Abstract
A new version of the ontological argument for the existence of God is outlined and examined. After giving a brief account of some traditional ontological arguments for the existence of God, where their defects are identified, it is explained how this new argument is built upon their foundations and surmounts their defects. In particular, this version uses the resources of impossible worlds to plug the common escape route from standard modal versions of the ontological argument. After outlining the nature of impossible worlds, and motivating the need for positing them, the new argument is delineated and its premises justified. It is taken for granted that the argument cannot be sound, since it would prove too much. However, its premises are all plausible, and their denial promises to have significant ramifications. Several intuitive lines of objections are then explored in order to illuminate their shortcomings. The puzzle that the argument poses is therefore not whether the argument is sound, for it clearly cannot be. Rather, it is to place pressure on its plausible premises, so some plausible account of how the argument fails can be identified, and that the devising of such an account promises to be insightful. In the process, we should gain an improved understanding of how such ontological arguments work.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
To believe a thing is impossible is to make it so.
—French Proverb.
In what follows, I will outline what I believe to be a new, and somewhat more compelling, version of the ontological argument for the existence of a God. However, I do not think that the argument is sound. I am not a theist, but even if I were, I do not think that I would want my theistic conviction to rest upon any mere conceptual grounding. Such grounds for how things are, seem, to me at least, insuppressibly suspect. Theists and non-theists alike should be sceptical of positing the existence of things, any things, on such bases. Why, then, you might ask, have I taken the trouble of constructing and presenting an argument that, if sound, would prove the existence of a God on purely conceptual grounds? It is because the premises of the argument are so very plausible, and their denial appears to have significant, and perhaps unwelcome, ramifications. This at least is my interest in the forthcoming argument.
Moreover, I think this new version of the ontological argument makes significant advances upon previous versions, and thus has the potential to help further this long-continued, and seemingly inexhaustible, dialectic. It is not part of my present purpose to explain where the argument goes wrong, for it inevitably must err somewhere, or so I will contend. But I will instead attempt to draw out some uncomfortable consequences of various options that initially seem promising, whilst counselling against journeys down blind alleys, pointing instead towards potentially more fruitful avenues (“Mission IMPOSSIBLE” section). The present work, nonetheless, should be construed as a piece of cartography, rather than navigation; there will be no conclusive destination, but instead, the aim is to establish and detail a challenge for better minds to contend with. That is, I am merely mapping some terrain; journey as you please.
However, before my exposition of the argument (“An impossible ontological argument” section), I will briefly describe the context from which the argument emanated (“The ontological argument” section). This serves to aid comprehension by making clear how my argument builds and improves upon earlier, and more familiar, versions. I will then introduce and outline the impossible worlds account of impossibility upon which that argument crucially depends (“Impossible worlds” section). Only once this solid foundation has been set will I build my argument upon it. So, it is to those foundations that we first proceed.
The ontological argument
Ontological arguments for the existence of a God tend to share the common theme of arguing from the mere concept, to the actual existence, of a God. The first known version originated in the work of the eleventh century C.E. theologian, St. Anselm of Canterbury.Footnote 1 He argued that we have in our minds the concept of a being than which none greater can be conceived. And accordingly, asserted that this greatest conceivable being exists in our minds. Anselm then asks us to assume (for reductio) that the greatest conceivable being exists in the mind only. And claims that to exist merely in the mind alone is a limitation. For it is greater to exist also outside of the mind, to exist in reality, than to exist in the mind only. Indeed, we can conceive of a being that is greater in this respect. Ergo, it is a contradiction to suppose that the greatest conceivable being exists in the mind alone. Hence, via reductio ad absurdum, the greatest conceivable being must exist in reality as well as in the mind.
Of course, this argument is very crude, and among other faults, it founders on its fallacious inference from our mere possession of the concept God, to the existence of the thing, God, in our minds. When we conceive of things, we do not have the conceived of things themselves in our minds, but rather, we have a mere representation of those things in our minds. Thus, it is not the case that the thing—in this instance, God—exists in the mind when we conceive of it. All that exists in our mind when we conceive of a thing is the representation of that thing.
In recent times, the argument has been reinvigorated in light of developments in modal logic. Most notably, Hartshorne (1941b, 1965), Malcom (1960), Plantinga (1974, Chap. 10) developed modal versions of the ontological argument that do not rely on many of the dubious assumptions afflicting some traditional versions of the argument, such as that assumption just shown to be fallacious—that when we conceive of a thing, that thing conceived of must itself exist in the mind, rather than a mere representation of it.Footnote 2 Here is one way to formulate a modal version of the ontological argument:
P1. The concept God is that of the greatest conceivable being.Footnote 3
P2. Conceivability entails possibility.
P3. It is part of the concept the greatest conceivable being that it is satisfied by a being, x, only if x has the essential property of being a necessary existent.Footnote 4
P4. If it is possible that an individual, x, exists and has the property of being a necessary existent, then x actually exists.
C1. A God actually exists
Call this argument ‘MODAL’. MODAL is clearly valid. However, despite its merits, it has been subject to much criticism. This is unsurprising, since many share my intuition that, ‘one cannot build bridges from the conceptual world to the real world.’ (Plantinga 1974, p. 196).Footnote 5, Footnote 6 Some have disputed P2, and argued that despite the conceivability of a God, conceivability does not entail possibility, so that the conceivability of a God does not itself entail that something satisfies that concept. This might be because it seems we can separately conceive of things, x and y, such that x and y are essentially necessary existents, yet are mutually incompatible with each other. Or likewise, it seems we can separately conceive of the truth of incompatible propositions, p and q, such that, if true, they are necessarily so.Footnote 7
Rather than surrendering the entailment from conceivability to possibility, several philosophers have taken this point to indicate that, only if something is conceivable in sufficient detail does its conceivability entail its possibility.Footnote 8 And that these problematic cases are symptomatic of a lack of strictness in the level of detail to which they are conceived, such that they do not blight the entailment construed with the stricter level of conceivability.Footnote 9
One common example from the literature concerns Goldbach’s conjecture, which states that all of the positive non-prime even integers are expressible as the sum of two primes. To date, there is no known proof to show whether or not the conjecture is true. Now, I can conceive of a mathematician waving a few sheets of paper with alien-looking scribbles on them, and claiming to have proved that conjecture. Accordingly, it might be said that I can conceive of Goldbach’s conjecture being true. But this surely cannot be the level of detail required for the conceivability of something—in this case, the truth of Goldbach’s conjecture—to entail its possibility. Since, I can equally conceive of that same mathematician claiming to have disproved Goldbach’s conjecture. And by the same lights, it might be said that I can conceive of Goldbach’s conjecture being false. Yet Goldbach’s conjecture cannot be both possibly true and possibly false, since, as a conjecture of number theory, it bears its truth-value necessarily.
However, it may also seem too strict to some to claim that for conceivability to entail possibility, the level of detail required in that which we conceive is such that we must conceive of the proof itself.Footnote 10 If that were the case, the conceivability of Goldbach’s conjecture would be redundant to our assessment of its possibility, since the actual proof alone suffices to establish its possibility—because what is actually the case is possibly the case. It should be evident, then, that there are varying degrees of detail to which we can conceive of something. And that those who want to hold onto the principle, that conceivability entails possibility, could still dispute the soundness of MODAL by arguing, that the level of detail at which we can conceive of a God is not sufficient to entail the possibility of a God. Accordingly, this would provide a reason to be sceptical of the inference from P1 and P2 to:
P5. It is possible that something satisfies our concept God.
There would seem to be some degree of equivocation on the notion of ‘conceivable’ between P1 and P2, where the notion of ‘conceivable’ used in P2 is stricter than that used in P1, on pain of preserving the truth of the two premises—or so the objection runs.
But for those who dislike P2, for whatever reason, the argument could be run instead by replacing P2 with the perhaps weaker premise, P5.Footnote 11 As just noted, P5 is in fact a consequence of P1 and P2, and to that extent is justified by those premises. By removing P2 from the argument, we would thus make the argument dialectically weaker, since we would need to find alternative grounds for believing P5. The absence of such grounds would promote agnosticism about the soundness of the ensuing argument (cf. Le Poidevin 1996, pp. 25–27). However, I do not want to dismiss the possibility that there could be some alternative justification for P5. The plausibility of the argument depends on the extent to which you find P5 more plausible than:
P5*. It is possible that nothing satisfies our concept God.
Given that our concept God is such that whatever satisfies it would have the essential property of being a necessary existent (from P1 and P3), if it is possible that nothing satisfies our concept God (i.e. P5*), then it follows that nothing actually satisfies our concept God. Since, if a God were to actually exist, it would exist in every possible world, contrary to P5*.
Perhaps those who would reject the ensuing version of MODAL that supplants P2 with P5, would do so on the grounds that they are generally sceptical of the possibility of necessary existents. Such philosophers would either take issue with P3, claiming that our concept God is not such that it is satisfied only by entities that have the property of being a necessary existent. Or they would reject P5 by contending that our concept God is such that nothing could satisfy it.
