Printer Friendly

Semantic realism.

Michael Devitt has argued that Michael Dummett unsuccessfully attacks realism because Dummett does not address the traditional, and perhaps more interesting, doctrines that have been called by the name "realism." Dummett, will balk at the charge that his writings on realism, truth, and the theory of meaning do not bear on the traditional metaphysical issues of realism. Indeed, he thinks that his most singular philosophical achievement has been showing that different realisms (about universals, mental states, physical objects, numbers) have a common characteristic: each involves the claim that the principle of bivalence holds for the relevant class of statements. Since he thinks that bivalence holds for a class of statements just in case those statements possess truth-conditions which transcend their conditions of verification, and since he thinks that the meanings of statements are either their truth-conditions or their conditions of verification, Dummett thinks he has succeeded in transforming persistently thorny questions of metaphysics into more tractable questions of meaning. His argument against realism, then, centers on his critique of truth-conditional semantics.

In this paper I shall set forth the major claims of one form of truth-conditional semantics, which I call "semantic realism." After investigating criticisms of this doctrine, I conclude that it remains unscathed by Dummett-style objections. En route I will defend four major theses. First, a truth-conditional theory of meaning is compatible with a verificationist theory of understanding. Second, the warrant for attributing truth-conditions to statements derives from the nature of the justification-conditions for those statements. Verificationist semantics, properly understood, justifies truth-conditional semantics. Third, metaphysics is prior to semantics. Dummett's strategy to treat metaphysical issues as issues of meaning begs certain metaphysical questions. His position about meaning crucially depends upon certain metaphysical commitments. Thus verificationists cannot allow relevant metaphysical issues to remain unsettled pending the outcome of the debate over the nature of meaning. Fourth, skepticism, which is supposed to create trouble for truth-conditional semantics, creates equal trouble for any plausible account of truth that is acceptable to the verificationist.

Section 1 contains a sketch of semantic realism and in sections 2 through 4 I defend this view against objections. Section 2 concerns Dummett's claim that the theory of meaning is nothing more than a theory about competent linguistic usage. In section 3 two objections from undecidable statements are defused and in section 4 alleged skeptical difficulties for any theory dependent upon substantive truth-conditions are examined. I conclude, in section 5, that not only do these objections fall to pose difficulties for semantic realism, but Wittgenstein's rule-following considerations give it further credibility.


What is Semantic Realism? Two doctrines constitute semantic realism: one semantic, one metaphysical. The semantic doctrine is the well-known view that truth is central to a correct theory of meaning and that the meanings of statements are their truth-conditions.(2)

The content of truth-conditional semantics is obscure in the absence of some characterization of the nature of truth-conditions. Hence the need for a metaphysical doctrine. The metaphysical component of semantic realism is that truth-conditions are robust, objective structures of the world that render particular statements true. They are often concrete complexes commonly constituted by medium-sized dry goods and their attributes. As such they are not identifiable with any set of evidential conditions because they transcend any relevant conditions of justified belief or assertion. This transcendence allows for the truth of the relevant statement(s) in the absence of any attending evidence and, in a range of cases, it allows for the truth of some statements in the absence, in principle, of evidence sufficient to warrant assertion. Thus, these robust structures are distinct from the justification- or assertibility-conditions which might render a belief rational or an assertion appropriate. For the purposes of this paper, semantic realism involves only these claims about the natures of meaning and truth-conditions.(3)

These two doctrines are independent of one another. Obviously, truth-conditional semantics does not entail a theory of metaphysically robust truth-conditions. Such conditions might be characterized evidentially. Nor does a theory of robust truth-conditions entail truth-conditional semantics. Statements might be rendered true or false by conditions in a largely impersonal world that is typically not created by our conceptual scheming without their meanings being their truth-conditions. Concrete properties of the stuff that falls from the sky onto Antarctica might render "Snow is white" true, while other-worldly Platonic forms or Fregean senses provide the meanings for "snow" and "is white." Perhaps meanings are not ontological posits at all but socially- and contextually-constrained features of linguistic usage that determine their conceptual role in a cognizer's functional organization. Neither alternative to truth-conditional semantics requires the denial of the semantic realist's claim that, truth-conditions are objective worldly conditions. Furthermore, the combined semantic and metaphysical components of semantic realism still admit of many theoretical options. They are compatible with holistic, molecular, and atomistic approaches to meaning and do not determine whether entire languages, statements, or individual terms are the primary units of meaning. Semantic realism has the appearance of a moderate theory that occupies a middle ground between abstract ontological theories of meaning and nonreferential use theories of meaning. This appearance makes it a plausible theory to the extent that such metaphysical moderation is plausible.

The semantic component of semantic realism straightforwardly fills out the theory of meaning for statements. The meanings of statements are their truth-conditions. It is not obvious, though, how such a theory can account for a competent speaker's ability to understand a language. Without doubt, any adequate theory of meaning must accommodate a theory of linguistic competence - a theory which tells us in virtue of what we understand language. Even if Hilary Putnam is correct that meaning is not exhausted by what a competent speaker carries in the head (or even by what a group of competent speakers carry around in their collective heads), showing how the contents of the head provide for linguistic competence is a reasonable constraint on any adequate theory of language.(4)

Michael Dummett and Crispin Wright think that the flaw in semantic realism is that there is no accounting for semantic competence with statements which have metaphysically robust truth-conditions as their meanings. Allegedly, semantic realism is deficient precisely because it is incompatible with any plausible account of linguistic understanding. Both Dummett and Wright presume that the semantic realist has the burden of proposing a peculiarly realist account of linguistic competence, a uniquely realist theory of how we understated language, one that makes reference to our knowledge of transcendent truth-conditions rather than to our knowledge of mere assertibility-conditions.(5) In the remainder of this section I shall argue that while semantic realism specifies only a theory of meaning, it is compatible with an adequate theory of understanding. In particular, the theory of understanding with which it is compatible is none other than the verificationist theory of understanding proposed by Dummett and Wright themselves. This theory explains linguistic understanding in terms of knowledge of justification- or assertibility-conditions rather than in terms of knowledge of truth-conditions. I shall further argue that, this theory of understanding serves as the basis for legitimately inferring a truth-conditional theory of meaning.

