Giving the Ontological Argument Its Due

Abstract

In this paper, I shall present and defend an ontological argument for the existence of God. The argument has two premises: (1) possibly, God exists, and (2) necessary existence is a perfection. I then defend, at length, arguments for both of these premises. Finally, I shall address common objections to ontological arguments, such as the Kantian slogan (‘existence is not a real predicate’), and Gaunilo-style parodies, and argue that they do not succeed. I conclude that there is at least one extant ontological argument that is plausibly sound.

Ontological arguments are a priori arguments for the existence of an orthodoxly conceived monotheistic god. Typically these arguments move from the concept of God to his actual existence, given some set of (allegedly) necessary truths, e.g. ‘existence is a perfection.’ There is almost universal agreement that these arguments are bad arguments; that they are unsound or unmotivated. Despite this near-universal agreement, I shall argue that there is at least one extant ontological argument that, at the very least, makes theism rationally acceptable, given the plausibility of its premises. I shall argue that the common objections that philosophers have defended over the centuries do not warrant the standard position on these arguments.

I understand by ‘God’ the perfect being. By a ‘necessary being’ I mean a being whose non-existence is metaphysically impossible. Consider the following two accounts:

• (G) ☐(∀x)(x is perfect ↔ x has all perfections essentially and x lacks all imperfections essentially).

• (N) ☐(∀x)(x exists necessarily ↔ x exists in all possible worlds).

Then the ontological argument can be stated as follows:

• 1. Possibly, there exists a being that is perfect.

• 2. Necessary existence is a perfection.

Hence,

• 3. There exists a being that is perfect.¹

Premises (1) and (2) logically entail (3), given (G) and (N). For suppose (1) and (2) are true. Then, by (G), there is a possible world W in which a being that has all perfections essentially and lacks all imperfections essentially exists. Furthermore, given (2), this being has necessary existence essentially. But, by (N), a being has necessary existence in W only if it exists in every possible world; that is what it means to have this property. Consequently, the being that exists in W exists in all possible worlds, including the actual one. It follows that (3) is true. That is, it follows that there is a perfect being, and, as St. Thomas might have concluded, this is what everyone means by ‘God.’²

The Possibility Premise

Maydole (2003, 2012) defends the following proposition:

• 4. Perfections entail only perfections.

He argues for (4) as follows:

‘Suppose X is a perfection and X entails Y. Then it is better to have X than not, and having Y is a necessary condition for having X. But it is always better to have that which is a necessary condition for whatever it is better to have than not; for the absence of the necessary condition means the absence of the conditioned, and per assumption it is better to have the conditioned. Therefore, it is better to have Y than not. So Y is a perfection.’³

According to Maydole (2012, p. 580), a perfection is a property that is necessarily better to have than lack.⁴

How might we argue from (4) to (1)? The proof is simple, and valid in second-order modal logic. Let PJ = ‘J is a perfection,’ A = the property being perfect, and B = the property being evil. Then we have the following argument, whose premises are as follows:

• 5. (∀J)(∀K) [(PJ & ☐(∀x)(Jx → Kx)) → PK]

• 6. PA

• 7. ~PB

Assume, for reductio, that,

• 8. ~◊(∃x) Ax

• 9. (∀K)[(PA & ☐(∀x)(Ax → Kx)) → PK] 5, UI

• 10. [(PA & ☐(∀x)(Ax → Bx)) → PB] 9, UI

• 11. ☐ ~ (∃x) Ax 8, MN

• 12. ☐(∀x) ~ Ax 11, QN

Since, necessarily, A is not exemplified, it follows trivially that,

• 13. ☐(∀x)(Ax → Bx)

• 14. PA & ☐(∀x)(Ax → Bx) 6, 13, Conj.

• 15. PB 10, 14, MP

• 16. PB & ~PB 7, 15, Conj.

• 17. ~~◊(∃x) Ax 5-16, RAA

• 18. ◊(∃x) Ax 17, DN

The conclusion is that it is possible that there exists a perfect being.

