Donald Davidson
Donald Davidson; drawing by David Levine

There is no more creative or systematic philosopher at work in America today than Donald Davidson, but most people would find his essays formidable. This is not because they are long, convoluted, or technical, or because they are obscure or self-indulgent. On the contrary, it is because the prose is so plain, sparse, pruned, a prose long in preparation and short in delivery.

These eighteen pieces, averaging fifteen pages each, represent twenty years of thinking about how people communicate. They are matched by fifteen more in a companion volume called Essays on Actions and Events.1 Davidson proceeds by composing small talks that he delivers over and over again in lectures and seminars. Their first publication may be nearly anywhere. I have on my desk one talk recently published in Lawrence, as the Lindley Lecture at the University of Kansas, and another offprinted from a journal published in Graz, on the border between Austria and Yugoslavia. I doubt whether a dozen public libraries have both of these.

After Davidson’s work has had its long and careful larval life, it passes into a dormant chrysalis stage of obscure publication. Inquiries into Truth and Interpretation is the occasion for the final metamorphosis into a readily available sequence of sustained arguments. Davidson’s two volumes range over many topics other than language: causation, weakness of the will, belief, human action, mind and body, events, ontology (what kinds of things there are in the world). Language, however, dominates the present collection.

Davidson’s title is exactly right, for truth and interpretation are his fixed points of departure. Yet those very words could mislead many a follower of humanistic studies. “Interpretation” has become a vogue word that rightly goes along with talk of texts and hermeneutics. Those who write about interpretation usually have a thoroughly pluralistic vision, whether they are literary critics, anthropologists, historians, or semioticians. This is because they take it for granted that other people think with concepts that have no place in our own lives. Interpretation becomes the art of moving from one vision of the world to another. Truth becomes relative to a scheme of thought.

Davidson, in contrast, thinks that we need an objective and non-pluralistic theory of truth to understand how any understanding is possible. He denies that it makes sense even to imagine alternative realities, each with its own truths, that cannot be translated into another way of thinking. Thus Davidson differs fundamentally from many of the streams of thought current throughout the humanities today. This is partly because he comes from the heartland of analytic philosophy, which traditionally has been doubtful about non-Western rationality. His book is dedicated to the distinguished analytic philosopher W.V. Quine, “without whom not.”

Davidson is almost a cult figure for many young philosophers, and not only in the United States. His impact on Oxford has been called “the Davidsonic boom.” Last April a four-day conference at Rutgers was dedicated to Davidson, and it was notable both for its high jinks and for the presence of a majority of the most respected philosophers in the English-speaking world who came to talk about Davidson’s work.

Davidson wants to show systematically how a speaker can understand even the most trifling utterance of another person. It is not for him to start with weighty texts. “Snow is white” is his most often repeated handy example of something you can say. But Davidson is no mere technician, striving to be exactly right about a point or two of no importance. The chain of his deductions mercilessly sweeps across the entire field of traditional philosophical concerns.

He begins with deceptively innocent questions. Two people in conversation do more than exchange sounds that stand for words. They understand each other. How do their words mean what they mean? By way of answer, Davidson invites us to think about “radical interpretation,” the problem of two mutually alien people trying to figure out each other’s language from scratch. In this situation, guesses about the meanings of particular words alone are not enough. An interpreter must have some idea about the beliefs and wants of the other.

If I play “interpreter” and you play “informant,” I may take a sentence of yours—“the tides are dangerous here”—to mean that the tides are dangerous here. I do so in part because a roaring riptide is obviously ahead of us; I have to assume that you see that as well as I do. I cannot propose that you think that the waters are calm and that your words mean that. If on other evidence I did infer that you are trying to tell me that the waters are calm, I will tack on some other beliefs about you—that you have a death wish, or more likely are trying to lull me into paddling to destruction in the maelstrom. To generalize: any account of what your words mean goes along with my expectations about your beliefs and desires. Radical interpretation demands not only that we try to match up words, but that we try to form a picture of all pertinent aspects of the person with whom we are interacting.


Such commonplaces would not seem to take us very far until we read the second paragraph of a piece entitled “Radical Interpretation.” “The problem of interpretation is domestic as well as foreign: it surfaces for speakers of the same language in the form of the question, How can it be determined that the language is the same?” When my old English crony says, “The sea is calm today,” how do I know that her words mean that the sea is calm? Davidson forestalls the obvious answer, that she is speaking English, that we speak the same language. For how do I know that? It is just as much interpretation, says Davidson, when I understand her words, “The sea is calm today,” to mean that the sea is calm today, as when I venture to translate the words of a Bulgarian whom I meet at a beach on the Black Sea.

