Every other argument for God starts by looking at something: motion, causes, the appearance of design, the fact that anything exists at all.

This one does not look at anything. It claims you can establish that God exists by examining the concept of God and noticing what is already in it — the way you establish that triangles have three sides without measuring any triangles.

If it worked, it would prove God with the certainty of arithmetic. Almost nobody thinks it works. Nobody has managed to make it stay dead either.

Anselm's move

Anselm wrote it in 1078, in a prayer, addressed to God.

Take God to mean that than which nothing greater can be conceived. Even someone who denies God understands the phrase — otherwise they would not know what they were denying. So the concept is in their understanding.

Now: if that being existed only as a concept and not in reality, you could conceive of something greater — the same being, existing. But nothing greater than that than which nothing greater can be conceived is conceivable. That is what the phrase means. So the supposition breaks, and the being must exist in reality.

The whole argument turns on one premise: that existing is greater than not existing. Everything since has been an argument about that premise or about what kind of move is being made with it.

The perfect island

The objection arrived immediately, from a monk named Gaunilo, in a piece titled On Behalf of the Fool.

Conceive of the most excellent island — more perfect than any island, than any land at all. By Anselm's reasoning, if it existed only in your understanding it would not be the most excellent, since an actually existing island would beat it. Therefore the island exists.

It does not. So something is wrong with the reasoning.

Anselm's reply is better than it is usually credited as being. The island has a built-in ceiling: islands are the kind of thing that can always be improved, and there is no intrinsic maximum of islandness. Greatness without restriction is not a property islands can have. God, defined as maximal greatness as such, is the one case where the form of the argument applies.

Whether that genuinely blocks the parody, or just asserts that God is the one exception, has been argued about for nine hundred years.

How Kant killed it

Aquinas had already rejected it, on the ground that we do not have the kind of access to God's essence that the argument assumes. But the objection that settled matters for two centuries came from Kant, and it is the sharpest thing in the whole debate.

Existence is not a real predicate.

Say the triangle is three-sided and you have added something to the concept — that is information about triangles. Say the triangle exists and you have added nothing to the concept at all. You have said that something falls under it. The concept of a hundred real dollars, Kant notes, contains not one cent more than the concept of a hundred possible dollars. What differs is whether your pocket instantiates it.

If that is right, existence cannot be packed into the concept of God and then unpacked, because existence is not the sort of thing concepts contain. On that account the argument commits a category mistake, and that is where the matter sat through the nineteenth century.

It comes back

Norman Malcolm in 1960 and Charles Hartshorne in 1962 noticed that Anselm has a second argument, in chapter three of the Proslogion, which is about necessary existence rather than existence — and Kant's objection does not obviously touch it.

Alvin Plantinga gave the canonical modern version in 1974, in modal logic. Roughly: maximal greatness means being maximally excellent in every possible world. So if a maximally great being exists in any possible world at all, it exists necessarily, and therefore in this one. The whole argument reduces to a single question — is a maximally great being possible?

Notice what has happened. Kant's objection is sidestepped, because nothing is being analyzed out of a concept. But the cost is that everything now rests on a premise about possibility that is exactly as hard to establish as the conclusion.

The parody, again

Gaunilo's move returns in modal dress, and it is the strongest objection going.

Define maximal anti-greatness — necessarily existing, maximally evil. Concede it is possible. The same logic delivers a necessarily existing evil being in every world, including this one.

Both arguments are valid and their conclusions are incompatible, so at most one possibility premise is true. And nothing inside either argument tells you which. The modal version, like Anselm's, works only if you already have a reason to grant one possibility and deny the mirror image.

Plantinga does not really dispute this. His considered verdict is that the argument shows belief in God to be rationally acceptable, not that it demonstrates God to anyone who does not already grant the premise — a notably honest assessment of one's own argument.

Why it is worth your time

Not because it proves anything. It is worth your time because it is the cleanest available demonstration that you cannot define something into existence — and of how genuinely hard it is to say why you cannot.

The intuition that Anselm has cheated is immediate and nearly universal. Nine centuries of first-rate philosophers have found it remarkably difficult to locate exactly where. Kant's answer is the best one anybody has, it took seven hundred years to arrive, and it still did not close the question.

That gap — between knowing an argument is wrong and being able to show it — is worth sitting with. It recurs whenever an argument's conclusion is absurd and every premise looks fine.

Further reading