Still others have questioned whether MODAL shows that there is something wrong with the axiom ‘◊ p → p’ of the modal logic system B.Footnote 12 Consider the following remarks by Kane:
The modal ontological argument is valid in some standard systems of alethic modal logic, e.g., S5 (which is assumed by advocates of the argument like Hartshorne and Plantinga), and in at least one weaker system [i.e., B]. But it is not valid in all. For example, it is not valid in T (Von Wright’s M) or any Lewis system weaker than S5, including S4. What should all this mean for our assessment of the argument? Is choosing a modal system in theological contexts as innocent as choosing hors d’oeuvres? Can it be so with a proof for the existence of God at stake? (1984, p. 336)
Those philosophers, who are tempted to undermine MODAL by rejecting the modal logic system B, must take issue with P4. But this manoeuvre is ill-advised as a response to MODAL, since MODAL can be modified so that it does not rely on modal system B. This is done by replacing P3 and P4 with:
P3*. It is part of the concept the greatest conceivable being that it is satisfied by a being, x, only if x has the essential property of being an actually necessary existent.
P4*. If it is possible that an individual, x, exists and has the property of being an actually necessary existent, then x actually exists.
Here we have made a subtle alteration to P3 and P4, where the important property is not that of being a necessary existent, but rather that of being an actually necessary existent. The term ‘actually’, here, and elsewhere, unless stated otherwise, should be read as expressing the rigid, rather than the non-rigid, notion of actually. That is, that ‘actually’ always takes us back to the perspective of our world, so that we consider which worlds are accessible from our world when considering what is actually possible or impossible.Footnote 13 When we consider whether some possible entity satisfies the concept God, then we need to look to all the worlds accessible from our world, not those worlds accessible from the world of the entity in question. Given these two modifications to MODAL, the resulting argument would rely only on the extremely weak modal system T to support the conditional expressed in P4*.Footnote 14 This is because we require the inference rule ‘ p→ p’ to get us from the actual necessity of a God, to its actual existence.
Furthermore, it might be that this meagre assumption could be done away with also if we modify P3* and P4* accordingly:
P3**. It is part of the concept the greatest conceivable being that it is satisfied by a being, x, only if x has the essential property of being an actual existent.
P4**. If it is possible that an individual, x, exists and has the property of being an actual existent, then x actually exists.
Here we have exchanged the essential property of being an actually necessary existent, to that of being an actual existent. Admittedly, this does appear prima facie to be an improvement to the argument; the resulting argument would be simpler. However, I worry that those following in the footsteps of Kant, Frege, et al., might have ingrained in them the belief that existence is not a (first-order) property, and may well find the property of being an actual existent equally suspect. It is unclear to me that the arguments offered for thinking existence is not a property are at all adequate, but I will assume there is some substance to them for the sake of this essay.
Finally, an objector might take issue with P1. That is, they might claim that our concept God is not in fact that of the greatest conceivable being. Perhaps this is because they cannot make sense of terms like ‘greatest’, ‘best’, ‘perfect’, ‘supreme’, and others of their kind, without their being combined with some qualifying determinate. This qualifying determinate would be required to provide the relative ordering of things with respect to it. These bare superlative terms, when lacking a qualifying determinate, appear to be semantically incomplete. So though I can understand what is meant by ‘the greatest trampolinist’, or ‘the best botanist’, or ‘the perfect sphere’, I cannot grasp what is meant by ‘the greatest’, or ‘the best’, or ‘the perfect’ simpliciter (i.e. without qualification). Perhaps the qualification is meant to be that of ‘being’. But if so, I see no reason to think that any one existent is better with respect to being than any other.Footnote 15 So even this charitable interpretation of the expression, ‘greatest conceivable being’, appears to be incoherent. Or maybe an objector will contest P1 on the grounds that the concept God is such that we can conceive of a greater being. For example, it might be thought that a God’s power would be limited with respect to the ability to sin, and we can conceive of a being that, ceteris paribus to a being that satisfies our concept God, has in addition the ability to sin.
Again, both these strategies seem fruitless,Footnote 16 since we can easily alter the argument to avoid the problematic phrase, ‘greatest conceivable being’, seemingly without loss to its persuasiveness. Instead of P1 and P3*, we could have:
P1*. It is conceivable that a being satisfies the concept God.
P3***. It is part of the concept God that it is satisfied by a being, x, only if x has the essential property of being an actually necessary existent.
Nevertheless, it remains probable that some objectors will dispute even P1*. They might do this by claiming that the concept God is incoherent. Perhaps this will be because the concept God is such that some of the properties of any being satisfying it would have to be incompatible. Or alternatively, that our concept God is relational, such that the defining characteristics are mostly or wholly negative. This would then allow us only negative conceivability—that the concept is coherent (cf. Chalmers 2002, Sect. 2). It might then be argued that the relevant notion of conceivability contained in P1*, that concords with the use of conceivability in the antecedent of P2, is that of positive conceivability—whereby conceiving here means roughly to ‘…imagine (in some sense) a specific configuration of objects and properties.’ (Ibid., p. 150). I make no judgements about the success of these responses. This merely serves as a brief introductory sketch to a version of the kind of ontological argument that I intend to develop in this paper.
Impossible worlds
An impossible world is a world at which an impossibility obtains; where a proposition that cannot be true, is true. If we think of possible worlds as maximal consistent sets of propositions—as I take it most, or at least many, philosophers do—then we can think of impossible worlds as maximal inconsistent sets of propositions.Footnote 17 Whilst possible worlds are ways the world could have been, impossible worlds are ways the world could not have been.Footnote 18 Both kinds of ways are plentiful. Just as there are many ways things could have been, it intuitively seems as though there are also many ways that things could not have been. In this respect some have thought it inadequate to lump impossibility into one broad category of that which is not possible—impossibility is more fine-grained than that. For example, not only could Socrates not have been a frog, he could not have been a rhombus either.Footnote 19 Instead of just asserting that neither of these statements expresses a possible truth—that there is no possible scenario that represents them—if we accept an impossible worlds account, we can claim that the two statements describe distinct impossible scenarios, each of which is represented by a distinct impossible world.
Another reason to believe in impossible worlds is that they can be employed to great effect in offering an account of the semantics of counterpossible conditionals (cf. Nolan 1997; Yagisawa 2010, Sect. 8.3). On the standard Lewisian semantics, counterfactuals have the following truth-schema: ‘If it were the case that P, then it would be the case that Q’ is true at some world, w, if, and only if, there is some metaphysically possible world, w′, at which P and Q is true, and w′ is closer than any metaphysically possible world, w, where P and not-Q is true, or there is no metaphysically possible world where P is true (cf. Lewis 1973).Footnote 20 As per the second disjunct of the truth conditions for the Lewisian semantics, counterfactuals with impossible antecedents are vacuously true. This may or may not have some intuitive plausibility for counterpossible conditionals with logical or mathematical falsities in the antecedent. However, things are less satisfactory for other kinds of antecedent impossibilities. Consider the following counterpossible conditionals:
-
(A)
If Hobbes had squared the circle, then geometricians would have been amazed.
-
(B)
If Hobbes had squared the circle, then Socrates would have been an astronaut.
-
(C)
If the futon were red-all-over and green-all-over, then it would have been a spectacle to see.
-
(D)
If the futon were red-all-over and green-all-over, then the moon would have been hollow.
Whilst A intuitively seems true, it does not give the impression of being vacuously so. Its truth appears to be owed to something else. On the other hand, and to support this assessment of A, intuitively, B seems false. Likewise, while C intuitively seems true, it too does not give the impression of being vacuously so; whereas D gives the impression of being false. The Lewisian semantics cannot adequately accommodate these intuitions; it cannot accommodate the fact that we want to offer a different assessment of conditionals like A and C, compared to those like B and D. However, with the tool of impossible worlds at our disposal, the Lewisian semantics for counterfactuals can be modified to accommodate our intuitions. We can offer the following alternative truth-schema: ‘If it were the case that P, then it would be the case that Q′ is true at some world, w, if, and only if, there is some metaphysically possible or impossible world, w′, at which P and Q is true, and w′ is closer than any metaphysically possible or impossible world, w″, where P and not-Q is true.Footnote 21 This account has the added plausibility of giving a less disjunctive account of the semantics for counterfactual conditionals. For these reasons, among others,Footnote 22 impossible worlds are becoming increasingly popular.
Now, just as the advancement of modal logic provided the capacity to develop some sophisticated versions of the traditional ontological argument, I think the adoption of impossible worlds, whatever they may be, suggests a way of developing certain modal versions of the ontological argument.
An impossible ontological argument
Let us now proceed by delineating that new version of the ontological argument; a version that makes vital use of the impossible worlds described in the previous section. The proposed argument has the following form:
P1*. It is conceivable that a being satisfies the concept God.
P6. Conceivability entails either possibility or impossibility.Footnote 23
From P1* and P6, via modus ponens:
I-C1. It is either possible or impossible that a being satisfies the concept God.
P7. It is part of the concept God that it is satisfied by a being, x, only if x has the constitutive property of being an actually necessary existent.
A1. It is possible that a being satisfies the concept God.
P4*. If it is possible that an individual, x, exists and has the property of being an actually necessary existent, then x actually exists.
From A1, P7, and P4*, via modus ponens:
I-C2. There actually exists a being that satisfies the concept God.
