Truth and the Past

Michael Dummett

• Publisher: Columbia University Press
• Year: 2003
• eBook ISBN: 9780231506458
• Topic: Philosophy
• Subtopic: Philosophical Metaphysics

Michael Dummett's three John Dewey Lectures—"The Concept of Truth," "Statements About the Past," and "The Metaphysics of Time"—were delivered at Columbia University in the spring of 2002. Revised and expanded, the lectures are presented here along with two new essays by Dummett, "Truth: Deniers and Defenders" and "The Indispensability of the Concept of Truth."

In Truth and the Past, Dummett clarifies his current positions on the metaphysical issue of realism and the philosophy of language. He is best known as a proponent of antirealism, which loosely characterizes truth as what we are capable of knowing. The events of the past and statements about them are critical tests of an antirealist position. These essays continue and significantly contribute to Dummett's work.

1

THE CONCEPT OF TRUTH

The distinction between linguistic utterances and what they express is borne in on us by many common experiences, for instance, that of finding out how to say something in another language, and that of rephrasing something we have said to make it clear to our hearer. What a sentential utterance expresses is a proposition. Is “true” to be taken as predicated primarily of sentences or of propositions? Although a number of type sentences, such as “Eels swim to the Sargasso Sea to mate,” qualify as true or as false without relativization to any particular occasion of utterance, it is well known that we cannot regard truth as in general being an attribute of type sentences, if only because these may contain indexical expressions whose reference varies from one utterance to another. We must therefore emend our question to: Is “true” to be taken as primarily predicated of propositions or of linguistic entities of some kind?

What reasons are there for opting for one alternative or the other? Frege was strongly in favor of taking truth to attach to propositions, which he called “thoughts” and regarded as being expressed by sentences, in fact as being the senses of declarative sentences; the same thought can be expressed by sentences in different languages, or, indeed, by different sentences in the same language. More exactly, a thought, for Frege, is the sense of a declarative sentence considered independently of the assertoric force that may be attached to it. It need not be a sentence used on its own to make an assertion, but can be a clause within a more complex sentence; but it must be a sentence that could be used on its own to make an assertion, and thus not one containing a pronoun governed by an expression of generality in some other clause. (It can also be an interrogative sentence used to ask a question demanding the answer “Yes” or “No”—a Satzfrage.) A thought, for Frege, is either true or false: it cannot be true on one occasion and false on another. From time to time he somewhat casually acknowledged the occurrence of indexical expressions in some sentences, and was content with saying that in such a case the identity of the speaker or the time of utterance contributed to determining the thought expressed.

Davidson, by contrast, has taken truth as attaching to linguistic items, that is, to actual or hypothetical token sentences. Like Frege, however, he treats indexicality as the only reason, at least within a theory of meaning for a language, why truth cannot be taken as attaching to type sentences: he conceives his token sentences as triples, each with a type sentence, a speaker, and a time as its three terms. The reason for taking truth to be an attribute of linguistic items of some kind is obvious. We must do so if we want to use the notion of truth to explain meaning: whether the meaning of a particular expression or the concept of meaning in general, and whether informally or by means of a systematic theory of meaning for a whole language.

Two things must be accomplished if we are to explain the meaning of a word or sentence ab initio: we must explain the concept or proposition that that word or sentence expresses, or at least what it is to grasp that concept or proposition; and we must explain in virtue of what that word or sentence expresses that concept or proposition. Any account purporting to provide a full explanation of what is involved in understanding the word or sentence must cover both these aspects. That is not to say that it must cover them separately, first explaining the concept or proposition and then stating in virtue of what the word or sentence expresses that concept or proposition; it will do much better to explain the use of the word, or of the words making up the sentence, in such a way as to make manifest what concept or proposition the word or sentence expresses. Plainly, if the explanation is to be given by appeal to the notion of truth, we shall accomplish only the first part of the task if we take truth as attaching to propositions, for no linguistic expression will then be mentioned. If the meanings of linguistic expressions are to be explained in terms of the notion of truth, then truth must be taken as an attribute of linguistic expressions of some kind, and not of nonlinguistic entities such as propositions. To take truth as an attribute of propositions is to take meaning as given antecedently to truth and falsity, since it depends on the meaning of a sentence what proposition it expresses: it is therefore to forswear the project of explaining meaning in terms of truth.