But at least this may seem obvious: in chatting with my friend or guessing what the Bulgarian is saying, I know or conjecture what the words mean. Davidson doubts, however, that talk of knowing the “meaning” of words explains my understanding of them. It is not that he wants to cross the verb “to mean” out of his prose, but that he believes it should be possible to state how I understand another person without using the word “meaning” and its cognates. He is asking, what knowledge must I have in order to understand you? He wants to specify this knowledge without depending on some unanalyzed idea of meaning, for anything less would be circular or question-begging. It is unhelpful to say, for example, that I understand you because I know the meaning of “The sea is calm today,” for what is this “meaning” that I know?

Davidson is not posing the practical questions of a linguist or psychologist about the processes by which people learn and deal with words. He is putting an abstract question about how to represent the knowledge that a person must have in order to understand another person. The answer cannot usefully be that we know what the words of the other person mean, for it is just that kind of knowledge that we are trying to explain.

The explanation that Davidson is groping for has two critical elements. First, he discourages an atomistic, building-block picture of language. We don’t first master “sea” by pointing at huge expanses of water, and then “calm” by some restful examples, and then “today” (how?), and then build up to understanding a whole sentence, “The sea is calm today.” Following the late-nineteenth-century German logician Gottlob Frege, Davidson takes it for granted that the smallest unit of understanding is the whole sentence. More than that—and here Davidson is more radical—understanding cannot limp along, sentence by sentence. I can understand you only by having an implicit theory about how to understand a network of your possible utterances, and this theory, as we have said, must be accommodated to a larger view of your desires and beliefs. Philosophers call such a theory of language holistic, as opposed to atomistic.

After holism, Davidson’s second critical element in the explanation of understanding is his fixed point, truth. In what terms shall a theory about you and your speech be couched? In terms of truth, he argues. Here is the merest sketch of his view. A complete stranger, or my oldest friend, it matters not which, utters a sentence, s. I would know a lot if I knew the conditions under which this sentence would be true, especially if I could say that s is true if and only if ______, where the blank is completed in my language, in English. If my friend just said “The sea is calm today,” the completed formula would come out as the trivial-sounding statement, “In her language the sentence ‘The sea is calm today’ is true if and only if the sea is calm today.”

But that is not a trifling claim. It could be false because her language may include sentences that sound just like mine but are true only under different conditions. (In her language, the sentence, “The sea is calm today,” may be true if and only if the sun is round.) The assertion that her sentence, “The sea is calm today,” is true if and only if the sea is calm is just as substantive as one about the truth conditions of a Bulgarian sentence. In stating truth conditions, therefore, I do not use the nonexplanatory word “meaning,” but in a noncircular way I have in effect said what the words mean.


Moreover in careful talk of such truth conditions, we avoid an ambiguity. For example, to know the truth conditions of a sentence is not yet to know what the speaker intends, or in the usual sense “means,” on some occasion. If the sea is obviously rough and dangerous when she says, “The sea is calm today,” she may be ironic, mocking the macho canoeist. She may be giving a warning—what she means is that, we’re being observed, don’t trust anything I say from here on in. Thus we distinguish what the words mean in the language from what a person means by using them. What the words mean in the language is given by truth conditions: call that the literal meaning of the words. We can use words to do almost anything, but Davidson holds that a wide range of uses, including irony and metaphor, depend upon their literal meanings, i.e., upon their truth conditions.

We have to put this talk of truth conditions together with Davidson’s holism. When I interpret what another person is saying, I can’t just have a view about this one sentence, “The sea is calm today.” Consistent with his attack on the atomistic view of language, Davidson holds that our sentences are part of a web of sentences, whose truth conditions depend on one another. According to Davidson, therefore, I have, or am trying to construct, a theory about the person’s language as a whole, a theory that will have the truth conditions for this sentence as only one tiny part. Now most ordinary people use a home computer after memorizing a finite list of commands that tell it what to do; but the language of a person is not like that. There is no discernible upper limit to the number of sentences in your language. I could not memorize them all even if I wanted to. Moreover, you are constantly blurting out new sentences that I have never heard anyone say before, yet I commonly understand you without even thinking about it. So the knowledge that I have, when I understand you, cannot be compared to or simulated by a memorized list of truth conditions. I may memorize a vocabulary that is finite—even Webster’s comes to an end—but your possible sentences do not come to an end. So knowledge of your language must follow the model of a system for generating truth conditions of an indefinite number of declarative sentences.

