``Truth,'' says Davidson, ``is as clear and basic a concept
we have.''
And he goes on to ask: ``Why on earth should we
expect to be able to reduce truth to something clearer or more
fundamental?''
Davidson's remarks are based on the view that we can give the content of the truth-predicate in a simple way. If we take sentences to be truth-value bearers we can express the content of the truth-predicate in the following way:
In contrast to the notion of truth the notions of meaning and
understanding seem to many philosophers unclear. For instance
Quine holds that the theory of meaning is in a ``sorry state''. According to him the notion of meaning is
unclear and resist conceptual elucidation. Since Quine put forth his
sceptical attitude towards the concept of meaning attempts have been
made to define it in terms of possible worlds. But Quine and many
philosophers with him would not regard the concept of a possible world
as less problematic than the concept of meaning.
Given this background elucidations of the notions of meaning and understanding that use the notion of truth are very attractive. Problematic notions would be illuminated by or even reduced to a clear and basic notion.
Donald Davidson tries to give such an elucidation. Davidson has become famous for his project to develop a theory of meaning for a natural language which uses the notion of truth as it's conceptual basis. Davidson's work is inspired by Tarski's attempt to give a recursive definition of the notion of truth by appealing to the concept of translation. Davidson reverses the order of explanation:
[...] while Tarski intended to analyse the concept of truth by appealing (in Convention T) to the concept of meaning (in the guise of sameness of meaning, or translation), I have the reverse in mind. I considered truth to be the central primitive concept, and hoped, by detailing truth's structure, to get at meaning.''An account of the semantic competence of a speaker S of a language L would be an answer to the question: ``What must S know in order to understand utterances of sentences in L?'' Davidson thinks such an account will crucially rely on a truth-theory for L. But he doesn't exactly try to answer the just formulated question. He merely seeks sufficient conditions for the possession of semantic competence. A rough formulation of Davidson's main idea is: knowledge of a truth-theory for a language L suffices for understanding utterances of L sentences.
When we ascribe to a speaker such
knowledge, we credit him with a systematic method for interpreting
utterances of a language. This systematic method or theory can also
be described as a theory of meaning for the language under
consideration.
Before elaborating Davidson's sketch of his semantic program, I like
to make a remark about another methodological point. Normally one would expect
that a philosophical investigation of the concept of meaning takes the
form of an analysis of this concept, that is the epistemologist tries
to give analytic necessary and sufficient conditions for the
application for the expresion ``x knows that p''.
Michael Dummett has pointed out that nobody would expect an
epistemologist to engage in the development of a theory
which entails specifications of every state of an individual which
counts as knowledge.
Why are things different in investigations about meaning? Dummett's
answer is:
is precisely because, in this area of philosophy, we know even less what it is that we are talking about than we do in other areas, that the proposal to approach our problems by considering how we might attempt to specify the meanings of the expressions of an entire language does not appear the waste of time that an analogous proposal would seem to be within epistemology.''If one shares Quine's sceptical attitude towards the concept of meaning the truth- theoretic approach seems promising not only because the notion of meaning is characterized in terms of the basic and unproblematic notion of truth, but also because theoretic consideration force us to make the intuitive concept more precise. The demands of the theoretical treatment will lead to new distinctions which will result in a more secure grip on the notion of meaning.
The truth-theory Davidson envisages will have axioms for singular terms like the following one:
to Afla
by x iff x is a mountain
iff what ``a'' refers to satisfies ``F''
A truth-theory of this type will entail for every
sentence of L a theorem of the form ``x is true
if and
only if p''. We got from this schema a so-called ``T-sentence'', if
we substitute for ``x'' a singular term for a sentence of L and
for ``p'' a sentence of the meta-language. The substituend for ``x'' will be
canonical singular terms for the sentences of the object-language.
Tarski originally choose structural descriptions of sentences,
Davidson uses quotation-names. I will follow Davidson's practice.
If a truth-theory is a theory of meaning, as Davidson claims, the right-hand side of a T-sentence should specify the meaning of the sentence of the object-language the T-sentence is about. Now due to the logic of the bi-conditional, a T-sentence is true if and only if the denoted sentence of the object-language and the used sentence of the meta-language have the same truth-value. This reveals a central problem of Davidson's approach to characterize meaning in terms of truth conditions. There is a gap betweeen a theory of truth and a theory of meaning: sameness of truth-value does not guarantee sameness of meaning. Two well known examples illustrate that.
Imagine a truth-theory whose object-language is English and whose meta-language is German. A correct T-sentence of such a theory could be:
genau dann, wenn Blut rot ist.
genau dann, wenn Schnee
weiß ist und 2 + 2 = 4.