To begin, let us consider the purviews of theories of meaning and theories of understanding, respectively. A theory of meaning, on the one hand, concerns the nature of meaning, the existence and identity conditions for meanings. It should tell us what constitutes meaning and what things bear meaning. Prima facie, a theory of meaning is a metaphysical theory which specifies in what, meaning consists. A theory of understanding, on the other hand, concerns our knowledge of meaning. It should deliver a practical semantic epistemology, rather than a semantic metaphysics. It should provide an account of the proper representation of meaning rather than the nature of meaning. Since surface appearances indicate that theories of meaning and theories of understanding concern different issues, there is no pretheoretical reason for thinking that these theories are the same or even that, the core notions of the two theories are the same. On the contrary, there is an initial reason to think that, the core notions of these theories are not the same at all, since one concerns the nature of something, while the other concerns our epistemic access to that something and our ability to do things with words. Thus one who claims that the story of meaning must be told in epistemological terms rather than metaphysical terms owes us a substantive argument for that claim.

Initial appearances notwithstanding, the semantic realist is obliged to forge links between the theory of meaning and some theory of understanding. The apparently troublesome element of semantic realism is the metaphysical doctrine. Truth-conditions, which transcend evidential conditions insofar as they are not exhaustively characterized by evidential conditions, threaten to be epistemically inaccessible because no amount of evidence insures truth. The verificationist objects that we can never have access to such truth-conditions, and consequently we can never know that statements have the property of being appropriately related to some objective worldly situation. All we can know is whether we have sufficient justification for asserting a statement; we never independently get at the truth-conditions. Since truth-conditions are epistemically beyond us, they are not the meanings we grasp.

Dummett raises this concern in connection with the issue of language learning, dependent as it is upon the ways speakers manifest their grasp of language. He says:

We no longer explain the sense of a statement by stipulating its truth-value in terms of the truth-values of its constituents, but by stipulating when it may be asserted in terms of the conditions under which its constituents may be asserted. The justification for this change is that this is how we in fact learn to use these statements.(6)

We learn to use statements by observing the contexts in which they are used and the evidence available at the time of utterance. Evidence alone signals the appropriateness of utterance, leaving no theoretical work for truth-conditions to do. Thus semantic realism must give way to verificationist semantics.

What [Wittgenstein] is concerned to deny is that it is by appeal to the condition for such a statement to be true, grasped independently of our means of knowing it, that we apprehend the meaning of the statement. The account of meaning in terms of truth-conditions has to be replaced by one in terms of the conditions under which we are justified in making such statements, including ones when the justification may be over-turned; and what justifies a statement of this kind does so only in view of the fact that certain general connections hold.(7)

The semantic realist's task is to explain how we can both grasp transcendent truth-conditions and manifest that grasp in our linguistic practice.

The semantic realist must deny that truth-conditions are noumenal posits utterly disconnected from the evidence we have about the world. Otherwise, there could be no suitable linkage between truth-conditions and semantic knowledge. Fortunately, semantic realism's metaphysical doctrine does not require that truth-conditions possess any noumenal qualities. That truth-conditions are not fundamentally evidential conditions certainly does not entail that cognizers must fail to have adequate evidential relations with truth-conditions. The semantic realist's claim about truth-conditions is analogous to, indeed largely the same as, the common sense realist's claim about physical objects. According to common sense realism, the nature of ordinary physical objects is not given by evidential relations with cognizers. Facts about epistemic warrant do not determine what such objects are. Rather, facts about epistemic warrant determine only under what circumstances we are justified in making claims about these objects. Yet, because physical objects are constructed the way they are and because we are constructed the way we are, they typically can be objects of justified belief and assertion since we regularly have adequate epistemic access to them.

Likewise, whatever else is true regarding truth-conditions, they are metaphysical structures of the world whose natures are not given by evidential conditions. Because truth-conditions are constructed the way they are and we are constructed the way we are, truth-conditions, like the physical objects that sometimes constitute them, typically can be objects of justified belief, assertion, and knowledge.(8) Nevertheless, a realistically construed statement might be true or false regardless of what I, or anyone else very much like me, have reason to believe.(9)

With this said, it is clear that a semantic realist has the luxury of constructing a theory of understanding that is precisely the same as that of the verificationist, that is, linguistic competence is explained in terms of assertibility-conditions. Both may affirm that one understands a fragment of a language when one has learned under what conditions the statements in that, fragment are warrantedly as-sertible, that is, they are properly utterable. How does one learn this? Ask the vcrificationist. The semantic realist need not disagree. One learns by observing the linguistic habits of others and by noting the salient conditions of the world on the occasions when various expressions are used and different statements are made. This process typically results in a competent speaker who usually utters a statement only when there is sufficient evidence that it is true.

To see how this theory of understanding functions for the semantic realist, let us distinguish between knowing that certain conditions obtain and knowing that these conditions are the truth-conditions for a given statement. Consider a standard M-sentence:

(M) "Snow is white" means that snow is white.

In the case where the metalanguage in which (M) is stated contains the language of the mentioned sentence, knowing (M) is uncontroversial and trivial. The semantic realist owes an account of both how one learns the right hand side of (M) and also how one learns that the condition reported there is the truth-condition for "Snow is white." The semantic realist can certainly say that one knows that snow is white just in case one has the relevant familiarity with snow and whiteness - whatever the epistemologists tell us is necessary. Clearly, part of this story concerns evidence, as both the semantic realist and verificationist agree. Language learning takes us further and shows that statements like "Snow is white" have the property of being associated with a complex worldly structure which functions as its meaning. We observe sentences like this uttered by competent speakers when we have certain evidence about the temperature, consistency, location, and color of various things in the environment. Portions of the environment are apparently singled out by the use of "snow" and these portions are apparently attributed a property by the use of "is white." Through repeated exposure to linguistic usage by others in various evidential contexts and through repeated exposure to approvals and disapprovals when one uses these terms in, perhaps novel, contexts, one learns the proper evidence necessary for the proficient use of expressions of the language.

For statements about an object's color, we look to its appearance in the right sort of light. For statements about an object's mass we look to readings on a scale. For statements of character, we look to a person's actions. When we learn how to use expressions like "red," "weighs one gram," and "brave," we learn what conditions afford good evidence for making color, weight, and character judgments. We learn this by observing the environment. When we have certain visual experiences we say that something is red. Given other experiences, we say it weighs one gram, and from yet different experiences we say someone is brave.(10)

Not all of our observations regarding the use of various linguistic expressions, however, are limited to canonical evidential situations. We often observe speakers using these and similar expressions when the relevant objects are not in good light, not on a scale, not engaged in actions, or even visible at the moment. When speakers use expressions under such conditions, we learn that in order to be red, to weigh one gram, or to be brave, an object need not be currently in the appropriate canonical observational circumstances. We learn that many objects retain these properties even when we do not observe them at the time of assertion. By noting the evidence that attends the speakers' uses of the expressions of their language, we have reason to think that objects possess some of their properties independently of our observations. Some degree of objectivity for a wide range of objects and their situations can be extrapolated by the way competent speakers use the language on verificationist grounds alone. A verificationist theory of understanding yields a modest commitment to realism about the world, not yet semantic realism but ontological realism - a commitment to an ontology of objective, mind-independent objects and their properties. The assertibility-conditions, as complex and diverse as they are, belie the commitment of other speakers to independent objects.(11)

This story, so far, is acceptable to the verificationist. Language learning and understanding is explained solely in terms of justification-conditions. Significantly, though, there is no reason for the semantic realist to deny any part of this account. To understand a statement is to know under what, circumstances that statement is warrantedly assertible. What the semantic realist must deny is that the meaning of a statement is exhaustively given by such circumstances. The semantic realist, then, must provide some reason for positing truth-conditions as the meanings of statements, when it is admitted that linguistic behavior is driven primarily by evidence and not by truth.

