SEMANTIC COMPETENCE AND TRUTH-CONDITIONAL SEMANTICS

SEMANTIC COMPETENCE ANDTRUTH-CONDITIONAL SEMANTICS

Davidson approaches the notions of meaning and interpretation with the aim of characterizing semantic competence in the syntactically characterized natural language. The objective is to provide a truth-theory for a language, generating T-sentences expressed in the semantic meta-language, so that each sentence of the object language receives an appropriate interpretation. Pro-ceeding within the constraints of referential semantics, I will argue for the viability of reconstructing the notion of linguistic meaning within the Tarskian theory of reference. However, the view proposed here involves a re-vision of Davidson’s conception of the object of semantic investigation. Taking (idealized) language-theories as the proper object of semantic charac-terization, provides solutions to outstanding problems in Davidson’s views, better approximates the practice in standard model-theoretic semantics, and incorporates the elements of semantic competence sought for in traditional theories of lexical analysis. Sources of evidence beyond those emphasized by Davidson will be invoked in order to allow for the selection of interpretive T-sentences. In the final section, possible Quinean objections will be consid-ered.

1. Background for Meaning as Truth-conditions“To give truth-conditions,” says Davidson, “is a way of giving the meaning of a sentence.”1 He thus takes up the challenges of a recurrent, appealing, but difficult perspective on linguistic meaning. Considering only this fragment, he appears to echo a long line of similar claims in the analytic tradition. Moritz Schlick, for example, writing in the mid-thirties, held that when we ask for the meaning of a sentence, “we want a description of the conditions under which the sentence will form a true proposition, and those which will make it false”2 But Schlick goes on to equate truth-conditions with verifica-tion-conditions, while Davidson is not tempted in the least to make this step.

Slightly revised from Erkenntnis 28, 1988, pp. 3-27.1. Davidson 1967, “Truth and meaning,” p.24.2. Schlick 1936, “Meaning and Verification,” see Schlick, 1970, p. 100.

2

The contrast serve to highlight some of the ambiguities of the notion of truth-conditions in the analytic tradition.

That we see Schlick’s formulations as ambiguous—he equates truth-conditions both with verification-conditions and with instructions for useis due to the influence of other philosophers, and Tarski’s work on behalf of the concept of truth is quite important here. Schlick saw himself as following Wittgenstein who held in the Tractatus that “to understand a proposition means to know what is the case if it is true.”3 There are surely important similarities among the three quotations here. However, Schlick, under Wittgenstein’s influence, but lacking an alternative to a verificationist con-ception of truth-conditions, reads verificationism into Wittgenstein. David-son, under the influence of Tarski’s semantic conception of truth, in contrast, need not take a view according to which every statement of truth-conditions must be a statement of verification-conditions. Yet, it is open to him to hold that in some cases evidential relations will “verge on being constitutive of meaning.”4

Looking back to Frege’s influence, though the concept of truth is crucial to his semantic work, one finds that he leaves it undefined. “It seems likely,” he says in the Logical Investigations, “that the content of the word ‘true’ is sui generis and undefinable.”5 Such a view could not and did not appeal to the empiricist element in subsequent analytic philosophy. Thus, though Frege did not equate truth-conditions with verification-conditions, the impressive appeal of using Frege’s work on the quantifiers and truth-functional connec-tives in an account of meaning tended to lend plausibility to verificationism. Only after Tarski does it become possible to retrace the steps of theory and attempt a non-verificationist version of truth-conditional semantics. As might be expected, Davidson’s views along these lines have come into question at a fundamental level where there is serious doubt concerning his use of the concept of truth.6 Sentences will count as meaningful even where we cannot decide on their truth-values, non-logical vocabulary will count as meaningful in spite of room for doubt or questions regarding its reference or satisfaction conditions. This much realism comes along with Tarski’s treatment of truth.

Since Tarski’s semantic conception of truth has been primarily an influence upon the formal semantics of logical theory and special-purpose

3. Wittgenstein 1922, Tractatus Logico-philosophicus, 4.024.4. Davidson 1980, “Toward a Unified Theory of Meaning and Action,” p. 7.5. Frege 1977, Logical Investigations, p.4.6. Thus Dummett’s doubts in “What is a Theory of Meaning,” (1974, 1976)

appear to be rooted in intuitionistic scruples concerning attribution of truth-value to sentences incapable of demonstrative proof. Cf. Dummett, 1974, p. 138; Root and Wallace 1982, “Meaning and Interpretation,” pp. 169-73.

