donald davidsonOverview

A greatly simplified introduction to one influential theory of meaning: its approach, successes and difficulties. Provided language is seen as non-metaphorical (i.e. there is an essential skeleton of meaning regardless of how expressed), and the logical formulation is not expected to be entirely comprehensive and watertight, then Davidson's theory refutes the wilder speculations of Postmodernism.

There are certainly shortcomings (exceptions, qualifications, alternative formulations are the bread and butter of philosophy) but there are no grounds for asserting that language is an endless web of self-referencing signifiers.


Davidson's theory of meaning begins with Alfred Tarski's approach to logical paradoxes like All Cretans are liars. Tarski's solution was to consider the primary sentence as written in an object language, and to propose another, higher level, metalanguage that could handle object languages without being tangled up in paradoxes of self reference.{1} Superficially, the two may seem the same — both are formal and not natural languages — but only the metalanguage could incorporate and refer to the object language.

Tarski's Concept of Truth

Consider an example: It is true that "snow is white" iff snow is white — where iff stands for if and only if . There is nothing objectionable or difficult here, but what's the point? Even if the opposing sides are in two languages, object and metalanguage, the statement — called a T sentence — seems practically tautological. It doesn't tell us how the truth of the proposition snow is white was arrived at, the least we might expect. Agreed: but let's push on, and turn this apparent shortcoming into an asset, making the correspondence between two languages the point of interest.

Take a simple proposition.{2} It consists of a name (N — e.g. Lenin, which refers to something in the outside world) and a predicate (P — e.g. is bald, which describes the name in some way). Let us suppose that this proposition can be represented as NP in language L. Now take another proposition, completely different, in another language altogether. Represent this as np in language l. Both languages are formalized in Tarski's terminology, though we are not distinguishing here between object and meta-language. Our concern is with the translation process when we run the two languages together. Let us list the components of the two propositions, and how they appear when rearranged between the languages:

N in L refers to Lenin: N in l refers to Paris

n in L refers to Marx: n in l refers to Rome

P in L refers to bald things: P in l refers to French things

p in L refers to pink things: p in l refers to warm things.


Using Tarski's procedures we can say: PN is true in L iff Lenin is bald. PN is true in l if Paris is French. pn is true in L iff Marx is pink. pn is true in l iff Rome is warm. And that is all we can say. That exhausts the possibilities.

What do these T sentences say? They are partial definitions of the languages L and l. They spell out what we first asserted, namely that: Any sentence PN in a formalized language will be true if, and only if, the predicate applies to or is satisfied by whatever it is that the name refers to. Very well, but what now?

The next step is twofold. Firstly, and crucially, we shall regard the full definition as the total of all these partial definitions. Secondly, we shall consider six partial concepts: reference-in-L, satisfaction-in-L, truth-in-L, reference-in-l, satisfaction-in-l, truth-in-l. What does Tarski mean by these partial concepts? Here is the answer in our language L:


X refers-in-L to Y iff: X is N and Y is Lenin, or X is n and Y is Marx.


Y satisfies-in-L iff: X is P and Y is bald, or X is p and Y is pink.


S is true-in-L iff: S is PN and Lenin is bald; S is pN and Lenin is pink; S is Pn and Marx is bald; S is pn and Marx is pink.

A similar list appears for reference-in-l, satisfaction-in-l, and truth-in-l. So? We have reached the end of our quest. This is how Tarski defines reference, satisfaction and truth — by the totality of these partial definitions. That is all. Of course our example is simple, even trivial. More useful sentences would generate very long lists — impossibly long of course, and Tarski devised recursive procedures to eliminate that need. He starts with sentential functions, which resemble sentences, but have gaps or free variables in which suitable terms and expressions have to be inserted. While the gaps or free variables are unfilled there is no sentence as such, and no certainty that the expression is true or false. But once the gaps are filled, a sentence is formed, and is either true or false.

Turned around, this requirement becomes a definition: a sentence is a sentential function containing no free variables. Consider the sentential function: x was the teacher of y. This is satisfied by Socrates was a teacher of Plato, i.e. by {{Socrates, Plato}}. The order in the multiple brackets is important; Socrates substitutes for x and Plato for y. It would also be satisfied if any number of other objects following Plato in the multiple brackets — but only the first two are needed to correctly substitute for the variables and make a true sentence. So truth will be defined as being satisfied by all such {{Socrates, Plato, P, Q, R...}} sequences, and its falsity defined as being satisfied by none. That, in essence, is the procedure, though it clearly becomes more difficult in sentences that are not straightforward assertions. {3}

But Tarski's stance should be clear. These are his definitions of reference, satisfaction and truth. Are they enough? Tarski thought so. If he hadn't strengthened the correspondence theory of truth, he had at least laid it out more plainly. Logicians, Logical Positivists and grammarians tend to agree, but most philosophers see the procedures as an evasion of the real problems. Tarski's truth is grounded in languages: it ends in or is lost in the logical procedures by which sentences are put together. Fine, but such an approach tells us nothing about the truth of the individual propositions themselves, the judgements we make in asserting such things to be the case. Nor anything about how language is used in real societies: how, to what ends, with what assumptions.