This language learning warrants the positing of truth-conditions, and not merely assertibility-conditions, in that by paying attention to justification-conditions we learn that property possession is not solely a function of current, evidence. By understanding the nature of objective property possession, one understands that the state of objects is distinct from the evidential-conditions required for justified assertion. This yields part of what the semantic realist wants: knowledge of the distinctness of truth-conditions and assertibility-conditions. Learning that "Snow is white" has the snow's being white as its truth-condition is explained quite simply. First, we note the typical conditions under which "Snow is white" is asserted. Second, we note that competent speakers take some common, salient elements in these conditions to warrant the assertion of "Snow is white." These involve the way things look, the way things feel, and so forth. Third, and most crucially, we note that affirmations about the existence and character of objects presuppose a theory about the world according to which objects and properties are not, generally constituted by our possession of current evidence. This presupposed theory involves the assumption that these conditions are both independent of and accessible to us. Some of this linguistic behavior is implicit, as in the reference to things that are not currently observed. Some is explicit verbal behavior, as in the answers to questions concerning the unobserved existence of objects, their (lack of) dependence on us, and so forth. Ordinary speakers typically explain their rudimentary metaphysic in a way that reflects the transcendence of the typical constituents of truth-conditions. It is this presupposed theory that leads us to "infer" the existence of certain conditions thai are supposed by speakers to be associated with declarative statements. Thus, when we learn the assertibility-conditions for a statement, we come to know what transcendent conditions serve as the truth-conditions for that statement.(12)

That these truth-conditions are the meanings of statements is discerned by the apparent referential function of expressions like "snow" and "is white." They refer to a substance and a property respectively, the objectivity of which is learned in the way suggested above. If our statements refer to these things and the statements attribute the property to the substance, then the very meaning of the statement must be the condition under which the substance, indeed, has the property. Since this truth-condition is transcendent and because it is a complex of the substance and the property, the meaning for "Snow is white" must transcend evidence in the same ways snow and whiteness themselves do.

Indeed, there is something quite odd about the verificationist's suggestion that the meaning of a statement is its evidential conditions. On this alternative, "Snow is white" means that under certain circumstances normal observers have specific, characteristic experiences and these experiences are taken by the speaker to warrant the utterance of "Snow is white." This is puzzling because what appears to be a sentence about the stuff which, as a matter of fact, is cold and falls on Antarctica, turns out to be a sentence whose meaning does not involve snow at all, but (human) observers and evidence they might possess. Accordingly, the meaning does not involve the characteristics of snow, but of (our) experiences. Verificationism, as an account of meaning, then, misrepresents the intensional character of statements made when "Snow is white" is uttered. Such statements are not about me at all, but about snow. They are not about characteristically human visual or tactile experiences, nor about the situations that, give rise to these experiences. They are about the attributes of some physical stuff. "Snow is white" is quite neutral about the vast array of experiences cognizers might have when perceptually confronting snow.

A verificationist, no doubt, will question whether the semantic realist is warranted in claiming, in effect, that the background theory to which use points is, indeed, a bivalent theory. The verificationist agrees that this theory involves the objectivity of certain features of the world, so long as that objectivity involves no more than the truth of certain counterfactuals. The existence of my car in the garage is objective insofar as some variation on the "were I looking, I would see it" (or "were a more competent cognizer looking, it would see it") theme is true. Objectivity of this sort applies to decidable cases, but how is this to be extended to undecidable cases?

The semantic realist should resist the claim that, there is one type of objectivity that applies to decidable statements which must be extended to undecidable cases. The sense of objectivity and independent existence is the same in all cases, regardless of whether the statement in question is (un)decidable. The counterfactual gloss on the claim that my car is in the garage does not accurately capture the thrust of the background metaphysical theory. One claiming that objects exist and have properties objectively is not thereby committed to some kind of counterfactual explication of that objectivity. Whether this gloss is accepted by a speaker depends on whether the speaker thinks that the relevant matter is decidable. If so, then dearly the gloss is acceptable, at least, as a partial account of the epistemological facts associated with objectivity, even if not as an account of the nature of objectivity. If the speaker thinks that the statement is undecidable, then the gloss should be rejected.

To insist on the counterfactual gloss which must be extended by appeal to the investigative powers of a superior intelligence is to insist on an epistemological account of objectivity, which is precisely the bone of contention between the semantic realist and the verificationist. Worse, this particular epistemological account involves an element which is itself undecidable by the verificationist's own lights, that is, the embedded counterfactual. So, pressing a counterfactual account of objectivity by the verificationist at best begs the question, and at worst is self-defeating.

Since a comprehensive semantic theory must capture the details of actual linguistic practice, no amount of desktop theorizing can settle the issue of whether the background theory presupposed by ordinary speakers is bivalent or not. Given that semantic realism can embed the verificationist's theory of linguistic understanding and thereby permit an account of how one comes to understand undecidable statements, there is no longer a purely theoretical obstacle for semantic realism. The issue between semantic realism and verificationism is now an empirical issue: Is the implicit background metaphysical theory that informs ordinary linguistic practice bivalent or not? I submit that for at least a wide range of philosophically naive speakers, the theory is bivalent. The verificationist is right that most behavior does not allow us to discern the subtle details of the background theory, but explicit affirmations about the natures of items mentioned often tip the balance in favor of a bivalent background theory. It might also turn out that there is no single theory that fits the implicit and explicit linguistic behavior of all speakers. Perhaps for some speakers the background theory is bivalent while for others it is not. Though semanticists do not dwell on this prospect, this outcome should be no more surprising than the fact that the linguistic behavior of speakers of different languages is best described by different grammars. In the same way form and content, and theory and experience, are not thoroughly independent, so theory and meaning are not.

