Religious Language

Every rung of this tier has been arguing about sentences of the form "God is F". God is simple. God is timeless. God knows what you will do tomorrow. The arguments have been conducted as though the meaning of those sentences were settled and only their truth were in question.

The classical tradition denies that. Aquinas holds that we do not know what God is, and he holds this not as a confession of ignorance to be remedied by further study but as a structural consequence of the doctrine of simplicity. If God is not a member of a kind, has no parts, and is not composed of essence and existence, then the machinery by which we ordinarily describe things has nothing to grip. The words are still used, and the tradition uses them constantly. What it denies is that they are doing what they do elsewhere.

That leaves a problem that has to be faced before the doctrines mean anything, and in the twentieth century the problem arrived from a second direction entirely, in the form of a general theory of meaning under which theological sentences did not merely fail to describe God but failed to say anything at all.

Why the problem is generated from inside

The pressure does not come from skeptics. It comes from the doctrine of divine simplicity, and it is worth seeing precisely how.

Take the sentence "God is good". If it works the way "Ahmed is good" works, then there is a property, goodness, and God has it. But rung 4 established that on the classical account God does not have properties, because having a property means being composed of a subject and something that could have been absent. So whatever "is good" is doing in the theological sentence, it is not attributing a property to a subject.

Take "God is omnipotent" and "God is omniscient". If these attribute two properties, God has parts. If they attribute one, the two sentences say the same thing, which they plainly do not seem to. The classical answer is that they pick out the same reality under different descriptions, and that the difference lies in us rather than in God, which is a coherent thing to say but leaves us needing an account of how a description can succeed in picking out a reality it cannot describe.

And take transcendence, which is the commitment underneath both. If God is not one item among the items in the world, then our vocabulary, which was developed to sort items in the world, is being applied outside the conditions under which it was given its meaning. The worry is not that we will get God wrong. It is that we will fail to say anything determinate about God at all, and mistake the resulting noise for reverence.

Saying what God is not

The oldest strategy accepts the difficulty and turns it into method.

The tradition runs from the writings attributed to Dionysius the Areopagite, particularly The Divine Names and The Mystical Theology, through Maimonides, who gives the strongest statement of it in the first book of the Guide for the Perplexed. Maimonides holds that the only attributes properly ascribed to God are negative ones and attributes of action, and that positive attributes of essence are strictly excluded. Saying God is powerful, on his account, is best understood as saying God is not powerless, and saying God is wise as saying God is not ignorant.

Maimonides also has an answer to the obvious complaint that negations are uninformative. The answer is that negations narrow. To say a thing is not corporeal, not divisible, not in time, and not dependent is to rule out enormous swathes of possibility, and to accumulate enough such exclusions is to converge on something even if you never describe it. He compares it, in effect, to identifying a thing by elimination.

That reply is available, but it makes the negative theologian dependent on natural theology in a way the mystical tradition would not have welcomed, and it concedes that pure negation is not self-sufficient. Most of the tradition therefore did not stop with negation.

What a word can be doing

Set out the options for a predicate applied to both creatures and God, and there are exactly three.

The second option is pure equivocation, on which the word means something entirely unrelated in the two cases, as "bank" does for a river and a lender. This is not a live position, and it is worth saying why. Under pure equivocation no argument about God could ever be valid, since every term would shift sense between the premise that mentions creatures and the conclusion that mentions God. Every argument in this tier, including the ones the classical theist wants, would collapse. Pure equivocation is the reduction to absurdity that motivates looking for a third option, not a view anyone defends.

The account is doing two things at once, and both are needed. It preserves enough sameness for inference to go through, and enough difference to keep God out of a common genus. The order of priority matters as well: on Aquinas's view the perfection is in God first and in creatures derivatively, even though our knowledge runs the other way, from creatures to God. That asymmetry between the order of being and the order of knowing is what allows him to say both that we name God from creatures and that God is not one more creature.

The standard objection is that analogy needs more than it can supply. To use a proportionality you need to know three of the four terms, and here the fourth is exactly what is in question: to know how God's wisdom stands to God, you would need to know God, which is what the doctrine denies. Critics conclude that analogy is a name for the problem rather than a solution to it, and that once its content is spelled out it collapses either into univocity, in which case simplicity is lost, or into equivocation, in which case meaning is lost.

The univocity of being

John Duns Scotus took the first horn deliberately. He argued in the Ordinatio that the concept of being must be univocal across God and creatures, and his argument is not obscure. Anyone who is certain that God exists while uncertain whether God is finite or infinite has a concept of being that is neutral between the two, and a concept neutral between two alternatives is one concept, not two. Without such a concept, he argued, natural theology cannot get started, because the arguments would equivocate at the point where they moved from creaturely being to divine being.

Scotus was not denying that God differs from creatures. His claim concerns the concept, not the reality: being applies univocally, while the modes, finite and infinite, differ absolutely. The dispute is about whether that is a coherent separation or a concession that lets the analogical distance collapse.

Rung 5 met the historical claim that this move was the beginning of a long decline, with God gradually turned into a very large item in a shared inventory rather than the source of the inventory. That narrative is associated with Radical Orthodoxy and is contested by historians of medieval philosophy, who dispute both the reading of Scotus and the causal story that follows. What is not contested is that the two options are live and that they exhaust the reasonable positions: either the concept is shared, with whatever that costs simplicity, or it is not, with whatever that costs inference.

The charge of meaninglessness

In the twentieth century the problem was reformulated by people who had no interest in it as a theological problem at all.

A. J. Ayer's Language, Truth and Logic (1936) is the most direct English statement of the position, and its treatment of theology is brief because on Ayer's account there is not much to say: theological sentences fail the criterion, so the traditional debate between theists and atheists is not a debate about anything. Ayer was explicit that this was not atheism. The atheist denies a proposition; on Ayer's view there is no proposition to deny.