A2. It is impossible that a being satisfies the concept God.
P8. If it is impossible that a being satisfies the concept God, then there is an impossible world, w, such that a being in w satisfies the concept God.
P9. If some individual, x, exists at an impossible world, and x has the property of being an actually necessary existent at that world, then x actually exists.
From A2, P7, P8, and P9, via reductio ad absurdum:
I-C3. It is not the case that it is impossible that a being satisfies the concept God.
And from I-C1, I-C2, and I-C3, via disjunctive syllogism:
C2. There actually exists a being that satisfies the concept God.
Call this argument ‘IMPOSSIBLE’. As should be clear from the above presentation, IMPOSSIBLE is valid. Nevertheless, I shall guide the reader through some of the inferences, as I examine and justify the premises. To begin with, let us grant P1* for the sake of argument. As we shall later see, nothing much rests upon it for the wider significance of the argument kind. The “conceivable” here is to be understood in a very weak sense of the term. Perhaps all that is needed for us to conceive of a God is that we grasp the concept God. That is, P1* merely requires an understanding of what it would take for something to satisfy the concept God. I take it that those who think our concept God is incoherent will find all arguments for theism unattractive. But if this is your response, independent reasons must be provided for the rebuttal. And given the slipperiness of the concept, this is no mean feat.Footnote 24
However, it is perhaps not entirely implausible that it is inconceivable that something satisfies the concept God. The ‘inconceivable’ here permits of further divergent readings. Firstly, it might be thought that something is inconceivable if it is inconceivable by us. This might be due to our limited cognitive capacities. Secondly, it might be thought that something is inconceivable because it is inconceivable in principle. Such things would be inconceivable even for an ideal conceiver. The worry that the existence of a being satisfying the concept God is inconceivable in the former sense poses no threat to the truth of P1* if we thought it were not inconceivable in the latter sense. Though our justification for P1* would rely on our actually being able to conceive of the existence of a being satisfying the concept God. Since if we could not conceive of such a being, due perhaps to our limited cognitive capacities rather than any manifest incoherencies in the concept, then we would not know whether or not such a being is inconceivable in principle. And thus, we would not know whether P1* is true.
If you have these concerns about P1*, fear not, as we can easily modify the argument by supplanting P1* and P6 for I-C1, and beginning the argument from there. My preference, however, remains with the argument that includes P1* and P6, for two reasons. Firstly, that way of presenting the argument lends more support to P7. Some philosophers may well attempt to reject P7 on the grounds that they do not think there is a property of being an actually necessary existent. Yet, all that need be under consideration for P7 in the preferred version of the argument is that that property is conceivable, in that weak sense of conceivable mentioned earlier. If that property is conceivable, as it must be if the concept God is conceivable—where it is constitutive of the concept God that a being satisfying that concept has this property, at least given the concept God that is currently under consideration—then it must be either possible or impossible that it has some instance. Once this much has been granted, there would seem to be no principled reason why this property cannot be built into the concept God, or why it would affect whether that concept is satisfied at a possible or impossible world.
Without the justification for I-C1 that comes from P1* and P6 combined, and assuming no alternative justification can be found, the doors would be open for a potential objector to deny P7, on the grounds that a being instantiating the property of being an actually necessary existent cannot constitute a part of our concept God, since the latter is, whilst the former is not, possibly or impossibly satisfied. And if this were the case, the theist would then be faced with the option of either rejecting P7, or rejecting I-C1 and admitting that our concept God has no instance. I assume that given this choice the theist would opt for the former option, unless they were to betray their theistic commitments. So the inclusion of P1* and P6 blocks one potential objection to the argument.
Secondly, an objector might similarly be tempted to reject I-C1 on the grounds that the concept God is neither satisfied at a possible nor an impossible world. The conceivability of a being satisfying the concept God (P1*), and the entailment from conceivability to possibility or impossibility (P6), allows the theist a foot-in-the-door to force the dilemma proposed in I-C1. It should be clear then that P1* and P6 provide good justificatory grounds for believing I-C1. Nonetheless, this does not close off provisions from alternative justificatory grounds for believing I-C1. Other forms of justification could only make the argument stronger. But to merely state I-C1 without appeal to P1* and P6, or any other justification for that matter, would result in a dialectically weaker position for a defender of IMPOSSIBLE.
And whilst we are drawing our attention towards P7, I should like to note an important difference that it bears to its parallel premise in MODAL, namely P3***. P3*** describes the property of being an actually necessary being as an essential property, whereas I have avoided that description in P7 in favour of describing that property as being constitutive of a being satisfying the concept God. It will be enlightening, then, to explain what I take a constitutive property to be, and how it might differ from an essential property. Philosophers have understood the notion of an essential property in different ways. Part of the reason for describing the kind of property I am concerned with as ‘constitutive’ rather than ‘essential’, is to prevent confusing it with one particularly well-known of those distinct ways of understanding ‘essential’. That way that essential properties have been understood that I especially want to contrast with what I am calling ‘constitutive properties’, is that an essential property of a thing or kind, is a property that is de re necessary of that thing or kind. That is, a property that a thing has in every possible world that it exists, or a kind of thing has in every possible world (and every instance of it in that world) where that kind of thing exists.Footnote 25 Whatever the merits of this way of understanding essential properties, it places no constraints on what things are like at impossible worlds. For the purpose of my argument, and perhaps for purposes more general, this clearly will not do.
The notion of a constitutive property that I am using is stricter than that of an essential property understood in the sense just described. A constitutive property of a thing or kind of thing, x, is a property of x, such that x would not be that thing or kind of thing respectively, if it lacked that property. So, to give an example in the case of a kind, a triangle would not be a triangle if it lacked three angles. It is thus constitutive of a thing being a triangle that that thing has three angles. And to give an example in the case of an individual, Socrates would not be Socrates if Socrates lacked the property of being identical to Socrates. Likewise, when P7 states that the concept God is such that it is constitutive of a being satisfying that concept that it possess the property of being an actually necessary being, if that being lacked that property, it would be a different thing and a different kind of thing. Since this notion of being a constitutive property is quite understandable, it seems to present no further difficulties to our ability to conceive of a being satisfying our concept God—in the weak sense of conceivable descried earlier. Accordingly, the objector would do well to find some other point of dispute.
There is perhaps an ambiguity in the concept God, as I have being using it, that might have here occurred to the reader, between it being a concept of an individual entity or of a kind of entity. But whether it is a concept of God, the individual, or God, the kind, should make little difference to our argument. However, I am not sure that I can make much sense of the former interpretation. And in so far as I do understand it, I worry, and hope that others might share this worry, that it presupposes the existence of God, the individual. For this reason, it is the latter interpretation of the concept God, as a concept of a kind of entity, which I have intended throughout this essay. This means that if the concept God were satisfied at an impossible world, that satisfying entity would possess the property of being an actually necessary existent at that world, lest it not be a God.
As already noted, from P1*—it is conceivable that a being satisfies the concept God—and P6—conceivability entails either possibility or impossibility—we arrive at I-C1: It is either possible or impossible that a being satisfies the concept God. If we can show that C2—there actually exists a being that satisfies the concept God—follows from one disjunct of I-C1, whilst the other disjunct leads to a contradiction, then we can validly infer C2 via disjunctive syllogism. Consider then the sub-argument from the first disjunct of I-C1, namely:
A1. It is possible that a being satisfies the concept God.
P7. It is part of the concept God that it is satisfied by a being, x, only if x has the constitutive property of being an actually necessary existent.
P4*. If it is possible that an individual, x, exists and has the property of being an actually necessary existent, then x actually exists.
From A1, P7, and P4*, via modus ponens:
I-C2. There actually exists a being that satisfies the concept God.
When combined with P4* and P7, we can validly infer IC-2 from A1. Since the conjunction of A1 and P7 gives us a particular case of the antecedent of P4*, and accordingly we are allowed to infer the corresponding particular case of the consequent of P4*, namely I-C2. Our justification for accepting P4* has already been given with respect to MODAL. Indeed, the reasoning from A1 to I-C2 is the same as that employed in that modal ontological argument. This was found to rest upon some quite plausible assumptions. It is likely then that the objector will reserve their contention for the inference from the second disjunct of I-C1 to I-C3.
Let us consider, then, the sub-argument that runs from the second disjunct of I-C1:
A2. It is impossible that a being satisfies the concept God.
P7. It is part of the concept God that it is satisfied by a being, x, only if x has the constitutive property of being an actually necessary existent.
P8. If it is impossible that a being satisfies the concept God, then there is an impossible world, w, such that a being in w satisfies the concept God.
P9. If some individual, x, exists at an impossible world, and x has the property of being an actually necessary existent at that world, then x actually exists.
From A2, P7, P8, and P9, via reductio ad absurdum:
I-C3. It is not the case that it is impossible that a being satisfies the concept God.