The objectivity of worldly situations plus the referential nature of our statements justify the semantic realist's contention that meanings are not justification-conditions; they are the worldly situations that function as the relevant truth-conditions for a given statement. However, the objectivity of worldly situations alone is sufficient to undermine the semantic antirealism of Dummett, who claims that (1) the traditional issue of ontological realism can be translated into the issue of the proper form of a theory of meaning and that (2) denying truth-conditional semantics is sufficient for denying the traditional doctrine of ontological realism - the evidence-transcendence of worldly situations. By modus tollens reasoning, defending the evidence-transcendence of worldly situations is sufficient for defending truth-conditional semantics. Anti-realists like Dummett cannot be concerned only with the semantic issue of whether statements have the property of being relevantly related to a possible situation or truth-condition to the exclusion of the ontological issues of the objectivity of worldly situations. If they are, then their clam to deal with the traditional issues of realism via semantics is nullified. Thus, while the referential character of our statements is required to undermine other forms of anti-realism, it is not necessary to undermine Dummett's.

I have been characterizing both how we come to know that snow is white and how we come to know that snow is white is the meaning for "snow is white." Both require the proper kind of epistemological contact with a particular state of the world. Knowing that snow is white involves, among other things, an adequately justified belief about snow and whiteness. Nothing mysterious there. To know that this worldly condition is the meaning of "snow is white," in similar fashion, involves possessing an adequately justified belief about the way speakers use expressions like "snow" and "is white" and this adequately justified belief involves discerning the presupposed metaphysics that attends their use of these expressions and makes sense of them.

The major burden of this account of linguistic competence is simply to show both that a truth-conditional theory of meaning and a verificationist theory of understanding are compatible and that, linguistic competence is not mysterious, even if truth-conditions are transcendent. One consequence is that if semantic realism is supplemented with a verificationist theory of understanding, the result is that meanings are not completely internal to the speaker, or even a group of speakers. Meanings are, partially, beyond us and understanding stands in stark contrast to meaning. Even if linguistic understanding is internal and even if what drives our linguistic behavior is wholly internal, it does not follow that meanings themselves are internal. In the same way, the plausible clam that a theory of understanding must be primarily an epistemological theory, does not entail that a theory of meaning must likewise be an epistemological theory. The only hard and fast requirement is that meanings must be accessible to language users. Semantic realism as a two-part theory of meaning, satisfies this requirement and is justified by the very conditions of evidence and linguistic usage on which verificationists are so keen to focus. Thus my first two major clams are defended. My first claim is that there is no fundamental incompatibility between a robust truth-conditional theory of meaning and a verificationist theory of understanding. Critics are mistaken when they attack semantic realism on the basis of the alleged inscrutability of our linguistic competence with statements whose meanings are robust truth-conditions. Semantic realism can absorb the whole of the verificationist's theory of understanding lock, stock, and barrel and incur no theoretical damage.

My second claim is that linguistic use justifies truth-conditional semantics. The details of justified assertion point to an assumed metaphysics of independently existing objects and properties, which themselves constitute truth-conditions. What makes sense of apparently referential expressions is a background theory of transcendent objects, properties, and truth-conditions that serve as the meanings of names, predicates, and statements. Thus a truth-conditional theory of meaning is itself justified by the whole of the assertibility-conditions for statements: the theory that makes sense of the assertions, as well as the more obviously sensory bits of evidence one has in various circumstances. So, not only is a truth-conditional theory of meaning compatible with a verificationist theory of understanding, but a verificationist theory of understanding justifies a truth-conditional theory of meaning.


The "The Theories are Identical" Objection. A natural objection to this development of semantic realism is that the whole enterprise is misconceived because there are not two separate theories to supply, a theory of meaning and a theory of understanding, but one. The debate in semantics is only a debate about the way competent speakers understand statements precisely because, as Dummett says:

[A] theory of meaning is a theory of understanding; that is, what a theory of meaning has to give an account of is what it is that someone knows when he knows the language, that is, when he knows the meanings of the expressions and sentences of the language.(13)

This passage is ambiguous as it stands. If, on the one hand, it enjoins us to recognize that a theory of meaning should tell us about the things with which we must be acquainted when we have semantic knowledge, the semantic realist agrees. Truth-conditions are "what it is that someone knows [and knows about] when he knows the language." If, on the other hand, Dummett's point is (as I suspect) that semantic theory is nothing but a theory of semantic epistemology in which only epistemological notions figure, then Dummett poses a genuine challenge to semantic realism.

Adjudicating between the semantic realist and Dummett hinges on exactly how plausible it is to think of meanings ontologically. If semantic realism is correct, meanings are things with which one comes into either direct or indirect epistemic contact in understanding a language. Meanings are entities which are known (about) or grasped and associated with various sentences. It is quite natural to think that a theory about the nature of these entities will utilize notions distinct from a theory about how one knows about those entities. Distinguishing a theory of meaning from a theory of understanding is quite natural, then, on such a view. However, if meanings are not things, this bifurcation is less natural. If meanings are not objects in our ontology, then perhaps talk of meaning is merely structured talk about linguistic practice. Sameness of meaning is sameness of role played in the complicated structure of this practice. Hence, speaking of how one engages in such a practice is speaking of how one knows how to do various things with words. Once one has provided an account of understanding, one has said all that can be said about meanings and, thus, one has automatically provided a theory of meaning. So, on this alternative, there is nothing besides a theory of understanding.

At the initial stages of semantic inquiry, it is not at all obvious whether the theory of meaning ought to be identified with the theory of understanding. This is a substantive theoretical question which must be decided in the same way other theoretical questions are decided. We must determine which general theoretical program is preferable. If there are powerful reasons for thinking that meanings are elements in our ontology, then there are good reasons for thinking that the theory of meaning and the theory of understanding are very different theories utilizing very different sets of central concepts. If the weight of other considerations, however, tilts the balance against an ontological view of meanings, then the theory of meaning may be identified with the theory of understanding. Since this issue hinges on prior matters, this objection, in itself, is not effective against semantic realism. The most that can be said prior to an investigation of the comparative worth of competing semantic theories is that a theory of meaning must be amenable to an adequate theory of understanding. Neither semantic realist nor verificationist semantics are disqualified on this count. Identifying the theory of meaning and the theory of understanding can, at most, be a consequence of adopting a particular semantic program. It cannot by itself be an objection to semantic realism.