This conclusion does not undermine Frege’s practice, however much it is at variance with his repeated declarations that it is of thoughts that truth and falsity are predicated. In his Grundgesetze der Arithmetik¹ he proceeds quite differently. He lays down stipulations intended to determine the reference (Bedeutung) of every expression, simple or complex, of his formal system. These stipulations make no mention of the senses of these expressions; rather, the notion of sense is explained in terms of them. The Bedeutung or reference of a sentence is its truth-value. We may here ignore Frege’s special doctrine that truth-values are objects, so that a sentence may be described as a “name of a truth-value,” as a numerical term is the name of a number. We ought not to interpret Frege’s notion of Bedeutung as “denotation”; it is better understood in general as “semantic value”—the contribution an expression makes to determining the truth or falsity of a sentence in which it occurs. Frege’s stipulations thus serve to determine the truth-value each formal sentence has: and they do so independently of the notion of sense, and in advance of its being introduced at all. The stipulations lay down, inductively, what the semantic value of each expression is to be; the sense of an expression consists in the way in which these stipulations combine, in accordance with how that expression is made up out of its parts, to determine it as having a certain semantic value. The thought expressed by a sentence therefore consists in the condition, as given by those stipulations, for it to have one or the other truth-value; since it will have the value false just in case it does not have the value true, we may identify it with the condition for it to have the value true. To arrive at this account, it was necessary to take the notion of a sentence’s having the value true as prior to that of the thought it expresses.

In a formal language such as Frege’s the type sentences do not vary in truth-value from one context to another; he dreamed of a language for all purposes which would have that property and so be suited to scientific use. We have no such language. Of what items of natural language should truth and falsity be regarded as attributes? Indexicality is far from being the only bar to treating truth as an attribute of type sentences: for one thing, the references of demonstrative expressions are not determined simply by the time of utterance and the identity of the speaker. It is certainly not enough to consider type sentences, together with references associated with their indexical and demonstrative components, as that to which truth-values are attributed. Our objective being to pick out only those forms of words that express definite propositions, we must exclude those that are devoid of truth-value because they do not express any proposition at all. We might try requiring that the sentence be well-formed in the sense that it is grammatically correct in some language, and contains only words that are meaningful in that language and that occur in appropriate contexts. In natural language it is inappropriate to predicate loudness of a chemical substance, so the English sentence “Hydrogen is loud” violates the last condition.

We need, however, to make sure that we pick out only forms of words for which it is determinate in which circumstances what they say may be recognized as true; those, that is, which express quite specific propositions. Plainly, the stipulations we have so far made are insufficient to guarantee that we shall do this. We may try excluding all sentences involving unfamiliar metaphors. A metaphor may be reckoned familiar if it is recorded in dictionaries; but the line between familiar and unfamiliar metaphors is very imprecise (different dictionaries might give different results). We might also exclude all ambiguity, ruling out sentences ambiguous in construction or containing ambiguous words. But this would go too far. In certain contexts no competent speaker would understand an utterance in more than one of the senses it could in principle bear. Can we not recognize this fact by suitably relaxing our ban on ambiguous sentences? A similar problem relates to vagueness. We certainly cannot exclude all sentences containing vague expressions without ruling out the great majority of sentences of the language. Often we can pronounce a statement to be certainly true or certainly false despite the presence in it of vague expressions: we should like to exclude only those whose truth or falsity in given circumstances vagueness would be an obstacle to judging. A particular variety of vagueness is provided by predicates that have a quite determinate theoretical application, but are frequently, and quite legitimately, used in a much looser sense: an excellent example of Peter Unger, adopted by the late David Lewis, is the adjective “flat.”² It depends on the context how a statement involving the word “flat” is to be understood.

