More on Merricks on Truthmaking

I've now finishing my re-reading of Trenton Merricks' book Truth and Ontology and I'm still quite unconvinced by his contention that what's true does not depend on what exists in any substantive way. A couple posts ago I noted my main reasons for dissatisfaction with Merricks' arguments:

1. He loads down truthmaker principles, which affirm a substantive dependence of truth on being, with extraneous commitments.

Most significantly, he takes the raison d'etre of such principles to be that of catching all sorts of metaphysical "cheaters"--those who either (a) try to have their truths on the cheap, without paying an appropriate metaphysical price, or (b) invoke "suspicious" properties to do their truthmaking work. I deny that these concerns are fundamental to truthmaker principles. Such principles are merely attempts to articulate the driving intuition behind correspondence theories of truth, namely, that truth depends on being in a substantive way and, therefore, that there must be some being corresponding to any given truth that "makes" that truth true. What properties count as "suspicious" is to be determined not by truthmaker principles but by explanatory considerations (e.g., Are such properties merely "ad hoc", or do we have independent reasons for positing them? Can they plausibly be regarded as primitives, or can they be cashed out in some explanatorily informative way?)

2. He never carefully defines the crucial notion of "aboutness" upon which his major arguments depend.

Merricks insists that truths can only be made true by what those truths are "about". He then argues that for many types of truths there are no existing things that those truths are "about", from which he concludes that truthmaker principles are false. It is clear that the notion of "aboutness" is crucial to this argumentative strategy, and it is surprising that a philosopher of Merricks' calibre doesn't define it carefully. In fact, he explicitly declines to give an analysis (pp. 33-34) and relies, instead, on a handful of examples (e.g., pp. 28-29) which indicate only that "aboutness" has something to do with relevance. Here's one such example:

[Y]our thumb fails to be a genuine truthmaker for FLT [Fermat's Last Theorem]. . . . Even though your thumb's existence necessitates FLT's truth, FLT is not about your thumb. (p. 27)

I'll grant that a thumb is not a truthmaker for FLT. And I'll grant that a thumb's inadequacy as a truthmaker for FLT can be explained, to a first approximation, by the observation that FLT is not relevantly "about" a thumb. But Merricks needs to go a lot further than that. If he wants to refute truthmaker principles (and he does) then he needs to give an analysis of what the relevant "aboutness" relation is. Were he to provide such an analysis, of course, fans of truthmaker principles could subject it to scrutiny. Leaving the notion nebulous allows Merricks to rest his argument on impressionistic declarations that this or that type of truth isn't relevantly "about" any existing thing. He thus arrives at his conclusions by a certain amount of theft over honest toil.

In the rest of this post I'm going to help Merricks out a bit by clarifying the relevant sense in which truthmakers must be what truths are "about".

First, we need to distinguish between connotative and denotative senses of "about". In the connotative sense, to speak "about" (say) FLT is to say something about the content of the theorem. This requires some understanding of what the theorem means. But I can say, "I heard that some dude proved FLT a few years ago" and thereby talk "about" FLT without having any clue what FLT stands for. That's a strictly denotative sense of "about". Now, truthmaker principles are metaphysical, not epistemological in nature. They require that there exists something in virtue of which a truth is true, but they don't require us to know what that something is. Hence, the relevant sense of "aboutness" is denotative, not connotative.

Second, truthmakers must relevantly necessitate their truths. What that means is not simply that it must be impossible that the truthmaker exist and the truth in question fail to be true. Your thumb, after all, necessitates FLT in that sense. No, it must also be the case that the truth in question could be derived from knowledge of the truthmaker. In other words, if (hypothetically) someone were fully acquainted with a putative truthmaker and only that truthmaker, would that person thereby know with certainty that the truth in question is true? Your thumb is not a truthmaker for FLT because someone acquainted only with your thumb (however fully) would not thereby know FLT.