It is tempting to think that the plausibility of use theories of meaning automatically lends credence to the collapse of theories of meaning and understanding. Use theories do not, however, entail this collapse. If the use that is meaning has a significant social character and is not merely a private use by individual speakers, the bifurcartion of theories of meaning and understanding remains quite natural. According to such a use theory, meanings are not as ontologically independent of language users as Forms, senses, or robust truth-conditions, but they are features of the relevant linguistic community which are no less external to a particular speaker and no less distinct, from a given speaker's evidence regarding proper rise. The presence and nature of communal practices are no more a function of the evidence one speaker has for thinking that the community uses a term in a given way than truth-conditions are a function of a speaker's evidence for asserting a statement.

These remarks undercut Dummett's suggestion that traditional metaphysical issues about realism can be recast as issues regarding language.(14) He claims that realism about any of these subject matters can be undermined by showing that semantic realism is an inadequate semantic theory for the relevant portion of discourse and that all realisms can be undermined by showing that semantic realism is inadequate for any area of discourse. In effect, Dummett suggests thai semantics is prior to metaphysics, since metaphysical issues can be settled by settling the more basic semantic issues. The argument of this section supports my third major claim: that the recasting of metaphysical issues into semantic issues is misguided. In the end, metaphysics is prior to semantics. If it is plausible to think that there are Forms, senses, or robust truth-conditions and it is plausible to think that our discourse makes reference to these things, then the form of our semantic theory is partly determined by the need to distinguish the theory of meaning from the theory of understanding. If the metaphysics of Forms, senses, or robust truth-conditions are all implausible, then there may be no choice but to identify the theory of meaning and the theory of understanding. Whatever the correct form for a semantic theory, the metaphysics must first be settled. Otherwise, the claim that a theory of meaning just is a theory of understanding simply begs the question against theories like semantic realism, and exposes no flaw in them. Question begging can be avoided only if there is some principled objection to ontological theories of meaning. It is alleged, for example, that truth-conditional theories of meaning face an insurmountable problem regarding our mastery of undecidable statements. It is to this objection that I now turn.


Undecidables and Explanatory Simplicity. Decidable statements are those for which there is an effective procedure for determining their truth value in every circumstance.(15) Ignorance of the truth value of a decidable statement is always a matter of "in practice" rather than "in principle." Undecidable statements, those for which there is no such effective decision procedure, are supposed to pose the most pressing problem for semantic realism.

Precisely which statements are undecidable depends on the relation between truth and justification. Critics of semantic realism adopt a relatively weak account of decidability, so that most ordinary statements turn out to be decidable. Dummett, for instance, restricts undecidables to quantification over infinite domains, counterfactual conditionals, and statements about inaccessible regions of spacetime. His account is weak enough that inductive evidence is sufficient for conclusive justification for some statements. According to this standard, then, most physical object statements count as decidable. The line between these categories of statements, however, might be placed so that decidable statements are only those for which truth is entailed by justification. There are few such statements. Perhaps, only some statements about introspective states. The objection from undecidables does not depend on any particular way of demarcating the class of undecidable statements.

Semantic realism faces two objections from undecidables. The first is that if semantic realism is true, there is some class of statements whose truth value outstrips our ability to effectively determine that value, regardless of how one draws the distinction; we never have sufficient access to the truth-conditions of these statements to determine their truth value. If we have insufficient access to the truth-conditions to determine truth or falsity, and meanings are truth-conditions, semantic realism seems to entail that we have knowledge of things we never fully confront, that is, the truth-conditions for undecidable statements. Thus semantic knowledge, which involves the knowledge of which worldly situations are the truth-conditions for a given statement, is impossible. In short, the semantic realist cannot give an adequate explanation of how we could gain semantic competence with undecidable statements.

This is, at most, a limited objection from undecidables. Decidable statements are supposed to present no problem for either semantic realism or verificationism because their truth-conditions are not damagingly transcendent. Conclusively verifiable statements leave no epistemological gap between their evidential- and truth-conditions. For example, if we are infallible regarding some first-person utterances, then these are statements for which there is no epistemological gap between assertibility-conditions and truth-conditions, even if these conditions are distinct. The truth of these statements need not be a function of evidence, only that, given the evidence, there is no possibility of error.

Semantic realism's adequacy for decidable statements has two consequences. First, semantic realism is not incoherent. If it is only for the undecidable fragment of a language that semantic realism fails, then the program is feasible for languages limited to decidable statements. Second, any argument which hangs on semantic realism's inability to handle undecidable statements cannot properly show that the whole of an acceptable semantics for a natural language is verificationist. A doubly bifurcated theory would do: semantic realism, as developed above, for the decidable fragment and wholesale verificationism for the undecidable fragment. Such a theory is rather like a theory of science which is (ontologically) realist about observables but (ontologically) anti-realist about unobservables. Whether a doubly bifurcated theory is preferable to a unified verificationist theory depends on the plausibility of the requirement that a simpler theory is preferable to an equally adequate but more complex theory, and whether a unified verificationist theory really is the equal of the bifurcated theory. To the extent that semantic realism for decidables is justified by the considerations of section 1, the unified verificationist theory is not the equal of the bifurcated theory and so a limited application of semantic realism is justified, even if this first objection from undecidables cannot be answered successfully.

This objection can be answered, however. To the extent that this objection assumes that understanding a statement, S, requires its verification, it is implausible. In the same way sophisticated verificationism allows for the acquisition of an understanding of S even though S is never actually verified, the semantic realist can allow that one might understand S even when one has not actually been epistemically intimate with the state of affairs which is S's truth-condition. So long as we can know what S's truth-condition is, and so long as we can know what state of affairs is required for S to be true, then the objection fails and the principled inaccessibility of S's truth-condition is a red herring.

More importantly, semantic realism needs an account of how one could understand the components of S and understand how they work together to yield a statement that requires a certain condition in order to be true. The general appeal to a verificationist theory of understanding suffices for undecidables as well as it does for decid-ables. One gains understanding of a statement by gaining knowledge of the conditions for appropriate assertion. Consider:

(T) There are infinitely many twin primes.