When combined with P7, P8 and P9 we can validly infer I-C3 via a reductio ad absurdum on A2. This can be derived by first inferring the first conjunct of the antecedent of P9 from premises A2 and P8 via modus ponens. We can then combine P7 together with the consequent of P8 to fulfil a particular case of the antecedent of P9. Again, via modus ponens this allows us to infer the corresponding particular case of the consequent of P9. But since the consequent of the particular case of P9 is that there actually exists a being that satisfies the concept God, it must be possible that a being satisfies the concept God. This is of course assuming that the actual world is accessible from the actual world, and as such is possible relative to the actual world. But this assumption is hardly contestable; if anything is actually possible, then surely what is actually the case must be; we are dealing with an alethic modality here. However, the claim that it is possible that a being satisfies the concept God, contradicts our initial assumption, A2, that it is impossible that a being satisfies the concept God. And as such, we can reject A2 on the grounds that it leads to a contradiction when combined with other plausible premises.
The justification for P8 crucially depends on a commitment to impossible worlds. I gave some motivations for believing in impossible worlds in the previous section. The thought is that, if it is impossible that a being satisfy the concept God, as per assumption A2, then there must be an impossible world where a being does satisfy the concept God. This just falls out of the commitment to impossible worlds. But for it to be a God at that impossible world, that entity must possess the constitutive property of being an actually necessary being, in accordance with P7, lest it be some other kind of entity besides a God. However, if there were no impossible world where a being satisfied the concept God, then it would also not be the case that it is impossible for a being to satisfy the concept God, contrary to A2. Yet, now it seems that the inference from a God’s having the property of being an actually necessary existent in some world, to a God actually existing, is no more problematic where that world is of the impossible variety than if it were of the possible variety. It relies merely on the very weak modal logic system T, plus the meaning of the predicate ‘…is an actually necessary existent’. Accordingly, P9 also appears to be very plausible.Footnote 26 Again, it appears that an objector would make better use of their time looking elsewhere in the argument for weaknesses. From this we can then derive A1 via disjunctive syllogism, and reiterate the reasoning that brought us to I-C2. But, this time I-C2 does not merely rest on the assumption that A1 is true. A1 has been deductively derived, and as such, we can assert C2 without assumption.
Alternatively, if you dislike disjunctive syllogism as an inference rule,Footnote 27 the same result can be achieved by showing that C2 follows from each disjunct of I-C1, applying the uncontroversial inference rule of constructive dilemma. Here we follow the same line of reasoning from the A2 disjunct of I-C1, but instead of going on to show that A2 leads to a contradiction, we merely stop at the derivation of P9’s consequent. One corollary of adopting this alternative route to C2 is that, although we arrive at our destination—the conclusion that there actually exists a being that satisfies the concept God—we do not demonstrate along the way that,
Antitheist: There actually does not exist a being that satisfies the concept God.
is false. Since those who do not like disjunctive syllogism may not think that we can move from the truth of C2—that there actually exists a being that satisfies the concept God—to the negation of Antitheist, it remains an option to accept both C2 and Antitheist. This, however, would be a deeply unsettling result.
I suspect that few, if any, would be happy with this consequence as a result of accepting the seemingly reasonable premises of our argument. It is one thing to accept that something does and does not exist in an impossible world, and quite another to accept that something actually does and does not exist. At the very least, a deeply unattractive consequence of our premises would have been drawn out. But I think that most would think it quite reasonable to accept that things cannot both actually exist and actually not exist (at the same time). And this can be added as a further premise if need be. For the typical reasons given for rejecting the law of non-contradiction—the underlying reason for rejecting disjunctive syllogism—do not support the possibility of something actually existing and actually not existing (at the same time). We have then, here, a justification for the premises of the argument. And overall we have found that, at least prima facie, the premises seem rather plausible. Let us dig deeper and make a few critical comments in the final section on some potential responses to the argument.
Mission IMPOSSIBLE
I imagine that the instinctive reaction of many to IMPOSSIBLE will be that, if successful, it would open the gates that hold back the impossible, and allow us to postulate the actual existence of many other intuitively impossible beings.Footnote 28 This is clearly an unwanted conclusion for a theist hoping to defend IMPOSSIBLE, and mirrors Gaunilo’s famous ‘Perfect Island’ objection to the traditional ontological argument. If the theist were to defend IMPOSSIBLE, they would need some way of restricting the variety of molds subject to the argument form so that it can only be employed to demonstrate the existence of a being satisfying the concept God, and not the existence of the intuitive impossibilia.
St. Anselm relied on the concept God as that of the greatest conceivable being to make his case for treating the concept God differently from other concepts. The superlative nature of any being satisfying that concept ensured that the property existence was essentially instantiated by anything satisfying that concept. But unlike other superlative concepts, it was conceivable alone of the concept God that it be satisfied. It was thought of our other superlative concepts, like that of greatest island or greatest painting, that it be inconceivable that they be satisfied, since all potential satisfiers of those concepts are superable without limit; or in St. Anselm’s terminology, they have no ‘intrinsic maxim’. For these other kinds of superlative concepts, there simply is no greatest that is conceivable; there could always be a greater one conceived. It was claimed that this was not so for our concept God. Though, it is in fact hard, at least for me, to understand why the concept God alone avoids this charge.
However, the construal of the concept God as that of the greatest conceivable being was earlier found to be problematic. The superlative interpretation of the concept God was found to be, at best, of no real help to the theist, since it is unclear what it takes for one being to be greater at being than another. At worst, the superlative conception was found to be semantically incomplete; the notions of ‘greatest’, ‘best’, ‘perfect’, ‘supreme’ etc. all require some determinate to fix the subject of the relative ordering. Whilst someone can be the best gladiator, or the greatest poet, whereby the determinates, here, are ‘gladiator’ and ‘poet’ respectably. But the notion of something being the best, greatest, or perfect simpliciter is nonsensical.
Let us not concern ourselves with the difficult task of trying to distinguish the concept God alone as that which has constitutively built into it the condition of being satisfied only by a being that is actually necessary. I leave this difficult task to those with a greater inclination. Nevertheless, an objector who criticises IMPOSSIBLE merely on the grounds that it sets a precedent for arguing many apparent impossibilia into existence, will have missed the point of the exercise entirely. All they would have established is that IMPOSSIBLE is not a sound argument. But this was freely admitted. The difficulty arises from pinpointing where exactly the argument errs. As such, this objector’s response offers us little headway on the real issue. Yet, perhaps it does lend some support to the idea that rejecting P1* is not the way to go. For even if the concept God is incoherent, a parallel argument might potentially be employed using the concept of some other kind of being for which it is constitutive of that kind of being that it is an actually necessary existent.Footnote 29
Notably, IMPOSSIBLE will not be a problem for those accounts of modality that do not invoke impossible worlds. For this reason, P8 stands out to me as the most vulnerable premise. And accordingly, it does not seem unreasonable, for someone to present IMPOSSIBLE as a case for favouring, ceteris paribus, those accounts of modality that do not camber one towards a commitment to impossible worlds.Footnote 30, Footnote 31 But let us not be too hasty here. There are perhaps some alternative responses for those interested in defending impossible worlds. One interesting response would be to claim that I have overestimated the scope of modal space. Underlying P6, the claim that conceivability entails possibility or impossibility, is the following two principles:
Proposition: Conceiving is a propositional attitude, such that there is some propositional content to that which is conceived, which is capable of being true or false.
Ubiquity: For every conceivable proposition p, p represents something that is either possible or impossible.Footnote 32
Both seem to be very plausible. But if one were to be denied, P6 would be false, and our justification for believing I-C1 would be in jeopardy. Proposition seems to me to be quite plausible.Footnote 33 However, I suspect that Ubiquity might become a target for some objectors. They would likely claim that, despite our ability to conceive of a being satisfying the concept God, the proposition that there is something that satisfies the concept God represents a situation that has no modal status, it is amodal—the existence of a God is neither possible nor impossible.Footnote 34 As things stand, Ubiquity is deeply integrated into our modal logic, such that, if a proposition, p, is not true at any possible world, then p’s truth is impossible. And under the impossible worlds assumption, Ubiquity demands that there are impossible worlds at which propositions that cannot be true, are true. The impossible worlds assumption makes a distinction between that which is not possible, and that which is impossible. That is, impossibility is not given a relational or negative characterisation, as the absence of possibility, but rather is given a non-relational or positive characterisation, namely, being so at an impossible world.Footnote 35 Though it is typical, and indeed natural, to accept, in accordance with Ubiquity,Footnote 36 that not being possible entails being impossible, given the positive characterisation of the impossible, there opens up conceptual space for a view that denies this.Footnote 37
However, given the sheer intuitiveness of Ubiquity, there would need to be sufficient justification for denying it. Indeed, it sounds incredibly odd, if not straightforwardly absurd, to say that something is not possible, but it is not impossible.Footnote 38 Though, the awkwardness of such a statement perhaps could be explained in terms of it being a category mistake to apply any modal notions at all to that represented by a certain proposition. So let us look briefly at how you might go about trying to undermine Ubiquity. One line of argument would draw its influence from certain a posteriori necessities. It would be argued that we can conceive of Hesperus being non-identical with Phosphorus, but that this situation is neither possible nor impossible. Since, the scenario is conceivable, there must be some proposition that represents the content of that which is conceived (from Proposition). But if what is represented by that proposition is neither possible nor impossible, as per this line of argument, then Ubiquity must be false.
