Monday, August 20, 2012

Counterexample to Russell's theory of descriptions?

To: Graham C.
From: Geoffrey Klempner
Subject: Counterexample to Russell's theory of descriptions?
Date: 3rd March 2008 12:50

Dear Graham,

Thank you for your email of 15 February, with your essay in response to the University of London Logic question, 'One can understand the claim, "the baby has been sick all day", without supposing there is one and only one baby in the world. So Russell's theory of definite descriptions is wrong.' Discuss.

I've taken you at your word and allowed this to remain in my in-tray for a little longer than it would normally do. I hope you enjoyed your holiday!

You give a clear explanation of Russell's theory, and you propose an initially possible refinement which would deal with the non-unique baby objection, namely to 'demote uniqueness from universal uniqueness to contextually relevant uniqueness'. You also explain the difference between refining Russell's theory and rejecting it, as Strawson and Donnellan do.

I'm not sure, however, about your second alternative for analysing 'The present King of France is bald'. The second, negative existential conjunct effectively states that for all y it is not the case that (if Fy then y is not equal to x), i.e. for all y it *is* the case that Fy and y=x, from the equivalence of P->Q and not-(P and not-Q). This states that there is one and only one object in existence and that object is the present King of France!

As I have rather a soft spot for Russell's theory of descriptions, I would at least like to have a go at defending the theory against the objections which you raise.

First, the question of contextual relevance. You say, 'Once the idea of contextual relevance is introduced, we no longer seem to be talking in terms of "denoting" but are back to referring (naming, pointing), which is precisely what Russell wants to avoid.' This isn't very convincing as it stands. Why does Russell want to avoid, 'naming, pointing'? Russell notoriously restricted demonstrative reference to objects whose existence has an epistemological guarantee, i.e. sense data, and he can well be criticized on this account (as Evans does in 'Varieties of Reference'). However, if we bracket that question and consider Russell's theory in combination with a more sensible account of indexical reference, then the question we have to consider is how statements including a definite article can have truth conditions.

The big objection to the alternative, i.e. Strawson, is that his initially more intuitive account of 'referring' is paid for by having to allow truth-value gaps, in cases where the presupposition is not satisfied. In the context of constructing a theory of meaning for a natural language, this is problematic.

If one could give a plausible account which did not have the consequence of truth value gaps, that would be preferable. Does contextualising definite descriptions do the trick?

Your second objection is that, 'The baby has been sick all day' would be analysed as follows:

1. There is at least one contextually relevant baby.
2. There is at most one contextually relevant baby.
3. Whatever is the contextually relevant baby has been sick all day.

The problem you raise is that the third clause uses 'the' and thus fails as an analysis as it contains the term that it was intended to analyse.

However, it is clear from 1. and 2. that 'the' in 3. is redundant. All you need is:

3. Whatever is a contextually relevant baby has been sick all day.

So I don't get the force of this objection.

Surely, there is a problem with contextual definite descriptions, as a way of resisting Strawson, in that this seems to give rise to the possibility of a failure of contextual relevance, analogous to Strawson's failure of presuppositions. If I write to you, 'I'm glad to tell you that the art therapy session went well' (e.g. as a result of mixing up email addresses) have I stated a truth, or is there a failure of reference owing to lack of relevant context? If the latter, then we have re-introduced truth value gaps.

This raises the question of the point of view from which we are analysing statements, i.e. the speaker/ writer or hearer/ reader. The idea of a theory of meaning in terms of truth conditions is based on the assumption that we can identify 'the' truth-conditional content of a statement, and yet clearly, the author and the recipient are not always in a position to agree on what this content is.

An alternative to contextualization which you might have considered is the much simpler explanation of ellipsis. When I say, 'the baby has been sick all day', I mean, 'the baby which (blah blah blah) has been sick all day.' The context gives the clue to what this is elliptical for, but reference to context is not part of the analysis of the statement concerning the definite description.

I think that this is the alternative which Russell would prefer. We often truncate the things we say because we know that the missing element will be supplied by the recipient. What we 'mean', however -- the content identified by speaker and hearer -- is given by the non-elliptical, non-truncated version. The hypothesis of ellipsis is not like claims about 'underlying logical structure', which raise difficult questions about whether the analysis is an analysis of actual language or a proposal for an 'improvement' on natural language (as, e.g. in Quine's account of 'regimentation'). We are all familiar with the practice.

The problem remains, however, that Russell's theory itself is precisely an example of such a problematic 'analysis'. Do we need it? Yes, some would say, if we are serious about the project of constructing a theory of meaning for natural language in terms of truth conditions; no, if one is sceptical about the prospects, or point, of a systematic theory of meaning in terms of truth conditions. (I'm not saying that these are necessarily the only alternatives.)

All the best,