We do not know whether (T) is decidable in Peano Arithmetic. Suppose that it is undecidable quite generally. It follows that there is no effective procedure for determining (T)'s truth value. Yet, we manifestly understand (T). Dummett thinks that the realist must now tell a story about how an ideal cognizer, who could survey an entire infinite domain, would come to know whether (T) is either true or false.(16) But the semantic realist can eschew any appeal to a being with powers very different from our own. It would not help explain how we gain and manifest our competence with such statements. One who genuinely understands the claim that there are infinitely many twin primes and knows that it is undecidable may nevertheless manifest competence by refusing to baldly assert that there is such an infinite sequence and by refusing to assert that there is not. Such a refusal is conditioned on the speaker's understanding of the undecidability of the matter for us. A speaker can enter the relevant linguistic practice through decidable cases in precisely the way verificationists already admit for undecidables. We learn that 3 and 5, 11 and 13, 17 and 19 are twin primes by virtue of each being prime and separated by only one natural number. One has the concept of infinity when one knows that whenever a series is infinite and the question is put "Have we come to the end?" the answer is "No." Simple observation of the conditions in which others assert, deny, and refuse either to assert or deny is sufficient to teach one that this is the appropriate response. There is no problem about gaining the understanding of this question and the needed answer. There is no need to appeal to an induction based from decidable cases to the undecidable. There is no leap from finite cases to infinite cases. Undecidables pose no greater problem for the semantic realist than the verificationist. Verificationists explain the use of undecidable statements and since the semantic realist can adopt their account of this understanding wholesale, the semantic realist has a story to tell about how we gain competence with undecidables and the first objection from undecidables has been answered.

The second objection from undecidables is that semantic realism suffers from an embarrassment of explanatory riches. How many resources are needed to explain our competence with undecidables? If it turns out that all the relevant semantic epistemology can be done by appeal to knowledge of justification-conditions alone, then simplicity constraints make verificationism preferable to semantic realism, since semantic realism incorporates a gratuitous appeal to truth-conditions. The problem is not that semantic realism has nothing to say about our knowledge of undecidable statements; rather, it says too much. All the data regarding linguistic understanding is adequately explained by attributing to us knowledge of justification-conditions alone. So, truth-conditions are semantic danglers. Semantic realism is insufficiently simple.

Two semantic realist responses are at hand. First, this objection moves the issue from some alleged intrinsic difficulty with semantic realism to whether there are any general theoretical reasons for appealing to transcendent truth-conditions. If there are, then semantic realism poses no additional burden on a more general philosophical theory. Of course, if we restrict our scope severely enough, we find that certain things enter explanations which are not strictly necessary to explain those phenomena. We do not need to postulate physical objects to explain my percepts. A demon and my spirit will do just fine, but we justify the postulation of physical objects on the basis of wider considerations. Quite often we justify the introduction of certain elements into a theory on the basis of how those elements contribute to the whole theory, not only to the explanation of one phenomenon. It might be that one set of phenomena can be explained by mechanism A while a second is explained by mechanism B. It might also be that mechanism C provides a less elegant explanation of their respective subject matters than either A or B, yet nonetheless C is preferable because it provides a more elegant single explanation of everything. What may lead us to accept the comprehensive explanation as the correct one is not how it fares in the narrow contexts when compared with the other, more limited, explanations. We accept it, if the overall theoretical package is better. Thus while simplicity considerations may properly guide theory construction, it is not legitimate to invoke them in every context. Sometimes less elegant local explanations should be tolerated if they allow for a more elegant global theory. Therefore, the significance of semantic realism's alleged lack of simplicity can be evaluated only in the context of a more general philosophical verdict about the nature of truth-conditions. The context of semantic theory may be too narrow and limited in scope.

Second, and more substantively, semantic realism itself already provides the semantic grounds for ascribing robust truth-conditions to statements. These grounds arise from the specific content of the conditions of justified assertion for various kinds of statements. These conditions highlight the referential character of names and predicates as well as a background theory involving the objectivity of these conditions. Not only is this objection one that cannot be properly evaluated in the absence of evaluating more comprehensive philosophical theories, it is already obviated by semantic realism's rationale for positing robust truth-conditions. Until it is argued on independent grounds that undecidable statements, in virtue of their undecidability, do not even purport to have the same referential character as their decidable cousins, there is no reason at all to think that undecidable statements do not have transcendent truth-conditions as their meanings. In the end, undecidables pose absolutely no threat to semantic realism.


Trouble with Truth. Without the epistemological objection from undecidables, semantic realism is threatened only if there is some problem with its metaphysical doctrine regarding truth-conditions. In particular, skeptical difficulties seem to arise for theories that embed a notion of transcendent truth-conditions. The trouble supposedly involves the epistemic gap between evidence and truth: in some range of cases, evidence is insufficient to insure truth. A critic might object that it is preferable either to adopt an alternative epistemic account or to give up truth altogether. Either alternative, says the critic, is better than being saddled with skeptical problems. Knowledge of truth-conditions is elusive at best, and unobtainable at worst. According to this line of criticism, skeptical difficulties undermine the very possibility of propositional content.

Skeptical difficulties are not so near at hand, however. Semantic realism's metaphysical component does not itself entail the existence of an epistemic gap between justification and truth, nor does it entail that, no amount of evidence is sufficient to insure the correctness of the belief for which it is evidence. Truth-conditions are not, by their nature, special cases of justification-conditions, so the two conditions belong to distinct classes. This is perfectly compatible with superlative justification guaranteeing the truth of some propositions. This doctrine is even compatible with coextensiveness between truth and a given high degree of justification. Semantic realism's metaphysical doctrine serves to highlight the fact that such coextensiveness does not entail identity and that evidential conditions are not constitutive of truth-conditions. Semantic realism, then, entails no inevitability of skeptical problems. More is needed than just the semantic realist's metaphysical doctrine.

Though skeptical problems are not inevitable for semantic realism, the evidential conditions that typically warrant assertions do not guarantee the truth of those assertions. Plainly, both semantic realists and strict verificationists agree that justification sometimes fails, that error sometimes results even in the face of strong justification. The parties describe this failure differently, but both recognize it. Semantic realists say that at certain times we are justified in believing something that is false. Verificationists say that at certain times we are justified in believing something and at other times we are justified in believing its negation. The evidence for each is the same: justification has been overturned sometimes. For the semantic realist, even now we might be justified in believing some false propositions. For the verificationist, we do not now know whether our current justification will continue to prevail. The nature of truth plays no crucial role in generating a problem about the failure of justification.

Verificationists claim to have a ready explanation for our knowledge, be it semantic or otherwise, that the semantic realist does not. If truth is exhaustively characterized by epistemic notions like coherence, simplicity, explanatory power, and so forth, knowledge is certainly possible, however rare it may be. Truth is defined by justification-related concepts and so cannot outstrip justification in any systematic way.

