Skip to main content

Modal Epistemology and the Cosmological Argument

William Rowe formulates the Leibnizian cosmological argument (roughly) as follows:

1. Either everything can be a dependent being, or there is a self-explanatory being.
2. Not everything can be a dependent being.
3. Therefore, there is a self-explanatory being (from 1 and 2)

A common objection is that the argument depends on the Principle of Sufficient Reason (PSR), and that there are successful undercutting and rebutting defeaters for PSR. We've discussed those sorts of worries on other occasions (here, for example), but here I want to discuss a variation on a different criticism that goes back at least to Hume: suppose we grant, at least arguendo, that the argument's sound. The worry is that the argument still doesn't get us to theism. For the necessary or self-explanatory being might, for aught we know, be the universe, and not the god of theism.

In reply, a number of proponents of the argument add premises that, when conjoined to the other premises, entail that the necessary or self-explanatory being is the god of theism. For our purposes, we can keep things simple and add two more premises to get such a conclusion:

1. Either everything can be a dependent being, or there is a self-explanatory being.
2. Not everything can be a dependent being.
3. Therefore, there is a self-explanatory being. (from 1 and 2)
4. If there is a self-explanatory being, then it is either the universe (or the stuff of which it's composed) or the god of theism.
5. The self-explanatory being is not the universe (or the stuff of which it's composed).
6. Therefore, the self-explanatory being is the god of theism. (from 3-5)[1]

Since we're simplifying in the ways mentioned above, we'll avoid questions about (4). What about (5)? Why should we think the universe isn't a self-explanatory being? In answering this question, the proponent of the argument relies on some substantive assumptions that fall within the sub-field of modal epistemology. In particular, the proponent of the argument relies on a substantive thesis about the connection between conceivability and possibility. Thus, she argues that we're able to conceive (or imagine, or intuit) the universe failing to exist. And since (here comes the crucial assumption) conceivability is prima facie evidence of metaphysical possibility, we're prima facie justified in thinking that it's possible for the universe to fail to exist. And since nothing that can fail to exist is a necessary or self-explanatory being, the universe is not such a being.

Richard Taylor's use of the conceivability-possibility inference is typical. Thus, he argues that for any object in the universe, we can imagine that it fails to exist (e.g., a six-foot-in-diameter translucent sphere). But if imaginability is evidence of possibility, then this is evidence that, for any arbitrary object in the universe (whether a stamp or a solar system), it's possible for it not to exist. But we can just as easily imagine the whole universe failing to exist. Therefore, we can say with equal justification that the universe can fail to exist, in which case it's a dependent being.

Is the line of reasoning above for the contingency of the universe a good one? One might think not, on the grounds that Taylor conflates evidence for the possible non-existence of a material object (a stamp, a solar system, etc.) with evidence for the possible non-existence of the stuff of which it's composed (matter-energy).

William Lane Craig is aware of this sort of worry. However, he thinks he can make legitimate use of a conceivability-possibility inference in support of premise (5) by cutting to the chase and asking us to imagine the most fundamental constituents of reality -- quarks (assuming the string theorists are wrong) -- failing to exist; alternatively, he asks us to imagine a universe composed of different quarks. Given this modification of the thought experiment, he assumes that we can adequately imagine this, and further that this is sufficient prima facie evidence that such things are possible.

Has Craig sufficiently addressed the worry that plagues Taylor's inference? Not obviously. For one might think (as Hume and others have thought) that the line of reasoning proves too much. For one might think that the same goes for God: we can conceive of his non-existence, in which case we should conclude that God, too, is a contingent or dependent being.[2] Thus, consider Peter van Inwagen’s “knownos”.[3] A knowno is a being who knows there are no necessary beings. Now I fail to see any incoherence in the notion of a knowno. Alternatively, I see no incoherence in a purely physical, contingent, yet metaphysically independent or “free-standing” universe: there are the fundamental particles, and all else logically supervenes on that. So if conceivability is a guide to possibility, it’s also reasonable for me to believe that there are no necessary beings.

I think this Humean criticism has a lot of force. Furthermore, it should be noted that Craig fails to even mention it, let alone address it, in the piece linked to above. For Craig to adequately address this worry, then, he'll need a principled basis for saying that the conceivability-possibility inference for the No Quarks and Different Quarks thought experiments are justified, while the conceivability-possibility inference in the No God thought experiments are not.