I claim that the above two points are all there is to the relevant "aboutness" relation. A truth is "about" one of its truthmakers in the relevant sense if and only if (a) there exists something such that (b) full acquaintance with that thing and only that thing would enable one to know with certainty that the truth in question is true.

Merricks, undoubtedly would want to insist on further constraints, at which we could have a healthy debate about them. My complaint is that he should have gone at least as far as I have here, and he could easily have done so.

In my next post, I'm going to use my "aboutness" criterion and show that on metaphysical assumptions that Merricks accepts, there are truthmakers for all of the truths for which says there aren't any.

Gratuitous Evil and Divine Providence

I'm really happy right now because a paper of mine ("Gratuitous Evil and Divine Providence") was just accepted at Religious Studies. In my experience, at least, the turnaround at that journal is phenomenally fast. This is the second paper I've submitted there. On both I received an acceptance notice within a week or less.

I'll post a digital version of the paper to my website once I've sent the final version off to the journal. Here's the abstract:

Discussions of the evidential argument from evil generally pay little attention to how different models of divine providence constrain the theist’s options for response. After describing four models of providence and discussing theistic strategies for responding to the evidential argument, I articulate and defend a definition of “gratuitous evil” that renders the theological premise of the argument uncontroversial for theists. This forces theists to focus their fire on the evidential premise, enabling us to compare models of providence with respect to how plausibly they can resist it. I then give an assessment of the four models, concluding that theists are better off vis-à-vis the evidential argument if they reject meticulous providence.

Merricks on Truthmaking

I've recently been rereading Trenton Merrick's book Truth and Ontology (Oxford, 2007) in which he argues against the substantive dependence of truth on being. Thus, he rejects theses like

• TM - Every truth has a truthmaker, a parcel of reality the existence of which necessitates, and thereby grounds, that truth.
• TSB - Truth supervenes on being. I.e., every possible difference in truth corresponds to a possible difference in reality (i.e., in what things exist and in what properties have and/or what relations they stand in), and vice-versa.
• Correspondence theory of truth - A proposition (or truth-bearer) is true iff what it represents is the case (i.e., iff reality is as the proposition depicts).

Merricks' strategy is to saddle the above theses with as much baggage as he can, thereby imposing steep restrictions on what an adequate theory of substantive dependence would have to look like, and then arguing that such theories fail to satisfy those restrictions.

His favorite restrictions are (1) that it is the purpose of theses like TM and TSB to "catch cheaters" by criticizing (a) those who fail to endorse an ontology robust enough to ground or necessitate the truths they feel free to invoke, and (b) those who try to discharge their grounding obligations by positing "suspicious" properties; and (2) that substantive grounds for truths must consists of what those truths are "about", in some rather loosely defined sense of that term.

For example, Merricks argues (ch. 6) that presentism (the theory that whatever exists simpliciter exists now) is true, that it violates TM and TSB, and therefore, that TM and TSB should be rejected. While he discusses some attempts by presentists to satisfy the demands of TM and TSB, Merricks rejects these either because they trade in "suspicious" properties or because they violate the "aboutness" constraint.

I'm not persuaded by Merricks. In part, that's because I think he's wrong that a purpose of TM and TSB is to catch cheaters of the second sort, namely, those who invoke allegedly "suspicious" properties. (All things equal, we ought to try to avoid "suspicious" properties where possible, but it's not TM and TSB's job to say what those properties are.) Also, I don't think he's nearly clear enough on the relevant sense of "aboutness". He admits it has something to do with "relevance", but apart from a handful of examples that he claims to find intuitive, he nowhere defines the notion.

Still, it is a fair question what someone like myself who is partial to TM and TSB should say about the truthmakers or the supervenience base for various kinds of truths. In particular, necessary truths, general truths, modal truths, negative existentials, counterfactuals, and metaphysical theses like presentism, actualism, and nominalism have all been thought to raise serious problems for TM and TSB. In the next few posts, I'm going to look at some of these problem cases and argue that they are less worrisome for theists than they might be for those of different metaphysical persuasions.