Is this a good argument? That is going to depend on whether or not there are good reasons to suppose that the premises are true. Premises (6) and (7) are plausible. It seems to be the case that being perfect, viz. having all perfections essentially and having all of the complements of all imperfections essentially, is a perfection, on Maydole’s account. If something were to have this property, it would be greater than otherwise, and anything that has this property’s complement, viz. the property of not being perfect, would be imperfect in virtue of its having it. It is also clear, given his account, that being evil is not a perfection, for it is better to lack than have.

What about (4)? We have already seen one argument in favour of it, and it has some prima facie plausibility. But, as Oppy (2004) points out, there is a straightforward counterexample to it. The property being omnipotent is a perfection, for it is necessarily better to have than lack. But this property tautologically entails the disjunctive property being omnipotent or else a mass murderer. If (4) were true, then it would follow that being omnipotent or else a mass murderer is a perfection. But it is not necessarily better to have than lack, for both Hitler and Stalin had this property (both had the property being a mass murderer), and it would have been better for them to have lacked it. The reason, of course, is that both Hitler and Stalin are essentially not omnipotent, so their having this property entails their being mass murderers; had they lacked it, they would not have been mass murderers.

Maydole (2005) responds by arguing that it is not true that it is necessarily the case that the necessary conditions of properties that are better to lack than have are themselves better to lack than have. For example, free will is necessary for moral evil, but it is not better to lack than have. This is quite true, and quite irrelevant. As Oppy (2007) points out, being omnipotent or else a mass murderer is not necessarily better to have than lack because it would have been better for Hitler and Stalin to have lacked this property. But this is enough to falsify (4) because, according to Maydole, a perfection is a property that is necessarily better to have than lack; no matter what has it, it adds to that thing’s greatness. I conclude that perfections can entail non-perfections, whence it follows that (4) is false.

Is this, as Oppy concludes, ‘the end of the story’? Here I should like to make a few distinctions that may help to repair this argument. I shall stipulate that there are three categories into which all properties fall: properties are either great-making, neutral, or else lesser-making. I understand by ‘great-making properties’ (GMP) properties that are necessarily better to have than lack (e.g. being good). I understand by ‘neutral properties’ (NP) properties whose instantiation does not necessarily add to or subtract from the greatness of the beings in which they inhere; properties that are neither necessarily better to have than lack nor lack than have (e.g. being a person or else crunchy). I understand by ‘lesser-making properties’ (LMP) properties that are necessarily better to lack than have (e.g. being a mass murderer).

Given these distinctions, it is clear where tautological properties, including being omnipotent or else a mass murderer, fall: they are neutral properties. Consider the following, amended premise:

• 4* Perfections entail only GMPs or NPs.

Oppy’s objection does not apply to (4*), for (4*) admits that perfections entail non-perfections, including being omnipotent or else a mass murderer. If (4*) is false, then it is false in virtue of some other reason(s). But there is a plausibly sound argument for (4*). The argument is as follows: Suppose it were possible for a perfection P to entail a LMP, or imperfection, L. Then L is a necessary condition of the instantiation of P. L is a necessary condition of the instantiation of P only if P’s instantiation is sufficient for the imperfection of anything in which it inheres. For anything that has P also has L, and L is a lesser-making property. So anything that has P is imperfect, and it is imperfect in virtue of its having P. But clearly this is a feature only of LMPs. Perfections add to, but do not subtract from, the greatness of the things in which they inhere. If this were false, then there would be some perfections that are not great-making, which is absurd; all perfections are great-making. Hence, it cannot be the case that the instantiation of L is a necessary condition of the instantiation of P. We can generalize and conclude that it cannot be the case that perfections entail imperfections, or LMPs. Since perfections do entail GMPs (e.g. omnipotence entails the property being powerful) and neutral properties (e.g. omnipotence entails being omnipotent or else crunchy, which is had by anything that is crunchy, e.g. apples), but not LMPs, and since these are the only categories into which properties fall, it follows that (4*) is true.