SEMANTIC COMPETENCE AND TRUTH CONDITIONS 3

artificial languages, it was a bold and yet elegant step for Davidson to adapt Tarskian truth-theory to the semantics of natural language. In doing so, he had to question the long-standing Quinean distinction between theory of meaning and theory of reference.7 For, what Davidson proposes, essentially, is that the resources of the theory of reference suffice to explicate the notion of meaning. Moreover, though Davidson is indebted to Quine in many ways, and there are large areas of agreement between the two (including the rejection of the analytic-synthetic distinction), Davidson must work in the face of Quine’s critique of traditional notions of meaning and synonymy. Quine’s positive comments on Davidson’s approach to meaning8 signal a new constructive attitude. Moreover, they underline the appeal of reconstruct-ing theory of meaning within the theory of reference. Still, this much admira-tion from Quine’s direction has not prevented some important criticisms.9

Quine’s remarks emphasize the question of how the notion of synonymy might be expected to enter into Davidson’s construal of linguistic meaning: What will insure that Davidsonian T-sentences are interpretive rather than being merely true?

To meet this kind of objection, I will urge, Davidson’s account of mean-ing must be more thoroughly holistic, and thus in a way more thoroughly Quinean. Beyond the bare bones of logical form, a sentence has truth-conditions only relative to an embedding theory which elaborates its logical consequences in concert with those of other sentences. In “Two Dogmas” Quine lamented the persistence of the notion that “each statement taken in isolation from its fellows, can admit of confirmation or information at all”10

He concludes that “it is misleading to speak of the empirical content of an individual statement—especially if it is a statement at all remote from the experiential periphery.”11 If, in spite of such points, one persists with the aim of interpreting sentences, then this will only be viable if such interpretations are given in relation to a theory expressible in the object language.

In contrast with the thrust of Quine’s work, Davidson’s project holds out hope of support, through formal methods, to both the notion of linguistic meaning and the notion of belief. A theory of meaning rooted in exten-sionalistic semantics, even from the standpoint of contemporary linguistics apparently, is more appealing than no theory of meaning at all—and more

7. Cf. Quine 1953, “Notes on the Theory of Reference,” p. 130, and Davidson 1967, “Truth and Meaning,” p. 310. Similar views are expressed by Hintikka in the opening pages of 1969, “Semantics for Propositional Attitudes.”

8. Cf. Quine 1974, “On the Very Idea of a Third Dogma.”9. Cf. Quine 1977, “Review of Evans and McDowell,” p. 226.10. Quine 1951, “Two Dogmas of Empiricism,” p. 41.11. Ibid., p.43.

4

appealing than a completely behavioristic ersatz. But, if Davidson’s program is to be even initially plausible, it must bridge the perceived gap between semantic theory for formal languages and formal semantics for natural language, a problem which has doubtlessly bothered many of Davidson’s sympathetic readers. Only by bridging this gap can Davidson hope to influence semantic theory in linguistics. For, where a formal truth-theory appears to provide at best mere paraphrase, rather than interpretations of expressions of natural language, it will claim little attention from those who seek to characterize semantic competence of native speakers.

2. Description or Paraphrase?It is important to ask, then, if a good fit can be expected between Davidson’s program and the empirical objectives of semantic theory in linguistics. Davidson’s intent is clear. He is “not interested in improving on natural language, but in understanding it.” He views “canonical notations as devices for exploring the structure of natural language”12 Davidson himself empha-sizes, however, that we must expect a truth-theory “to rely on something very much like Tarski’s sort of recursive characterization of satisfaction, and to describe the object language in terms of familiar patterns created by quantification and cross reference, predication, truth-functional connectives and so on.”13 This quite naturally leads one to the question of whether there are, in natural language, such semantic structures to be described. Can we plausibly regard canonical notations which are suitable for construction of a truth-theory as providing anything more than loose paraphrase? Again, could the logical tradition from Frege through Quine have really been so wrong in thinking that it was revising natural language for scientific purposes generally and logical purposes in particular?

In spite of such question, there has been some positive response to Davidson’s program from linguists and those more influenced by semantic theory in linguistics. Most noteworthy is Fodor’s view in The Language of Thought. In the central argument of the book, Fodor invokes the assumption that “one understands a predicate only if one knows the conditions under which sentences containing it would be true.” He goes on to recommend reading Davidson, “for a useful introduction to the general program of analyzing meaning in terms of truth”14 Like Chomsky, Davidson views empirical semantics as a contribution to a psychological theory of speaker’s

12. Davidson 1977, “The Method of Truth in Metaphysics,” p. 203; cf. also Davidson 1970, “Semantics for Natural Language,” p. 63.

13. Davidson 1974, “Belief and the Basis of Meaning,” p. 151.14. Fodor 1979, The Language of Thought, pp. 59-60.


competence. For example, writing of an interpretation theory for an alien lan-guage, Davidson says, it may “be used to describe an aspect of the inter-preter’s competence at understanding what is said.” Moreover, “we may maintain that there is a mechanism in the interpreter that corresponds to the theory.” Still, in contrast with Chomsky, Davidson does not insist “that all interpreters know the theory,” rather, “it is enough if they know the required consequences.”15

The basic problem here is to understand the description of expressions of natural language in Davidsonian terms. Questions regarding the internal representation of speaker’s semantic competence or the psychological reality of a theory of meaning constitute a further problem. Though Chomskian formalisms are specifically designed for the task of describing natural lan-guage, many a theorist has gasped at the prospect of attributing Chomskian linguistic descriptions to language in the mouth of the native speaker. It may seem even less plausible, then, to hold that what counts as proper logical form, by the lights of contemporary logic (and further results of Davidson’s program!), depending as this does upon the syntax of formal languages, has been there all along “in the head” or otherwise. Can we really expect to describe the semantic features of natural language accurately in such terms?

