22. The criteria for testing a claim of revelation

The question: by what scale is the claim that God has spoken to be tested?

The answer in two lines: five criteria, and they are handed over to you to weigh with them whatever you please — including what we ourselves say in this book. Most of them rule out and do not establish, and that is their limit and not their fault: a scale that rules out clears what is set before you, and then the verdict is left to you by other instruments.

And they were four and became five, and the fifth was added after a doubting reader told us that all our criteria are procedural, and that most of the people he knows did not leave over cosmology but left over a text they found repugnant. And what he said was true, and the criterion will come in its place.

And why do we not apply them here to a religion in particular? — and it is a sound objection

The same reader said: 'you gave me a scale, and told me that it does not establish, and refused to show me how it is held, and then counted it the third of the things I come out with'. And it is an objection well placed, and we answer it with three things:

1. Applying them to a religion in particular is work that cannot be done well in a chapter. It calls for manuscripts and histories of transmission and texts in their own tongues, and every branch of it is a book. And whoever has carried it out in a few pages has led you to suppose that the scale is lighter than it is.

2. But we do not leave you without practice: we shall apply them shortly to a claim that is not religious, in which the instrument can be seen at work, without deciding a matter the book promises not to decide on your behalf.

3. And we grant the unevenness: this chapter holds back from naming, and the two chapters after it — 23 and 24 — do presuppose a religion in particular and speak of 'whom to consult' and 'the environment'. And our reader said: neutrality where it costs nothing, and naming where it costs. And he is right in his description of the state of things, and our answer is that 23 and 24 are two practical chapters addressed to whoever has already chosen, and are not part of the scale. And we shall preface each of the two with this declaration so that it is not read as a verdict.

And ranking the criteria is a claim, so we do not rank them

Our draft made the integrity of the transmission 'the strongest of them' and its effect on the one who bears it 'the weakest of them'. And the reader said: and those are exactly the best thing your claim scores and the worst thing it is met with, and your own rule forbids this — 'if the criterion does not occur to you until after you know the answer required, then it is a conclusion dressed in the clothes of a premise'. And ranking after the answer is known is the very fault itself, not a branch of it.

So the ranking was struck out. The criteria are set out unranked by strength, and what is for each of them and against it is stated. And one criterion is left whose weakness we declare — its effect on the one who bears it — because its weakness is stated with reasons that can be examined and not by a taste that is merely asserted. And it is a weakness that tells against us when we argue from a good effect, and tells for us when a bad effect is argued against us — and we use it in both directions in this book (see chapter 24). So let it be read with both its pans, and we do not pretend that declaring it is a pure cost to us.

The criterion: the integrity of the text

Does it exist in its first language, or in a translation of a translation? Has it a path of transmission that is known, or did it break off and then appear? And how many differences are there among its copies, and of what kind?

The example: you are in dispute over a will and you go to court. The first thing the court asks for is the signed original, not a copy taken from a copy. And were the original missing, the will would not be void automatically, but its weight in the scale is reduced, and the dispute becomes one about transmission before it is one about content.

And its limit: the integrity of the transmission does not establish the truth of what is transmitted. A letter preserved to the letter for a thousand years may have been false from the day it was written. So this criterion tells you: are you arguing with the text or with its ghost? And it says no more.

And what distinguishes it is that it is examined and not merely asserted: manuscripts are counted, differences are classified, chains are traced. And we do not say that it is the most reliable of the five — that is a ranking we struck out a few lines ago for a fault that no change of wording removes, and it is the most useful of the five to us, so we shall not raise it up again by a fresh form of preference. And 'it is examined' is not 'it tells more of the truth' — and we said two lines ago that it does not establish the truth of what is transmitted at all. And the two are run together, and whoever runs them together comes out with other than what he holds.

And we give notice against ourselves: this very criterion favours a text like ours. So whoever rejects our putting it forward has a case, and let him weigh with the five unranked. And whoever accepts in this matter the testimony of the claimant about himself has left the scale and taken the advertisement — and we are claimants, so do not accept our testimony in it.