I Made the List

of 100+ living philosophers of religion.

Back to Blogging

I haven't done much in the way of blogging for the past several months. In large part this is because my energies have been focused on being a father to my now 18-month old daughter and on finishing several papers that I've been working on. Now that my desk is starting to clear up, I'm going to resume blogging again.

I don't promise to blog regularly. Blogging is something I do when I have spare time. It's not something I want to get absorbed into. So when I feel like I'm spending too much time at it, I'll back off again for a while. We'll see how it goes.

I'll put up a post sometime tomorrow with thoughts on truthmaking, theism, and modality. Right now, however, I've got to go home and spend time with my wonderful wife and daughter.

Truth Accessibility Relations among Times

In my two previous posts, I have explained the notion of 'truth at a time' and argued that for a (contingent) proposition to be true at T it must supervene upon what exists at T. In short, truth at T requires a truthmaker at T.

A question that remains concerns accessibility relations among times. If a proposition is true at some earlier time, is it necessary that it be true at all later times that it was true at that earlier time? And if a proposition is true at some later time, is it necessary that it be true at all earlier times that it will be true at that later time? In common tense logical notation, these questions concern the theses:

1. p → FPp
   
2. p → PFp

(1) says that if p is true at some time, say T, then it will be true at all times subsequent to T that p was true at T. (2) says that if p is true at T, then it has been true at all times prior to T that p will be true at T.

I believe that (1) is true and (2) is false. In other words, I believe that earlier times are accessible to later times, but I don't think that later times are accessible to earlier times. This antisymmetry means that I take S4 and not S5 to be the right system of modal logic as far as times are concerned.

S4 ensures reflexivity and transitivity among times. Reflexivity means that each time is accessible to itself. Hence, if it is true at T that p, then it is true at T that it is true at T that p, etc. Transitivity means that if T1 is accessible to T2, and T2 is accessible to T3, then T1 is accessible to T3. Hence, if it is true at T1 that p, and if it is true at T2 that it is true at T1 that p, then it is true at T3 that it is true at T1 that p.

To reflexivity and transitivity, S5 adds symmetry. This means that if T1 is accessible to T2, then T2 is accessible to T1. Hence, if it is true at T1 that p, then it is true at T2 that it is true at T1 that p, and it is true at T1 that it is true at T2 that it is true at T1 that p, etc.

Now, if S5 is right, then both (1) and (2) have to be accepted - what's true at any time is mirrored at all times. If S4 is right, however, then we cannot accept both (1) and (2), for that would violate antisymmetry. The questions, then, are whether we should accept antisymmetry or not and, if so, whether we should go with (1) or (2).

My argument turns on the relation between time and causality and on the falsity of determinism. Earlier events can cause later events, but not vice-versa (or so most of us think). That motivates antisymmetry, but to establish it we have to appeal to the idea that what's true at a time depends upon what exists at that time. Suppose event E1 at T1 indeterministically causes event E2 at T2. It is true, therefore, at T2 that E2 is occurring. Given symmetric accessibility between T1 and T2, it is true at T1 that it is true at T2 that E2 is occurring. But this is false if the causal relation between E1 and E2 is indeterministic, for in that case we do not have a truthmaker at T1 for "E2 is occurring at T2".

If determinism is true, then a unique sequence of later times is implied by earlier times. It can be true at T1, therefore, that "E2 is occurring" will be true at T2. Hence, T2 is implicitly accessible from T1. But if determinism is false, then no unique sequence of later times is implied by earlier times. Let p represent the occurrence of some event that is causally contingent from the perspective of T1 (e.g., "E2 occurs at T2"). From that perspective, it may be that p is true at T2, and it may be that p is false at T2. Moreover, there is nothing at T1 sufficient for making it true either that p will be the case at T2 or that p will not be the case at T2. Since what's true at T1 depends on what exists at T1, it cannot be true at T1 either that p will be true at T2 or that p will not be true at T2. Given indeterminism, therefore, it follows that (2) is false. If, for example, there being a sea battle at T2 is causally contingent at T1, then even if a sea battle does occur at T2, it does not make "A sea battle will occur at T2" true at T1, contrary to (2).

