To: Graham C.
From: Geoffrey Klempner
Subject: Donnellan on referential and attributive descriptions
Date: 25th April 2008 11:29

Dear Graham,

Thank you for your email of 10 April, responding to my comments on your essay on truth, your email of 17 April with your University of London Logic essay in response to the question, 'Can the sentence ‘The man over there drinking Martini is a philosopher’ be true even though nothing satisfies the definite description? Justify your answer.' and your follow-up email of 19 April.

My take on the 1949 Analysis article is that Strawson was using the vernacular of the day. The crucial point is, 'What commends the word as, e.g., a confirmatory device is its economy. By its means we can confirm without repeating.'

Worry about ideology is somewhat simplistic. Why should philosophers agree? It is better that we are working with multiple paradigms than if we all agree on the same paradigm. You try one thing then another, until something works. (And I'm not proposing 'pragmatism' as a paradigm either.) Philosophy is not a science.

Regarding the parochial 'the'. Russell and Frege shared the view that natural language is not a reliable guide, it is messy and somewhat chaotic, having evolved over time to meet practical needs. This was a bone of contention between Russell and the early Wittgenstein, whose view was that every statement of natural language has a precise analysis, and indeed this is how it gets its meaning (so when I say, 'the cup is on the table' I am in fact asserting a massive disjunction of atomic propositions, giving every possible alternative location for 'the' cup).

When Russell wrote in the introduction to the Tractatus that Wittgenstein was interested in describing a logically ideal language, Wittgenstein strongly objected. He was describing the logic of our actual language, which lies hidden beneath the deceptive surface, the actual machinery that makes it work.

This is relevant to understanding current debates over definite descriptions. Russell knew that descriptions are needed for logic and maths, where for example one makes a statement about 'the' class which satisfies such and such a condition. You never know when you might encounter a paradox (!) which proves that the class you were referring does not exist. To preserve logic (i.e. truth conditions) Russell's analysis is indispensable.

A relatively contemporary proponent of this approach is W.V.O. Quine who talks of 'regimentation' of natural language for this or that purpose, rather than analysis. Many would regard Quine's 'Two Dogmas of Empiricism' as having kicked into touch any idea that there are definitive answers to questions about 'the' meaning of this or that expression in natural language.

This is very much a 'potted account'. There is much, much more to say. But let's now look at your essay:

I had to laugh about your example of 'The cat is on the mat' meaning 'I want a divorce'. But, seriously, we are in very different territory if the question is one of the significance of an utterance in this sense. Consider a psychoanalyst overhearing a conversation where there is a great deal to read between the lines.

We want and need an account of the content of an utterance, based on the actual words used together with any indexical elements ('this', 'that', 'now', 'I', 'she' etc.). That much is agreed. As soon as you bring in a 'pragmatic' dimension then you have taken a step towards the cat on the mat case. We want this too, but first we need a clear account of the semantics.

Pragmatics is not an adequate substitute for an inadequate semantic theory.

Here's an even better example of referential use of a description. I happen to know that the man in the corner is drinking a Oxyiac, a rare non-alcoholic herbal essence made my Monks in Tuscany (I made this up). However, as it is extremely unlikely that you have even heard of this beverage, I deliberately use the false description 'Martini' because all I'm doing is trying to secure a reference, and the drink has the same pale yellow colour. So what if you go home musing over the fact that philosophers like to drink Martini? Did I lie? he is a philosopher, isn't he?

Doesn't that strike you as extremely weird? have we so little regard for the truth?

Without going into the business of rival semantic accounts (and if you read, you will discover there are further alternatives besides simple Russell and simple Donnellan) it is not difficult to produce examples which push our intuitions the other way. Of course, someone who has a full grasp of the context would know whether or not the information the speaker intended to convey is correct or otherwise. It may indeed be the drink not the profession of the drinker which is the focus of interest, but you couldn't tell just from hearing one sentence.

What a person says -- the content of their utterance -- and the information they intend to convey are not always the same, as your cat on the mat example shows.

However, Donnellan might reply that we are missing the point. All he was concerned to do was show that natural language has a strong contextual/ demonstrative element. Whereas the use of words can be described by rules, human gestures escape the net of formal theory. One conclusion to draw from this is that there can never be a fully adequate 'semantics'. Even so, I would argue, this is no excuse for ditching Russell.

All the best,

Geoffrey