Davidson is well aware of the extensionality problem.
He argues convincingly that a truth-theory which entailed (T1) could
not give the right truth-conditions for all English sentences. For
instance such a theory
would fail to give correct T-sentences for the English sentences
``This is red'' and ``This is white''.
The counter-example (T2) seems to be amenable to a similar treatment.
Prima facie it takes one step more to derive a T-sentence like (T2)
from a truth-theory than the meaning-giving T-sentence (T2')
`` `Snow is white' ist wahr
genau dann, wenn Schnee weiß ist.''
These considerations suggest that only a truth-theory which entails a
true T-sentence for every sentence of the language in combination
with a canonical proof-method for its theorems will yield
meaning-giving T-sentences. A canonical proof of a T-sentence will be
the shortest proof of such a sentence. The T-sentence of
a truth-theory fulfilling the above mentioned requirement whose truth can
be ascertained by a canonical proof are the meaning-giving ones.
This is one aspect of Davidson's position in his essay ``Radical
Interpretation''.
But considered as a solution for the extensionality problem this is not sufficient. If we had a truth-theory T1 for English which would entail true T-sentences for all English sentences, we could generate a truth-theory which inherited this property by replacing the clauses in T1 that determine the reference and application of the non-logical vocabulary of L with extensional equivalent clauses. For instance we could replace the axiom for the predicate ``is white'' with the extensional equivalent axiom:
genau dann, wenn
x weiß ist
Davidson has so far imposed no restrictions on
meaning-theories for natural languages which rule such
truth-theories out.
How can this be done in a way that is consistent with Davidson's main idea? Davidson can't appeal to the intuitive notion of meaning because this notion should be characterized by the truth-theory. So which theoretic constraints can Davidson use to exclude the truth-theories which are from a meaning-theoretic perspective unfruitful? This brings us to another aspect of Davidson's position.
To solve this problem Davidson appeals to what he calls ``holistic'' or
``empirical constraints''. Davidson holds that
a truth-theory for a language L is an empirical theory:
its T-sentences are confirmed by empirical generalizations like
x)(
t) (if x belongs to the English speech
community then (x holds true ``It is snowing'' at t if and only if it
is snowing near x at t)
But is it really unproblematic to say
that an interpreter can identify that a speaker holds true a certain
sentence? If the interpreter does not yet understand any of the
speaker's utterances (does not know what truth they express) he can
hardly discriminate between holding a sentence true and other
attitudes directed towards a sentence like believing the sentence to be
probable, believing it plausible and so on.
After pointing out this problem I will expand further on Davidson's approach.
The second aspect of Davidson's position can now be stated as follows:
Only a truth-theory which ``optimally fits evidence about sentences
held true by native speakers''
can serve as a theory of meaning in the intended sense. The theory
should assign truth-conditions to the sentences of the language in a
way that makes the native speakers statements come out true as often
as possible. Together with the first aspect we get the following
statement of Davidson's position:
We should note that Davidson started out to give a theory of meaning for a natural language. What he arrived at is an empirical theory of the speech-dispositions of speakers of such a language. Thus the question arises whether Davidson has really characterized what he wanted to characterize, namely the meaning of natural language expressions.
But even if we accept Davidson's empirical constraint on truth-theories, there is a further problem. John Foster has pointed out that someone may know a truth-theory for the language L that met all empirical and formal constraints but yet couldn't understand utterances of L sentences. He could simply fail to recognize that the theory he knows met the relevant constraints, i.e. that it's theorems interpret sentences of L. Knowing a truth-theory that meets all the constraints is therefore not sufficient for being a competent speaker of L. Davidson has accepted this point. In his ``Reply to Foster'' Davidson illustrates Foster's argument with an analogy:
get a precise parallel if we ask what someone must know to be a physicist. A quick answer might be: the laws of physics. But Foster would say, and I agree, that this is not enough. The physicist must also know (and here I speak for myself) that those laws are laws---i.e. that they are confirmed by their instances, and support counterfactual and subjunctive claims.''
With this in mind, we can now formulate a sufficient condition for the possession of the ability to understand utterances of sentences in L:
We have finally arrived at Davidson's substantial semantical thesis. (DT) will serve as a folio for our further discussion.
As I already pointed out in the introduction the semantical problems which indexicals raise for Frege will be in the centre of interest in this paper. Therefore it is reasonable to discuss the way Davidson handels indexicals in his theory of meaning. I will do this in the next section. The considerations of this section will prepare the discussion of section 5 and 6.
Home Page
CSLI Home Page
Stanford Home Page