We can argue from (4*), and a couple of other plausible premises, to the conclusion that God’s existence is possible. Again, let PJ = J is a perfection, A = the property being perfect, and B = the property being evil. Furthermore, let GK = K is great-making, and NK = K is neutral. The argument’s premises are as follows:

• 4* (∀J)(∀K)[(PJ & ☐(∀x)(Jx → Kx)) → (GK v NK)]

• 19. (PA & ~GB) & ~NB

Assume, for reductio, that

• 20. ~◊(∃x) Ax

• 21. (∀K)[(PA & ☐(∀x) (Ax → Kx)) → (GK v NK)] 4*, UI

• 22. (PA & ☐(∀x)(Ax → Bx)) → (GB v NB) 21, UI

• 23. ☐~(∃x) Ax 20, MN

• 24. ☐(∀x) ~Ax 23, QN

Since, necessarily, A is not exemplified,

• 25. ☐(∀x)(Ax → Bx)

• 26. PA & ~GB 19, &E

• 27. PA 26, &E

• 28. PA & ☐(∀x)(Ax → Bx) 25, 27 Conj.

• 29. GB v NB 22, 28, MP

• 30. ~NB 19, &E

• 31. GB 29, 30, DS

• 32. ~GB 26, &E

• 33. GB & ~GB 31, 32, Conj.

• 34. ~~◊(∃x) Ax 4*-33, RAA

• 35. ◊(∃x) Ax 34, DN

This argument has two premises. I have already defended (4*). What about (19)? Again, having all perfections seems to be a perfection. It is clear that being evil is neither great-making nor neutral. Since there are independent reasons for each of the premises, the argument is not question-begging. At the very least, both of the premises of this argument are defensible, and so we have a defensible argument from plausible premises for the truth of (1), the possibility premise in the ontological argument.

It is, however, possible to parody this argument. Before I get to the parody, it will be helpful to understand exactly what parody objections are supposed to accomplish. A parody argument A is an attempt to undermine another argument A* by using A*’s logical structure to get to an obviously false conclusion. The parody ought to be a compelling objection only if the argument against which it is directed has a premise with about as much support (or lack thereof) as its corresponding premise in the parody. That is, for example, if the central premise in an argument has no support, and the central premise in its parody has no support, then we have little reason to reject the latter and not the former, or to accept the former and not the latter. But since the parody’s central premise must be false (because otherwise its conclusion would not be false), we have reason to be suspicious of the central premise in the parodied argument.

Consider the following parody against the above argument. Let us say that a being is supremely imperfect just in case it has all imperfections essentially and lacks all perfections essentially. Then add MP:

• (MP) Imperfections entail only LMPs or NPs.

We can run a similar argument to the conclusion that it is possible that something is supremely imperfect. This is not a conclusion classical theists will want to accept. According to classical theism, God exists necessarily, in every possible world. If it is possible that there exists a supremely imperfect being, then there is a possible world W in which both God and this being exist. Since the supremely imperfect being has all imperfections essentially, and since the conjunctive property being omnipotent and evil is an imperfection, in W there exists an omnipotent being that is not identical to God! So if it is true that the reasoning of this parody and the argument to the possibility of God are parallel, and (4*) and MP have about the same support (or lack thereof) in their favour, then the parody successfully refutes my argument.⁵

The problem is that MP is palpably false. It is true that being evil is an imperfection. This property clearly entails the properties being a moral agent, and being a person. Necessarily, something is evil only if it is a moral agent and a person. But both of these properties are (plausibly) perfections. Consequently, (MP) is false: there is at least one imperfection that entails a perfection. Hence, the parody fails, for there is a very good reason to suppose that (MP) is false, and not a comparably good reason to suppose that (4*) is false. I conclude that this parody fails to refute the above argument.

In Bernstein (2014) I argued for (1) as follows. Suppose it is impossible that something has all perfections. Then there is some perfection that entails the complement of a perfection. But a property entails the complement of a perfection only if that property is sufficient for the imperfection of anything in which it inheres. Hence, there is some perfection that is sufficient for the imperfection of anything in which it inheres. But no perfection is sufficient for the imperfection of anything in which it inheres; this is a feature only of imperfections. Contradiction! So the assumption is false: it is possible that something has all perfections.