The criterion: the coherence of the conception

Is there a contradiction in the description? Is an attribute affirmed of God and then its like denied in another place with no difference between them?

The example: you look at the hospital rota and find the same doctor in two departments at the same hour. You do not need to visit the hospital to know that there is a mistake in the sheet — the fault is plain in the sheet itself.

And its limit: coherence is a condition that does not suffice. A carefully invented account is perfectly coherent, and a truthful testimony may look confused until it is understood. So this is a criterion that rules out and does not establish — a sieve that stops the gravel, and tells you nothing about the salt that passes through it.

And its second limit: it is internal only. It examines the sheet against itself and does not carry it out into the world or to the conscience. And that is the way in to the fourth criterion.

The criterion: falsifiability

Does the claim say something whose contrary could turn out to be the case? Or is it framed so that every possibility agrees with it?

The example: two sellers. One of them says 'my goods are good', and the other says 'if it breaks down within a year I will refund the price'. The second has put something on the table that he can lose, so his words carry information. And the first has said what every seller says in every case.

And the idea became known through Karl Popper, and honesty in transmission requires its limit: he put it forward as a line between science and what is not science, and stated that an unfalsifiable claim may be meaningful and indeed true — only it is outside science in this sense. And extending it to metaphysical claims is a disputed matter. So a claim's being unfalsifiable does not make it false — it makes it untested by this instrument, and that is another thing.

And this criterion cuts in both directions, and that is the most dangerous thing about it. Whoever requires of his own claim that it be falsifiable is bound to accept the falsification if it comes.

And we tried it on ourselves in a project earlier than this book: widespread numerical claims attributed to a text we hold in reverence, in which the forms were counted mechanically, and the count went against three of them, and they were dropped from display. And the heaviest thing in that experience is a lesson prior to the count: a claim that does not specify what it counts cannot be tested at all — and 'yawm' is not 'yawman' and not 'al-yawm'. So its fault lies in its framing before the number goes against it.

And we do not make this bigger than it is. What was dropped is a set of appended claims on which we had built nothing and which nobody held to be a pillar of his religion, not the parent claim. And confessing to a sin you did not commit is no test. And the question that example ran away from is the real question:

So what, if it turned up, would fell the central claim with us?

We answer, and do not plead that the answer is heavy. And these are cases such that if one of them were established the claim would fall with us, or most of it would fall:

• That a material alteration in the transmission be established in the text — not differences of script and not readings, but the addition of a ruling or its removal in a later layer with a chain that is known.
• That an internal contradiction be established in it that admits of no reconciliation except by an interpretation invented to fend it off — the condition being that the interpretation be of a kind that would not be accepted in another text had it occurred there.
• That it lay down a report about the world which what we know for certain belies — not an understanding of ours read into it, but an explicit report that bears no other sense.
• That it command something we know for certain to be a wrong, by the measure of the fourth criterion to come.

And we grant two difficulties: that the condition in the second and the third opens the door to interpretation, and whoever wishes to interpret will not want for a way — and that is a fault in the criterion on our part, not a credit to us. And that many of the holders of the claim do not accept any case of falsification at all, and by that they are outside this whole criterion, and whoever is such should say so plainly and know that his claim is untested by this instrument.

The criterion: the moral content — and this is the one that was added

Is there in the content of the claim something you know for certain to be a wrong?

Not 'what your taste finds hard to bear', nor 'what goes against the custom of your age', but what, had you seen it from a man, you would not have hesitated to call a wrong, and your verdict on it does not change with the change of the agent.

And this last condition is a disputed premise and not a conceded one, and we say so and do not hide it: that the moral term run uniformly between creator and creature, so that 'a wrong' is taken in one sense in the two places. And it is disputed on the ground that God's transcendence makes our measuring him by ourselves unreliable — and that is Michael Rea's reply in another matter, which we set out in chapter 25. So we adopt the uniformity and do not claim that it is conceded: whoever does not accept it has not accepted this criterion at all, and that is a price we state in its place and do not fold away. And it will come in chapter 25 that the answer 'our scale is deficient in this matter' annuls the criterion at its root — so both faces of it are known to us and standing.