The case assumes that the identification of Hesperus and Phosphorus is stronger than necessity. More would need to be said to justify this claim and spell out just what level of strength this identity relation has.Footnote 39 Still, prima facie, this seems to be an interesting case to explore. It is, nevertheless, not at all obvious to me that the ensuing investigation would be successful in disposing of P6. There appears to be two ways of interpreting the case in question:
-
1.
Conceptual: This interpretation states that we can conceive of our concept Hesperus being satisfied by something distinct to that which satisfies our concept Phosphorus.Footnote 40
-
2.
Metaphysical: This interpretation states that we can conceive of the individual actually picked out by our concepts Hesperus and Phosphorus, being distinct from itself.
Now, the former interpretation appears to be conceivable. This explains how prior to the empirical discovery that Hesperus and Phosphorus are actually numerically identical to Venus, people were able to comprehend their distinctness. However, interpretation 1 presents no problem to Ubiquity, since there accordingly appears to be no good reason to think that the concepts be applicable to distinct things in some (possible or impossible) world. The second interpretation, on the other hand, would seem to challenge Ubiquity, if it were at all conceivable. But why should we believe that we can conceive of that? The absurdity of the situation that interpretation 2 describes prevents our being able to conceive of it. Just as I cannot conceive of square circles, I cannot conceive of an individual being distinct from itself.Footnote 41 It seems plausible that those who think they can conceive of 2, are really conceiving of 1, and mistaking it for 2. The objection equivocates between discussing the concepts in the conceivability claim—that we can conceive of our concept Hesperus being satisfied by something distinct from that which satisfies our concept Phosphorus—and the individual that satisfies those concepts in the propositional claim—that the proposition (assuming on behalf of the objector that it is a proposition) that the individual that satisfies our concepts Hesperus and Phosphorus is distinct from itself is not true in any (possible or impossible) world.
Yet, even if this last avenue were to reveal something suspicious about P6, there would still be the further task of showing that the reasons for rejecting that conceivability entails possibility or impossibility in any suggested counterexamples are transferable reasons for believing that our conceiving of a being satisfying the concept God does not entail that the satisfying of the concept God is either possible or impossible. And it would also need to be shown where the weaker argument, that bypasses P1* and P6, and instead begins at I-C1, goes wrong. Nevertheless, it would be both a significant and surprising result if we were to discover that conceivability did not entail possibility or impossibility. One consequence of this outcome would be that there would then be counterpossible conditionals where the antecedent is neither possible nor impossible, such that the impossible worlds semantics for counterfactuals would either be unable to deal with, or produce a counter-intuitive result for, them. This would undercut one of the main motivations for believing in impossible worlds in the first place.Footnote 42, Footnote 43
Another objection that initially seems promising, but ultimately goes nowhere, targets P9. This objection states that one way that a world can be impossible is by misrepresenting modal space. So from the perspective of the impossible world we can follow the sub-argument that proceeds from A2, to the conclusion that at some impossible world, w, it is true that a being satisfying the concept God actually exists, but still maintain that it is not the case that at the actual world, @, that a being satisfying the concept God actually exists. The proposition a being satisfying the concept God actually exists is in fact world-indexed. Though such world-indexed propositions, if true at some possible world, are necessarily true,Footnote 44 this objector claims that there are some impossible worlds where world-indexed propositions are not true, despite being true at every possible world. Importantly, this result would show that iterated world-indexing is not redundant (contrary to Yagisawa 2010, Sect. 3.5), at least if we are allowing impossible worlds into our modal space. Again, this would be a significant result.
However, I do not think that I can make any sense of this proposal. The proposition that a being satisfying the concept God actually exists is about how the actual world is. It is constitutive of a world that it is the way that it is. Worlds are defined by their members.Footnote 45 If things were different at a world, it would not be that world. It would be some other world. But then, if a being satisfying the concept God does not exist at the actual world, then the proposition that a being satisfying the concept God actually exists is surely not going to be true at any world, lest the proposition not be about the actual world. Modal space, it would seem, cannot be misrepresented in the suggested way, even at impossible worlds. The confusion here comes about because when we conceive of things actually being different to how they are, the ‘actually’ here is the non-rigid ‘actually’. We thus understand things being actually different to how they are by “looking” at how things are at some other world where things are that way (cf. Yagisawa 2010, Sects. 8.10, 8.11.2, and 8.11.3). But I have explicitly stated that the ‘actually’ I have been using is the rigid notion of actually.
Yagisawa (2010, Sect. 8.10) in fact proffers a way to understand things being (rigidly) actually different from how they (rigidly) actually are.Footnote 46 According to Yagisawa, to do so we must consider how things are with respect to some ‘non-local metaphysical space’, where the modal facts are different: ‘Since in our metaphysical space, Twin Earth does not exist at @ (as we assume), we need an alternative metaphysical space, a metaphysical space in which Twin Earth exists at @. For any such metaphysical space M, it is true to say of any world w in M that w is m such that Twin Earth exists at @ in M.’ (Ibid., p. 220).Footnote 47 Let us grant Yagisawa for the sake of argument these alternative metaphysical spaces. IMPOSSIBLE concerns what is actually impossible, not what is impossible in some alternative metaphysical space. So Yagisawa’s way of understanding things being (rigidly) actually different from how they (rigidly) actually are, helps our fictional objector not.
Likewise, given P7—that it is part of the concept God that it is satisfied by a being, x, only if x has the constitutive property of being an actually necessary existent—if the entity in the impossible world were not an actually necessary existent, then it would not satisfy our concept God. That is, if, from the perspective of our world, i.e. the rigidly actual world, the proposition that a being satisfying the concept God actually exists is false, then any entity we care to arbitrarily select at an impossible world would not satisfy the concept God. This is regardless of whether or not there are other modal spaces, where at the actual world in one of those modal spaces, a being satisfying the concept God actually exists. This is because when we describe an actually impossible world, we use our actual concepts, not the concepts we would have if we were at that impossible world. Even if the actual world is different in alternative modal spaces, our use of the term ‘actual’ picks out our actual world, in our modal space. It should be clear then that this line of objection to IMPOSSIBLE does not hold up under scrutiny.
This response should now put us in good stead to reply to another objection to P9. It has been suggested to me that P9 will only work to establish the existence of possible beings, and not impossible beings, that exist at an impossible world. On the one hand, this is obviously true, but merely supports IMPOSSIBLE. That argument shows that a being that satisfies our concept God, is not an impossible being, since it is constitutive of a being that satisfies that concept that it instantiates the property of being an actually necessary existent. As was earlier explained, if a being does instantiate the property as being an actually necessary existent, it instantiates that property in virtue of being actually necessary. And since the actual world is an actually possible world, that being must also be an actual being. Eo ipso, it must be a possible being, i.e. it will not be an impossible being.
But there is a further element to this objection. An impossible being may well instantiate incompatible, or even inconsistent, properties. The thought seems to be that the properties of actually existing and of actually not existing can be jointly instantiated by a being at an impossible world. This objector claims that what IMPOSSIBLE establishes is that a being satisfying the concept God both actually exists and actually does not exist. Which are we to believe; that a being satisfying the concept God actually exists, or that it actually does not exist? The objector would claim that IMPOSSIBLE moves us neither one way nor the other, and uses this contradictory result to suggest that nothing has been established regarding the actual existence of a being satisfying the concept God.
However, as we have just seen in our response to the previous failed objection to P9, worlds appear to be constrained in their representation of modal space. It was shown that if a world, w, be it impossible or not, represented an individual, x, as (rigidly) actually existing, as the antecedent of P9 demands, given the proviso that the actual world is one of the actually possible worlds, then the actual world must be such that x does in fact exist. Likewise, if that same world, w, further represents that same individual, x, as also (rigidly) actually not existing, then the actual world must be such that it is not the case that x exists. But how can this be?
One way of interpreting the result being proposed here is just that option, discussed at the end of section three, of accepting both C2, that there actually exists a being that satisfies the concept God, and Antitheist, that there actually does not exist a being that satisfies the concept God. It was noted there that the negation of Antitheist, cannot strictly be proved from the premises without the law of non-contradiction: ‘¬ (p & ¬p)’. And that those accepting impossible worlds may well be motivated to deny that logical truth. Yet, since we are able to show via IMPOSSIBLE that C2 follows from the premises, C2 itself can be used as sufficient reason to reject Antitheist. To reiterate, it is one thing to accept that something does and does not exist in an impossible world, and quite another to accept that something actually does and does not exist. And the objector’s proposal would have us accept the latter. However, most, I think, would not entertain this result.
An alternative, and more interesting way of interpreting the result, is to offer a coherent interpretation of an entity in an impossible world both actually existing and actually not existing without there being an entity at the actual world that actually does and does not exist. This strategy relies on the contingent identity thesis, and serves to highlight a real hole in the argument that we shall shortly plug. According to Gallois (1998, Chap. 6), things can be identical in one world but non-identical in another. I shall not delve into the complicated details of the account, since this would take up too much space. The reader is referred to Gallois’s excellent book for further details.