This suggestion, though, affords no solution to skeptical problems. Suppose we are confronted now not with the skeptical challenge - the challenge to demonstrate one's knowledge to a skeptic's satisfaction - but with skeptical hypotheses, like the hypotheses that really there are only disembodied spirits who are deceived by a deceitful demon, or that I am a brain in a vat, and so on. When we compare these hypotheses with other hypotheses like realism about physical objects, phenomenalism, idealism, and so forth, an ontological realist can claim that our experiences make it seem as though there really are tables and chairs, people and cars, cliffs and seas. Experience alone surely does not adjudicate between realism about physical objects and some competitors such as phenomenalism and idealism, to take two examples, which disagree over the nature of tables, chairs, people, cars, cliffs and seas. However, these experiences do offer prima facie warrant for any hypothesis claiming that these objects exist. Those with no experiences which seem to involve disembodied spirits and deceitful demons should give no credence to the Cartesian skeptical hypothesis. Those with no experiences of being a brain in a vat should give no credence to the relevant vat hypothesis. We have no evidence for any of the skeptical hypotheses taken as theories about the nature of the world. All we know is that the skeptical hypotheses are consistent with all the epistemic warrant we might have for the other hypotheses. However, if the mere fact that a skeptical hypothesis is consistent with all the data one possesses for a given belief is sufficient to undermine the justification for that, belief, then there is nearly no justification at all, on any theory of truth. For nearly any interesting proposition, our justification for believing that proposition is consistent with its falsity.

If the warrant for an hypothesis is undermined simply because the evidence is consistent with bizarre hypotheses, the verificationist, who normally is thought to fare better in these matters, is equally threatened. The verificationist thinks that some epistemic conditions warrant some statements while warranting the denial of others. If, however, such warrant is undermined because it does not entail the hypothesis for which it is warrant, then the verificationist is no more able to defeat the skeptic than is the believer in transcendent truth-conditions. Plausible epistemic theories of truth tighten the relation between justification and truth by saying that certain internal, epistemic conditions in the ideal limit of inquiry entail truth.(17) But the only inquirer for whom this entailment between justification and truth is of any value is one placed in the ideal circumstances. Entailment is of no use, if the entailing conditions do not obtain and they rarely obtain for we who are not at the ideal limit of inquiry.

The semantic realist, having distinguished theories of meaning front theories of understanding, can agree with some sophisticated verificationists that norms of evidence and canons of reasoning play a constitutive role in linguistic understanding. Certainly one has inadequately mastered a concept if one does not understand the proper conditions that warrant its application. This constitutive role for evidence is not, however, inconsistent with semantic realism's claim that truth-conditions are robust, objective worldly structures. These norms and canons are just not infallibly truth-directed. Since all parties agree that justification can be overturned, the semantic realist is not disadvantaged by this admission. Once theories of meaning and understanding are clearly separated it is easy to see that epistemic factors naturally play a large role in the theory of understanding and, so long as truth-conditions are epistemically accessible, there is no need for epistemic factors to play a constitutive role in the theory of meaning. While no party avoids the skeptical worry that for all our best epistemic efforts we might nonetheless be wrong, both major parties avoid the need to take seriously various skeptical hypotheses. Thus epistemic accounts of truth are no better at saving us from skepticism than theories in which robust truth-conditions play a major role. The advantages of epistemic accounts are present in semantic realism's theory of understanding. Thus my fourth major thesis: skepticism is a double-edged sword that cannot be used safely against semantic realism by verificationists. There is no trouble with truth that is peculiar trouble for the semantic realist. Any trouble presents itself to the semantic realist and verificationist alike.


Conclusion. Semantic realism's separation of theories of meaning and theories of understanding yields an externalist account of meaning. Surprisingly, Wittgenstein's rule-following considerations play into the semantic realist's hands. According to Wittgenstein, no finite amount of data or justification suffices to determine the meanings of terms or statements. The semantic realist embraces this and suggests that Wittgenstein's observation supports the claim that meanings are external to speakers and not simply a matter of the justification one has for making an assertion. Were justification sufficient to determine meanings, sameness of meaning between speakers would be relatively unproblematic and some of Wittgenstein's worries about meaning would not arise. Furthermore, the semantic realist can also maintain that the inference to meaning skepticism, made by Kripke's Wittgenstein, is valid only if one conflates theories of meaning and theories of understanding.(18) Once the distinction between these theories is clear, we can see that knowledge of meanings is no more problematic than knowledge of anything else which is outside of our head. Semantic competence involves the ability to use expressions of the language in the right sort of way, and knowledge of truth-conditions involves the correct association of external conditions with statements that are known to be justified in certain salient ways. So long as there is access to meanings, their external features are not troublesome and statements can possess determinate meanings even if this determinate meaning cannot be straightforwardly extracted from actual linguistic practice. Thus, semantic realism explains both Wittgenstein's observation that no amount of ostension fixes the meaning or referent of a term and Putnam's claim that speakers on Earth and Twin Earth agree in their internal understanding of "water" even though the speakers from the different planets mean different things by "water."(19) Both cases highlight, in rather vivid ways, the semantic realist's claim that the theory of meaning is distinct from the theory of understanding. Wittgenstein and Putnam simply point to special cases of what the semantic realist suggests is quite general, that is, meanings are not epistemic conditions. If meaning is woefully underdetermined by the evidence any speaker or group of speakers can bring to bear on any given set of statements, and if meanings are not in the head, then meanings cannot be satisfactorily characterized by way of evidence alone. The core notion of a theory of meaning should not be epistemic. In contrast, our successful participation in social linguistic practices can be described in terms of the evidential conditions that we associate with a statement.(20) Thus Wittgenstein's and Putnam's considerations support semantic realism.

Not only is a truth-conditional theory of meaning of the sort sketched above compatible with a verificationist theory of understanding, careful attention to the justification-conditions for statements provides the theoretical justification for claiming that truth-conditions are the meanings of statements. Attempts to collapse the two theories hide metaphysical issues that must be settled first. Thus Dummett's attempt to transform traditional metaphysical issues into issues of the proper form for a theory of meaning are misguided. Undecidable statements pose no particular difficulty for semantic realism, precisely because truth-conditional theories of meaning are compatible with verificationist theories of understanding for undecidables. Dummett mistakenly thinks that such a theory of understanding places severe constraints on the nature of truth-conditions. Finally, robust truth-conditions generate no more difficulties regarding skepticism than do truth-conditions understood in epistemic terms. I submit that there are no good verificationist reasons for thinking that statement meanings are anything other than good old-fashioned truth-conditions.(21)

The University of Western Australia

1 Michael Devitt, Realism and Truth (Princeton: Princeton University Press, 1984).

2 Wittgenstein defended truth-conditional semantics in the Tractatus, while Donald Davidson provided his classic statement of this theory in his "Truth and Meaning," Synthese 17 (1967): 304-23. He has set forth his views more recently in the John Dewey Lectures, "The Structure and Content of Truth," The Journal of Philosophy 87 (1990): 279-328. Davidson thinks that the "radically non-epistemic" account of truth my semantic realist wants is trivial because there is only one fct to which statements could correspond. Peter Menzies has adequately refuted Davidson's (Frege's/C. I. Lewis's/Strawson's/Quine's) argument for this in his "A Unified his Account of Causal Relata," Australasian Journal of Philosophy 67 (1989): 59-83. Davidson claims that we must incorporate human interests, intentions, desires, and beliefs into the nature of truth, but he actually argues for the much weaker claim that the interpretation of someone's utterances requires an appeal to truth, the determination of which requires reasonable conjectures about the speaker's beliefs, desires, and intentions.