Some fragments of an obvious system are easy enough. Knowing the truth conditions for “The sea is calm today” and “The sharks won’t pass this way for a week,” I know the truth conditions for “The sea is calm today and the sharks won’t pass this way for a week.” The long sentence is satisfied if—and only if—the two component clauses are. So I have a system for generating the truth conditions of conjunctions. But from there on in, Davidson’s model becomes difficult and technical. Obviously the sentence “Every shark is ruthless” has truth conditions somehow connected with the sentence, “That shark is ruthless,” but how should the connection be spelled out? Davidson here falls back on the formal theory of truth constructed by the great Berkeley logician, Alfred Tarski. Many of the finer points in Davidson’s philosophy are connected with this logic, but it seems to me that in his later papers Davidson demands neither fidelity to, nor much knowledge of, Tarski. The allusions to Tarski in this book, although plentiful enough, should not hinder anyone who wants the drift of Davidson’s reasoning.2 The essential point is that Davidson thinks that knowledge of the language of another—the ability to interpret another—should follow the model of a systematic method for generating truth conditions of declarative sentences.

Now let us put truth and interpretation together. A radical interpreter, trying to grasp a language from scratch, forms conjectures about what the other people are saying. In practice this will be a hodgepodge of hunches and analogies, but we can form an abstract model of the end product: a theory of truth for the language. Any such theory ought to be empirical. What would be the evidence for it? It must be the behavior, including the actual speech, of the other person. But what is said will not by itself settle whether the theory of truth is adequate for this person’s language. As is urged in my earlier discussion of interpretation, we need a fairly complete picture of the other person, of belief and desire.

Davidson, as I have said, energetically opposes a building-block, atomistic view of language. Understanding is a matter of understanding a network of whole sentences, and it is the theory of that network that must be matched against the behavior of the other person. But of course we do not literally see the beliefs and desires of other people; these are speculative constructs that we attribute to a person when we propose a theory of truth. Belief, desire, and truth conditions are of a piece.

This conclusion incidentally bears on the old question of whether there can be thinking without language. Davidson would recast that question as one of “propositional attitudes”—to think that…, wish that…, doubt that…, intend…. He argues that we can attribute such propositional attitudes to other people only if we also attribute to them a language for which we propose a theory of truth. In imprecise shorthand: dumb beasts don’t have thoughts.

That inference is an example of the single-mindedness that characterizes so many philosophers. Davidson insists on being taken literally, and on drawing the consequences for doing so. Could there, for example, be a people whose language employs a scheme of ideas fundamentally different from ours, which cannot be expressed by us? Not on Davidson’s account. Here he offends against many humanistic views of interpretation. If other people have a language that we profess to interpret, we shall be matching their sentences with our truth conditions, for we shall be proposing, in our language, a theory of truth for their language.

We will be able to do so by attributing to the other people beliefs and desires in which a vast number of commonplace concerns run together with ours, such as that oranges are round and that having a cold makes us sniffle. If we could not do this, we would be hard pressed to say that these people have a language at all. Naturally they and we can both say exotic or preposterous things on the fringes of our talk, and what they think is exotic we may find lunatic. But you can, for example, attribute to me the ridiculous belief that if I ate six hundred oranges a day, I would never catch a cold, only because you assume that we share more commonplace attitudes toward the shape of oranges and the discomforts of nasal drip. If your interpretation of my language came out with all my utterances being as odd as the one about six hundred oranges, you would have the best of reasons for thinking you had got matters wrong and were misinterpreting me. In short, to attribute too much disagreement or idiocy to another is to cast doubt on the adequacy of the proposed theory of truth for the other. Just as we cannot attribute thoughts to dumb beasts, so we cannot attribute a wholly alien scheme of thought to articulate people.

Long ago Evans-Pritchard, the British anthropologist, implied that the ideas of causality current among people in the Southern Sudan cannot be expressed within the concepts of our way of thinking. Similarly, B.L. Whorf, the American fire insurance adjuster and student of the Southwest, depicted Hopi Indian notions of space and time as carving up the world entirely differently from the way in which we do it. Davidson is not denying that the Nuer or the Hopi live very differently from us. He is claiming that the interpreters, Evans-Pritchard and Whorf, notwithstanding their conclusions, must have been treating those peoples as much more like us, in their beliefs and desires, than is commonly made out. Many more recent ethnographers would agree.

Davidson is, however, after bigger game. The idea of fundamentally different organizations of concepts arises, he thinks, from the notion of carving up reality in alternative ways. It invokes what he names “the dogma of scheme and reality.” That is, the idea that there is a given reality, and that there are various human schemes of thought for presenting this reality. Davidson doubts that this makes sense. We may be misled by something that does make sense, for example ways of mapping the earth. We have numerous projections—Mercator, stereographic, and so forth. There is nothing dogmatic about contrasting a scheme (the map) and reality (earth), for we have different methods of mapping an independently identifiable planet. But there is no independent way of identifying “reality” except by talking about it and interacting with it, and Davidson has been arguing that there cannot be essentially different conceptual schemes. Hence map-and-earth bears no analogy to scheme-and-reality. The latter is a mistaken dogma that has obsessed us at least since the time of Kant. There is, according to Davidson, nothing to it.

Davidson propounds a related thesis in the paper “Reality Without Reference.” He opposes one commonsense attitude to our surroundings that runs like this. The world is full of things, such as people, rivers, and amino acids. There are relations between these: I am sitting beside the river eating a sandwich of cheese, which includes amino acids. What I’ve just said is true because “the river” refers to this very river, the St. Lawrence (etc.), and my sentence correctly picks out a relation between me, the river, the cheese, and the molecules. True sentences correspond to relations between the things to which I refer. This attitude brings a comforting antidote to relativism and anti-objectivity. Just look, I’m beside the river, and my sentence is a copy of that relation, thanks to the fact that the words “the river” refer to the St. Lawrence (etc.).

Davidson dissents. He thinks that this is entirely the wrong way to defend the objectivity of our assertions—for it commits the very error that makes antirealism possible. He thinks that there is no unique way in which individual words latch on to things, and that it does not matter in the least. Reference of words, to use Quine’s aphorism, is inscrutable. A theory of truth—the very model of interpretation—does not at all work at the level of words referring to things. A theory of truth will not commit me to the claim that “the river” here refers to the St. Lawrence, nor even that the word “river” should pick out rivers. Truth conditions need never reach down below the level of whole sentences in order to grab the river. The very metaphors in which we try to portray such a feat display their own absurdity.

Sometimes Davidson is read as a kind of pragmatist. He does not welcome this but he does try to dismiss contrasts between pragmatists and others. For more than a century philosophers have debated whether truth is what merely hangs together, what works in practice, or whether true sentences somehow correspond to the way the world is. The most accessible pragmatist is William James, who cheerfully said that truth is what works, and delightfully jeered at what he called the copy theory of truth, the doctrine that true sentences are copies of the world. Davidson subtly rejects the entire debate, and thereby proposes to transcend generations of a certain style of philosophizing. He defends “what may as well be called a coherence theory of truth and knowledge which is not in competition with a correspondence theory, but depends for its defense on an argument that purports to show that coherence yields correspondence.”3

No simple summary of that argument will capture Davidson’s apparent precision, but here is what I judge the idea to be. In conjecturing a theory of truth about the speech of another person, we must (on pain of suspecting mistranslation) arrange for him a coherent bundle of beliefs and utterances, coherent by our lights, the only lights we have. But what we call reality is not something that can be identified independently of how we identify it. Hence that which we, by using the standard of coherence, call true, unsurprisingly, and vacuously, “corresponds” to the world.

To put the result in William James’s businesslike terms, Davidson magnanimously finds a place for correspondence and for coherence, but only by selling off the unprofitable, if sensational, subsidiaries that managers of both enterprises held most dear. Thus Davidson’s coherence theory of truth has little to do with what works in my life, as James would have it; it is about making sense of other people. Likewise Davidson’s correspondence, shorn of reference, does not copy a given reality. A great many Davidsonian themes are in conspiracy here. For example in an early essay he debated whether true sentences match up, individually, with facts. But facts cannot be identified except by sentences, and Davidson, consistent with his holism, concluded that although individual sentences do not correspond to individual facts, the whole web of sentences has to be true to the one fact, that is, the fact of everything. There is no local correspondence, “the river” to the river. There is only total correspondence of all true sentences to the fact of everything; but this fact, the world, has no autonomy beyond what we say. That is one way in which “coherence yields correspondence.”

The present collection of essays ends with a group labeled “Limits of the Literal.” A Davidsonian theory of truth will provide, in the form of truth conditions, what Davidson calls the literal meaning of a sentence. Some dictionaries define “literal” as using words with their standard denotation, but as we have seen, Davidson does not want to go down to the denotation of individual words. He always works at the level of the sentence. He is concerned with the contrast between literal and metaphorical uses of whole utterances. For an analogy, we might think of a brick as having the literal function of building walls, but we can also use a brick as a weapon, a weight, for concealing a birthday present, and, in the case of an old and crumbling brick, for writing pink messages on black asphalt. But the imaginative use to which we put the brick does not in general depend upon the original purpose of bricks for construction. By contrast Davidson thinks that although we can do endless things with sentences, a central class of those doings employs or is parasitic upon the literal meaning, the standard truth conditions.

Thus in “What Metaphors Mean” he repudiates the idea “that a metaphor has, in addition to its literal sense or meaning, another sense or meaning.” When Job wonders, “Is my strength the strength of stones, or is my flesh of brass?” he is not attaching a new sense to the sentence, “My flesh is made of brass.” He is relying on the standard truth conditions of that sentence in order to draw attention to his suffering. There is not a momentary new sense to the words “flesh” and “brass.”

In this opinion Davidson takes issue with most literary critics, linguists, psychologists, and philosophers. To use a metaphor is not to attach a new meaning to the words, but to use the standard meaning for a surprising effect. This “makes metaphor a more, not a less, interesting phenomenon.” It is instructive that Davidson begins his philosophy by pedantically insisting that “meaning” is not a concept with explanatory value. Then he provides a precise substitute for the idea of meaning, in terms of truth conditions. This very precision enables him to come back and say that metaphors are not new meanings, just new uses of old meanings. This is also an elegant twist of Wittgenstein’s often quoted maxim, “Don’t ask for the meaning, ask for the use.” Davidson would have us ask for the meaning (truth conditions) and then ask for the use.

A more far-reaching application of the same insight is given in another essay at the end of the book, “Communication and Convention.” It is a commonplace that language is conventional, but Davidson doubts that there is much helpful truth in this commonplace. Certainly we have the convention of greeting each other with words like “Hello,” and, in some places, parting with “Have a nice day,” or “Hurry back.” There are also innumerable historical facts, more dubiously called conventions, exemplified by the fact that the sound “brick” serves us much as the sound “brique” serves the French. Such facts are mere prerequisites to our ability to state anything. Davidson disputes the more controversial suggestion that there is a fundamental “convention” that we use declarative sentences to tell the truth.

Here he is taking issue with, for example, the Oxford philosopher Michael Dummett. He is also cancelling the suggestion, which might seem natural, that a philosopher who uses truth conditions as a surrogate for meaning, must think that the point of talking is first of all to tell the truth. Not at all. To repeat, we can do almost anything with our words: lie, procrastinate, bully, incite, amuse, delight. What we do depends on there being truth conditions for our sentences, but it does not depend on our ever using those sentences in order to tell the truth. We could exaggerate Davidson’s position by saying that once, as a matter of sheer human and local history, we have words and sounds in place, it is almost never a matter of convention what we do with our words. We can do anything. As usual, Davidson, who runs the tightest ship of language afloat today, gives intensely free play for any kind of creativity.

Old-fashioned hagiographies used to begin by warning that the life of the saint was more for our admiration than for our emulation. Admire St. Simeon for his thirty years living on top of a sixty-foot pillar in the desert, but don’t try to do it yourself. Many philosophers feel this way about Davidson’s work. Some argue that there is no place for such systematic development in our times. Moreover, just because Davidson does have a system, premises and assumptions are laid bare for the doubting. For my part, I do not think language is anything like so unitary as Davidson implies. I query his holism and imagine language and life as a bundle of activities that fit together much more loosely than he supposes. I am suspicious of vestiges of the Wild West in American philosophy, the frontier spirit that dares to model all speech according to the requirement of radically interpreting the newly encountered natives. And as is the way with systems, when parts are called into question, the whole may be put at risk.

Nor is Davidson one to let things rest. In as yet unpublished work, he doubts that there is such a thing as a language, at least as commonly understood by philosophers, psychologists, and linguists.4 That would seem to put the cat among Davidson’s own pigeons. Such complex issues are a matter for long, hard, cautious thinking. Whatever conclusions we shall finally reach, it is evident that Davidson has already constructed one of the most remarkable pillars of sustained philosophical reasoning to be found in any era.

This Issue

December 20, 1984