The example: the constitution of a society you are being invited into is put before you. You do not ask first about the soundness of its copy nor about the coherence of its articles; you read the article that says what is to be done with the dissenter. And if there is in it something your conscience will not accept, it is not saved by its copy's being an original and by its articles' not contradicting one another.

And why was it added? Because all four criteria are procedural or internal, and there was not one criterion among them that looks at what is said. And most of what reaches us is that doubt comes by way of a verse its holder found repugnant and not by way of the manuscripts — and that is noticed and not measured, and we have no survey of it to refer to. And a scale that does not weigh this matter leaves a door open to come in by.

And its limit — and it is heavy, so this criterion is not to be taken unexamined:

• Knowing a wrong for certain is a claim. People have differed enormously over what a wrong is, and what one century counted justice the next counted tyranny. So whoever uses it, let him use it in what scarcely two men differ over, not in what two ages differ over.
• And it is a criterion that works on our religion as it works on any other. And we say that while knowing that it is the hardest of the five upon us, and that bringing it in opens against ourselves a door we were well able to leave shut. And in chapter twenty-five we stopped at a place in this matter that we did not settle.
• And this criterion does not tell you 'so the claim is false'; it says: here is a site of weight that must be answered, and whoever passed over it without answering has passed over the heaviest thing on the road.

The criterion: its effect on the one who bears it — and we declare its weakness

Not in its institutions, nor in those affiliated to it by inheritance, but in the man who acted on it.

And I declare its weakness plainly for three reasons:

• Selection. We take the best of our followers and set them against the worst of other people's, and then call the difference a result. And the opponent does exactly the same.
• The sample selects itself. A school whose results are excellent — is it the method, or does it admit none but high achievers in the first place? You do not know until you know who came in, not who came out.
• Tangled factors. The good man is rich or poor, educated or unlettered, raised in a warm house or a cold one. And ascribing his effect to his belief alone is a verdict we do not have the instruments for.

And its scale has two pans: whoever argues from a good effect is bound to hear the man who argues from a bad one, by the same measure. And whoever refuses one of the two pans has broken the scale, not used it.

Practice: the scale at work on a claim that is not religious

Take a manifesto attributed to the founder of an old political movement, and people are being called to fall in under it. How is it weighed?

And see what you have come out with: you have not come out with a verdict of 'true or false'. You have come out with three of the questions having answers that can be asked for, and two that stop you at a site of weight. And this is the use of it: the scale turns 'I believe or I do not believe' into a list of questions each of which has an answer to be asked of the claimant. And whoever gives you a verdict outright has given you his opinion and not his scale.

And three remarks that keep the scale sound

1. Do not tailor a criterion to the measure of a claim in particular. If the criterion does not occur to you until after you know the answer required, then it is a conclusion dressed in the clothes of a premise.

And this remark stood in our draft as: 'do not invent a fifth criterion' — and a doubting reader said: the effect of it is that it forbids adding the one criterion I was going to weigh with. And he was right, and the old wording shuts the door on the number and not on the fault. So what is forbidden is tailoring to measure, not adding to the number. And we have added a fifth that works against us, and that is an indication and not evidence: a criterion tailored to measure is expected to serve the man who framed it, so when someone adds what harms him he has cleared one charge and has not cleared every charge — for the addition itself may be what buys assent. And the decisive thing is that the criterion be applied and not announced, and we have not applied it to our own text in this book, and we say so (see chapter 25, the sixth place).

2. Passing the criteria does not yield certainty. It yields that the claim is still standing after the examination. And that is less than enthusiasm supposes, and more than rejection supposes.

3. And even the result of the examination is conditional. Plantinga — among the staunchest defenders of the rationality of belief — states plainly that his book puts forward no argument for the truth of belief, and that its conclusion is conditional: *if* it were true then it is warranted. And the difference between this and 'a proof of God' is a difference that must not be lost in the transmission.