The two premises of this argument are as follows:

• 36. A property is sufficient for the imperfection of anything in which it inheres only if that property is not a perfection.

• 37. A property entails the complement of a perfection only if that property is sufficient for the imperfection of anything in which it inheres.

These two propositions are true only if all perfections are compossible. I shall now argue for them.

We may argue to (36) as follows. A property P is sufficient for the imperfection of any being in which it inheres just in case P detracts from the greatness of any being in which it inheres. Assume, for reductio, that (36) is false. That is, assume that P is sufficient for the imperfection of any being in which it inheres and that P is a perfection. Since P is sufficient for the imperfection of any being in which it inheres, it follows that P detracts from the greatness of any being in which it inheres. But then it is false that P is a perfection, for this is a feature only of imperfections: necessarily, some property detracts from something’s greatness only if that property is not a perfection. It follows from our assumption that P is a perfection. Since this is a contradiction, we must reject our assumption. Hence, it is false that < P is sufficient for the imperfection of any being in which it inheres and P is a perfection>. This, of course, is equivalent to (36).

A few clarifications will help us to see this (37) is true. First, a property is sufficient for the imperfection of any being in which it inheres just in case it entails the property being imperfect. Second, a being has the property being imperfect just in case it has at least one imperfection, where an imperfection is a property that detracts from the greatness of any being in which it inheres. Finally, any being that lacks a perfection is an imperfect being, because it would then have the property of not having that perfection, which is an imperfection.

Now we are ready to see that (37) is true. Where a property P* is a perfection, P is sufficient for a being’s lacking P* just in case P entails P*’s complement. Since the complement of P* is an imperfection, it entails the property being imperfect. But since P entails the complement of P* and since P*’s complement entails being imperfect, it follows that P entails being imperfect. So P is sufficient for the imperfection of any being in which it inheres. Hence, a property is sufficient for a being’s lacking some perfection only if it is sufficient for that being’s being imperfect. This just is what (37) says.

If this is right, it follows that all perfections are compossible. But since having P essentially, where P is any perfection, is a perfection, and since having P* essentially, where P* is any complement of an imperfection, is a perfection, it follows that it is possible that something has all perfections essentially and lacks all imperfections essentially. That is, it follows that (1) is true.

I should like to address one last objection. Perhaps the skeptic is committed to the proposition that the traditional divine attributes—omnipotence, omniscience, and moral perfection—are necessarily uninstantiated. Since the above arguments entail that a property is a perfection only if it is possibly instantiated, the above properties are not perfections (in my sense). If this is right, then the arguments in this section do not support the possibility premise in the ontological argument in a way that is required in order to get to the conclusion that God, classically conceived, exists.⁶

But the point is that it is very plausible to suppose that omnipotence, omniscience, and moral perfection are perfections—viz. properties that are better to have than lack—independent of the project of either atheism or theism. That they are perfections is consistent with both views, and would be inconsistent only with a strong version of naturalism⁷ that rules out, a priori, even the mere possibility of something’s having any of these properties. That these properties are evidently perfections is a reason for both atheists and theists to think all such views are false. If one is strongly committed to a position that entails that it is impossible that these properties be instantiated, then they will not share the relevant intuitions. At that point, we are at an impasse. But I emphasize that it is consistent with atheism that the standard divine properties are perfections.

Necessary Existence

Before we get to the arguments, it is important to state clearly what I mean by a few concepts. I understand by a ‘metaphysically necessary being’ a being whose non-existence is impossible. We might put it this way: a being exists necessarily if and only if it exists in every possible world; if and only if all possible states of affairs include the existence of that being. I understand by a ‘contingent being’ a being whose non-existence is possible. Alternatively, a being is contingent if and only if it exists in some, but not all, possible worlds; if and only if some, but not all, possible states of affairs include its existence. To say that something is contingent, then, is to say that it is possible that it fails to exist; or that, if it exists, it just happens to exist, insofar as it might have failed to exist. Metaphysical necessity is to be distinguished from what I shall call ‘factual necessity,’ which is the property of not being causally contingent on anything. Something is factually necessary if its existence is not the result of some other thing’s causing it to come into being, and if it will never cease to exist. Some say that, although God is not metaphysically necessary, he is nevertheless factually necessary, and factual necessity is the only type of necessity that is a requirement of perfection. This is compatible with his being contingent.⁸

One can argue for (2), viz. that necessary existence is a perfection, by arguing for the following proposition:

• 2* Being contingent is an imperfection.