Thus, we seem to come to a dilemma for Davidson’s program. Davidson aims to be descriptive, and a truth-theory must therefore bring out something that is already there, including certainly the logical form of sentences of natural language and reference or satisfaction conditions for non-logical vocabulary. But, there are reasons to doubt that sentences of natural language have the kind of systematic logical form they must have in order to be captured by a truth-theory. Discounting the prospects of a pre-established logical harmony among all languages, then, it appears that Davidsonian semantics for natural language must be either formal and not genuinely descriptive or descriptive but not fully formal.

While I believe this problem can be answered, focusing on it will serve to emphasize the radical character of Davidson’s construal of the notion of meaning and highlight crucial elements. Thus, Davidson suggests, by way of answering Tarski’s similar doubts concerning formal semantics for natural language, that we need only consider fragments of natural language where formal difficulties do not arise. We are not to reform natural language where it is incapable of formal treatment (because it contains its own truth predicate, for example, and thus generates the semantic paradoxes), but rather we are only to deal with a language in so far as a formal treatment is possible.

15. Davidson 1974, “Belief and the Basis of Meaning,” p. 141.

6

“Much of what is called for,” Davidson says, “is just to mechanize as far as possible what we do by art when we put ordinary English into one or another canonical notation.” The objection to this point will likely be that the logician’s paraphrase to canonical notation often ignores much of what it sets out to paraphrase. Still, Davidson insists that “if we know what idiom the canonical notation is notation for, we have as good a theory for that idiom as for its kept companion.”16 This is just to say, then, that for Davidson “adequate” paraphrase to canonical notation preserves meaning.

What counts as adequate paraphrase, and even what counts as correct canonical notation is open to question and development on Davidson’s view, so that the picture of meaning he proposes is to that degree vague. Still, it remains clear that on Davidson’s views we capture the meanings of sentences and other expressions of a natural language when we are able to relate such expressions to canonical notation in a thorough and systematic way. Thus, where it is objected that not all of natural language can be put into canonical notation and that this is exactly what leads one to hold that the logician’s language constitutes a reform of natural language, Davidson must evoke or reiterate the fact that he is attempting to explicate the notion of linguistic meaning. Davidson’s reform is not a reform of natural language, rather he reforms the theoretical notion of linguistic meaning. When natural language does not come over into canonical notation (even with Davidson’s expansions, e.g., his analysis of adverbs or indirect discourse), then this merely shows the extent of reform involved in Davidson’s construal of meaning. This reply will not suffice to answer all problems in Davidson’s program, but it is important to see that it does provide answers to problems related to the unsystematic character of natural language.

Seen from this perspective, the question of whether a Davidsonian truth-theory can be regarded as genuinely descriptive comes to depend upon the acceptability, in overall terms, of Davidson’s conceptions of meaning and logical form. In a very important sense, this result is to be expected. To know if we have an adequate semantic description, we must know what counts as falling within the domain of semantics. To know if we have captured the meaning of an expression, we must also know what meaning is. What to count as an adequate description is a question which cannot be totally divorced from the considerations which enter into decisions between theore-tical frameworks. It should, therefore, be no surprise if we employ our best theory of logical form, quantification, truth-functional connectives, variables and so forth to describe an alien language.

16. Davidson 1967, p. 29.


It is not a question of a pre-established logical harmony among all languages. Davidson’s approach does presuppose, I think, that natural languages serve certain common purposes and that they have identifiable structures suited to serve such purposes. To view a natural language in terms of Davidson’s conception of meaning is to view it with a selective eye, no doubt. In spite of all outward variety and variation, one aims to abstract those features which enable a language to express a potential infinity of inter-related truths, making use of finitely statable means. In fact this is substantially the same perspective we take on ordinary language if we set about to reform it for logical and scientific purposes. Thus, it is no accident if interpretations of natural language can be given in an idiom approximating the logician’s canonical notation. Given Davidson’s conception of meaning, this is just about what we should expect.

Thus, in so far as native idiom comes over into canonical notation, based on empirical evidence, then there will be grounds for describing native expressions in terms of their logical form or other standardized semantic features. But what does not come over into canonical notation, once sufficient evidence has been considered, can be counted as inessential from a semantic perspective. It is not after all the native theory of logical form or meaning that we want to capture. It is rather the actual practical employment of language as this is relevant to our own best picture of logical form and of meaning.

