Cosmological Arguments

People speak of the cosmological argument as though there were one of them. There are at least three, they have different premises, they are vulnerable to different objections, and they do not have the same conclusion.

One starts from the claim that the universe began. One starts from the claim that the universe did not have to exist. One starts from what is happening at this moment, and would run in exactly the same way if the universe had always been here. Refuting any one of them leaves the other two standing, which is why arguments about this subject so often consist of two people demolishing positions the other has not taken.

The shared shape

The name comes from cosmos, the world, and marks the starting point rather than the method. What distinguishes these arguments from design arguments is that they proceed from the bare fact that there is a universe, or from something true of absolutely everything in it, rather than from any particular feature such as its order or its suitability for life. That matters for what can defeat them. A design argument can be answered by supplying a better explanation of the specific feature. A cosmological argument cannot be answered that way, because there is no more specific fact to explain away.

All three versions have the same three-part skeleton.

Something about the world stands in need of a cause or an explanation.

That cause or explanation cannot be found within the world, or cannot be found there without the same demand arising again forever.

Therefore something outside the world satisfies the demand.

Where they differ is on the first step. The Kalam version says what needs explaining is that the universe began. The Leibnizian version says what needs explaining is that it exists at all when it might not have. The Thomistic version says what needs explaining is that anything is exercising causal power right now. Those are three different demands, supported by three different principles, and a reader who does not keep them apart will not be able to follow the objections.

One limitation applies to all of them and should be fixed at the start rather than sprung at the end. Even a fully successful cosmological argument delivers a first cause, or a necessary being, or a sustainer. It does not by itself deliver a being that is knowing, good, or personal. Closing that distance takes further argument, and the beginner tier called this the gap problem. It comes back at the end of this article.

The argument from a beginning

The name comes from kalam, the tradition of Islamic speculative theology. Its best known statement is by al-Ghazali, who lived from 1058 to 1111 and who set it out in Tahāfut al-Falāsifa, The Incoherence of the Philosophers, written in 1095 against the Aristotelian philosophers of his tradition, Avicenna among them, who held the world to be eternal. William Lane Craig revived it for contemporary philosophy in The Kalām Cosmological Argument of 1979, and most current discussion descends from that book.

The argument is short enough to state completely, which is part of its appeal and part of its trouble, since everything then depends on two premises that carry a great deal of weight.

The causal premise

Whatever begins to exist has a cause. Defenders treat this as more evident than anything that could be produced against it, and note that denying it commits you to the view that things can simply appear, with no explanation and no constraint on what appears or where.

Two objections are pressed.

The first is from quantum mechanics, where events occur without determining causes and particles behave in ways that resist description as one thing making another happen. The reply is that indeterminacy is not the absence of causal conditions, since these events occur only in specific circumstances and with specific probabilities, and that on some interpretations the indeterminacy is a limit on what can be known rather than a feature of what is there. The dispute is genuinely open and turns on the interpretation of quantum mechanics rather than on anything about God.

The second is more damaging and less often answered. Every case that supports the premise is a case of something coming to be out of material that was already there, in a time that was already running. A statue begins to exist when marble is shaped. What the argument needs is a principle covering something else entirely: the coming into being of matter, space, and time together, with no prior material and no prior time. A generalisation drawn from rearrangements is being applied to a case that is not a rearrangement, and the article on creation in the beginner tier gave the reason to think the two are not the same kind of event. The defender has to argue that the causal principle holds for both, and cannot simply read it off the cases that motivate it.

The premise that the universe began

This premise is supported in two independent ways, and the two supports are in an uncomfortable relationship with each other.

The philosophical support says that if the universe had no beginning, then an infinite number of past events would have occurred, and that this is impossible.

The distinction is the crux. Nobody disputes that you can always add one more to a finite collection. The claim is that a collection whose members are all there, and are infinitely many, cannot exist in reality.

The standard case for this is by illustration. Imagine a library holding infinitely many books. Remove every second book and you have removed infinitely many, and infinitely many remain, and the shelves hold exactly as many books as before. Add a further infinite collection and the number is unchanged again. Craig's contention is that a collection behaving this way is not merely surprising but absurd, and that the absurdity shows no such collection can be instantiated.