If (2*) is true, then the complement of being contingent is a perfection. But this property self-evidently entails being necessary. Consequently, something has all perfections only if it is a necessary being.

Hartshorne (1962, p. 61) produced the following argument for the truth of (2*):

‘To exist contingently is to exist precariously, or by chance (for to say, by cause or intention is to prompt the query, is the cause or intention necessary—achieving its result necessarily?—nor non-necessarily?). But to exist precariously or by chance is an imperfection, appropriate only to imperfect individuals. That “humanity” might not have been at all means that each of us exists, and has any particular excellence, thanks to accidental factors only. This total dependence on the way things happen to be, this radical “iffyness” or precariousness of our existence and nature…is a defect in principle from which various limitations follow, for instance, having a kind of temporal beginning of existence. If, then, an individual of a kind which can only exist contingently is necessarily imperfect, no such individual could exemplify perfection.’

The argument can be stated as follows:

• 38. Necessarily, something is contingent only if it exists by chance.

• 39. Necessarily, to exist by chance is an imperfection.

Hence,

• 40. Necessarily, something is contingent only if it has an imperfection.

Depending on what is meant by ‘chance’, (38) is true. If we suppose that x exists contingently, then there are worlds, indeed infinitely many non-actual worlds, in which x does not exist but that might have been actual. So it then seems right to ask why this world should be actual and not a world in which x does not exist. The answer seems to be that this world’s being actual is due to modal chance: this world just happens to be actual and there is no further explanation for why it is actual. But then x exists by chance in the sense that its existence depends on what appears to be a modal accident.

Premise (39) is much more difficult to motivate. Hartshorne attempts to do it, but his reasons seem to me to be unsuccessful. The reason he gives to suppose that (39) is true is that ‘various limitations follow’ from existing by chance. The only example he gives for such a limitation is the having a temporal beginning of existence, which presumably is a limitation insofar as it would then require causal dependence: something begins to exist in time only if there is some other thing that causes it to come into existence. But it is hard to see why existing by chance (in the sense described above) entails the having of a temporal beginning of existence. Consider the following. There exists an S such that:

• (a) S is contingent.

• (b) S is essentially eternal (where something is eternal in a world W if there is no state of affairs that obtains in W that includes S’s non-existence).

One traditional reason why is due to Leibniz’s Principle of Sufficient Reason (PSR), which I shall state as follows:

• (PSR) ☐(∀x)(x is contingent → possibly, x has a sufficient explanation of its existence).

According to (a) and (b), S is contingent and essentially eternal. This entails that there is no possible world in which S is caused to begin to exist. Since S is contingent, it follows from PSR that,

• 41. Possibly, S has a sufficient explanation of its existence.

(41) entails that there is a possible world W in which S exists and in which S’s existence is sufficiently explained. Now, there are two kinds of explanation: a thing explains its own existence or something else explains a thing’s existence. A thing explains its own existence only if it is necessarily existent: the explanation for why, say, the number seven exists (if it does exist) is that its non-existence is impossible. This provides a sufficient explanation for why seven exists. A contingent being cannot have in itself the explanation of its existence, for it is part of what it means to be contingent that it might not have been, and no essential property that it has explains why it happens to exist. Consequently, if S’s existence in W has a sufficient explanation, then it is to be found in another contingent being (or set of contingent beings) or a necessary being. Hence, if S’s existence has a sufficient explanation in W, then there is another being (or set of beings) in W, S*, without whose existence S would not exist. But then, in W, S has the property being dependent on another being. It follows that, given PSR, factual necessity (where (a) is included) is incoherent, for something is a dependent being (in this sense) only if it is caused to come into being, which entails the complement of being eternal. Since S was just any contingent being, it follows that, necessarily, something is contingent only if it is not essentially eternal.