But what about (1)? It would be very counterintuitive if (1) were false. It's falsity would allow, for example, that "Caesar crossed the Rubicon" might be true in, say, 1000 A.D., but not true in 2000 A.D. So I want to insist that (1) is true, but I emphasize that it's truth is not a mere matter of logic. If we accept that truth supervenes on being, then (1) commits us to a substantive metaphysical position. Since what's (contingently) true at T2 depends upon what exists at T2, for it to be true at T2 that some event E occurred at T1, there must be something at T2 sufficient for making that true. E's occurring at T1 is not enough. There must also be something that carries E's occurring at T1 forward in time and that constitutes at T2 E's having occurred at T1. As to what that "something" is, I refer the reader to this forthcoming paper of mine.

In summary, if determinism is false, as I believe, then (2) is false. (1), however, is true - it would be intuitively quite odd (to say the least) if it weren't - though it comes at what some will regard as an unacceptable metaphysical price. Those folks either have to deny that truth supervenes on being or reject (1). I think either is a much bigger price to pay.

Truth at a Time and Truth Simpliciter

As a follow-up to my previous post, I'd like to say a bit about how truth at a time and truth at a world relate to truth simpliciter.

Unlike 'truth at a world' and 'truth at a time', truth simpliciter is not evaluated from the perspective of any given time or world, but absolutely. In other words, it is evaluated from the proverbial "God's eye" or absolute perspective, that is, in relation to what exists simpliciter. Consequently, what is true simpliciter depends on the correct metaphysics.

Actualism is the thesis that the only possible world that exists simpliciter is the actual world, whereas possibilism is the thesis that all possible worlds exist simpliciter. Presentism is the thesis that only what exists now exists simpliciter, whereas eternalism is the thesis that all past, present, and future world states exist simpliciter. If actualism and presentism are correct, then what exists simpliciter is coextensive with what exists now. Moreover, truth simpliciter coincides with what is true now. Hence, "Obama is now President" is true simpliciter.

In contrast, if actualism and eternalism are correct, then all instantaneous world states (however foliated into past, present, and future) exist simpliciter. As a result, truth simpliciter and truth at a time can come apart. Omnitemporal truths, which are true at all times, hold distributively across the entire time series and are true simpliciter. So are transtemporal truths, which hold collectively of the entire time series. These can, but arguably need not, be true at all times in the series. For example, "At some time or other, Obama is President" is true simpliciter. It is also true now. But it may not have been true at all earlier times because the fact that makes it true may not be accessible from all earlier times. One way in which this could happen (defended by Peirce, Prior, Hartshorne, and Ryle) is through reference failure. Arguably, before Obama was conceived, there was no such individual for propositions to refer to. Finally, there may be (though this is controversial) atemporal truths, truths about matters that stand in no relation to time whatsoever. These truths, if they exist, are true simpliciter, but fail to be true at any time. (Some people think mathematical truths fall into this category. I doubt this. I say that '2+2=4' is omnitemporally true, not atemporally true. It sounds rather odd to say that it isn't true now.)

Finally, if possibilism and eternalism are correct, then all worlds and all instantaneous world states in all worlds exist simpliciter. As before, truths that hold distibutively over all world-time pairs (i.e., at all times in all worlds) will be true simpliciter. Similarly, anything that is true at all worlds (i.e., necessary truths) will be true simpliciter. Further, truths that hold collectively of the entire world/history ensemble (e.g., "At some time in some world Obama is President") will be true simpliciter. This latter truth is true now in alpha (the actual world), but it need not be true true at all times in all worlds for it may not be accessible from all times in all worlds.

