To: Michael Y.
From: Geoffrey Klempner
Subject: Michael Dummett on anti-realist theories of meaning
Date: 28 October 2008 13:52

Dear Michael,

Thank you for your email of 21 October, with your third essay for the Metaphysics program, in response to the question, 'What do you understand by Michael Dummett's idea of an 'anti-realist theory of meaning'? Give an example illustrating how such a theory of meaning would be applied to one particular class of statements,' your email of 22 October, with your notes on unit 10 and your email of 23 October with your notes on unit 11.

Dummett does in fact consider a multiple world version of anti-realism in 'reality of the past', distinguishing it from genuine anti-realism. However, he doesn't consider the possibility of a 'global' multi-worlds theory. In general, Dummett would be hostile to the approach to the realism/ anti-realism debate taken here, although given his great regard for Wittgenstein, one would have thought that he ought to be prepared to consider a 'dialectical' approach.

According to my anti-realist, what we term the 'actual world' is not a world, but an indeterminate *subset* of all possible worlds, namely those worlds which are consistent with possible knowledge. This preserves a distinction between 'actual world' (as a subset) and 'non-actual world' as the remainder when the subset is removed.

Truth as evidence and verification

You have taken on a very difficult topic, and also a difficult writer. In addition to that, there are issues which Dummett assumes his reader to be familiar with, and therefore doesn't attempt to explain.

In his article 'Truth' Dummett is doing several things. First, he considers the problem of empty reference and also the problem of conditionals with false antecedents (the paradoxes of material implication), in relation to the idea of bivalence. His intention here (which I think you have misunderstood) is to defend a Russellian analysis. Russell's theory of descriptions was famously attacked by Strawson in 'On Referring'. In that article, Strawson argues that in cases of reference failure, a condition which was 'presupposed' by the making of the assertion fails, and therefore we can't speak (as Russell wants to do) of the assertion being false.

Dummett's reply, in effect, is, 'If presupposition failure isn't a kind of falsity, what is it?' In other words, the principle of bivalence is too important and central to our understanding of truth, to give up merely on this basis.

A similar argument (perhaps more controversially) applies in the case of material conditionals. In the conditional P->Q, according to the truth table for '->', the statement is false when P is true and Q is false, and true otherwise. Giving up bivalence is too high a price for reconciling the paradoxes that result from this.

In other words, Dummett's strategy is to defend the principle of bivalence against unjustified attacks, before attacking the principle by raising far deeper worries about the role of the concept of truth in relation to an account of the nature of a speaker's understanding of his/her language.

It is this argument that is crucial. In 'Reality of the Past' Dummett attacks the attempt to use the systematic linkage between the truth values of statements uttered at different times to defend realism about the past. For example, suppose you ask me what it means to say that it is true that an oak tree stood here one million years ago. I reply, 'It means that if the statement, an oak tree stand here' had been asserted one million years ago, it would have been true.' But this isn't how we *understand* statements about the past. It doesn't explain (according to Dummett) how it is that we grasp the very idea of a 'truth' which is not here and now.

We have a picture in our heads, say, of an oak tree standing there, but this picture doesn't suffice to account for the actual knowledge of how to *use* the statement in question. To quote Wittgenstein, it is a 'wheel that turns, although nothing turns with it.' Looking at the actual way in which we learn language, and manifest our understanding of language, the crucial thing is the *rules for use* of the terms in question, which cannot be derived from truth conditions.

I don't accept that Dummett is committed to making the statements you make about the example of the '150 people shot by the Chinese secret police in January 2008.' Any judgement about truth is a judgement made now. If we say what 'will be' true, or what 'was' true, we are still talking from the present point in time. You can speculate about what people will say in the future, in this or that circumstance, but this isn't the same as speculating about the truth value that the statement 'will have'. What I am saying here is just a corollary of the principle of truth value linkage which Dummett accepts. What he is denying is the realist's attempt to use the truth value linkage to defend realism.

All the best,

Geoffrey