It is also easy to show that this argument establishes the truth of (2). Consider the following plausible premises:

• 42. ☐(∀φ)(φ is an imperfection → being possibly ϕ is an imperfection).

• 43. Being dependent on another being’s existence is an imperfection.

• 44. Being caused to exist is an imperfection.

• 45. Being not eternal is an imperfection.

Why think that (42) is true? Say that ϕ is any imperfection, and that S lacks ϕ contingently. Then S has the property of being possibly ϕ. The reason why this is an imperfection is that S is merely accidentally perfect in respect to lacking ϕ, which is less great than essentially lacking ϕ; a being who essentially lacked ϕ would be perfect in that respect, not merely by accident, but of necessity, which is obviously better than merely happening to lack ϕ.

Given the previous argument, all contingent beings, of necessity, possibly have each of the properties in (43)–(45). Given (42), it follows that all contingent beings are, of necessity, imperfect, for all contingent beings, of necessity, have imperfections. Since being contingent is sufficient for the imperfection of anything that is contingent, being contingent is an imperfection. But if being contingent is an imperfection, then its complement, not being contingent, is a perfection, for it is always better to have the complements of properties that are better to lack than have. It follows that, if it is possible that there is a perfect being, then there is a possible world in which this being exists necessarily. So it is in this way, by employing Leibniz’s PSR, that one can defend Hartshorne’s argument from the challenge of the possibility of factual necessity.

Of course, this argument relies on a very controversial thesis in PSR, and defending it here is beyond the scope of this paper.⁹ But the point is that there is a defensible argument to the truth of (2) from PSR, and the weak version of PSR to which I appealed seems, at least to me, to be more plausibly true than false. At the very least, believing its negation is not a requirement of reason. Neither is belief in PSR contrary to reason. Hence, if, as I have argued, PSR’s being true gives good reason to think that (2) is true, then a person can be rational in her acceptance of (2).

Leibniz (1964) famously argued for the possibility of the existence of God in defense of Descartes’ ontological proof. Maydole (2012) reconstructs his argument as follows.

• 46. All perfections are compatible.

• 47. Every essential property of a perfect being is a perfection.

• 48. If something’s essential properties are perfections and all perfections are compatible, then its essential properties are compatible.

• 49. If the essential properties of something are compatible, then it is possible that it exists.

Hence,

• 50. It is possible that a perfect being exists.

Maydole thinks that (47)–(49) are ‘self-evident.’ Suppose he is right about (47). Consider the following argument. A property is a perfection only if it is a perfection for anything that has it. It would be queer to suppose that a property is a perfection for me and not for you. Suppose it is possible for there to be a contingent being that is perfect. Since it is impossible for something to have the property being contingent contingently, it follows that, necessarily, something is contingent only if it is contingent essentially. So, necessarily, if a perfect being is contingent, it has the property being contingent essentially. By (47), it follows that being contingent is a perfection, for all of the essential properties of a perfect being are perfections. But this is clearly false, for it is evident that contingency is not a perfection for every being that has it. It is implausible to suppose that a snail has a perfection and that it is its being contingent; or that a snail has as one of its perfections being contingent. Would the snail be less great if (counterpossibly) it lacked contingency?

I earlier in this paper distinguished between great-making, neutral, and lesser-making properties. Let us state these categories as follows:

• (PERFECTION) (∀ϕ)(ϕ is a perfection ↔ ϕ is not greatness-detracting and ϕ’s complement is greatness-detracting).

• (IMPERFECTION) (∀ϕ)(ϕ is an imperfection ↔ ϕ is greatness-detracting).

• (NEUTRAL) (∀ϕ)(ϕ is a neutral property ↔ ϕ is not an imperfection and ϕ is not a perfection).

Intuitively, it seems that contingency falls into one of the two latter categories, but not the former category. The reason is that it does not seem that contingency’s complement—the property not being contingent—is greatness-detracting. Plausibly, then, contingency is not a perfection. Hence, it is not possible for there to be a contingent being that is perfect. But since all of the essential properties of a perfect being are perfections, and since, necessarily, something is not contingent only if it is necessary, it follows that the property being necessary is a perfection. If so, then its complement—not being necessary, viz. being contingent—is an imperfection.

