Friday, March 18, 2011

The truth of 'snow is white'

To: Ryan S.
From: Geoffrey Klempner
Subject: The truth of 'snow is white'
Date: 25 April 2001 09:53

Dear Ryan,

Thank you for yesterday's e-mail and your second essay for the Metaphysics program (16 April) which I'll come to in a minute.

Since writing the program, I have come to modify my views about privacy and subjectivity. Have you read my paper for the Shap Conference, 'Truth and Subjective Knowledge'? You will find it on the Glass House Philosopher site, at:

A 'private object' is not the same as what I term 'subjective knowledge'.

All feeling and experience involves subjective knowledge, things being given to us in a certain way. But this given is not 'Cartesian mental events', private objects which (via the 'cogito') I know to exist irrespective of how things are in the physical world. It is a brain state.

In the Shap paper, I put forward the conjecture that 'a brain state is accessible only to the organism whose brain it is'. A brain state is accessible when it has some relevant input on the organism's behaviour. Subjective knowledge may therefore be seen as a form of irreducibly practical knowledge. For example, the acrobat's sense of balance, qua, subjective knowledge is what enables him to walk a tightrope. If the acrobat loses his balance and falls, that is not because he made a false judgement. You could call it a 'failed action'. (It is possible, of course, to fall as the result of making a false judgement. E.g. 'The wind is not too strong now'.)

I still do not know what to say about the stoic. When the stoic says, calmly, 'I am suffering terrible hysterical pain' and we can discover nothing wrong, then we still do not know how to assess the truth value of that judgement. It is a problem worth looking into. I don't, however, think that it poses any threat to the reality principle.

Now to your essay on '"Snow is white" is true if and only if snow is white.'

This sentence is a particular instance of the general schema:

'P" is T if and only if P.

According to Alfred Tarski, who first formulated this 'Convention T' as a condition on the adequacy of any proposed truth definition, we are dealing with two languages, an 'object language' and a 'meta-language'. ('On the concept of Truth in Formalized Languages'). In fact, we can see that there is a whole hierarchy of languages, because we can say things like,

'"Snow is white" is true if and only if snow is white' is true if and only if "Snow is white" is true if and only if snow is white,

and so on.

The original "snow is white" is a sentence from the object language. '"Snow is white" is true if and only if snow is white' is a sentence from the meta-language. The example I have just given is a sentence from the meta-meta-language. The previous sentence is a sentence from the meta-meta-meta-language.

What is the point of Tarski's Convention T? He is not setting out to define truth. He is merely saying, 'Any definition of truth must meet this requirement'. This has profound consequences. For it turns out (I would argue) that no informative definition of truth can possibly meet this requirement. Truth is indefinable.

Now you make the point that this account simply does not make any sense if we think of a sentence as just marks on paper. The response is to say that a 'sentence' is always taken as part of a language.

When we talk about a fragment of an object language in our meta-language, such as "snow is white", that fragment is identified as being the fragment it is, having the meaning that it has because it is taken in the context of that object language. So it matters whether the object language is English or Blinglish.

In our meta-language, 'snow' refers to snow and 'white' refers to white. while "snow is white" refers to a certain sentence in the object language. (Notice that in order to state this, I have had to use a meta-meta-language.) The result is that we are able to make a statement about the world and at the same time about a certain sentence from a language which can itself be used to make statements about the world.

So you are wrong to say, "If 1 contains a concept, then A is true, but it's a very boring 'a=a' truth." The slippery term here is 'contains a concept'. The words in the object language 'contain concepts' in the sense that they are words from a language, not meaningless marks. But in the meta-language in which we express the equivalence, they are mentioned, and not used. That is why the result is not an a=a truth but a statement both about words and about the world.

All that I have said does, I agree, beg one very big question, which is how there can be such a thing as 'a language' in the first place. What are the kinds of facts in virtue of which mere marks on paper or sounds acquire 'meaning', acquire the capacity to make statements which are true or false? That is not a question which concerned Tarski. It is a question which does concern Donald Davidson in his famous paper 'Truth and Meaning' where he proposes to use Tarski's Convention T as the basis for a theory of meaning for natural language. The paper is in his collection of papers on Truth and Meaning. Strongly recommended.

All the best,