The mathematician's reply is that nothing here is absurd. That a proper part can be matched one to one with the whole is the defining property of an infinite set, established by Cantor, and the arithmetic of infinite cardinalities is consistent and well understood. Behaving unlike a finite collection is what an infinite collection is; calling that a contradiction assumes that intuitions formed on finite collections must transfer.

Craig's response is that consistency within a formal system does not establish that anything answering to it could exist, and that this is a metaphysical question rather than a mathematical one. That reply is not obviously wrong, and it is also not obviously right, and the honest report is that the dispute has not moved much in decades. What can be said is that the burden here is heavier than the illustration suggests. To get the conclusion, the defender needs it to be impossible that an infinite collection exists, not merely strange, and strangeness is what the illustration establishes.

The scientific support says that physical cosmology describes a universe with a finite past, expanding from an early hot dense state, and that a universe with a finite past began.

Two cautions apply, and the first is one the beginner tier already raised in a different setting.

Cosmology is revisable, and an argument resting on it inherits that. Models without a first state have been proposed and have not been ruled out. A premise that current cosmology supports is a premise that a future result could withdraw, and someone whose case for God rests on it has accepted that exposure.

And the two supports pull in different directions for a theist. The philosophical support, if it worked, would show that an eternal past is impossible. The scientific support ties the doctrine of creation to a datable physical event, which the article on creation gave reasons to avoid, and which Aquinas explicitly declined to do when he held that the world's having had a beginning cannot be demonstrated by reason. The Kalam is the one argument in this family that a classical theist can lose on empirical grounds, and the tradition it comes from was content with that. It is worth knowing which trade you are making.

The argument from contingency

The second version drops the question of a beginning entirely. It would run unchanged in a universe that had always existed, and this is its principal advantage.

Leibniz gave the strong statement of it and built his argument for God directly on it, in the Monadology of 1714 and elsewhere, and Samuel Clarke argued in the same style in England shortly after. The argument runs as follows.

Everything in the universe is contingent, in the sense that it exists but might not have. Each such thing has an explanation of its existing, and that explanation lies in other things.

Now take the whole collection of contingent things, however many there are and however far back they run. That collection is itself contingent, since a collection all of whose members might not have existed might itself not have existed. So it too has an explanation.

The explanation cannot be any member of the collection, since no member explains the collection that contains it. And it cannot be another contingent thing, since every contingent thing is already in the collection. So the explanation is something that is not contingent.

Two senses of necessity have to be distinguished, because the argument is sometimes attacked in one sense and defended in the other. On the logical sense, a necessary being is one whose non-existence would be a contradiction. On a weaker sense sometimes called factual necessity, a necessary being is one that did not come into existence, cannot cease, and does not depend on anything. The argument as stated needs something in the neighbourhood of the second, since what it establishes is a terminus of explanation. Whether the terminus must also be logically necessary is a further question, and the ontological argument in a later article is what would be needed to settle it.

The objections are the best in this family.

The fallacy of composition. Russell pressed the charge that the argument infers a property of the whole from a property of the parts, which is invalid in general: every part of a wall being small does not make the wall small. Every contingent thing having an explanation does not obviously mean the collection of them does, and every member being contingent does not obviously make the collection contingent.

The reply is that composition is not always fallacious and that this instance is one of the sound ones. If each member of a collection might have failed to exist, then it might have been that none of them existed, in which case the collection would not have existed either. That reasoning has force. What it does not do is establish that the collection needs an explanation, which is a separate step supplied by the principle rather than by the composition point.

The van Inwagen objection, which is the sharpest thing in the literature on this argument. Take the conjunction of every contingent truth, a single enormous fact. The principle says it has an explanation. If the explanation is itself contingent, it is one of the conjuncts and so explains itself, which is circular. If the explanation is necessary, then it necessitates what it explains, and everything contingent turns out to be necessary after all, which is absurd.

The reply that most defenders take is that a necessary being can freely produce a contingent effect, so the explanation is necessary and the effect is not. That costs something. The being's making that choice rather than another is itself a contingent fact, and the principle now has to be weakened so that free choices count as explained without being necessitated. Whether the weakened principle still supports the argument is the live question, and there is no settled answer.

Hume's objection to the principle itself. It is not a truth of logic, since we can conceive of something existing with no explanation without contradiction. It cannot be established by induction, since every instance we could survey is drawn from inside the universe and the principle is being applied to the universe as a whole. Defenders reply that it is self-evident, or that it is presupposed by every inquiry including scientific inquiry, and that abandoning it makes the demand for explanations arbitrary. The objection to that reply is that the practice of looking for explanations does not require that everything has one, only that it is worth looking.