But what could such a basis be? Presumably, it'd be a substantive account of modal epistemology, and one that underwrites the legitimacy of the No Quarks and Different Quarks thought experiments, but precludes the legitimacy of the No God thought experiments. As a matter of fact, he has elsewhere endorsed Charles Taliaferro's account.[4] Taliaferro states his account as follows:

"If one can conceive (picture, visualize, imagine) that a state of affairs obtain and one has carefully considered whether the state of affairs is internally consistent (self-consistent at a minimum) and consistent with what one justifiably believes, then one has prima facie reason to believe it is possible for the state of affairs to obtain."[5]

Does this account of the epistemology of possibility ground a principled basis for saying that Craig's Quark thought experiments are justified, while No God thought experiments are not? Unfortunately, no. For the states of affairs depicted in the No Gods thought experiments are also conceivable in Taliaferro's sense. So, for example, I can coherently conceive of a universe where there are a set of metaphysically contingent yet independent or "free standing" fundamental particles, and all else in the universe logically supervenes upon them. This state of affairs seems internally consistent, and consistent with what I justifiably believe. Therefore, according to Taliaferro's account, I have prima facie reason to believe it is possible for this state of affairs to obtain. But such a state of affairs is incompatible with any state of affairs involving a necessarily existent God. Therefore, Craig's modification of Taylor's defense of premise (5) is unsuccessful.

What are the prospects for an adequate account of modal epistemology that could underwrite Craig's thought experiments, yet preclude the legitimacy of the No God thought experiments? My own worry is this. I wrote my doctoral dissertation on modal epistemology and thought experiments. And while I don't want to give away my bag of tricks while I have some papers on them out for review, let me just say that it's extremely dubious that our knowledge of what's metaphysically possible is extensive enough to underwrite Craig's inference from conceivability to possibility here. The basic worry is that while we do have knowledge of a significant range of possibilities, such knowledge is constrained by our knowledge of actuality in various ways. Unfortunately, though, we don't have relevant actual-world knowledge (or at least not enough of it) to ground the justification of the claim that there could be no quarks (or strings), or that there could've been a universe composed of different quarks (or strings). And if that's right, then we don't know -- nor can we have sufficiently justified belief -- that the states of affairs depicted in Craig's thought experiments are possible.

At any rate, whether I'm right about our knowledge of possibility or not, Craig hasn't given a principled basis for distinguishing between the conflicting thought experiments. Therefore, as it currently stands, Craig's modification of Taylor's defense of the Leibnizian cosmological argument is subject to a dilemma: either his account of modal epistemology is legitimate or it isn't. If it is, then it equally justifies the possible non-existence of God. But if it isn't, then he loses his basis for believing in the possible non-existence of quarks. Either way, his modification of Taylor's defense of the Leibnizian cosmological argument is inadequate (at least when offered to the antecedently unconvinced).
[1] Here I compress instances of Modus Ponens and Disjunctive Syllogism to minimize the number of explicit steps.

[2] Now one might attempt to get around the criticism by offering a cogent ontological argument. If one could do that, then one could argue that while the universe's non-existence is conceivable, God's is not. But of course, if one had such an argument in hand, the cosmological argument would be superfluous. In any case, the ontological argument itself relies on a substantive connection between conceivability and possibility, just as does the cosmological argument under consideration here. And as I have argued elsewhere, our modal knowledge doesn't extend to matters as remote as that of the metaphysical possibility of Anselmian beings. My current task in this post is to begin to argue that, perhaps surprisingly, the relevant modal intuitions involved in the Leibnizian cosmological argument are just as dubious as those involved in (say) Plantinga's ontological argument.

[3] Van Inwagen, Peter. Metaphysics, 2nd ed. (Boulder, CO: Westview Press, 2002), pp. 107-108.

[4] "The Cosmological Argument", in Copan, Paul and Moser, Paul. The Rationality of Theism (Routldege, 2003), pp. 112-131, esp. p. 115, and p. 130, fn. 8.

[5] "Sensibility and Possibilia: A Defense of Thought Experiments", Philosophia Christi 3, 2 (new series) (2001), pp. 403-420. The account is given on p. 407.


Josh said…
Interesting and thoughtful post. I'm curious what you might think of the thought that the large but finite number of fundamental particles (assuming the "universe" is ultimately built up from fundamental particles) could have been greater or lesser. Might it strike one as unlikely that the exact number of gravitons, say, is a matter of broad logical necessity? Why that number rather than one more?

