Is “Truth” a Redundant Word?

I prepared this piece for the site; it is not a course assignment. In the introduction I wrote out, one by one, where the texts in this section come from.

---

The witness rises. The judge has the formula read out: I swear on my honour and my conscience that I will tell the truth, only the truth.

Now let us run an experiment. Remove the word “truth” from that sentence.

What is left? “I swear that I will tell.” Nothing. The oath evaporates.

Yet one of the most serious claims in the philosophy of language is that this word adds nothing. An abbreviation. A convenience of language. There is no property in the world that corresponds to it.

How can both be true at the same time?

What Ramsey noticed

In 1927, at the age of twenty-four, Frank Ramsey said this: to say “p is true” is to say no more than p.

See how simple it is. Between “‘Snow is white’ is true” and “Snow is white” there is no difference in information. They are true in the same situations and false in the same situations. In passing from the one to the other you lose nothing and gain nothing.

If that is how things stand, looking for a property called truth is pointless. It is not something like a table's being brown; it is not a hidden quality of sentences. The word is a tool, not a description.

Ramsey died at twenty-six. After him, this idea was debated for eighty years, and the debate is still not over.

So why does the word exist

The deflationist does not say that the word is useless. It is useful but not descriptive, the deflationist says. And there are two places where it really is indispensable.

The first is generalization. You want to say, “Everything Newton said is true.” You cannot list all of Newton's sentences one by one; your lifetime would not be enough. The word “true” squeezes an infinite list into a single sentence. Like a sack.

The second is endorsement. You want to accept what someone said without repeating it: “What you said is true.” You sign off without rebuilding the sentence from scratch.

Horwich has been the name that defended this view most clearly: the whole of the concept of truth consists in that equivalence schema. There is nothing else to say.

I like this economy. Not producing unnecessary entities is a good discipline in philosophy.

The door Tarski opened and the door he closed

The technical ground of this debate was laid in the 1930s, in Warsaw.

Alfred Tarski asked: can we define the predicate “true” for a language, within that language itself, without generating contradiction?

The problem is the liar paradox. Say, “This sentence is false.” If it is true, it is false; if it is false, it is true. You are left with a revolving door.

Tarski's solution was a separation into levels: the truth predicate of a language is defined not in that language but in a metalanguage that speaks about it. In this way a sentence cannot speak of its own truth. The door is locked.

And his criterion became famous: a definition of truth must yield, for every sentence of the language, the equivalence “p” is true if and only if p.

Now pay attention, because one of philosophy's most beautiful moments takes place here. Tarski gave a method of definition; he did not give a metaphysical answer to what truth is. Two people stepped into that gap and drew opposite conclusions.

Hartry Field said: this definition is meaningful only if it is completed with a physical account of reference; Tarski has done half the job.

Quine, and later Horwich, said: if the equivalences already give everything, there is nothing left to explain; Tarski has done the whole job.

The same technical result. Two opposite lessons. If you want to understand why this debate does not end, the answer is here: the quarrel is not over the data but over what the data requires.

Let us really remove the word

The best way to test a claim is to apply it. Let us take five sentences and pull the word out.

“‘Snow is white’ is true.” Remove it: “Snow is white.” No loss. The cleanest ground for the deflationist.

“What you said is true.” To remove it you would have to repeat your interlocutor's sentence. No information is lost, but there is a practical cost. Here the word is a shortcut.

“Everything Newton said is true.” Here the removal operation gets stuck. You would have to build an infinite conjunction. The word is indispensable.

“Science is approaching the truth.” Here something starts to crackle. Try removing it: “Science is approaching…” What? The sentence announces a direction, and it seems there must be something at the end of the direction. The deflationist can translate this too (successive theories are yielding more and more true predictions), but in the translation something in the sentence goes dim.

“I swear to tell the truth.” And here the work comes to a complete halt. The translation would be: for every p, if p, I promise to say p. Try having that read out in a court.

My conclusion from this: in descriptive contexts the word really can be removed. In normative contexts, when it is removed, something jams.

Dummett's objection

The most serious objection to deflationism is that the concept of truth stands at the centre of the theory of meaning.

Michael Dummett's point is simple. To understand a language is to know under what conditions its sentences are true. If “true” is only an abbreviation, how does it play this central role? If you explain the meaning of a sentence as “it is true under such and such conditions,” and “true” adds nothing, has the explanation not undermined itself?

The deflationist has two ways out. Either the deflationist gives up explaining meaning through truth conditions and moves to a use-based theory, which means rewriting half of the programme. Or the deflationist says that the real work is done by subsidiary notions such as reference and satisfaction, and that the predicate “true” merely sums them up, which weakens the claim that “there is nothing to be said about truth.”

Both are coherent. Both have a price.

Three sentences, three different truths

Put these side by side:

Water boils at one hundred degrees. Seven is prime. Torture is wrong.

All three may be true. But are they true in the same way?

In the first there is a kind of touching of the world. In the second there is no physical thing to touch; mathematical truth is intertwined with provability. In the third, the debate itself concerns what truth means there.

From this Crispin Wright and Michael Lynch draw a proposal: truth is not a single property but a role. Different properties from domain to domain take on this role. What is common is what truth does: being the aim of inquiry, being the standard of assertion.

I find it attractive. I also have an objection: if truth is something different in every domain, what does the thing we can say “every true sentence shares” come down to? If the answer is “the role,” it is still unclear to me how far this is from deflationism. In the end both say that truth is not a thick substance.

Davidson's surrender, or his diagnosis

In 1996 Donald Davidson said something to those who were trying to wrap this business up: trying to define truth is folly.

His reasoning is this: truth is one of the most basic concepts, and basic concepts cannot be explained by more basic ones. Every word you would use to explain it already presupposes truth.

I do not read this as giving up. I read it as a diagnosis. Some concepts are the ground we keep standing on; you cannot lift the ground and look beneath it, because while you look you are standing on it.

Back to the courtroom

Let me say where I stand.

Deflationism describes the everyday workings of the word largely correctly. In most sentences, when you remove it, really nothing is lost, and that is not a datum to be easily ignored.

But it does not explain what is going on at the witness stand.

What is done there is not a description. It is taking up a stance. The person who says “I will tell the truth” is not giving information about the world; they are placing themselves under an obligation. They are accepting in advance that they will correct themselves if they are wrong, give grounds if asked, and back down if evidence comes in.

The word's semantic contribution may be zero. Its pragmatic contribution is not zero.

And perhaps this is why the whole debate has lasted eight hundred years: philosophers looked for what the word describes, whereas the word was not describing. It was promising.

It is the same in everyday life. When you say “this is true” in an argument, you are not transmitting information. You are taking on responsibility. That is why it is often possible to remove the word from a sentence; but not to remove what you have taken on.