Importantly, one consequence is that, if entities are identical at a world, they share all their properties, including modal properties, at that world. This follows from the following restricted version of the principle of the indiscernibility of identicals: ‘(LLw): (x) (y) (w) [in w: x = y → (in w: Φx → in w: Φy)]’ (Ibid., p. 145). This means that if the thing that stands on one side of the identity relation at some impossible world, w, actually exists, whilst the thing that stands on the other side does not actually exist, as the contingent identity theory permits, then a being existing at w can instantiate at w both properties of actually existing and actually not existing, since they are not contradictories under the contingent identity thesis. The problem this raises is that, the being in the impossible world that satisfies the concept God may actually exist at that world, not in virtue of satisfying the concept God—that thing does not actually exist—but in virtue of the kind of entity that stands on the other side of the identity relation.
Accordingly, there may be some philosophers who are inclined to find the argument less startling than it initially appears to be. Consider the following criticism that Plantinga offers against Hartshorne and Malcom’s version of the modal ontological argument:
As it is stated, however, there is one fairly impressive flaw: even if an essence entailing is maximally great in W is exemplified, it does not so far follow that this essence entails is maximally great in α. For all we have shown so far, this being might be at a maximum in some world W, but pretty insignificant in α, our world. So the argument does not show that there is a being that enjoys maximal greatness in fact; it shows at most that there is a being that in some world or other has maximal greatness. (1974, p. 213)
It would seem that IMPOSSIBLE, as it presently stands, is vulnerable to an analogous version of this ‘fairly impressive flaw’. The conclusion of IMPOSSIBLE, C2, is that there actually exists a being that satisfies the concept God. But all that the argument so far establishes is that an actually existent being satisfies the concept God in some world or other, be it possible or impossible. Perhaps in the actual world, that being fails to satisfy the concept God. Maybe, I satisfy the concept God in an impossible world. Yet, if this were so, it would be less than clear that I actually deserve homage for my otherworldly deeds or Godliness. What is needed to captivate this objector is not C2, but the stronger claim:
C3. There actually exists a being that actually satisfies the concept God
That is, what is demanded by this objector is that the satisfaction of the concept God be actual satisfaction, the satisfaction of the concept God in the actual world. For those with resolute dispassion for anything less than C3, fear not. We need only insist that,
P10. It is part of the concept God that it is satisfied by a being, x, only if x has the constitutive property of satisfying the concept God in the actual world.Footnote 48
But once this has been granted, it should be clear that, C2—there actually exists a being that satisfies the concept God—can be combined with P10 to yield the desired claim, C3. Indeed, this stratagem is in fact analogous to the one that Plantinga’s himself employed to resolve this issue. And neither is it hard to find support for P10. I follow Plantinga in drawing attention to the remarks of Findlay in support of something close to P10:
Not only is it contrary to the demands and claims inherent in religious attitudes that their object should exist “accidentally”: it is also contrary to those demands that it should possess its various excellences in some merely adventitious or contingent manner. It would be quite unsatisfactory from the religious stand point, if an object merely happened to be wise, good, powerful, and so forth, even to a superlative degree, […] And so we are led on irresistibly, by the demands inherent in religious reverence, to hold that an adequate object of our worship must possess its various qualities in some necessary manner. (1948, pp. 180–181)
We also find in this passage additional support for P7. And yet I see no obvious additional worries that would threaten P10 that has not yet been addressed elsewhere in this essay.
This issue now resolved, an alternative line of response to IMPOSSIBLE that might be suggested is to question whether we can really conceive of an entity, or kind of entity, such that, if it were not actual, or there were no actual instance of that kind, then we could not have conceived of that thing, or kind. Here it seems that our ability to conceive is constrained not merely by internal cognitive states, processes, and events, but also, by how the actual world is, external to our mind. And as such, the argument might appear to assume an unwarranted externalism about conceptual content, and should be rejected on those grounds.
But this will not have got things quite right. Firstly, the property in question is that of being an actually necessary existent. This property does not in itself commit you to the actual existence of a being satisfying the concept God, but does so only under the condition that there is an actual world and that it is accessible to itself. There does not appear then to be any commitment to externalism about conceptual content here. There still remains conceptual room, despite its diminutive size, for our conceiving of a being satisfying the concept God not entailing the actual existence of any such being. A fortiori, this line of criticism gets things backwards. We arrive at the actual existence of a being satisfying the concept God in virtue of the conceivability of such a being; a move from the conceptual to the actual, not from the actual to the conceptual.
Finally, a related form of objection to IMPOSSIBLE would be to argue from an externalism about conceptual content. According to this view, our having a concept would depend partly on the way reality is. The objection would run such that if a being satisfying the concept God were not actual, then we could not possess that concept God. And since we do not know whether a being satisfying the concept God actually exists, we do not know whether we possess the concept God—or at least not the particular concept God that has been employed in IMPOSSIBLE. As such, the objector would claim that P1* begs the question, since our ability to conceive of a being that satisfies the concept God would presuppose our possessing the concept God, and as a result presupposes the very point of contention, that a being satisfying the concept God actually exists.
I have very little to offer in response to this objection other than to state that this externalism about conceptual content seems clearly false. We seem able to conceive of all sorts of things that are not reflected in the structure of reality. Nor does it seem plausible to suggest that I do not know what it is that I am conceiving of. I may sometimes be unable to describe what I conceive of, or be confused about just what it is that I am conceiving of. But that is another matter entirely. My concept God certainly seems clear enough that I am able to explicate the important characteristics of the concept that are required to get the argument off the ground. So once again, the response looks to be quite counter-intuitive, and would be very surprising if it were right.
It should now be clear, IMPOSSIBLE permits no easy refutation. It has shown itself to be unusually resilient. But despite this resiliency, we should not let ourselves get carried away and think that the argument is sound. At the very least, I think that I have shown that the result of rejecting one of the premises promises to be a significant, interesting, and perhaps surprising outcome. Your mission, should you choose to accept it, is to reveal the indiscretion in my argument, and settle this matter in whatever way you deem fit.
Notes
In particular, see his Prosologion (1077–1078, Chap. II–III). It is unclear to me that the argument I present in the text is in fact the one Anselm makes. But my concern here is not exegesis, and it is a version of the argument that is at least sometimes attributed to him (cf. Le Poidevin, 1996, Chap. 2).
This version of the argument also avoids the cross-hairs of the renowned Kantian retort, that existence is not a property.
Throughout I am at pains to avoid naming God. Doing so would have risked begging the question of theism by presupposing an entity that answers to that name. Instead, the arguments run from the mere concept God, to its being satisfied by an actual existent. I apologise for the cumbersome, and perhaps at times clunky, prose that results from this decision.
Roughly, a concept, C, is satisfied by a being, x, iff the ascription of C to x expresses a true proposition.
Plantinga bemoans, rather than advocates, this intuition.
Or at the very least, one cannot build bridges from the conceptual world to the “concrete” world. Many in fact do seem happy to build bridges from the conceptual world to the “abstract” world. This would leave the door open for the theist to claim that God is an abstract entity (see Leftow 1990, for a discussion of this proposal). Indeed, on some conceptions of what it is to be an “abstract” entity, God would likely fall into that class. Take for example the construal of “abstract” entities as non-spatial and non-temporal. If God is conceived as transcending space and time, as many do, then God would be one of the abstracta.
It is not clear to me that there is any natural distinction between concrete and abstract entities. For those who hold a disparate attitude towards whether conceptual bridges can be built towards the abstract, compared to the concrete, world, some good reason(s) for treating the abstract and the concrete differently will need to be given. Otherwise, it will be evident that their differential treatment will be resting upon an unprincipled bias. I doubt that a principled way of distinguishing the two kinds of entities, a way that justifies a differential attitude towards building bridges from the conceptual world, can be had.
Though, of course, this is controversial. For example, we might be mistaken about what it is we are in fact conceiving of, thus creating the illusion of possibility.
In Chalmers’ (2002) terminology, it might be claimed that it is ‘ideal conceivability’, as opposed to ‘prima facie conceivability’, that entails possibility.
Consider also the prima facie conceivability of certain time travel stories that are popular in the science fiction literature, but upon closer inspection are revealed to be incoherent.
It only ‘may’ seem too strict, since some might be happy to assign this level of strictness to conceivability for the entailment at a cost of delimiting the number of cases we can conceive of in sufficient detail, such that the conceiving entails the possibility of the conceived. This would add a, perhaps insurmountable, barrier to our ability to acquire knowledge of what is merely possible by using our faculty of conceiving.
The ‘perhaps’ here is because it is plausible that there could be some things that are possible, but not conceivable. And maybe a God is such a thing.
The ‘◊’ and ‘ ’ are modal operators on sentences or propositions ‘p’, to be interpreted respectively as ‘It is possible that…’ and ‘It is necessary that…’.
If you believe in transworld identity, and as such do not think that the ‘our’ in ‘our world’ helps us, then the demonstrative ‘this’ can do the important work of singling out the actual world in the expression ‘this world’, by this or that utterance or inscription by us of that demonstrative expression.
It will not rely on an actuality operator, since I am treating the property of being an actually necessary existent as a brute extrinsic first-order property of things. Though I do not think that it will affect the argument much if we were to analyse the property of being an actually necessary existent as a complex property accordingly: ‘λx.@ E!x’. Of course, this will rely on using lambda-abstractions to analyse the complexity of the predicate, as well as both an actuality operator and an existence predicate. I prefer not to analyse the property this way to keep the argument simpler and making as few commitments, and hence targets, as possible.