Another thought: I wonder if it might be useful here to think about modal properties of types rather than tokens. I'll explain. Take the type, being maximally great (or being God). That type might very well entail necessary existence, which is merely to say that were an x maximally great, x would have necessary existence. Someone might suspect that one cannot even conceive of there being a maximally great being that is not a necessary being. Note: this doesn't give us an ontological argument because the thought is not that one cannot conceive of the property, being maximally great, as lacking exemplification. It's rather that one cannot conceive of somethings being maximally great without conceiving of it as having necessary existence.

It might be further thought that the type, being material, is not like that: one can very well conceive of a material thing lacking necessary existence. Same for the fundamental kinds of material things: e.g., one can conceive of somethings being a graviton without conceiving of it as having necessary existence. All this may suggest that the fundamental particles are not themselves necessary, whereas if there were a maximally great being, it would be necessary.

On the other hand, the fundamental particles may have "hidden" properties that do entail necessary existence. Still, one may wonder whether that is likely in light of a principle of modal continuity: e.g., if there can be 10^22 tokens of being a graviton, then there can be 10^22 + 1 tokens of being a graviton.

I'm curious what you might think?
exapologist said…
Hi Josh,

Great questions!

There is a long story to tell here, as an adequate reply involves a defense of an observation-based modal epistemology, along with a critique of extant accounts. I have offered such a case in the first half of my dissertation. My views have changed a little since then, but I could email you a copy, if you're interested.

exapologist said…

Sorry -- I was a bit busy when you last commented, and didn't really give it a careful read, so I just gave a promissory note for a reply. I'm still busy, but I had a moment to read your comment more carefully. I'm glad I did!

Re: your Quarks thought experiments: Yeah, I've been meaning to revise my post about this. I recently revisited Craig's post, and his remarks suggest other thought experiments I've failed to address: a Fewer Quarks thought experiment and a More Quarks thought experiment. I think the former thought experiment shares problems similar to the No Quarks thought experiment. However, the latter is what you had in mind in your comment, and I tend to think that it has a good deal of force. On my own account of modal epistemology, our knowledge of possibility traces back to our knowledge of actuality in various ways, and one of them is rational or intuitive induction from one or more actual tokens of a given type to the possibility of other tokens of the same type. Therefore, on my own account, I'm justified in believing that more quarks of the same type are possible.

I'm not sure I agree with your principle of modal continuity though, at least not in this case. The reason is that, for all I know, there's an inherent sustainability limit to the number of fundamental particles. Thus, while I may be justified in believing that other tokens of actual quarks are possible, I'm not justified in believing that all possible tokens (or in any case, denumerably infinitely many tokens) of actual-world-type quarks are co-instantiable. However, that quibble doesn't affect the force of a More Quarks thought experiment that involves, say, just one more quark in the universe.

What about the God case? I agree with your conditional, but I'm not sure what help that is, as I don't know how to justify the antecedent. For all I know, it's a non-trivial counterpossible, on the order of Daniel Nolan's example: "If Hobbes had squared the circle, sick children in the mountains of South America at the time wouldn't have cared." Also, wouldn't your principle of modal continuity imply that if there can be one maximally great being, then there could be two such beings, and in general, if n Anselmian beings are possible, then n+1 such beings are possible? If so, then given Axiom S5, won't we have a plenitude of Anselmian Beings?

Finally, what about the No Quarks and Fewer Quarks cases? I don't yet see a reply to my dilemma: either Craig's modal epistemology is legitimate or it isn't. If it is, then it equally justifies the possible non-existence of God. But if it isn't, then he loses his basis for believing in the possible non-existence of quarks.

What do you think?
Josh said…
Good thoughts, EA.

I will be interested to see your dissertation ideas. I have a feeling I will be seeing them in published articles soon enough. :)

I don't know if Craig's modal epistemology would equally justify the possible non-existence of God. You suggest that we can conceive of a universe where there is a set of metaphysically contingent yet independent or "free standing" fundamental particles, such that all else in the universe logically supervenes upon them. Call the state of affairs of such a universe's existing, S. I wasn't sure how to understand S in such a way that a necessarily existent God would be incompatible with it. For example, God's existence might logically supervene on the existence of the free standing particles (as well as on their non-existence). It seems like we'd have to build into S something like "and there are no other entities." S would still be conceivable, in some sense of conceivable. It's just not presently clear to me that it would be conceivable in Taliaferro's sense. Can the non-existence of God be pictured, visualized, or imagined? What about the non-existence of the number 4?

By contrast, I certainly can visualize there being an extra pillow case sitting on my bed. Question: would our universe still exist if there were an extra pillow case in it?