3. Language and Language-TheoryAs might be expected, there are difficulties concerning synonymy in Davidson’s extensionalistic approach to meaning. The overall question is how close to an intuitive notion of synonymy and interpretation we can approximate, making use of the resources of the theory of reference. My main conclusion will be that it is possible to get somewhat closer than Davidson has so far indicated. To see why and how, we must first continue the investigation of semantic competency within truth-conditional semantics by looking at some objections to Davidson’s approach.

“For many words,” Putnam argues, “an extensionally correct truth-definition can be given which is in no sense a theory of the meaning of the word.” Thus consider the following which might appear as part of the finite base of a truth-theory:

‘water’ is true of x iff x is H2O.

Putnam’s point is that if we are giving a truth-theory for the language of a pre-scientific community, “most people don’t know that water is H2O.” Thus,

8

he concludes, “this formula in no way tells us anything about the meaning of the word.”17

If speakers cannot be said to know that water is H2O, then surely we misinterpret them if we make use of ‘H2O’ in our T-sentences and truth-theory. If, for instance, we claim to understand a language L when we have a mass of T-sentences such as the following,

‘Lakes are full of water’ is true-in-L iff lakes are full of H2O.

and supposing that no speaker of L know that water is H2O, then surely a mistake has been made. If Davidson is correct in holding that a theory of meaning can be given which proceeds from a specification of reference or satisfaction conditions for the non-logical vocabulary of the object language, then it seems clear, in light of Putnam’s objection, that not just any exten-sionally equivalent specification of satisfaction conditions will be equally of interest. Still Davidson can be found to say that if, in our truth-theory, the object language predicates and the meta-language predicates “have the same extensions then this might be enough.”18 The reason this is not enough is that we expect an adequate interpretation to capture the native speaker’s conceptual system in some fashion. If the native community knows no chemistry, then it is a mistake to equate their ‘water’ with out ‘H2O’, for the latter word does indeed involve some chemistry. Moreover, it seems clear that there will be a “fact to the matter” on the point. Whether chemical theory is present in the native culture is not something that could be easily ignored. What stands out here is the need to indicate the kind of evidence which will lead to a correct specification of satisfaction conditions. For, indicating the reference of native expressions, from the standpoint of our own ontology or our own best theory, may well be part of what we want from a theory of meaning, as Putnam stipulates; however, it is certainly not all that we want. Rather, in accordance with the concept of semantic competence, we want to state satisfaction conditions for native expressions, so far as possible, without extinguishing the distinctive traits of the native conceptual system, their belief system or their ontology.

So apparently, giving truth-conditions for all the sentences of a language is not the same as giving a theory of meaning for that language. Davidson, moreover, is well aware of the general problem sketched here. This shows in his concern to find “further reasonable and non-question begging constraints” on the acceptability of T-sentences.19 Clearly, one will hope for empirical

17. Putnam 1975, “The Meaning of ‘Meaning’,” p. 259.18. Davidson 1974, p. 151.19. Ibid., p. 152.


constraints in order to meet the demands of semantics viewed as a branch of empirical linguistics.

In constructing a truth-theory, according to Davidson, we are first of all to construct T-sentences on the basis of the empirical evidence. “The evidential base for the theory will consist of facts about the circumstances under which speakers hold sentences of their language true.”20 The logical particles of grammar must be identified and in the end, we must construct a finite axiom system which will logically imply an infinite number of T-sentences, including all the attested T-sentences. One way to put the problem pointed out by Putnam, then, is to ask what constraints on the acceptability of T-sentences are going to lead the theorist to make a correct choice between extensionally equivalent correlations of object language predicates and meta-language predicates. We need further constraints, because taking note of facts about the circumstances in which speakers hold sentences of their language true will be necessary but not sufficient. To obtain acceptable T-sentences, we need to be able to choose between various possible characterizations of the observable circumstances. If Karl says ‘Es regnet’ and we observe what are (in fact) fully appropriate circumstances for the use of this expression, we might formulate a T-sentence the right-hand side of which makes use of ‘It is raining’. But, perversely, on the basis of the same evidence, we might instead equate ‘Es regnet’ with ‘Liquid H2O is precipitating from the atmosphere’. Making a choice here cannot depend merely on evidence concerning the circumstances in which the expression ‘Es regnet’ is held true. So, further constraints will be needed. To say that these must be non-question-begging constraints is to say that they should not make use of problematic semantic terms. It will not do to simply require that the sentence on the right of a T-sentence must translate or interpret the native sentence. We want to know what evidence will allow us to approximate this goal.

A quite similar problem arises from J. A. Foster’s contribution to Truth and Meaning. For present purposes, I will borrow the more concise formulation Quine provides in his review of the book. Thus, consider the general form of a T-sentence:

S is true-in-L iff p.

