matter much to the central argument, so I’ll be mostly skipping over those parts.
(2.a) The argument for theological “fatalism”
I put “fatalism” in quotes because Craig routinely conflates fatalism (i.e., the denial of future contingency) with a certain sort of incompatibilism, namely, settled future–future contingency incompatibilism. I’ll reference that as “SF-incompatibilism” in what follows. Because of this conflation Craig often mischaracterizes anyone who endorses SF-incompatibilism as a “fatalist” even if the person—an open futurist, say—explicitly affirms future contingency. Bear in mind, then, that when Craig speaks of “fatalism” he sometimes means fatalism proper and he sometimes means SF-incompatibilism.
It should also be noted that when Craig speaks of “theological” fatalism, he sometimes means (a) fatalism arrived at by way of the assumption that God has EDF and sometimes means (b) EDF–future contingency incompatibilism. When (a) is in view we should describe it more accurately as a “theological argument for fatalism from EDF.”
Now, according to Craig, the “most elementary” theological argument for fatalism runs as follows:
(5) Necessarily (If God knows p, then p).
(6) God knows p.
∴
(7) Necessarily (p).
Thus, “[i]n virtue of God’s foreknowledge everything is fated to occur” (p. 227).
Craig correctly observes that this argument is a non sequitur. What follows from (5) and (6) is not (7) but simply
(7*) p.
Pointing out this non sequitur is the so-called “modal fallacy” charge against fatalism. As I detail in another blog post, however, Craig badly misrepresents the core structure of the fatalistic argument. Serious fatalistic arguments don’t commit this fallacy or any logical fallacies. Some of their premises may be false, but the logic is solid.
In addition to the modal fallacy charge, Craig has a long footnote on p. 228 wherein he accuses fatalists of “conflating certainty with necessity” (emphasis his). Certainty, he says, “is a property of persons” whereas necessity “in a property of propositions.” In short, “[p]eople are certain; propositions are necessary” (p. 228). Craig thinks that by means of this conflation the fatalist wrongly supposes that God’s certainty about what will happen entails the necessity of what will happen when it allegedly only entails that it does happen. Contrary to Craig, however, both the distinction and the alleged conflation are bogus. There are two kinds of certainty—subjective and objective. Subjective certainty is feature of persons (“I am certain that …”), whereas objective certainty is feature of how things are (“It is certain that …”). The latter entails some sort of necessity, but which sort has to be assessed on a case-by-case basis, as there are many different types of necessity (logical, metaphysical, causal, epistemic, moral, etc.). Of those only logical necessity is truly a feature of propositions.
Ironically, Craig immediately undercuts his own modal fallacy and certainty/necessity conflation charges by tacitly admitting that the argument above is a straw man because the fatalist can easily repair it by replacing (6) with
(6*) Necessarily, God knows p.
He notes, correctly, that the inference from (5) and (6*) to (7) is “logically valid” (p. 228). So far so good, but even this is not yet a fair representation of how real fatalistic arguments work, for the kind of necessity that figures in (6*) and (7) needs to be further specified. More on that below.
Craig next launches into a rambling discussion of Arthur Prior’s and Nelson Pike’s more sophisticated arguments for SF-incompatibilism, culminating with a recent formulation by David Hunt (pp. 228–234). Some inaccuracies come up in the course of this discussion, but I’ll leave those for others to identify. One thing worth noting is that Craig employs a “muddy the waters” strategy by bringing up Pike’s and Hunt’s arguments. These arguments are unnecessarily complex in that lose sight of the core issue of whether EDF is compatible with future contingency by bringing in all sorts of tangential assumptions, such as God’s relationship to time, the nature of human freedom, whether morally responsible action requires alternative possibilities, and so on. Interesting as these topics are, they are quite beside the point for, as I explain in a previously referenced blog post, all the fatalist needs to assume is (a) that there are “now-unpreventable” or “fixed” facts and (b) that these facts collectively specify or entail a unique actual future. Everyone already believes (a), so the only issue that matters for fatalism is whether (b) is true. Moreover, there are only two ways (b) can be false. One way, which I call open futurism, denies that there is any collection of facts, fixed or otherwise, that specifies a unique actual future. For the open futurist there is no such thing as a unique actual future. The other way, which I call preventable futurism, says that, yes, there is a unique actual future, and there are collections of facts sufficient to specify such a future, but no such collection is fixed.