The universe as the necessary being. If the argument needs a terminus, why should the terminus not be matter and energy, which on the conservation principle are neither created nor destroyed? The reply is that indestructibility is not the necessity required, since matter having the properties it has, in the quantity it has, looks like exactly the sort of thing that could have been otherwise. That reply is reasonable and it does not close the question, because it turns on a judgement about what could have been otherwise, and that judgement is contested.

The argument from what is happening now

The third version is the one most often misdescribed, and the misdescription is that it argues for a first cause at the beginning of time.

Aquinas's second way, in the Summa Theologiae at part one, question two, article three, argues from efficient causation to a first efficient cause. He also held, as the article on creation set out, that reason cannot show the world to have had a beginning. Those two positions are consistent because the series he says cannot be infinite is not a series running backwards in time.

The contrast is with a series ordered per accidens, where each member causes the next and then is no longer needed. A grandfather causes a father who causes a son, and the grandfather can be long dead while the father does his causing. A series of that kind can run back forever without difficulty, and Aquinas said so.

The stock illustration of the other kind is a hand moving a stick that moves a stone. Take away the hand and the stick moves nothing, and this is true at the moment, not merely earlier. The illustration is worth using and worth distrusting, because sticks do have causal powers of their own, and a critic can reasonably say that the case is not as clean as the argument needs.

The version that does not depend on the illustration runs through existence rather than through pushing. At this moment a thing exists. Its existing now is not something it accounts for by itself, on the account of conservation given in the beginner tier. So something is conferring existence on it now. If that something is in the same condition, the question repeats. The claim is that a series of this kind, however long, confers nothing unless there is a member that has the power in its own right, because what is being passed along is not present anywhere in a series of borrowers.

The objections are two.

The first is a direct denial: why can the series not be infinite, each member receiving from the next forever? Defenders answer that an infinite series of things each of which has no power of its own is an infinite series of things with no power, and that adding members does not produce what none of them has. The objector answers that this presupposes what is being argued, namely that causal power has to originate somewhere rather than being a feature of the whole arrangement. This is the fundamental disagreement in the argument and it is not resolved by either side restating it.

The second is that the framework is Aristotelian and that nothing in modern physics looks like a per se series. Physics describes states evolving under laws, not a hierarchy of things holding one another up. The reply is that the series claimed is not a series of pushes but of existential dependence, which physics neither describes nor denies, since physics takes the existence of its subject matter as given and explains how it behaves.

That reply is fair, and it has a consequence the defender should accept openly. This argument and the doctrine of divine conservation are the same claim in two directions. If nothing needs holding in existence, the argument has no premise. If the argument works, conservation follows. Neither can be used as independent support for the other.

From a first cause to God

Grant the best case. What has been established is a cause of the universe's beginning, or a necessary being, or a source of existence. Nothing so far is knowing, good, or personal, and classical theism asserts all three.

The standard route across is a series of further inferences. The cause of space and time cannot itself be in space or time, so it is timeless and immaterial. It brought about the whole physical order, so it is enormously powerful. Then the step that carries the weight: a timeless cause with unchanging conditions would produce an unchanging effect, since nothing about it would explain why the effect began at one point rather than another, so the production of a temporal universe by a timeless cause requires something that can act rather than merely operate, which is to say something with a will.

That last step is where a critic should press, and it is where Kant already pressed. Nothing in the argument shows that the alternatives are exhausted by mechanism and choice, and the notion of a timeless being making a choice is not evidently coherent, since choosing looks like something that takes place. The advanced tier takes up whether a timeless God can act at all.

The right summary is that the cosmological arguments, at their best, get you to something the universe depends on. They do not get you to God on their own. On the framework the beginner tier set out, that is not a refutation, since these are contributions to a cumulative case rather than proofs. It does mean that an argument presented as delivering God is being oversold, and that the overselling is common.

Next

These arguments proceed from the bare existence of the universe and take no interest in what it contains. The next article turns to the arguments that do: from the apparent adaptation of living things, and then from the narrow range of physical constants that permit any complexity at all, together with the two replies that have shaped the modern debate.