Perhaps, it might be thought, that some sense could be made of ‘greatest conceivable being’ if like Plato, Plotinus, Descartes, or more recently Quentin Smith (2002), you believed in varying degrees of being. However, it seems intuitive to many that being simply does not permit degrees; it is a categorical notion. Something either is, or it is not (though, see McDaniel 2013). But even if we grant that there are degrees of being, it remains to be explained why unqualified greatness should be positively, rather than negatively, correlated with higher degrees of being. Any pull in either direction seems more like prejudice that privileges one qualifying determinate that relativises greatness in one directed-ordering rather than its opposite.
I imagine that some might retort that one qualifying determinate for greatness, one directed-ordering of things, simply is privileged. Maybe so, but then we lack any epistemology for this hidden structure of reality. Nor is it clear what a privileged directed-ordering would need to be like for it to be a directed-ordering of unqualified greatness, rather than some other privilege. The simple fact is that there seems to be no non-redundant theoretical role for greatness to play, whether or not greatness is taken to be reductively analysable or primitive. And this leaves little room for understanding. I should like to emphasise that the target of my criticism here is perfect being theology.
Though, as we shall later see, there is an important upshot for these criticisms for my version of the ontological argument.
“Maximal” here means that for every p, ‘p v ¬p’ is true in all worlds. This does not of course rule out it being the case that for some p, ‘p & ¬p’ is true at a world, though, given the falsity of dialetheism, those worlds would be impossible worlds. Though, there is a worry here that a contradiction at an impossible world would infect the actual world (cf. Stalnaker 2002). Some might want to drop this “maximality” constraint on impossible worlds, perhaps because they think that such worlds can have gaps, as well as gluts. Nothing that follows will hang on this matter.
By “the world” I mean the world from which the possibility/impossibility of another world is being considered, i.e. the world which other worlds need to be accessible from to be possible relative to that world. This world would be non-rigidly actual relative to itself.
The examples are not crucial. If you think these scenarios are possible—Mackie (2006) is an example of this contention—then pick your own ways that things cannot be.
Roughly, a world w′ is closer than a world w″ to a world w, if, and only if, there are fewer differences in what propositions are true between w′ and w, than between w″ and w. There are clearly problems with this account of ‘closeness’, such as, for any slight difference between worlds w and w′, there would seem to be infinitely many propositions that have a different truth-value. So for example, if p is true in w, and false in w′, then there will be another proposition p*, that is the proposition that p is true, and yet another proposition p**, that is the proposition that p* is true, ad infinitum… that will all be true in w, but false in w′. But then there would appear to be infinitely many differences between every possible world. Hopefully, the intuitive idea of ‘closeness’ should be reasonably clear, or clear enough for present purposes, that we can set aside such tangential issues here.
The situation, as presented, is not quite as straightforward as I have made it seem. For the more attentive reader might complain that the closest world to the actual world where an impossible antecedent is true is just a world which is otherwise exactly like the actual world. Usually, when assessing the closeness of two possible worlds where the antecedent of a counterfactual is true, we can appeal to the preservation of certain important dependencies or laws to trump brute similarity. After all, brute similarity between worlds on its own would not yield an interesting analysis of counterfactuals. The counterfactuals are meant to be getting at some sort of dependency of the consequent on the antecedent.
However, let us assume an absolute notion of possibility (perhaps metaphysical possibility) where relative possibilities (e.g. nomic possibility) place extra constraints, or dependencies, on the constituents of the worlds than occur in the merely absolutely possible worlds. Each kind of relative possibility corresponds to the obtaining of some extra dependency relations, not contained in merely absolutely possible worlds, between the constituents of that world. We then take the common assumption that the relatively possible worlds are a mere subset of the absolutely possible. That is, that those extra dependencies only hold at worlds within that subset of the absolutely possible worlds. A problem then arises when we leave the domain of the absolutely possible worlds, as happens when consider counterpossible conditionals. It seems like no dependency relations obtain to trump brute similarity. And as a result, the closest worlds with an impossible antecedent would just be those that are otherwise exactly like the actual world. Accordingly, the impossible world theorist, not implausibly, needs to give up this picture. They must contend that the relative possibilities are not subsets of the absolutely possible. Indeed, Yagisawa (2010) does reject absolute possibility, opting only for relative possibilities, perhaps precisely for this reason.
Indeed, this picture is intuitive. For example, it gives a nice explanation of certain inconsistent time travel stories, such as in the Grandfather’s Paradox, as being nomically possible, even though absolutely impossible. This is the reverse of Lewis’s (1976) consistent time travel stories that are absolutely possible, even if nomically impossible. Our unease about the consistent time travel cases is that it seems like nomic dependencies do hold in those worlds, and that they would determine the changing of what happens at a time. That is, they would seem to permit the occurrence of contradictions. The obvious explanation is that they do. But all this means is that the nomically possible is not a subset of the logically or absolutely possible.
This is not to suggest that all possible and impossible things are conceivable. Indeed, many impossible things will surely not be conceivable, like circular-squares, and other analytic impossibilities. Nor am I suggesting that all that is inconceivable is neither possible nor impossible.
Indeed, part of what makes theism so difficult to assess is that there is really a cluster of, more or less, closely related God concepts in the vicinity, such that a charge against one often fails to spread to all others. As a result, a common move for the theist to escape criticism is to slink about between the concepts in that cluster. Though, it is unclear to me that this is in fact a defect of the dialectic, or as philosophical manoeuvre, illegitimate, it does make the ensuing debate particularly tough on atheists. I have tried to leave the concept God employed in IMPOSSIBLE relatively unspecified so that the reader may swap and change, within limits, as they please.
See Fine (1994) for a criticism of this view of essence.
Note, for those that find P9 suspect, it is an insufficient answer to the challenge that I am presenting merely to reject this premise. You need to explain how you can permissibly reject this premise given the justification it has received.
I am thinking here of those who like certain paraconsistent logics, such as relevantist logics. My guess is that such logics will be quite popular amongst those accepting impossible worlds. Though, the commitment to impossible worlds does not in itself entail a commitment to such logical systems.
By ‘impossible beings’ and ‘impossibilia’, I mean beings that do not exist at any possible world and whose existence is impossible. Impossible beings should be contrasted firstly with beings that do not exist at any possible world, but are also not impossible—perhaps because the notion of that being is incoherent (That is not to suggest that there are beings of this kind). And secondly, with beings which exist at an impossible world, but also exist at a possible world.
Some might think that this is fine for certain abstract entities, like numbers or sets, as long as those entities in fact exist. But you might be dissatisfied even if those abstract entities did exist, for the same reason that a theist might be dissatisfied with an argument for God’s existence that rest purely upon conceptual grounds (see fn. 6).
As Stalnaker notes, ‘[Some] have defended impossible worlds as a dialectical move, agreeing with Lycan that possibilities and impossibilities stand or fall together, but tying them to each other only to lend weight to the rejection of both.’ (2002, p. 55).
A plausible speculative diagnosis, I suggest, is that worlds bear an inextricable tie to actuality, and modal space at large—a relation that situates worlds in modal space, and thus making them worldly. And that it is this definitive relation between worlds that makes them wholly unsuited for capturing the entire illimitable scope of the impossible. I hope that these, somewhat cryptic, remarks become clearer in the light of the proceeding examination of responses.
This principle is underwritten by the more general ubiquity principle, Ubiquity*, that: For every proposition p, p represents something that is either possible or impossible. However, P6 only requires the weaker principle, Ubiquity.
Yagisawa (2010, p. 213) seems to have reached a similar conclusion.
Importantly, this is not to claim that our concept God is incoherent, but rather that whilst the concept is coherent, modal notions are not applicable to an entity’s satisfaction of this concept. Perhaps to do so would be a category mistake. However, it is doubtful that this proposal of the non-applicability of modal notions to our modally loaded concept God, in particular, can be made sense of. But let us ignore here this fact that the concept God that I have been employing throughout this essay is such that it is constitutively the case that specific modal predications must be applicable to all entities that satisfy that concept, lest they fail to satisfy the concept.
The distinction here between these competing conceptions of the impossible resembles the debate in philosophy of perception between positive and negative disjunctivism, and their characterisations of hallucinatory perceptual experiences (cf. Martin 2004).
Or, more generally, in accordance with Ubiquity* (see fn. 32).
This would also require rejecting certain commonly held inference rules of modal systems, at least in an unrestricted domain, such as: ‘¬◊¬p → p’.
This is potentially a point of disanalogy between modality and tense. Certain things have been thought to be atemporal. If this possibility is accepted, then one might deny the corresponding principle to Ubiquity for tense. Alternatively, one might state that such atemporal entities always obtain, in the same way that a moral error theorist might claim that everything is permissible. This would allow the preservation of the corresponding Ubiquity principles. However, the temporal connotation to “always” and the moral connotation to “permissible” puts a strain on these reconciliations.
For example, Lowe (1998, Chap. 6) seems to be relying on a notion of identity stronger than necessity when he gives his hyperintensional account of ontological dependence between things as the identity of depended entity/entities metaphysically determining the identity of the depending entity/entities. He calls this kind of dependence ‘identity dependence’. Importantly, this metaphysical determination of the identities is not something that can be captured modally. So Socrates’ singleton ontological depends on Socrates, but not vice versa, even though they are necessarily co-extensive.