The principle did not survive, and it is important to be clear that it collapsed for reasons internal to philosophy of science rather than because theologians objected. Every attempt to state it precisely either excluded scientific generalisations, which are not conclusively verifiable, or admitted the metaphysics it was meant to exclude. And the principle itself is neither analytic nor empirically verifiable, which makes it meaningless by its own standard. That objection is often presented as decisive and is sometimes waved away as a debating point, but no satisfactory reformulation was ever produced, and by the 1960s the project had largely been abandoned.

Its influence on philosophy of religion outlasted it, for a reason worth noting. The demand it left behind was not the letter of the criterion but a question: what would have to be the case for this claim to be false? That question is harder to dismiss.

Antony Flew put it to work in "Theology and Falsification", presented at the Oxford Socratic Club and published in New Essays in Philosophical Theology, edited by Flew and Alasdair MacIntyre (SCM Press, 1955). He adapted a parable from John Wisdom's "Gods", in the Proceedings of the Aristotelian Society for 1944 to 1945: two people find a clearing that looks tended, and one insists there is a gardener. Every test comes back negative, and with each negative result the claimed gardener is qualified further, until it is invisible, intangible, and undetectable. Flew's question is what is left of the original claim after so many qualifications, and whether an assertion that nothing could count against still asserts anything.

The reply is not that theology has no answer but that it has several, and the volume printed two of them. R. M. Hare argued that some of our deepest commitments are not assertions at all but frameworks through which experience is interpreted, unfalsifiable because they are what falsification is conducted in terms of, and he coined the word "blik" for them. Basil Mitchell offered a parable in which a partisan meets a stranger who claims to lead the resistance and who then behaves in ways that look like betrayal. The partisan keeps trusting, and Mitchell's point is that he has not abandoned evidence; he has weighed evidence against a commitment formed on other grounds, and he knows perfectly well that the ambiguous incidents count against him.

John Hick's response was to deny that theism is unfalsifiable at all. In "Theology and Verification", in Theology Today in 1960, he argued that religious claims are verifiable in principle even if not falsifiable, because if they are true a future experience will settle the matter. He asks us to imagine two travellers on the same road, one expecting a destination and the other not; nothing along the way distinguishes their positions, but the road has an end, and at the end one of them is right. The proposal turns on survival of death being coherent, which is why this rung depends on more than semantics, and its critics have pressed both that point and the asymmetry between verification and falsification: a claim that can only be confirmed and never refuted is still in an odd epistemic position, since no one can ever be shown to have been wrong.

Giving up the description

A different family of responses accepts that theological sentences are not descriptions and denies that this is a loss.

R. B. Braithwaite argued in the Eddington Memorial Lecture of 1955 that religious statements are neither analytic nor empirical, and that their function is to declare an intention to follow a way of life, with the associated stories serving to sustain that intention rather than to report facts. The account is a form of non-cognitivism about religious language, and it has the structure of the non-cognitivist accounts of moral language that were being developed at the same time and for the same reasons.

The Wittgensteinian line is more common. The later Wittgenstein's treatment of meaning as use, developed in the Philosophical Investigations (1953), suggests that religious utterances belong to their own practice, with their own standards of correctness, and that assessing them by the standards of scientific description is a category error rather than a discovery. D. Z. Phillips is the best-known philosopher of religion in this line. He is usually classified as a religious non-realist, and he rejected the classification, insisting that he was describing what believers actually do with their language rather than reducing it to something else.

The objection to all of these is the same, and it is not a subtle one. Ordinary believers take themselves to be asserting that something is the case. When someone prays for a recovery, they are not expressing an attitude toward suffering; they are asking someone to do something, and they think there is someone there. An account of religious language on which the sentences were never assertions in the first place has not defended religious belief against the verificationist charge. It has agreed with the charge and recommended continuing anyway, and it is fair to ask why the practice should survive the concession.

Phillips's answer is that the objection assumes the very picture in dispute, and that the believer's own use of the language is the evidence, not the philosopher's reconstruction of what they must have meant. Whether that answer is a diagnosis or an evasion is still argued.

Where the argument stands

The revival of philosophy of religion described in rung 5 happened because the verificationist objection collapsed, and the analytic philosophers who conducted it proceeded as though the classical worry about predication had collapsed with it. It had not. The two are independent, and only one of them is dead.

William Alston is the most sustained attempt to bring the two back together. In Divine Nature and Human Language: Essays in Philosophical Theology (Cornell University Press, 1989), he argues that certain predicates can be applied to God literally, without collapsing into anthropomorphism, if their meaning is functional rather than tied to the way they are realised in us. To say God knows something can be to say that a certain functional condition holds, without importing the neural or temporal machinery by which knowing is realised in a human being. The proposal claims a limited univocity: not that the words mean exactly what they mean of us, and not that they mean something merely related, but that a describable core carries over while the mode of realisation does not.

The classical theist has an obvious reply, which is that functional descriptions are still descriptions, and a God who satisfies functional conditions is a God with a structure, which is composition again. That reply returns the whole question to rung 4. This is not a stalemate that anyone has engineered. It is the same disagreement about independence and description surfacing wherever it is pushed.

Next

Suppose the language problem is manageable and the sentences do say something. There is still a question about what entitles anyone to accept them. The evidentialist demand is that belief be proportioned to argument, and the natural theology of the earlier rungs is an attempt to meet that demand. The next rung examines a well-developed refusal to meet it, on which belief in God can be rational without any argument at all, provided it arises from a faculty working the way it was meant to.