3 Paul Horwich argues that "'Snow is white' is made true by the fact that snow is white" is compatible with his minimalism about the truth relation. See Paul Horwich, Truth (Oxford: Basil Blackwell, 1990), 110-12. Nothing I argue in this paper casts any doubt on this claim. While Horwich is concerned with the nature of the truth relation and not, the nature of truth-conditions, I am concerned with the nature of truth-conditions and not the truth relation. Thus a minimalist about the truth relation can nevertheless hold that truth-conditions are robust objective worldly structures that render statements true via a trivial relation. In what follows I argue only that truth-conditions are the meanings of our statements. Horwich denies a semantic role for truth-conditions because he does not adequately distinguish the (quasi-) metaphysical theory of meaning from the more straightforwardly epistemic theory of understanding. Thus he mistakenly thinks that the semanticist must choose between truth-conditional and use theories. Even if a trivial truth relation can play no semantic role it does not follow that substantive truth-conditions cannot. Since no choice must be made between truth-conditional theories of meaning and use theories of understanding and linguistic competence (since they are compatible, as I will argue below), a minimalist may be a semantic realist as well.

4 Hilary Putnam pioneered work showing that a semantic theory cannot be adequate if it deals only with what is "in the head" in "The Meaning of 'Meaning'," in Mind, Language and Reality: Philosophical Papers (Cambridge: Cambridge University Press, 1975), 2:215-71. First published in Language, Mind and Knowledge: Minnesota Studies in the Philosophy of Science 7, ed. K. Gunderson (Minneapolis: University of Minnesota Press, 1975).

5 This is explicitly presumed by Crispin Wright in the introduction to Realism, Meaning and Truth (Oxford: Oxford University Press, 1987), 1-41; esp. 10-23.

6 Michael Dummett, Truth and Other Enigmas, (Cambridge: Harvard University Press, 1978), 17-18.

7 Dummett, Truth and Other Enigmas, xxxvii.

8 This analogy between the common sense realist's understanding of physical objects and the semantic realist's understanding of truth-conditions begs no questions against the relevant antirealists. The present context is one of merely exhibiting how the semantic realist can understand truth-conditions so that there is no obvious difficulty in allowing that meanings are truth-conditions. Argument for why one might hold to semantic realism follows. The justification for positing the objects of common sense and robust truth-conditions turns out to be similar; they are posited to explain certain features of our evidence.

9 This independence from epistemic relations with the world is difficult to characterize. If it is necessary that all facts are knowable, then truth is extensionally equivalent to whatever could be known by an omniscient being. If it is possible that there be facts which are in principle unknowable, then there is no such extensional equivalence.

10 All of these remarks are subject to the various pragmatic conditions which partially determine when it is appropriate to assert "Jones is brave."

11 Furthermore, this commitment is enhanced by the speaker's own experience of the world, characterized as it is by a wide array of regularities. It is easy to reconcile one's own regular experience with the linguistic behavior of others. The learner's own regular experience is easily explained if there are objective enduring realities, and the linguistic behavior of others is sensible on the assumption of this modest ontological realism.

12 I do not pretend that these few remarks adequately settle all age-old questions about realism. I merely sketch a way of defending semantic realism against the attacks of semantic verificationists like Dummett and Wright.

13 Michael Dummett, "What is a Theory of Meaning?" in Mind and Language, ed. S. Guttenplan (Oxford: Oxford University Press, 1975), 99; cf. also Michael Dummett, Frege: Philosophy of Language, 2nd ed. (London: Duckworth, 1981), 92.

14 Dummett seems to think that rejecting semantic realism undermines traditionally realist metaphysical doctrines. See Michael Dummett, "What Does the Appeal to Use Do for the Theory of Meaning?" in Meaning and Use, ed. A. Margalit, (Dordrecht: D. Reidel, 1976), 135; Michael Dummett, introduction to The Logical Basis of Metaphysics (Cambridge: Harvard University Press, 1991).

15 This is Dummett's account of undecidability in his "What is a Theory of Meaning? (II)," in Truths and Meaning, ed. G. Evans and J. McDowell (Oxford: Clarendon Press, 1976), 80-2.

16 Michael Dummett, "What is a Theory of Meaning? (II)," p. 98-9.

17 Putnam adopts Peirce's account of truth: truth is what a comprehensive theory of the world says in the ideal limit of inquiry. He contrasts this with a transcendent theory of truth. See Hilary Putnam, Meaning and the Moral Sciences (London: Routledge & Kegan Paul, 1978), 1, 36, 125; Hilary Putnam "Vagueness and Alternative Logic," in Realism and Reason: Philosophical Papers (Cambridge: Cambridge University Press, 1983), 2:272. The gloss that the relevant circumstances are the ideal limit of inquiry is necessary, since it is clear from the history of overturned justification that not just any old amount of justification will do. Any coherence theory of truth must likewise build some ideality conditions into the account to avoid relativism about truth.

18 Saul Kripke, Wittgenstein on Rules and Private Language (Cambridge: Harvard University Press, 1982).

19 Putnam's classic defense of meaning externalism is "The Meaning of 'Meaning'."

20 I view Putnam's work as playing into the hands of the semantic realist. His early work justifies the externalist claims regarding the nature of meaning while his later work justifies internalist claims about understanding.

21 I am indebted to Andrew Brennan, Michael Detlefsen, Kosta Dosen, Aron Edidin, Richard Kirkham, Michael Levine, and Cindy Stern for their valuable comments on earlier drafts of this paper.
COPYRIGHT 1995 Philosophy Education Society, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1995 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Shalkowski, Scott A.
Publication:The Review of Metaphysics
Date:Mar 1, 1995
Previous Article:Some peculiarities about musical aesthetic qualities.
Next Article:Common sources for the semiotic of Charles Peirce and John Poinsot.

Terms of use | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters