Davidson's Inquiries into Truth and Interpretation: Truth-Conditional Semantics
Donald Davidson’s 1984 collection develops truth-conditional semantics and asks how an interpreter can work out what a speaker means from what they hold true.
PhilosophyAll chapters Reading tools Notes & comments
Truth and understanding a speaker › Listening to someone who says it is rainingFull chapter · Sections in this chapter1 / 8 ↓
Listening to someone who says it is raining
Davidson asks us to imagine interpreting a person called Kurt. Near Kurt it is raining. It is Saturday at noon, and Kurt holds the sentence “Es regnet” true. If we already know German, the interpretation is easy. He takes it to be raining. Now suppose the interpreter knows neither German nor which belief Kurt is expressing. The connection between the weather and the sentence is still available as evidence. But Kurt could hold the sentence true because he means something other than what the interpreter guesses, or because he has a belief that turns out to be wrong. One observation does not settle which. Davidson, “Radical Interpretation,” pp. 322–324.
The two problems go together. Interpreting an utterance requires some idea of what the speaker takes to be the case. Ascribing a belief on the basis of the utterance requires some idea of what the sentence means. Donald Davidson's Inquiries into Truth and Interpretation, first collected in 1984, asks how an interpreter can get to grips with both. “Truth and Meaning” proposes a role for a formal theory that states truth conditions. “Radical Interpretation” asks how someone could have evidence for such a theory without already knowing the interpretation it is meant to supply. Inquiries into Truth and Interpretation, contents and edition information.
Building a theory from a finite stock
You understand “The copper kettle on the third shelf is slightly dented” although you may never have heard the sentence before. You know words that occur in it, can pick out the phrase describing a particular kettle, and can see what the rest of the sentence says about it. You could extend the sentence or understand further sentences about other objects. A theory that lists a separate interpretation for every possible sentence would offer little explanation of how a person can do that. Davidson asks for a theory with finitely many resources that can yield results for indefinitely many sentences. Davidson, “Truth and Meaning,” pp. 304–308.
A common response is to assign meanings to the individual words and ask how their meanings combine. That names the task but leaves a question about the work done by the assignment. Suppose you say the name “Mara” contributes Mara and the words “has left” contribute some entity called the meaning of “has left.” How does that pair of contributions determine what would be true when the sentence is true? Davidson examines why adding putative entities can leave the job just as hard as it was. He proposes treating the semantic work as a matter of deriving the truth condition of the sentence through systematic rules sensitive to its parts. “Truth and Meaning,” pp. 304–309.
A tiny language can make this easier to see. Suppose it has names, some words for properties, and a construction that makes the equivalent of “and.” Give the theory a rule explaining when a named object satisfies a predicate and another explaining when a conjunction holds. It can now derive that “Mara has left and the door is open” is true exactly when Mara has left and the door is open. The rules for a simple language may sound trivial. They can nonetheless apply repeatedly to longer sentences, so there is no need to enter those sentences separately in a phrasebook. A theory for a real language has more constructions to account for and would have to preserve this systematic feature.
What the truth theory supplies
Alfred Tarski's work shows how to use one language to formulate a definition of truth for another. The language being examined is the object language. The language in which the theorist formulates rules and proves consequences is the metalanguage. They may overlap, as when we use English to discuss a sentence in English. If we use English to investigate German, the difference is more visible. The German inscription is mentioned and the English sentence is used to state the condition under which what it expresses is true. “Truth and Meaning,” pp. 308–311.
A typical result says that the sentence “Snow is white” is true if and only if snow is white. The quotation marks on the left name the sentence. The words on the right specify a condition without quotation marks. A Tarski-style theory derives such results, called T-sentences, using the structure of the expressions involved. Tarski's Convention T is a condition of adequacy for a truth definition. It requires that the definition yield the appropriate biconditionals. In setting up a truth definition, one can already understand the object language or know the translations one needs. Davidson asks how a theory of that form could guide interpretation when an adequate translation is what we are trying to discover. “Truth and Meaning,” pp. 308–311. “Radical Interpretation,” pp. 317–321.
Knowing that two claims have the same truth value on one occasion does little to show what either means. The informative work comes from a truth theory whose rules derive the truth conditions for the whole language and which survives testing against the use of that language.
Davidson's own example brings out the gap. If snow is white and grass is green, a biconditional can pair “Snow is white” with the condition that grass is green and come out true. That does not make the pair an adequate stand-alone interpretation. Davidson asks whether it would be possible for a strange pairing to follow from a systematic theory that gets all the results right. Its acceptability would depend on what the theory does across the language and how it derives them. A reader can thus reject coincidence as a test of meaning without replacing the theory with a separate list of synonymies. “Truth and Meaning,” pp. 310–312.
The job is more visible when the truth condition varies with context. “I am tired” can be true when one person says it and false when someone else does, or false for its first speaker in the morning. A rule can state that such an utterance is true for the relevant speaker at the relevant time exactly when that person is tired then. The theory has to account for expressions such as “I” and “that book” through the way reference varies with use. Davidson takes this to give a useful test for an interpretation because the relevant person, time, and object can keep changing. “Truth and Meaning,” pp. 317–319.
What the radical interpreter has to go on
“Radical” signals what the interpreter is allowed to assume. The interpreter starts without a translation, a dictionary, or detailed prior knowledge of what the speaker believes. A useful starting point is that a person takes some sentences to be true on some occasions. Recognizing that someone endorses a sentence can be easier than determining what they believe in endorsing it. Davidson treats this attitude as evidence, together with the circumstances of the utterance. A cry of surprise or an overheard line from a play would call for a different diagnosis. Describing the evidence as what Kurt holds true marks a difference between these cases. “Radical Interpretation,” pp. 321–323.
Return to Kurt and the weather. The interpreter notices that Kurt accepts the sentence when it is raining near him, then watches what happens when there is lightning but no rain, when it rains elsewhere, and when the rain has stopped. The theory may come to assign a truth condition involving rain near the speaker at the time of utterance. That assignment is distinct from the report about Kurt at noon and from a generalization about when he in fact assents. The difference leaves space for Kurt to be mistaken. He might answer yes when he hears a sprinkler or fail to notice the rain starting. Repeated observations and the role of the sentence in other combinations can support a reading in spite of individual discrepancies. “Radical Interpretation,” pp. 322–324.
The interpreter also has to consider structure. Suppose they observe a speaker who uses a recognizable construction to reject a report about wind, the same construction to reject a report about rain, and the construction again to reject a claim about the time. Interpreting the construction as negation could unify those patterns. It also constrains further assignments. If the theory interprets each episode separately, assigning a weather term one day and an unrelated clock term the next, there is less reason to expect it to explain what the speaker does with a new sentence. Davidson thinks an interpretation is answerable both to formal constraints and to evidence about how the speaker takes sentences to apply. “Radical Interpretation,” pp. 320–324.
Radical interpretation is an explanatory test. It asks what an interpreter would need to know and how that knowledge could be supported. Readers need not imagine a child formulating axioms or ordinary speakers consciously solving an inference problem. A finite theory can describe relations someone has mastered implicitly. Davidson's project is to explain what can be known about a speaker from the public evidence and why that knowledge would suffice for interpretation. “Truth and Meaning,” pp. 310–312. “Radical Interpretation,” pp. 313–315.
Charity and the difference between error and different speech
Suppose you think someone has said that it is raining and you can see a cloudless sky. The person might have the weather wrong. They might be describing a place you cannot see, repeating a forecast, or using an expression you have misunderstood. There is no general instruction to prefer one explanation in every case. To treat the words as describing rain in the first place, you need reasons drawn from how they are used and from how the speaker responds to circumstances both of you can observe. With those reasons in hand, the theory can give a substantive account of why this particular judgment is an error.
Davidson calls attention to the role of charity in getting this process started. The interpreter tries to make the speaker's total pattern of assent broadly intelligible and to find considerable agreement about ordinary circumstances. Some amount of agreement helps the interpreter identify the subject of disagreements. It would be hard to make sense of a person as mistakenly denying every ordinary fact about the weather if the case supplies no independent reason to think weather is what their words are about. Davidson cautions against taking maximization literally as a count of individual agreement cases. Which agreements matter and under what conditions is part of making the best overall fit. “Radical Interpretation,” pp. 323–325.
There is room to worry that a commitment to agreement creates a bias toward how the interpreter already sees things. For example, an interpreter who assumes every color word has the same boundaries as a word in their own language might take several sound applications as mistakes. Better interpretations can make distinctions the interpreter did not expect. Davidson's framework gives an account of how this correction could happen. The evidence includes which judgments the speaker repeats in other circumstances, what counts for them as a reason for changing their mind, and the place of the disputed sentence alongside beliefs the interpreter can already recognize. Charity works at that larger scale. There is room for considerable difference in the assignments that make the exchange intelligible.
Why a whole scheme may fit better than a word list
Belief and meaning are interdependent in the method because holding a sentence true has both as determinants. A single assent cannot in general tell the interpreter which determinant to revise. Suppose a speaker tends to endorse a sentence near ice and a second sentence near steam. Later the same speaker endorses both after pointing to a kettle. A theory might first assign one phrase to temperature and the other to what the speaker expects to happen next. Further exchanges may suggest instead that one phrase refers to the water's present condition and the other to a setting on a thermostat. In each case later evidence can change the interpretation of an earlier remark. A word list assembled expression by expression may miss relations that would explain those revisions.
Davidson distinguishes the label we might use for a local part of an interpretation from the question whether different schemes could fit the same evidence. The formal structure of a theory and how it accounts for the evidence restrict the possibilities. Even so, two schemes can make the same predictions at the points the evidence can reach while making different assignments within the machinery. For Davidson such cases are a reason to ask what the difference consists in and what it would matter to get right. He does not need an additional private mark fixing a sentence's meaning between competing theories that do the same interpretive work. “Radical Interpretation,” pp. 324–328.
This also puts the relation to Quine in perspective. Davidson draws on the starting point of radical translation, in which a translator has no prior command of the language to be interpreted. He asks for a theory that yields an interpretation by assigning conditions of truth to the sentences, with a role for evidence of what the speaker holds true in the circumstances. The interpreter's task is more than matching a word in the new language to a word in the interpreter's language. Some of the best evidence involves sentence patterns both people can use to make repeated claims about things around them. “Truth and Meaning,” pp. 311–313 and 318–319. “Radical Interpretation,” pp. 321–324.
Where the program meets its limits
A truth theory for assertions can leave important features of speech unexamined. Suppose you repeat “It is raining” as a warning, as a line of dialogue, or while testing a microphone. The sentence can have an intelligible truth condition in each use. To understand what you were doing, an interpreter would also have to know the role it played on the occasion. Indexicals create a different issue, one Davidson addresses in “Truth and Meaning.” Who and when can enter into the truth condition itself. Determining which aspect of an utterance creates a difficulty helps specify what a proposed account has to explain. The collection develops applications to quotation, reports of what people said, and grammatical mood, each with problems of its own. “Truth and Meaning,” pp. 317–320. Inquiries, “Applications”.
There is also a question about how much the formal theory explains. A theorist might correctly derive that Kurt's utterance is true when it is raining near him without explaining which evidence first made the interpretation compelling. Another theorist might offer an insightful description of the evidence without showing how those guesses generalize to sentences Kurt has never said. Davidson's essays put those difficulties together. The truth theory has to explain how the parts and composition make sentences interpretable, and the evidence has to explain why these are the assignments that make sense of this speaker. What seems obvious once we know German can then become an instructive case of how understanding is possible.
Further reading
- Donald Davidson, “Truth and Meaning” (1967). See pp. 308–312 on truth theories, Convention T, and the example involving snow and grass, and pp. 317–319 on variation with time and speaker.
- Donald Davidson, “Radical Interpretation” (1973). Pages 321–324 develop the evidence for interpreting Kurt.
- Donald Davidson, Inquiries into Truth and Interpretation. Both essays are collected in this volume.
Join the discussionAsk a question or share what you learned.
Keep exploring
Reference
Citation details
Cite or share this article.

Comments
Add to the conversation
A question, a correction, or an example from your own work.
No comments yet. Be the first to share your thoughts!