But there is a problem with this argument. It seems to me that (47), as it stands, is false. The reason it is false is one that we have already considered. Recall, any perfect being will be omnipotent essentially, and it is tautologically true that, for any two properties ϕ and ψ,

• ☐(∀x)(ϕx → (ϕx v ψx)).

Consequently, omnipotence entails the property being omnipotent or else a mass murderer, which entails that any being that is omnipotent essentially has this property essentially. But then, by (47), being omnipotent or else a mass murderer is a perfection. As we have seen, this property is not a perfection. It follows that (47) is false.

Perhaps we can repair this argument and avoid this objection by amending (47). First, I shall distinguish between tautological and non-tautological essential properties:

• (T) A property P is a tautological essential property (TEP) of x just in case x has P essentially and x has P only in virtue of P’s following tautologically from another property, P*, that x has essentially.

• (NT) A property P is a non-tautological essential property (NEP) of x just in case x has P essentially and it is false that x has P only in virtue of P’s following tautologically from another property, P*, that x has essentially.

Given T and NT, consider our amended (47),

• 47* All of the NEPs of a perfect being are perfections.

The original argument will go through with (47*) because contingency is an NEP. Very clearly, though, (47*) is not susceptible to the same objection raised against (47), for the counterexample worked only by showing that a TEP had by the perfect being is not a perfection, which is compatible with (47*)

If the above arguments are sound, (2) is true.¹⁰

Common Objections

Almost invariably when philosophers talk about ontological arguments someone brings up the Kantian slogan, ‘existence is not a real predicate.’¹¹ Perhaps it is, perhaps it is not. The point I should like to make here is that, at least with respect to this version of the argument, the slogan is irrelevant, for nowhere do I assume that existence is a real predicate. I nowhere assume that existence is a property or perfection that something can have or lack. The only assumption I make is that necessary existence is a perfection and is, therefore, a real predicate. When we say of something that it exists necessarily we are not merely saying of it that it exists, we are saying of it that it could not have failed to exist; that its essence could not have failed to be instantiated. This contributes to our concept of a thing. For example, it tells us that the being in question is not a dependent being. If, then, Kant’s objection would apply to this putative property, then so much the worse for the objection. If one remains suspicious of this as a property, I can concede the point for the purposes of argument and instead say that contingency, which is a real predicate, is an imperfection. When we say of something that it is contingent we are saying of it that its essence might have failed to be instantiated. The ontological argument for which I have been arguing will go through with this assumption.

A much more formidable objection is Gaunilo’s parody objection.¹² Say that an island-perfection (IP) is a property that is necessarily better for islands to have than lack. Say that an island is perfect just in case it has all IPs essentially. The parody is as follows:

• 51. Possibly, there exists a perfect island.

• 52. Necessary existence is an IP.

Hence,

• 53. There exists a perfect island.

Plantinga (1974, p. 90–91) responds as follows. Having beautiful palm trees and having coconuts are examples of IPs. But these properties entail having a certain number of palm trees and having a certain number of coconuts. Let us suppose that O is a possible island that is perfect. Then it has the property of having coconuts. Let n be the number of coconuts had by O and s be the size of O. We can conceive of an island O* that has just as many IPs as O except that O* has n + 1 coconuts and whose size is s*, where s* ≥ s (to take into account any size increases necessary to prevent the island’s having an island imperfection, such as being cluttered). Since O* is greater than O, it follows that O is not perfect. But we supposed that it was. This can be done with any possible island. The reason is that many IPs, unlike the perfections simpliciter of the ontological argument, have no intrinsic maxima. It follows that perfect islands are impossible.

This reply does not work against the traditional divine attributes. Take just omnipotence and omniscience. Omnipotence is the ability to bring about any metaphysically possible state of affairs.¹³ It is impossible for something to be more powerful than something that has omnipotence. Omniscience is knowing the truth-value of every proposition. Nothing can possibly know the truth-value of more propositions than an omniscient being. Unlike the IPs above, the divine properties do have intrinsic maxima. So we have a reason to think that Gaunilo’s parody is unsound that does not obviously apply to the ontological argument.