“The trouble,” as Quine puts Foster’s point, “is that a true T-sentence remains true when the translation represented by ‘p’ is augmented by conjoining any irrelevant but true sentence to it; and this corruption can be worked into the

20. Ibid.

10

finite axioms themselves.” We thus produce, argue Foster and Quine, “a T-sentence which fails the translation requirement.”21

In his “Reply to Foster,” Davidson does not explain how we are to select the proper T-sentences or the proper truth-theory. It is clear, however, that he aims to have empirical constraints on the acceptability of T-sentences that will serve to eliminate whatever does not pass the translation requirement. “A theory which passes the empirical test,” Davidson says, “is one that can be projected to unobserved and counterfactual cases.”22 Apparently, then, Davidson hopes to eliminate the troublesome T-sentences by means of empirical tests which will show them to be unlawlike. For, it is only lawlike generalizations which we expect to stand up to such “unobserved and counterfactual cases.” Unfortunately, Davidson does not, to my knowledge, tell us enough about the appropriate constraints on T-sentences. Without this, his view appears seriously incomplete. We may hope to complete this aspect of the program, finding constraints on our T-sentences in the theory of action and /or belief, as Davidson has more recently suggested.23 However, Davidson’s appeal to these further domains, though well argued and interesting, is at the same time a retreat from his claim that a truth-theory “is tested by evidence that T-sentences are simply true.”24 All well and good. An orderly retreat is surely called for here. The objection due to Foster and Quine shows that Davidson must somehow draw on empirical evidence to select appropriate T-sentences and an appropriate truth-theory. Not just any true T-sentences will serve the purposes of interpretation; and a truth-theory must be selected which will generate appropriate T-sentences. It remains to be seen, however, whether we will have to retreat as far as the theory of action and decision theory in order to find the needed constraints on truth-theories.

What we have here is a methodological problem: How can we select appropriate T-sentences, and what evidence will allow for the selection of appropriate T-sentences? Yet, this methodological question reflects back upon Davidson’s conception of the proper object of semantic investigation. In speaking of a language, Davidson is most often concerned with a language taken as an infinite set of sentences specified by a finite number of formation rules. This is a syntactic conception of language, and he aims to provide semantic characterization of a language so conceived. But as I have urged elsewhere,25 in standard referential semantics, it is not language in this sense

21. Quine 1977, “Review of Evans and McDowell,” p.226.22. Davidson 1976, “Reply to Foster,” p. 174.23. Cf. Davidson 1980, “Toward a Unified Theory of Meaning and Action.”24. Davidson 1973, p. 134.25. Cf. Callaway 1981, “Semantic Theory and Language: A Perspective,” pp. 64-

67.


which is interpreted, but rather always a theory in a language. To put the same point in a slightly different way, in standard model-theoretic semantics, no interpretation of non-logical vocabulary will be neutral between all theories expressible in the syntactically specified language. Yet, the problems we have found in Davidson’s approach certainly are tied up with the need to interpret non-logical vocabulary. This much is certainly clear: we need to bring empirical evidence to bear in order to select between alternative interpretations of non-logical vocabulary, as in Putnam’s problem. I wish to suggest, then, that part of the problem concerning the selection of appropriate T-sentences and an appropriate truth-theory arises from placing too much emphasis upon a syntactic concept of language; it is not the syntactic language itself which calls for interpretation but rather discourse conducted in the syntactically specified language.26 But in order to develop the point more fully, I will need to backtrack a bit.

Thus, let Σ represent a theory, that is a proper subset of the syntactically specified language L, closed under deduction. If ‘F’, a predicate of L, is interpreted by assigning it to a subset of a domain D, then it will either be assigned to a non-empty set or to the null set. The interpretation thus either provides that ‘(Ex)Fx’ ε Σ or it will provide that ‘~(Ex)Fx’ ε Σ. My point is that no such interpretation of ‘F’ is neutral between all alternative theories expressible in the language L. Further, if we interpret two individual constants ‘a’ and ‘b’ and ID(‘a ’) = ID(‘b’), then ‘a=b’ ε Σ; and if ID(‘a’) ≠ ID(‘b’), then ‘a=b’ Σ. Similar points hold regarding the interpretation of any pair of predicates ‘F’ and ‘G’. An interpretation of the pair of predicates will correlate with theories or systems with either logically imply sentences such as ‘(x)(Fx only if Gx)’, ‘(x)(Gx only if Fx)’ and ‘(x)(Fx if and only if Gx)’ or do not. In the absence of this kind of interpretation it is difficult to see how it would be possible to attribute quantificational structure to the sentences of a language at all. For, identification of quantificational structure depends upon the identification of quantificational modes of inference, and this in turn de-pends upon identification of relationships between the interpretations of items of non-logical vocabulary. However, such relationships between the interpretations of items of non-logical vocabulary belong, properly, to a theory expressed in the syntactically specified language. Thus, since Davidson’s program is committed to identifying quantificational structure in the sentences of a language under interpretation, this brings along with it the further commitment to interpreting theories or belief-systems expressed in thelanguage. It is not sufficient to draw on evidence of what sentences are held