Truth-at-T Depends on What Exists-at-T

In my previous post I made the following claim without argument: "for a proposition to be true now, what it represents as being the case must correspond to ... what is the case now." I'm now going to give that argument.

Every argument for any categorical conclusion requires some categorical assumptions which are not defended within the scope of that argument. Here are some of my assumptions:

1. Truth supervenes on being (TSB). This means that what is true depends on what exists for its being true. Moreover, every difference in truth corresponds to a difference in being, such that if anything that is true had not been true, then there would have been a corresponding difference in reality.
2. Correspondence theory of truth (CT). A proposition is true if and only if the state of affairs (STOA) that it represents obtains. In other words, a proposition p is true if and only if the STOA that would be posited were p asserted obtains.
   
3. It makes sense to speak of a proposition's being true at a time. This notion is compatible with, but does not entail, the idea that truth is temporally invariant.

I won't defend these claims here, though if pressed I might be willing to restrict (1) and (2) to contingent truths.

Now, the key to establishing my claim that what is true at a time depends on what is real or what exists at that time is to understand the notion of 'truth at a time'. I want to suggest that we should think about this notion in a manner analogous to the idea of 'truth at a world'.

In metaphysics, possible worlds are typically distinguished in terms of what is true 'at' them. Thus, worlds W and W* are distinct if and only if something is true at W that is not true at W*, or vice-versa. To consider what is true at a world is essentially to ask what would be true if that world were actual. Alternatively, since each world is actual from its own perspective, we can ask how things stand from the perspective of that world. Finally, what is true at one world may be reflected in what is true at another world. Whether this is so depends upon the accessibility relations that hold among the worlds in question. Thus, if p is true at W, and if W is accessible to W* (that is, if one can "see" W from the perspective of W*), then it is true at W* that p is true at W.

Analogously, 'truth at a time' means truth as judged from the perspective of a given time. By a 'time' I mean a complete time-slice of reality, an instantaneous world state. To determine what is true at a time we ask what would be true if that time were present. Alternatively, since each time is present from its own perspective, we can ask how things stand from the perspective of that time. If T is such that Caesar is ruling in Rome, then "Caesar is ruling in Rome" is true at T. If T* is such that Nero (and not Caesar) is ruling in Rome, then "Caesar is ruling in Rome" is false at T*. As with worlds, what is true at one time may be reflected in what is true at another time. Whether this is so depends upon the accessibility relations that hold among the times in question. Thus, if p is true at T, and if T is accessible to T* (that is, if one can "see" T from the perspective of T*), then it is true at T* that p is true at T. If, in addition, T* is earlier than, simultaneous with, or later than T, then accessibility may be expressed in terms of tense. Thus, if p is true at T, and if T is accessible to T*, and if T is earlier than T*, then it is true at T* that it was the case that p is true at T.

The thesis I set out to defend follows from what I have said above. Just as what is true at a world depends upon what obtains at that world, so also what is true at a time depends upon what obtains at that time. This neither precludes nor entails there being accessibility relations among times such that what is true at one time either has been or will be true at another time. Accordingly, it cannot be true now, for example, that there will be a sea battle tomorrow, unless this truth supervenes upon present reality. Depending on accessibility relations, however, it may be true now that it will (probably) be true tomorrow that there is a sea battle.

A Recent Objection to the Alethic Openness of the Future

The future is 'alethically' open at time T with respect to possible state of affairs X and future time T* > T if and only if neither "X will obtain at T*" nor "X will not obtain at T*" is true at T.

I believe that the future is alethically open with respect to future contingent states of affairs. Thus, if the present state of the world is indeterministic with respect to whether a sea battle occurs tomorrow (i.e., if that's a future contingent), then on my view neither "A sea battle will occur tomorrow" nor "A sea battle will not occur tomorrow" is now true. Indeed, I submit that both propositions are now false.