So, I guess I wasn't completely convinced by the dilemma, though I surely might have missed something.

That said, apart from a little Chalmers reading, I haven't thought much about various kinds of conceivability and their potential guides to possibility. Perhaps Craig's modal epistemology is not that reliable-- I bet your dissertation would help us here.

There is the inductive method you alluded to. Question: does your inductive method give us prima facie justification for thinking that the type, being an even prime number, can have more than one token given that we know that there is a token of it, namely, the number 2?

Another question: if there can be more quarks, then doesn't that suggest that the type Quark doesn't entail necessary existence (otherwise, whatever "additional" quarks are possible would already be actual, given S5)? And if Quark doesn't entail necessary existence, then might that suggest that any given token quark is contingent? If so, then unless the non-existence of some quarks depends upon the existence of other quarks, both Fewer Quarks and No Quarks would be possible. Of course, in reply, one might wonder whether some quarks have additional properties that entail necessary existence. (In reply back, could one use your modal epistemology with respect to any "hidden" property P to suppose that there could be another token/instance of P, thereby revealing that P doesn't entail necessary existence?)

Turn to the God case. A couple reply thoughts:

1. Re: modal continuity. I suppose I was thinking that if there are n tokens of type T, and T entails necessary existence, then n isn't likely to be a large but finite number. Without additional data, I'd expect n to be 0, infinity, or perhaps 1. Still, I grant that strictly speaking, a completely unrestricted principle of modal continuity would suggest that there are an infinite number of Anselmian Beings (if there are any) as well as an infinite number of even prime numbers. And I admit that I've yet to come up with a principled way to restrict the principle.

2. You're right that for all that I said, the conditional--were there a maximally great being, it would have necessary existence--might be a counterpossible. My point was merely to contrast Being God with Being Physical (suggesting that being physical does not entail necessary existence) in order to suggest a reason to think that physical things are contingent even if God, were he to exist, would not be. In other words, I was suggesting a conceivability argument (concerning types) that might apply to physical things but not to God. I suppose I was suggesting possible support for (5) in your original post.

~EA fan
exapologist said…
Hi Josh,

Thanks for the kind words! I like your work very much, btw!

Sorry to take so long to reply to your nice questions and points. I hope to be done grading in the next couple of days or so, and to reply before I go on vacation.

All the best!
exapologist said…
Hi Josh,

Great questions as usual. Re: your first set of questions: One of my aims in discussing the different thought experiments is to indicate that if Craig’s Quarks thought experiments are to count as instances of adequate imaginings on Taliaferro’s account of modal epistemology, then there is pressure on his part to further elaborate on that account in a way that distinguishes between adequate and inadequate imaginings. Consider the Quarks cases. No one has seen a quark, in which case it would seem that a strict interpretation of Taliaferro’s pictureability criterion immediately precludes the legitimacy of Craig’s Quarks thought experiments. Thus, to apply the account to the Quarks cases, we’ll need to loosen the standards to allow for imaginative stand-ins, accompanied by stipulations to the effect that what one imagines is an adequate representative surrogate. But if so, then it’s not clear why my No Gods thought experiments are illegitimate, no?

Paul Tidman raises a related concern for imaginability accounts of modal epistemology: what he calls the “What Counts?” problem. The basic idea is that the mind can make anything represent anything via an act of stipulation. So, for example, I can imagine a Zeus-like being in space say, “let there be planets!”, followed by the immediate appearance of planets. One might then say that this is to count as evidence for the possibility of omnipotence, or of ex nihilo creation. But intuitively, this is pretty lousy evidence for such possibilities. But then there is pressure to give an account that distinguishes between adequate and inadequate imaginings. 

Furthermore, assuming this problem can be adequately addressed, there is what I have called the “Unexcluded Alternatives” problem raised by Yablo, and derivatively by van Inwagen. This problem occurs when what one imagines is compatible with the falsity of the modal claim it attempts to support. So, for example, consider the claim that, possibly, there is transparent iron (to take a case from van Inwagen). Now I can objectually imagine (e.g.) a scientist accepting the Nobel Prize before a cheering audience, who then thanks all those who helped him in his long and difficult journey to create transparent iron, and then holds up something that looks like a sheet of glass. But, intuitively, that wouldn’t prima facie justify the claim that transparent iron is metaphysically possible. For the imagining is equally supportive of other claims that are compatible with the impossibility of transparent iron. For example, the imagining equally supports the claim that, possibly, (despite the impossibility of transparent iron) a group of jokester scientists get together and fool the public that they’ve created transparent iron, using a piece of glass as their object of deception. And since my objectual imagining can’t rule out this alternative that’s compatible with its falsity, it seems wrong to say that the imagining justifies the claim.