26. Cf. Gochet 1986, Ascent to Truth, A Critical Examination of Quine’s Philosophy, p. 63.

12

true, when engaged in radical interpretation. Rather, the proper interpretation of non-logical vocabulary and with this the selection of an appropriate truth-theory depends upon and must be linked to a particular theory expressible in the object language. We must draw on evidence of logical relationships among native sentences in order to see what theories or belief-systems are held true. Since, as we have seen, interpretations of items of non-logical vocabulary will differ, in quite significant ways from theory to theory, there is every reason to believe that focusing, in this way, on native theory will allow for a selection between alternative matchings of object language vocabulary and meta-language vocabulary in a truth-theory.

If we must always interpret theories, in something like the logician’s sense of the term, then it seems to follow that what is wanted in interpreting a given linguistic community by means of a truth-theory is to match theory in the object language with theory in the meta-language. This is to say, first of all, that the patterns of inference observed among sentences of the object language are highly relevant to the task of providing a proper interpretation. This is so not merely because detection of such logical relationships provides a key to the identification of the logical particles, but also because one needs to identify non-logical axioms of the native belief-system(s). These can be expected to have a great semantic significance. Secondly, by matching theory in the object language with theory in the meta-language, we can reasonably expect to better capture the native conceptual system. For example, if we match the native ‘water’ with our own ‘H2O’, as in Putnam’s objection, this is a mistake. What the mistake consists in, however, is the fact that the latter word, but not the native expression is enmeshed in sophisticated chemical theory. The mistaken interpretation is not ruled out merely by consulting the observable circumstances in which the natives hold sentences containing ‘water’ true. But it will be ruled out if we find that there is no system of native sentences held true, logically related in the fashion that sentences of chemical theory are related. Observation of circumstances in which native sentences are held true will suggest hypotheses. However, decision between such hypotheses is better seen as a matter of seeing the place of a native sentence within a relevant belief system.

Consider again Foster’s objection. The problem is to find a principled way to exclude irrelevant material. If ‘S is true-in-L iff p’ conforms to the evidence provided by the observable circumstances in which speakers hold the sentence S true, then so will ‘S is true-in-L iff (p & the earth moves)’. We could construct a truth-theory to capture the latter T-sentence instead of the former, so we need an explanation of how evidence could lead us to select the interpretive T-sentence. Once again, we must draw on evidence of logical relations among native sentences. Having identified the native belief-system,


we will find that this provides a standard of relevance. Sentences are intuitively relevant to each other when logically related by some more or less viable theory. In order to show that a T-sentence is acceptable, one must seek to show, on the basis of evidence concerning patterns of inference among native sentences, that a given T-sentence relates two sentences with similar inferential roles. In short, we get closer to an intuitive notion of interpretation or translation if we set as a goal the formulation of a notational variant of native theory, making use of our own vocabulary and syntax.

The importance of interpreting theories in the object language, and taking the native language-theory27 as the object of semantic investigation, is emphasized if we consider the problem of interpreting non-observational native vocabulary. Since there is no directly discernible relationship between non-observational sentences and observable circumstances of their use, there is no way to correlate non-observational vocabulary of the object language with our own vocabulary, except by means of empirical evidence of logical relationships serving to relate non-observational sentences to each other and to observational sentences. When we reflect that all observational vocabulary is in some degree theoretical as well, the importance of interpreting theories in the object language is ramified.

The overall point is that drawing on evidence of patterns of inference among native speakers, in a specific speech community, we can hope to identify native theory, including non-logical axioms of the native belief-system. The perspective is that a semantic characterization of a language will only cohere when focused on the elucidation of a belief-system expressible in the syntactically specified language: that the selection of interpretive T-sentences, and an appropriate finite base for a truth-theory, must be guided by evidence of logical relationships among sentences as used by a particular linguistic community. Nor should one expect that the speech community relevant to a given truth-theory will always be so broad as to coincide with the community making use of the syntactically identified language. On the contrary, since a truth-theory is here focused on a belief-system (or, to idealize, on a theory), any difference in belief between speakers could, in principle, be used as a dividing line between different speech communities. However, the pragmatics of research will surely serve to focus interest upon belief-systems which are massively prevalent among important groups of speakers. One might, therefore, attempt to give a semantic theory for basic English for instance, though it will also be possible to focus more narrowly on sub-groupsfor instance the language-theory of a specific religious de-nomination, or one might select any group of speakers sharing a cluster of

27. I borrow this term from Follesdale 1973, p.291.

14

(mostly unstated) beliefs held in common. The focus in any case, will be upon just those beliefs which are unquestionable, or virtually unquestionable within the relevant speech community.28

A further adaptation of Tarski’s notion of a truth-theory suggests itself at this pointone which can be fairly said to build upon Davidson’s application of truth-theories to inter-linguistic interpretation. Given that evidence of logical relationships among sentences will suffice for the identification of non-logical axioms of the native belief-system, such non-logical axioms could be built into the finite base of a truth-theory and thereby simulate traditional theories of lexical analysis. Suppose, for example, that we have or want a truth-theory for a German-speaking community, the consequences of which include the following:

(1) ‘Ein Junggeselle ist ein unverheirateter Mann’ is truein Germanc iff a bachelor is an unmarried man.