(2.b) On reducing theological arguments for fatalism to “logical” arguments for fatalism
The next part of Craig’s strategy is to collapse so-called “theological fatalism” into “logical fatalism.” I’ve already explained about why the first term is a misnomer. So is the second. Neither of these is a distinct kind of fatalism. Instead, these labels represent different ways of arguing for fatalism. What differentiates them is the nature of the allegedly fixed, future-specifying facts that they posit. Theological arguments for fatalism based on God’s having EDF take God’s knowledge of the future to be a fixed, future-specifying fact. They present us an epistemic argument for fatalism. So-called “logical” arguments for fatalism, in contrast, present us an alethic argument for fatalism. That is, they posit a collection of fixed, future-specifying truths about the future. Craig’s goal in this subsection, then, is to collapse epistemic arguments for fatalism into alethic arguments for fatalism.
He attempts to do this by pointing out, first, that premise (5) is analytic. It’s truth is guaranteed simply by the definitional fact that knowledge entails truth. So we can drop out references to God and God’s knowledge to yield
(5**) Necessarily, if <x will happen> is true, then x will happen.
We can then replace (6*) with
(6**) Necessarily, <x will happen> is true.
And from (5**) and (6**) we can derive (7) just as before. In other words, we can replace premises about God’s knowledge with premises about truth and derive the same conclusion (p. 234). Hence, the “real issue” is not God’s foreknowledge but “the truth of falsity of future-tense statements” (p. 236).
To be at all plausible, however, the kind of necessity that figures in (6*) and (6**) has to be properly qualified. We don’t want to say that it’s logically or metaphysically necessary that <God knows x will happen> or that <x will happen> is true. Instead, Craig suggests, fatalists want to say that God’s foreknowledge and/or what’s true about the future is “temporally necessary,” having the unchangeable “necessity of the past” (p. 235).
This is inaccurate, however. Fatalism doesn’t require that the past be necessary in any sense. This is because fatalism is purely a thesis about the future, namely, that there is only one causally possible future, and consequently, that there is no future contingency. Instead of temporal necessity all the fatalist needs is the fixity or now-unpreventability of whatever future-specifying facts the fatalist posits. What makes a fact fixed is its causal necessity, the fact that no one (or at least no creature) has the power to prevent, undo, or overwrite that fact.
Moreover, Craig’s reduction of the epistemic to the alethic is problematic, for a fatalist could easily hold that there are no fixed and future-specifying epistemic facts while maintaining that there are fixed and future-specifying facts of another sort (e.g., alethic, causal, ontic, etc.). So, the kind of reduction Craig is looking for need not materialize. Different arguments for fatalism have to judged on their own merits. The main reason why Craig thinks he can pull off this reduction is because, as noted above, he doesn’t think God’s knowledge requires anything more than truth (p. 218). I argued above that this is wrong. God’s knowledge requires acquaintance with those features of reality that make truths true.
(2.c) On the “unintelligibility” of fatalism
Craig next tries to convince his readers that fatalism “must be wrong” because it “posits a constraint on human freedom which is altogether unintelligible” (p. 237).
I’m tempted to respond Inigo-to-Vizzini-style that “unintelligible” does not mean what Craig thinks it means. The word he’s looking for is “implausible.” It is massively implausible, he thinks, to suppose that our actions can be constrained “merely by the truth of a future-tense statement” or by God’s “merely knowing about it” (p. 237). In other words, Craig thinks fatalism must be false because it entails that foretruth and/or foreknowledge are per se causal when they obviously aren’t. It’s being true that it will rain tomorrow, or a forecaster’s knowing that it will rain tomorrow, obviously doesn’t cause it to rain tomorrow.
Now, while I agree with Craig that this consequence, namely, the per se causality of foretruth and foreknowledge, is massively implausible, his charge that fatalism entails this implausibility reflects significant confusion on Craig’s part. The charge is a straw man. No premise in any standard fatalistic argument says or implies that foretruth and/or foreknowledge are per se causal. Fatalistic arguments purport to show that the future must be causally determined because there are fixed facts that specify a unique actual future. They don’t claim that the future is causally determined by those very same facts. So, Craig’s attempt to show that fatalistic arguments “must be wrong” is a complete misfire.
What’s more, some fatalistic arguments are quite plausible, especially if one believes as Craig himself does that God has temporally prior EDF and is essentially infallible. After all, very plausibly–as Craig would agree (see p. 247)—the past is unchangeable. It cannot be undone or rewritten. And so the past content of God’s foreknowledge is fixed. No one can now bring it about that God knew or believed otherwise. Moreover, given EDF, God’s foreknowledge specifies a unique actual future. As I argue at length in my book, this is all one needs to construct a valid argument for fatalism for, if we had it in our power now to do anything otherwise than God has foreknown, then we would have the power to invalidate God’s foreknowledge, which is impossible given its infallibility.