The feature of the context on which it depends is the kind of thing that is being spoken of. A flat terrain is less flat than a bowling green, and less flat yet than a billiard table; we know from what is being spoken of how flat it must be to be rightly said to be flat. Does this make it like other adjectives that admit comparatives? The standard account is that a fat woman is one who is fatter than most women, a large mouse is one that is larger than most mice. Even if correct for these cases, this form of explanation will not do for “flat,” since there is such a thing as being absolutely flat, but no such thing as being absolutely fat or absolutely large. I do not know whether a semantic rule can be devised to govern the right way to understand “flat” according to context. It matters little whether such a rule can be formulated: it is implausible that, in interpreting a use of “flat,” we actually follow some rule; more likely that, as with an outright ambiguity, we go by how the speaker is likely to have wanted to be understood. In most such cases, the hearer exercises more than his understanding of the words uttered—also his common sense. To know what “flat” means, we must understand “flatter than” and “as flat as,” as well as “absolutely flat,” and must know that the acceptable application of the word varies with the type of thing of which it is predicated; but there is no need to suppose that we follow any rule laying down appropriate degrees of flatness, even if one could be formulated.

We try, without appealing to the notion of a proposition, to specify linguistic forms that express determinate propositions; and we stumble over one difficulty after another. How, then, can we ever have arrived at the idea that propositions (thoughts in Frege’s terminology) are determinately either true or false? We arrive at it because we grasp the notion of understanding a given utterance in a particular way. Most of the features we have been discussing have to do with the way in which a statement will be understood: not with a grasp of all its possible meanings, according to the senses of the words in the language, but with the way in which it is taken when it is made on a particular occasion. The hearer will ignore one possible meaning of a sentence that is in principle ambiguous in favor of another. He will treat the word “flat” as applicable in a manner that accords with what it is applied to. He may correctly construe an unusual turn of phrase as readily as if it were a standard idiom. In short, he will put a particular interpretation on it. Sometimes he will need to reflect in order to hit on the meaning the speaker probably intends; more often, he will adopt an interpretation without thinking or striving to attain it.

Aber nicht im Sinne Davidsons: this is not the sense of “interpretation” that figures in Davidson’s account of linguistic communication.³ A Davidsonian interpreter fashions an entire theory of meaning for the speaker’s language; but we are concerned with the understanding of a particular utterance in a language known to both speaker and hearer. The senses of the words of a natural language are extremely flexible. When we put them together in a sentence, these senses are greatly stiffened by the presence of the other words; they are stiffened still more by the whole context in which the sentence is uttered. Successful communication depends on its being our habitual practice to adopt that way of construing the utterance of a declarative sentence that renders it as determinate as possible which circumstances will warrant its being judged true and which its being judged false: in other words, as expressing a definite thought, in Frege’s sense of “thought.” We normally do this automatically, without having to review the different interpretations which the language would admit as possible. It is through our familiarity with this process that we come to grasp the concept of a thought or proposition.

It is impossible to specify those utterances of declarative sentences that express determinate thoughts purely by their linguistic form: we have to take account of how such utterances will be understood. We need to concern ourselves, not with how an utterance may be understood by an individual hearer, who may misunderstand in a variety of ways, but how it will be understood by the great majority of hearers competent in the language. It will be suggested that we should allow for this by conceiving of any statement as being made in a specific context. There is, however, no way to circumscribe the relevant features of the context, and no rule determining from the context, however widely conceived, how the statement will be understood by almost everyone. To know how an ambiguous statement is to be understood often requires a knowledge of what the speaker was likely to be saying; to know what a speaker meant by “flat,” one will need to know how close an approximation to absolute flatness would be appropriate in the circumstances. We are forced to consider statements, not as mere concatenations of words, but as subject to particular ways of understanding them, that is, as expressing specific thoughts or propositions. This is the ground for following Frege in regarding truth as attaching, not to linguistic items such as sentences, but to thoughts.