Generating the T-sentence will require a finite base which matches the appropriate vocabulary.

(A) x satisfies ‘Junggeselle’ iff x is a bachelor.(B) x satisfies ‘unverheiratet-’ iff x is unmarried.(C) x satisfies ‘Mann’ iff x is a man.

Suppose further that available evidence serves to establish that ‘Ein Junggeselle ist ein unverheirateter Mann’ is a non-logical axiom of the language theory or semantic dialect Germanc. If so, the point may be expressed within the truth-theory by augmenting the finite base with the following clause:

(D) x satisfies ‘Junggeselle’ iff x is a man and x is unmarried.

Such a clause has the effect of analyzing relevant items of non-logical vocabulary within the truth-theory, marking the non-logical axioms29 and exhibiting aspects of the native conceptual system. With the addition of (D), we will generate within the truth-theory not only T-sentences like (1), and (2) below,

28. ‘Meaning’ in a more intuitive sense, is plausibly a systematizing selection from this broader domain.

29. Which is intended here to include what would otherwise count as ‘meaningpostulates.’


(2) ‘Jacob ist ein Junggeselle’ is true-in-Germanc iff Jacob is a bachelor.

(3) ‘Jacob ist ein Junggeselle’ is true-in-Germanc iff Jacob is a man and Jacob is unmarried.

Given a good number of T-sentences like (3), generated by a specific truth-theory, that truth-theory has a better claim to serve as a theory of interpretation.

The example illustrates one type of sentence which might plausibly be regarded as involving a non-logical axiom of a given speech community. Non-logical axioms will certainly include many sentences that would be regarded as analytic, by the friends of the analytic-synthetic distinction. Still, this category of non-logical axioms will also include sentences which would never be regarded as analytic, e.g., ‘All matter has weight’, or ‘The earth moves’, or even false sentences such as ‘The earth stands always motionless’, given a society with a Ptolemaic astronomy. In identifying non-logical axi-oms, one focuses on what is treated as unquestionable, or virtually unquestionable, within the relevant linguistic community. There is no call here for a theory of necessary truth, and no justification for such a theory. On the contrary, supposed analytic truths assume a more modest and appropriate status, on this account, even as we pass over the notion of analyticity in theoretical linguistics.

4. Evidence for Language-TheoriesCertain general considerations strongly suggest that there is evidence available sufficient to identify such language-theories. Assume, contrary to Davidson, that different linguistic communities do employ conceptual systems which are alternatives to our own. It follows, given plausible assumptions, that we are, in fact, able to become aware of such alternative systems. This point alone insures that there is sufficient evidence available to identify the conceptual system of alien linguistic communities.

We expect, moreover, that systems of concepts quite divergent from our own will most likely be found where language is less observational and more theoretical in character. Thus, if we are able to detect and identify alternative conceptual systems, it appears to follow that even the non-observational aspects of a language-theory are semantically determinate enough to rule out their arbitrary interpretation. Otherwise, talk of alternative conceptual systems will fail of application.

Clearly, the observable circumstances in which theoretical or non-observational sentences are held true will not themselves determine an

16

interpretation, thus other sources of evidence become more crucial when it comes to interpreting non-observational language. We understand non-observational language by seeing how it fits into a system—we detect the logical relationships of theoretical sentences to each other and the relationship of such a theoretical system to observation sentences. A possible difficulty with this argument arises, however, from Davidson’s criticisms of the notion of conceptual scheme or system.30 Thus I will briefly consider Davidson’s positions on this notion before proceeding.

The basic problem in Davidson’s view is that he makes the criteria of difference in conceptual scheme much too strong. Detection of an alternative scheme, in Davidson’s critique, is made to depend upon “two kinds of cases that might be expected to arise: complete and partial failures of translatability.”31Davidson’s arguments are polemical. The target is “concep-tual relativism,” and responsibility for the assumption of untranslatable languages is attributed, not entirely without reason, to positions due to Worf, Kuhn and Feyerabend. But while Davidson is correct to reject the notion of untranslatable languages, and the positions he criticizes do appear to involve some such claims, its seems clear that Davidson has construed the target of his critique too widely.

Talk of alternative conceptual schemes does not always imply claims regarding failures of translatability, nor need it involve such claims. For instance, in his reply to Davidson’s paper, Quine offers “a measure of what might be called conceptual distance between two languages.” The proposal depends explicitly upon the possibility of acceptable translations. What Quine has in view here, evidently, is that an acceptable translation of a sentence of one language might be considerably longer than the sentences translated. A handy short term in the first language may require some elaborate description in the second language, for instance. I conclude, then, that Davidson has not demonstrated that talk of alternative conceptual systems requires com-mitment to untranslatable languages.