Davidson's Concept of Meaning

Can anything further be done? The American philosopher Donald Davidson made an enterprising attempt. His goal is meaning, a clear, unambiguous concept of meaning, and this he defined (audaciously) as the truth conditions of a sentence. Meaning becomes what needs to be true of its constituent parts if the sentence as a whole is to be true. Quite apart from such a novel redefinition {4}, Davidson has two difficulties to overcome. One is that Tarski's approach applies only to formalized languages, not to imprecise, ambiguous and elliptical natural languages. The second is that Tarski assumed identical meaning in making the translation from object to metalanguage, i.e. assumed the very thing that Davidson wishes to establish.

Davidson adopts Tarski's method, but relies on two supports: the top down approach and use of the radical interpreter.

By top down, Davidson is arguing for an approach that starts with the language as a whole and moves progressively into smaller components. "We can give the meaning of any sentence (or word) only by giving the meaning of every sentence (and word) in that language," says Davidson: a holistic view of language. A sentence has meaning only because of what its constituent words mean, and words only have a meaning by virtue of the contributions they make to the sentences in which they occur. According to Davidson we cannot give the meaning of one word without giving the meaning of all.

In the radical interpreter Davidson is looking for the means of translation between mutually incomprehensible languages. Quine's view was that, ultimately, we couldn't be sure of success in translation. Simply pointing and uttering the word was not sufficient: we needed other words to be sure that "sheep" indicated an animal and not wool-provider or grass-trimmer or mutton or part of a sheep. These other words would not be available prior to translation. Davidson finds something of a way round this, but has to accept a less demanding (charitable) view of the radical interpreter: that the native speaker is rational, not aiming to deceive us, and has a set of beliefs largely consistent with our own.

Given these two assumptions, however — top-down approach and radical interpreter — Davidson's approach is this: Suppose we have two languages, one natural and one formalized. We say in our natural language, to a logician speaking the foreign formalized language: snow is white. That is true in our language. He replies in his language: sun glare causes snow-blindness. That is true in his language. Since both sentences are true they could be assembled in a T sentence:

Snow is white-in-natural language is true iff sun glare causes blindness-in-formalized language.

Our interpreter is charitable. Both logician and natural language speaker are standing in a snow-draped mountain landscape, so that the two assertions presumably have something to do with each other. Without further conversation, we might suppose that sun-glare is the translation of snow, just as the predicate causes blindness is a translation of is white. That is what Tarski's partial definition listed above would suggest. S is true-in-L iff... But further conversation would soon disabuse us. Using snow in some other context would not return sun-glare but something very different. Eventually, a long time later, given sufficient exchanges involving words relevant to the context, and a well-intentioned interpreter, we should arrive at: Snow is white-in-natural language iff Snow is white-in-formalized language. No other result would avoid ludicrous mismatches somewhere along the line. And having made the translations of snow and white, we should go on with other words relevant to the situation — fresh falls, clean air, clear sunlight, etc. Our activities would gradually widen until we had made all the links between the two languages. At very long last our translation would be complete, and would indeed be able to express a natural language in a transparent, logical formalized language.


Is this achievable? Davidson has made great strides but the enterprise has hit snags with indexicals (pronouns and related expressions of time and place) and other complications. The programme has spread, ramified, and regrouped as new objectives, but none of these have been fully achieved. Davidson and his followers remain hopeful, but onlookers are less convinced.{5}

But even if success were to come, is this concept of meaning — the truth conditions in a formalized language — how we generally use the term? And what of the difficulties noted before with Tarski's definition? Davidson's approach counters the Poststructuralist view that language is an endless self-referencing web of signifiers, but does not correspond to how language is always used, either in literature or the everyday world.

This and other pages in the theory section have been collected into a free pdf ebook entitled 'A Background to Literary Theory'. Click here for the download page.


1. pp. 157-160 of A.C. Grayling's An Introduction to Philosophical Logic (1982) and pp. 99-119 in Susan Haack's Philosophy of Logics (1978).
2. Chapter 8 in Simon Blackburn's Spreading the Word: Groundings in Philosophy of Language (1984).
3. Chapter 8 and section 9.1. of Blackburn 1984. Also pp. 160-63 of Grayling 1982.
4. Chapter 12 in Ian Hacking's (Ed.) Why Does Language Matter to Philosophy? (1975).
5. Chapter 4 in John Passmore's Recent Philosophers (1988). Also Meaning entry in Tony Honderich's The Oxford Companion to Philosophy (1995).

Internet Resources

Donaldson's work is more wide-ranging and difficult than suggested by the introduction above (which takes some liberties with his thought). The following resources may help, but you'll probably need to read papers and books with a trained mind to appreciate what Davidson is really getting at — and, most particularly, those of his critics.

1. Alfred Tarski. Jan. 2004. Wikipedia entry dealing with the concept of truth in formalized languages, etc.
2. Tarski's Truth Definitions. Wilfrid Hodges. Nov. 2001. Technical and more rigorous presentation.
3. Donald Herbert Davidson (1917–2003). NNA. Links to three obituaries that provide an introduction to Davidson's ideas.
4. Donald Davidson. 2000. Brief sketch of the man and his importance.
5. Donald Davidson. Jan. 2004. NNA. Extended Wikipedia article.
6. Donald Davidson. Vladimir Kalugin. 2001. Relatively straightforward accounts of Davidson's position on Anomalism of the Mental and Causal Explanation of Action.
7. Donald Davidson. 2003.
. Listing of Davidson sources on EpistemeLinks.
8. Donald Davidson. Jeff Malpas. Nov. 2003. Stanford Encyclopedia of Philosophy entry: a more extended treatment, indicating the technical nature of Davidson's work.