(2.d) Challenging the fixity of God’s foreknowledge
Craig, however, thinks he can coherently deny premise (6*). God’s foreknowledge is not “temporally necessary.” Now, I’ve just finished pointing out (in section 2.b) that this is the wrong modality. Causal necessity is where it’s at. Craig, however, insists on distinguishing between temporal and causal necessity (p. 239), probably because he’s stuck on the wrong idea that fatalism relies on the necessity of the past. And so he embarks on a long rabbit trail (pp. 238–246) wherein he tries to unpack the notion of temporal necessity by drawing the work of Alfred Fredosso. I won’t bore my readers with the ins and out of this discussion. It’s both confusing and confused. Freddoso’s analysis of temporal necessity is, in fact, just plain wrong, in part because he supposes that “only logically contingent propositions can be temporally necessary” (p. 239)—logically necessary truths are obviously fixed in whatever sense the fatalist may require—and in part because he takes the logical possibility of doing otherwise to be sufficient for future contingency (p. 240) when the causal possibility of doing otherwise is what’s required. In the end, Craig concludes that a temporally contingent or “soft” fact is a past fact which entails that “were some future condition to be other than it will be, then that past fact would as a consequence have been different than it was” whereas a temporally necessary or “hard” fact is a past fact that would not have been any different “no matter how future conditions might vary” (p. 245).
Applying this to God’s foreknowledge, Craig says that
(8) God believed that p,
where p is a future-tense proposition, is not temporally necessary because “[p]ast facts are either hard or soft, depending on whether they are counterfactually dependent on future events or actualities” and (8) is “counterfactually dependent upon … the events or actualities described in p, such that if they were different, then God’s belief would have been different” (p. 246).
What Craig is implying is that none of God’s past beliefs about the logically contingent future are temporally necessary or hard facts. Because God’s propositional omniscience guarantees the truth of whatever God believes, simply by being about the future God’s past beliefs are counterfactually dependent on future events. So long as it is logically possible for future events to be otherwise, if those events had been otherwise God would have had different past beliefs. These beliefs only become temporally necessary or hard facts when, due to the passage of time, they are no longer about the future.
This is a genuinely bizarre result, for at least a couple reasons. First, the logical contingency of future events is hardly enough to establish their causal contingency, and if we want (as Craig does) to reconcile God’s foreknowledge with creaturely libertarian free agency, then it is the latter that we need. Second, counterfactual dependence is a myth. What Craig calls “counterfactual dependence” is merely a kind of necessary correlation. In this context it’s the necessary correlation between what God believes about the future and what happens in the future. But the fatalist would insist on just this correlation! According to the openly fatalistic position of theistic determinism, for example, if the future were to play out otherwise, then God would have foreknown otherwise, and He would have foreknown otherwise precisely because He would have deterministically caused things to play out otherwise. In short, then, the logical contingency of future events and their counterfactual correlation with God’s foreknowledge are perfectly agreeable to the fatalist. Craig’s long-winded analysis of temporal necessity therefore turns out to be irrelevant!
But, says Craig, fatalists haven’t adequately made their case, for they “have never provided an adequate account” of temporal necessity that “does not reduce to either the unalterability or the causal closedness of the past” (p. 246). I beg to differ. As I pointed out above, fatalism has no essential connection to the unalterability or causal closure of the past, but it has everything to do with fixity or now-unpreventability of future-specifying facts. Craig misses the point because he’s focused on the temporal sequence when what matters is the explanatory sequence. Is God’s foreknowledge explanatorily (not counterfactually!) dependent on future events? According to the preventable futurist, yes. If there are future contingents then God’s foreknowledge is explained and thus brought about by the occurrences of those events. That I freely choose chocolate over vanilla ice cream explains God’s having foreknown that I was going to choose chocolate over vanilla. Surprisingly, Craig rejects this approach. He says “there is nothing I can now do to cause or bring about the past” or “cause God to have had in the past a certain belief about my future actions” (p. 247). Indeed, Craig rejects both open futurism and preventable futurism. His rejection of open futurism is as question-begging as it gets: “By definition the future is what will occur”; hence, “the future … is just as unalterable as the past” (pp. 246–247). So, Craig implicitly concedes everything the fatalist needs to get a successful argument for fatalism off the ground! Craig doesn’t notice he’s lost the argument, however, because he’s apparently convinced himself that counterfactual “dependence” plus the logical contingency of the future is enough to secure future contingency. As I’ve already argued, that is not nearly enough, for nearly any fatalist will affirm as much.