There's an interesting challenge to my view that has recently been proposed by Alex Pruss of Baylor Univ. His argument is encumbered with the technical trappings of probability theory, but the basic objection can be stated quite simply as follows:

1. On the open future view that I espouse, there are no “will” or “will not” truths about future contingents.
2. But we have excellent (inductive) reasons for believing that there are “will” or “will not” truths about future contingents (e.g., “S will not win the lottery” - assume, e.g., a standard lotto and that S has only one lotto ticket).
3. Therefore, we have excellent reasons for believing that the open future view is false.

This is a potentially devastating argument against my position. The logic is clearly valid, I'm already committed to (1), and premise (2) looks very plausible. Clearly, I have to deny (2). The challenge is to make it clear why that's a reasonable thing to do.

In the first place, it is essential to get as clear as we can on exactly what proposition is expressed by the sentence "S will not win the lottery". This is important because linguistic expressions do not wear their meaning on the sleeve, so to speak. We have to interpret them as best we can in light of whatever contextual clues are available. Due to factors like vagueness, ambiguity, lack of information about the context, and so forth, interpretation is often difficult and frequently inconclusive.

In the second place, there is an important distinction that we have to make between modally inflected uses of "will" and "will not" and amodal uses. Interpreted in the former sense, "S will not win the lottery" may be read as saying something like "S will probably not win the lottery". Here the word "probably" attributes an objective probability to the event S's not winning the lottery. Different objective probabilities can be captured with different qualifiers (e.g., "will definitely", "will possibly", or, more generally, "will, with probability p"). Interpreted in the amodal sense, "S will not win the lottery" should be read as saying something like "Subsequently, S does not win the lottery". This says nothing about what it's objective probability is, other than implying that it is non-zero. Probabilities applied to this proposition have be read as epistemic probabilities, which qualify the attitude that a subject has toward a proposition without entering in the meaning of the proposition. The contrast, then, is between a modal, objective probability reading "S will probably not win the lottery" and an amodal reading "Subsequently, S does not win the lottery".

In the third place, there is a potential ambiguity in the scope of the word "not" in "S will not win the lottery". Does it apply just to the predicate alone or to the rest of the proposition in its entirety? On the first reading, "S will not win the lottery" means "S will not-(win the lottery)". On the second reading, "S will not win the lottery" means "Not-(S will win the lottery)". These readings are distinct since the former entails the latter, but not vice-versa. It is possible, after all, for "Not-(S will win the lottery)" to be true without "S will not-(win the lottery)" being true - for example, when S does not exist.

Crossing the above distinctions gives us four basic readings of "S will not win the lottery":

1. S will probably not-(win the lottery).
2. Not-(S will probably win the lottery).
3. Subsequently, S does not-(win the lottery).
4. Not-(subsequently, S does win the lottery).
   

Now, presumably, we have good (inductive) reasons to believe all of these to be true. (a) and (b), however, pose no difficulty for my position, since I am not committed to saying that "will probably" propositions are false when it comes to future contingents. It comes down, then, to the amodal propositions (c) and (d). To refute me the objector has to show that we have good reason for believing that either (c) or (d) concerns a future contingent and is true now. It is here that I make my stand. I deny that these conditions are satisfied by (c) or (d).

For a proposition to be true, what it represents as being the case must correspond to reality, to what is the case. Likewise, for a proposition to be true now, what it represents as being the case must correspond to present reality, to what is the case now. Now, (c) clearly represents what is, from our present perspective, a future contingent state of affairs. It represents S's not winning the lottery subsequent to a contextually specified index date. I deny, however, that (c) is true now or that we have good reason for believing it to be true now on the grounds that there is no present reality that it corresponds to. We have good reason for believing that it will probably come to be true, but not for believing that it is true. Thus, after the drawing of the lottery, either S will have won subsequently to the index date or not. At that point, assuming S does not win, (c) will be true. But not until then.