Brian Garrett (2013) argues that Plantinga’s response to Gaunilo’s parody of the ontological argument fails because IPs have intrinsic maxima. He does this by defining the perfect island as an island that has the ideal F, for any F that is an IP. On this picture, there is an ideal size, an ideal number of coconuts, and an ideal temperature for the island’s water. Moreover, Garrett argues that one cannot object to Gaunilo’s argument by asserting that islands are contingent, whereas God is not, because one can advance a parody that shows that islands exist of necessity, such as I have done above. I shall argue that Garrett’s replies fail to save Gaunilo’s parody.

It is questionable whether there is an ideal size of an island. Suppose we say that having moral agents who never go wrong as inhabitants is an IP. Let O be the perfect island, n* be the number of these agents on O, and s* be O’s size. We can conceive of an island O* that has n* + 1 agents and whose size is s**, where s** ≥ s*. But it seems that O* is greater than O, which is a contradiction. Consequently, there cannot be an ideal size, because it is better, everything else being equal, to increase an island’s size in order to comfortably accommodate more moral agents who never go wrong; the more such agents living happily on an island, the greater the island.

Garrett then argues that one cannot reply to Gaunilo’s parody by suggesting that islands are contingent, because one can produce a parody to ‘show that a perfect island exists of necessity.’ This move works only if there are not objections to it that do not similarly apply to the ontological argument. Consider the following objections that do not apply to it:

First, being an island entails being a material thing. But being a material thing plausibly entails being contingent. We have scientific evidence to suppose that this is the case, for there is evidence that space and time began to exist, and the existence of space and time are necessary conditions of the existence of material objects.¹⁴ Hence, there is evidence to suppose that there is a possible state of affairs—the timeless state of the actual world—that includes the state of affairs consisting in the non-existence of anything material, including islands. It follows that there is a world in which the timeless state obtains and no islands exist. Since being material is an IP, and since being contingent is a necessary condition of something’s being material, being contingent is also an IP, for it is always better for an island to have the non-tautological necessary conditions of IPs.

Second, it is possible that there exists a near-omnipotent being (or a very powerful being). This is plausible for reasons independent of the project of theism. But such a being could ravage the perfect island if she wishes. She can, for example, poison a lagoon, cause all the coconuts to vanish, create one more coconut such that the island no longer has the ideal amount, or cause the island to sink. Or consider an inhabitant of an island. Such an inhabitant can eat a coconut, in which case the island would lose the IP of having the ideal number of coconuts. Even if the island somehow replaces it immediately, there will still be a time, however short-lived, during which it lacks this IP. Hence, a perfect island can lose one of its IPs. If it is really an island, then this is a possibility. So a perfect island cannot exist necessarily, for it follows from this that no island has all of its IPs essentially, and an island is perfect in a world W at a time t only if it has all IPs at t.

Conclusion

Recall, the ontological argument for which I have been arguing is as follows:

• 1. Possibly, there exists a being that is perfect.

• 2. Necessary existence is a perfection.

Hence,

• 3. There exists a being that is perfect.

I have defended arguments from plausible premises to the truth of both (1) and (2). Furthermore, I have addressed the two most common objections to ontological arguments. I argued that Kant’s objection is not relevant to this version of the argument, since I nowhere assume that existence is a real predicate. I argued that Gaunilo’s parody objection fails because there are good reasons for thinking the parody unsound that do not similarly apply to this version of the ontological argument.

I conclude that there is at least one extant ontological argument that is plausibly sound (the argument is valid and the premises are plausibly true). I began by pointing out that most philosophers have thought that ontological arguments are bad arguments; that they are either unmotivated, irreparably problematic, or unsound . If what I have argued is correct, this view is unwarranted. Like most arguments to controversial conclusions, it is reasonable for someone to question the premises of this argument and the premises of the arguments presented in its defense. But it is also reasonable to believe in the premises of the arguments I have discussed, and, if that is the case, then it can be reasonable to believe that God exists.¹⁵