(2.e) The basis of divine foreknowledge
Craig wraps up his discussion of divine foreknowledge by distinguishing between two “models of divine cognition.” The perceptualist model “construes divine knowledge on the analogy with sense perception,” whereas the conceptualist model holds that “God’s knowledge is intrinsic,” like “a mind’s knowledge of innate ideas” (p. 250).
Craig rejects the perceptualist model and endorses the conceptualist model because he thinks the former raises genuine worries about how can know SFCPs. If presentism is true, for example, then there is no present reality God could be acquainted with that would reveal true SFCPs to Him. On the conceptualist model, however, this doesn’t matter: “God has essentially the property of knowing all truths; there are truths about future contingents [i.e., SFCPs]; ergo, God knows all truths about future contingents [i.e., SFCPs]” (p. 250). And if one persists in asking how can have innate knowledge of true SFCPs, then Craig’s response is that that’s a stupid question; it’s just innate (p. 251).
In my view this a really bad distinction on Craig’s part. He rejects the very plausible idea that God’s knowledge of creation is grounded (in part at least) in His de re acquaintance with creation because it threatens his metaphysical and theological commitments and replaces it with essentially … nothing. The so-called “conceptualist model” isn’t a model of anything. To the question of how God could know true SFCPs (assuming there are any), the conceptualist’s question-begging response is simply “He just does.” Wow, thanks Dr. Craig. That really clears things up! … not. What a “perspicuous basis” for God’s knowledge of SFCPs (p. 251)!
What’s more, despite what Craig says, the conceptualist’s “innate” knowledge isn’t innate. Genuinely innate knowledge stems from one’s acquaintance with oneself. One “finds” or “discovers” the information within oneself. For example, I know how I’m feeling right now because of my self-acquaintance. But when it comes to God’s alleged knowledge of true SFCPs, there is nothing God could be acquainted with within Himself that would reveal that information. God’s acquaintance with His own essence won’t reveal it because that’s absolutely necessary and this information is metaphysically contingent. God’s acquaintance with His will won’t reveal it because God has willed that the events in question be causally contingent. As far as God’s sovereign will is concerned, there are no true SFCPs but only true OFCPs. The only thing “innate” to God Craig can point to is God’s knowledge of true SFCPs, but the information that God (allegedly) knows isn’t innate. It’s just there. You can see now why Craig earlier claimed that truth is an intrinsic property of a proposition. On his account true SFCP information is just there. It’s not grounded in anything. It’s not innate to God. It’s just intrinsically “true” and therefore God knows it by acquaintance with the propositions themselves. Craig calls
It’s hard for me to overstate how nutty this “model” of Craig’s is. For example, if these SFCPs propositions are intrinsically “true,” how’d they get that way? As FCPs these propositions are logically contingent, so there can be no sufficient condition for their truth in the propositions themselves. But then how is their truth intrinsic? And if it is somehow both intrinsic and contingent, then what secures the necessary correlation between the truth values of these propositions and events in history? Intrinsic truths shouldn’t be sensitive to extrinsic events.
My Concluding Thoughts
This is a very poorly argued section of Craig’s omniscience chapter. Craig repeatedly straw man’s fatalism and begs the question against open futurism. And his antidote to fatalism is anything but. I’ve heard Craig say that he spent seven years studying the problem of foreknowledge and future contingents. Out of that study came three books—two scholarly and one popular. Nevertheless, despite all that effort, Craig never developed a clear and accurate understanding of the problem. Instead of starting, as I have tried to do, by reflecting on what it would take to construct a valid and plausible argument for fatalism, Craig has never taken fatalism seriously. He quickly convinced himself that it just had to be wrong and so contented himself with leveling all sorts of pseudo-difficulties: that it commits a “modal fallacy”; that it conflates certainty and necessity; that it makes foreknowledge and foretruth per se causal; that it requires some obscure and peculiar modality like temporal necessity; etc. Nor has Craig ever taken open futurism seriously. For him the future is settled by definition, and so questions about whether bivalence holds for SFCPs or whether they are all false don’t need to be carefully scrutinized because open futurism is wrong by definition.
Despite considerable discussion of both fatalism and open futurism over the past few decades, nearly the whole of this section of Craig’s chapter repeats verbatim stuff Craig wrote 35+ years ago. He apparently hasn’t learned much on these topic over the intervening years probably because he’s overly confident that he figured it all out 35+ years ago.