With respect to (d), I grant that it is true now and that we have good reason for believing it to be true now, but I deny that it directly concerns the future. It concerns, rather, a present fact, for it says it that it is not (now) the case that subsequently S does not win the lottery. This is provably correct. Future events do not obtain now.

In summary, I reject premise (2) of the objector's argument on the grounds that (i) the sorts of "will not" propositions about future contingents that we have good reasons for believing to be true now are not the sorts of propositions to which the doctrine of alethic openness applies, and (ii) the sorts of "will not" propositions about future contingents to which alethic openness applies are not the sorts of propositions that we have good reason for believing to be true now.

Could God Infallibly Know that He Is Omniscient?

I've been reflecting a bit about the following analogy proposed by Enigman:

Given someone who knows that she can move freely anywhere within an infinitely dimensional space, does she know that she has complete freedom of motion? The problem is that such a space is isomorphic to a hyperspace containing such a space as a mere slice.

This analogy is supposed to call into question whether God can be certain that there is no reality beyond his ken. Couldn't the entire space of God's knowledge simply be a mere slice of a larger hyperspace containing things of which not even God is aware?

Traditional theism answers with an emphatic NO. But can that answer be solidly justified, or is it simply a dogmatic prejudice?

One strategy for justifying the traditional answer is to go Thomistic. On Thomas's account, God is defined as ipsum esse subsistens. His essence is identical to his existence. That trait is necessarily unique to God. Hence, in everthing else there must be a real distinction between existence and essence. As such, beings other than God cannot account for their own existence. Hence, they must all owe their existence to God as their creator. And since God has perfect self-knowledge, he knows his own creative activity, and so knows of the existence of whatever else other than himself that there may be.

My problem with that response is that I think Thomistic metaphysics creates more problems than it solves. For one thing, it's hard to see how Thomas can account for real contingency. Identity is a necessary relation (if two things are identical then they are necessarily identical). It follows that God's essence = his existence, across all possible worlds. Now, either there are multiple, distinct possible worlds, or there are not. If not, then everything is necessary, which is counterintuitive in excelsis. If, however, there are multiple possible worlds, then they differ in some respects, which means that there are contingencies. Hence, what God has to know in order to know everything must vary from world to world, which means that the content of God's knowledge must to some extent be contingent. But if God's existence is really identical to God's essence, then everything about God must be essential and thus invariant across possible worlds in which God exists, including the content of God's knowledge, which is inconsistent with the assumptions that there are multiple possible worlds and that God's knowledge is exhaustive. In short, the following three assumptions are mutually inconsistent:

1. God has exhaustive knowledge of reality.
2. God's essence is identical with his existence.
3. Some facts are contingent.
   

Since 3 is undoubtedly true, either 1 or 2 is false. My preference is to deny 2.

Let me turn now to what I think is a better strategy. The key idea is that actuality is the sole delimiter of real possibility. What this means is that real possibility is always relative to some actuality. For example, what's physically possible depends upon what physical laws are actually in place. Now, suppose we stipulate that God is the sole delimiter of real possibility. If that be accepted, then there are no real possibilities that are not grounded somehow in God. For example, there could be no metaphysically possible world in which God did not exist. Now, if both God and actuality are delimiters of real possibility, then they must be related in some systematic way. Assuming that pantheism is false, that relation can't be identity. A theist, however, could subordinate the one to the other and say that God is the delimiter of possibility in virtue of being the delimiter of actuality. Thus, what's possible is grounded in what's actual, and what's actual is grounded either in God's nature or God's will. If that's right, then in knowing himself, God knows all actualities, and there is no possibility for there to be any actuality that is unrelated to God.

Does that solve the problem? Well, it's a good start at least. One might wonder how God can be sure that he is the sole delimiter of real possibility. If he is, then his perfect self-knowledge guarantees that he knows that he is, and the problem is solved. But if God is not sure, then he can conclude that either he isn't the sole delimiter of possibility or he doesn't have perfect self-knowledge.