Alternatively, if like Gallois (1998, Chap. 6) you think that the identity relation holds contingently between entities, then motivation for the claim that ‘Hesperus and Phosphorus being non-identical is neither possible nor impossible’, would be undercut.
In Kripke’s (1980) terminology, this amounts to the claim that there can be different epistemic counterparts of the individuals that actually satisfy our concepts Hesperus and Phosphorus.
A fortiori, self-identity is surely a constitutive property of a thing. To say of a thing, x, that it is not identical to itself, is to say that it is not that thing, that x is not x. But this is just to assert that what was initially claimed to be the subject of the proposition, namely x, is not the subject of that proposition. If x is not the subject of the proposition, then it seems that nothing is. As such, the statement that ‘x is not identical to itself’ does not even appear to express a proposition. And accordingly, interpretation 1 does not even seem to mount a challenge to Ubiquity* (see fn. 32). If I am mistaken about the inconceivability of 1, then it is Proposition, not Ubiquity that should be rejected.
Indeed, this outcome would appear to undercut most, if not all, reasons for believing in impossible worlds.
A reviewer for this journal has pushed this line of reasoning further, asking: ‘What should we say when asked about the "conceivability" of our different models of modal space?’ This is an excellent question. One of the reasons for introducing impossible worlds was so that modal space provides enough room to adequately accommodate all our thoughts, rather than appealing to strategies—such as Stalnaker’s (1999) diagonalisation strategy—to explain away supposed misapprehensions about what we think. But pressing this line of thought seems to suggest that those advocating impossible worlds can never fully escape the need for such reinterpretation strategies. For instance, what should be made of our ability to conceive of modal space (possible and impossible worlds included) differently from how it in fact is? Perhaps we can introduce higher-order ‘meta-modalities’ that capture possibilities and impossibilities of different lower-level modal spaces. Yet, if we can have rigid thoughts about this world (our world) being different, that world will not be located in those other modal spaces—it is located in our modal space. Rather, the best that can be achieved is to think of relevantly similar worlds in relevantly similar modal spaces, and explain away illusions that our modal space meta-can be different as confusing it for one of those relevantly similar modal spaces. This relates to our forthcoming point about misrepresentation.
Perhaps this might be seen as grounds for rejecting P6—that conceivability entails either possibility or impossibility—since what is conceived of is neither possible Should be ‘nor’. impossible in our modal space, but rather in some meta-possible or meta-impossible modal space. But all that really matters for the argument is that what is conceived of be at some world, possible or impossible, in the entirety of modal space (including all higher-levels of modality, with alternative representations of lower-level modal space). If this cannot be achieved—either because of a failure of Proposition or a broader version of Ubiquity locating propositions at least somewhere within the entirety of modal space—, then again, as promised, we have learnt something important and interesting about modal space.
See, for example, Plantinga’s discussion of world-indexed truths: ‘Indeed, for any proposition p and world W, if there is a world in which p is true-in-W, then p is true-in-W in every world. Propositions of the form p is true in α and p is true in W are non-contingent, either necessarily true or necessarily false.’ (1974, p. 55). Similarly, see Yagisawa’s discussion of α-transforms: ‘The α-transform of a property is the same property as indexed to @: for example, the α-transform of being a philosopher is being a philosopher at @. It is almost universally accepted that the operation of α-transformation produces necessity out of contingency, i.e. if a thing has a property contingently, it has the α-transform of that property necessarily. The widely accepted reason is that if x is F at @, then for any possible world w, x is F at @ at w, and vice versa.’ (2010, p. 59).
Yagisawa (2010) is perhaps an exception here, since he treats worlds as metaphysical indices, like times, or spatial points. Yet, on Yagisawa’s view, our standard modal notions are not applicable to worlds (pp. 52 and 90)—or propositions for that matter (p. 57). Instead, whilst rejecting Divers’ (1999) ‘redundancy account’ of advanced modalizing (see Yagisawa 2010, p. 203, n. 41), Yagisawa introduces what appears to be an infinite hierarchy of ‘über modal’ or ‘meta-modal’ notions to deal with advanced modalizing claims involving the existence and nature of worlds (i.e. modal claims about metaphysical/modal spaces). But as will soon be explained, Yagisawa’s proposal is ultimately of no help to our fictional objector.
He can do this because he treats worlds as entities, or as he calls them, ‘metaphysical indices’, analogous to the substantival treatment of times or spaces. That allows that a world is defined independently of its members. This is an interesting, but unusual, characterisation of worlds.
The ‘is m ’ here is a ‘modal tense’ that Yagisawa introduces to linking verbs in order to make clear what their relative domain of quantification is. This particular modal tense restricts the quantificational domain to what obtains at any particular metaphysical/modal space.
Or perhaps, it will be insisted that It is part of the concept God that it is satisfied by a being, x, only if x has the constitutive property of satisfying the concept God in the every, or every possible, world. Both of these options would entail P10. But the weaker claim, P10, would seem to yield the desired result on its own.
References
Anselm, St. (1077–1078), Proslogion. In M. Charlesworth (1965) (Eds.), St. Anselm’s Proslogion. Oxford: Oxford University Press.
Callender, C. (2002). Time, reality and experience. Cambridge: Cambridge University Press.
Chalmers, D. J. (2002). Does conceivability entail possibility? In T. S. Gendler & J. Hawthorne (Eds.), Conceivability and possibility (pp. 145–200). Oxford: Oxford University Press.
Divers, J. (1999). A genuine realist theory of advanced modalizing. Mind, 108(430), 217–239.
Findlay, J. N. (1948). Can God’s existence be disproved? Mind, 57(226), 176–183.
Fine, K. (1994). Essence and modality: The second Philosophical Perspectives Lecture. Philosophical Perspectives, 8, 1–16.
Gallois, A. (1998). Occasions of identity. Oxford: Oxford University Press.
Gendler, T. S., & Hawthorne, J. (2002). Conceivability and possibility. Oxford: Oxford University Press.
Hartshorne, C. (1941a). Man’s vision of God. New York: Harper and Row.
Hartshorne, C. (1941b). The necessarily existent. In Man’s vision of God (pp. 299–341). New York: Harper and Row.
Hartshorne, C. (1965). Anselm’s discovery: A re-examination of the ontological proof for God’s existence. La Salle, Illinois: Open Court.
Jago, M. (2014). The impossible: An essay on hyperintensionality. Oxford: Oxford University Press.
Kane, R. (1984). The modal ontological argument. Mind, 93(371), 336–350.
Kripke, S. (1980). Naming and necessity. Oxford: Basil Blackwell.
Le Poidevin, R. D. (1996). Arguing for atheism: An introduction to the philosophy of religion. London: Routledge.
Leftow, B. (1990). Is God an abstract object? Noûs, 24(4), 581–598.
Lewis, D. K. (1973). Counterfactuals. Oxford: Blackwell.
Lewis, D. K. (1976). The paradoxes of time travel. American Philosophical Quarterly, 31(2), 145–152.
Lowe, E. J. (1998). The possibility of metaphysics: Substance, identity, and time. Oxford: Oxford University Press.
Mackie, P. (2006). How things might have been: Individuals, kinds, and essential properties. Oxford: Oxford University Press.
Malcolm, N. (1960). Anselm’s ontological arguments. Philosophical Review, 69, 41–62.
Martin, M. G. F. (2004). The limits of self-awareness. Philosophical Studies, 120, 37–89.
McDaniel, K. (2013). Degrees of being. The Philosopher’s Imprint, 13(19), 1–19.
Nolan, D. (1997). Impossible worlds: A modest approach. Notre Dame Journal of Formal Logic, 38(4), 535–572.
Plantinga, A. (1974). The nature of necessity. Oxford: Oxford University Press.
Smith, Q. (2002). Time and degrees of existence: A theory of ‘degree presentism’. In C. Callender (Ed.), Time, reality and experience (pp. 119–136). Cambridge: Cambridge University Press.
Stalnaker, R. C. (1999). Context and content: Essays on intentionality in speech and thought. Oxford: Oxford University Press.
Stalnaker, R. C. (2002). Impossibilities. In Ways a world might be: Metaphysical and anti-metaphysical essays (pp. 55–67). Oxford: Oxford University Press.
Stalnaker, R. C. (2003). Ways a world might be: Metaphysical and anti-metaphysical essays. Oxford: Oxford University Press.
Yagisawa, T. (2010). Worlds and individuals, possible and otherwise. Oxford: Oxford University Press.
Acknowledgements
Thanks are due to Sarah Adams, Michael Bench-Capon, and Jason Turner for providing helpful feedback on earlier drafts of this work. Also to audiences at the Universities of Birmingham and Leeds, where I presented this work, and to fruitful conversations about this paper, again with Michael Bench-Capon, but also Richard Caves. The paper is better as a result of their generous feedback.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Pezet, R.E. An impossible proof of God. Int J Philos Relig 83, 57–83 (2018). https://doi.org/10.1007/s11153-016-9591-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11153-016-9591-0