Thus, it seems to me that, pending further elaboration on their part re: adequate picturings or imaginings, the Taliaferro/Craig account is susceptible to these sorts of problems. There is thus pressure for them to give an account that at explains why Craig’s Quarks thought experiments are legitimate, while my No Gods thought experiments are not.
Josh said…
Thanks for those further thoughts and elaborations. They were clear, and I agree with what you say. (I realize my example of the pillow case doesn't help with Quarks, as all I visualize is a pillow-cased shaped thing.)

Enjoy your vacation!
exapologist said…
Thanks, Josh!

I actually wrote several different versions of a reply, and kept deleting them. For a while, I just couldn't find a chunk of time sufficient to get clear on a reply. in any case, thanks for pushing me to get clearer on all of this.

It's Celeste! said…
I am studying Philosophy & Religion at Huston-Tillotson University in Austin, TX, and we studied about the cosmological argument. For me it is a very complex and often confusing subject for me. I am a Christian. I am also biased when it comes to my faith, because I believe in God. God, as the all-being who controls the entire universe, but gives humans the freedom to choose to believe or not believe. Would you say that the cosmological argument in your own life brought you to or away from whatever faith you have? You opinion would be greatly appreciated. Thank you in advance for your time.
exapologist said…
Hi Celeste,

I'm currently an agnostic, but there was a time when I was persuaded by a couple of versions of the cosmological argument (this one included), and they boosted my confidence quite a bit that theism was true. Conversely, when I later became convinced that they were defeated by certain criticisms, it deflated my confidence quite a bit re: the truth of theism.

Re: your remarks about free will: it's perhaps interesting to note that Christian philosopher Peter van Inwagen has argued that if PSR is true, then we lack libertarian free will.


Popular posts from this blog

Epicurean Cosmological Arguments for Matter's Necessity

One can find, through the writings of Lucretius, a powerful yet simple Epicurean argument for matter's (factual or metaphysical) necessity. In simplest terms, the argument is that since matter exists, and since nothing can come from nothing, matter is eternal and uncreated, and is therefore at least a factually necessary being. 
A stronger version of Epicurus' core argument can be developed by adding an appeal to something in the neighborhood of origin essentialism. The basic line of reasoning here is that being uncreated is an essential property of matter, and thus that the matter at the actual world is essentially uncreated.
Yet stronger versions of the argument could go on from there by appealing to the principle of sufficient reason to argue that whatever plays the role of being eternal and essentially uncreated does not vary from world to world, and thus that matter is a metaphysically necessary being.
It seems to me that this broadly Epicurean line of reasoning is a co…

CfP: Inquiry: New Work on the Existence of God

In recent years, methods and concepts in logic, metaphysics and epistemology have become more and more sophisticated. For example, much new, subtle and interesting work has been done on modality, grounding, explanation and infinity, in both logic, metaphysics as well as epistemology. The three classical arguments for the existence of God – ontological arguments, cosmological arguments and fine-tuning arguments – all turn on issues of modality, grounding, explanation and infinity. In light of recent work, these arguments can - and to some extent have - become more sophisticated as well. Inquiry hereby calls for new and original papers in the intersection of recent work in logic, metaphysics and epistemology and the three main types of arguments for the existence of God. 

The deadline is 31 January 2017. Direct queries to einar.d.bohn at

Notes on Mackie's "Evil and Omnipotence"

0. Introduction
0.1 Mackie argues that the problem of evil proves that either no god exists, or at least that the god of Orthodox Judaism, Christianity, and Islam, does not exist. His argument is roughly the same version of the problem of evil that we’ve been considering.
0.2 Mackie thinks that one can avoid the conclusion that God does not exist only if one admits that either God is not omnipotent (i.e., not all-powerful), or that God is not perfectly good. 0.3 However, he thinks that hardly anyone will be willing to take this route. For doing so leaves one with a conception of a god that isn’t worthy of worship, and therefore not religiously significant.
0.4 After his brief discussion of his version of the problem of evil, he considers most of the main responses to the problem of evil, and concludes that none of them work.

1. First Response and Mackie's Reply
1.1 Response: Good can’t exist without evil; evil is a necessary counterpart to good.
1.2 Mackie’s reply:
1.2.1 this see…