We appear thus to have good reasons to take truth to be attributable to sentences, and good reasons to take it to be attributable to propositions. The choice between sentences and propositions as bearers of truth-values is a false one, however: we should see truth as attaching to a token sentence, but one considered under a particular interpretation. The language determines what interpretations are in principle possible; when the interpretation that is to be put upon the sentence is obvious, it is the circumstances that determine which interpretation that is. The interpretation will select, for each word, one of the senses that the language allows it to bear; it will fix suitable ranges of application for vague expressions involved. The sentence so interpreted will probably not have a truth-value in every imaginable situation; with luck, it will have one in every situation that actually arises. It will therefore not express a thought according to Frege’s strict criteria; but we may take it as expressing a proposition that can legitimately be said to be true or false.

BY FAR the most popular type among theories of meaning is the truth-conditional variety. The central tenet of such theories is that the meaning of a statement is given by, or consists in, the condition for it to be true. This needs supplementation by an account of how the meanings of nondeclarative sentences are to be explained by reference to the meanings of correlative statements; but although this supplementation is not unproblematic, we may pass by the question how it is to be provided. The thesis also needs qualification, in that it applies, if at all, only to the central aspect of an expression’s meaning which Frege called its “sense” (Sinn), and not to those aspects which do not affect the truth or falsity of what is said, such as that which differentiates the words “dead” and “deceased.”

It was Frege who first clearly formulated the truth-conditional account of meaning, which was then adopted by Wittgenstein in the Tractatus. In Frege the theory has two layers. Each component of a statement has a semantic value, in accordance with its logical category: the semantic value of a singular term is an object, that of a functional expression a function, and so on. It is in virtue of its semantic value that an expression contributes to determining, of any statement containing it, whether or not it is true; according to Frege, the semantic value of a whole sentence is its truth-value, since it is, according to him, in virtue of its truth-value alone that it contributes to determining that of another sentence of which it is a subsentence.

But our understanding of an expression, including a sentence, never consists simply in a grasp of its semantic value. As Kant said, every object is given to us in a particular way. Otherwise expressed, an item of knowledge that someone has can never be completely described by saying that he knows, of some particular object, that such-and-such holds good of it (e.g., that it is what some term refers to). Likewise, when coextensive properties are identified, no item of knowledge can be completely described by saying that someone knows, of a certain property, that something holds good of it (e.g., that some things lack it). Quite evidently, every property that we can ascribe to objects must be given to us in a particular way; and the same applies to every relation and function of which we can speak. For Frege, the sense of an expression is the way its semantic value is given to us in virtue of our understanding of the expression; and this must be a feature of the language, rather than of any one individual’s grasp of that expression, since otherwise we should have no common basis for judging whether a statement involving the expression was correct.

There are, as we noted, two parts to knowing what a word or sentence means: we have to grasp the concept or proposition it expresses; and we have to know that that is the concept or proposition it expresses. Any account purporting to provide a full explanation of the meaning of the word or sentence must cover both these aspects. There are two opposite errors which reduce an explanation of meaning to only one of its two components. The proponents of what are known as modest theories of meaning maintain that it is no business of a theory of meaning to explain the concepts expressed by the words of a language, but only to specify which concepts they express: they thus dispense with Frege’s second layer. Plainly, there can be no theory which could convey to someone ignorant of them all the concepts expressible in a given language by individual words, since it would be necessary to grasp the concepts employed by the theory. Such a theory might well be able to convey many of those concepts; for every other concept expressible in the language, it must be able to explain what is needed for someone to be truly said to grasp that concept, namely by describing the use that he must make of a word that expresses it. A theory that fails to do this will omit an essential ingredient of the understanding of such a word.

The opposite error is most likely to be made in application to whole sentences, and it is usually advanced by those who display an insouciant attitude to quotation marks. The notion of truth, they argue, is needed only to state the general principle that to know the content of a proposition is to know the condition for it to be true. In any particular application of the principle the notion of truth is redundant: to know w...

Table of contents

• Cover 
• Half title
• Series Page
• Title
• Copyright
• Contents 
• Foreword
• Preface
• 1: The Concept of Truth
• 2: The Indispensability of the Concept of Truth
• 3: Statements About the Past
• 4: The Semantics of the Past Tense
• 5: The Metaphysics of Time
• 6: Truth: Deniers and Defenders
• Notes
• Index