On the contrary, Quine is correct in sensing “a conflation of truth and belief” in crucial portions of Davidson’s critique.32 An overly robust version of the principle of charity leads Davidson to think that an alien language expressing an alternative conceptual scheme must engender so much disagreement as to make translation impossible. “Given the underlying methodology of interpretation,” he holds, “we could not be in a position to judge that others had concepts or beliefs radically different from our own.”33

30. Cf. Davidson 1974, “On the Very Idea of a Conceptual Scheme,” pp. 183-198.31. Ibid., p. 185.32. Quine 1981, “On the Very Idea of a Third Dogma,” p. 39.33. Davidson 1974, “On the Very Idea of a Conceptual Scheme,” p. 197.


The lesson in this, however, is that Davidson’s version of the principle of charity is so strong as to threaten a linguistic version of naive realism. One can agree with Davidson that the possibility of interpretation depends upon finding areas of “massive agreement” with those who we wish to translate. However, it is unclear why Davidson thinks that finding such areas of massive agreement should be incompatible with fairly substantial dis-agreements and conceptual contrast. This problem in Davidson’s view may arise from too much emphasis upon the interpretation of sentences one at a time.

Consider, then the example of Newtonian and Einsteinian physics, discussed by Kuhn and Davidson. Any comparison of the two systems will depend upon the massive agreement in the experimental predictions of the two theories. But it seems clear that the disagreements and conceptual contrasts can also be understood. Einstein does introduce an alternative to the conceptual system of classical physics, just as Newton had introduced an alternative to the conceptual system passed down from Aristotle. We understand that Einstein had to disallow, for instance, familiar sorts of inference involving the notion of simultaneity. We understand that for Einstein the metric of space and time is not independent of the presence of matter and that the measurement of mass is linked with that of velocity. My point is just that massive agreement does not eliminate the possibility of disagreement and conceptual contrast in such scientific cases. Rather, Ein-stein’s revisions of classical physics could be understood as revisions, and understood as claims concerning a familiar domain when seen against the background of wide-ranging agreement concerning predictions and experimental results in relatively ordinary circumstances.

I take it that what is relatively unclear about the notion of conceptual system is the distinction between conceptual system and matters of fact and belief. I assume that this distinction cannot be made sharp. It is a matter of the relative importance of a change in theory and a matter of the systematic effect that a change has within an on-going system of beliefs. The present perspective is that we need not worry about just what differences amount to differences in concepts. Rather different conceptual systems are here to be understood in terms of different theories. I suppose that we can capture and characterize whatever theory we may want to investigate and do so in the detail needed to capture important or interesting contrasts.

The main conclusion, then, is that there is evidence available sufficient to interpret and understand the language-theory of an alien community, based upon investigation of the logical relations among native sentences. Initial interpretations of native observation sentences may be faulty in any given case, and such initial interpretations need to be checked from a perspective

18

informed by an appreciation of the systematic connections among observation sentences and sentences with a more theoretical character.

Consideration of domestic cases can be quite useful in illustrating the point. One needs to consider sentences alike, or substantially alike, in terms of the observable circumstances in which they are held true and then consider how logical relationships in a system of beliefs serve to distinguish such sentences. Take the example of the following two sentences:

(1) The man has been dismissed.(2) The man has been asked to leave.

In a given English-speaking community, we may find no difference, over a large range of cases, regarding observable circumstances which lead our speakers to hold the two sentences true. Yet, in spite of that, the non-logical axioms of the relevant belief system do allow for semantic discrimination between the two sentences and the two expressions ‘dismissed' and ‘asked to leave’. For example, we will find universal or near universal agreement on the truth of sentences somewhat like the following:

(A) If someone has been dismissed, then he has been given permission

to leave by a proper authority.

However, the corresponding sentence making use of ‘asked to leave’ will not command universal assent.

(B) If someone has been asked to leave, then he has been given permission to leave by a proper authority.

Thus, the required distinction between the interpretations of (1) and (2) might be arrived at by considering the logical form of the sentences (1) and (2) plus their logical relationship to (A) and (B)—the differing truth-values being attributed to (A) and (B), this aspect of the native belief-system, is the crucial point. The example is not as simple as could be desired, of course, since we presuppose an understanding of ‘permission’ and ‘proper authority’. But this is just to say that we expect evidence for speakers’ universal assent to something like (C).

(C) If someone has permission to act in a certain way, then permission has been given by proper authority.


The complication serves to emphasize the fact that we understand relatively observational language in terms of its logical relationship to an entire system of non-logical axioms.

Notice that given (1) plus (A) we get (3) as a logical consequence.

(3) The man has been given permission to leave by proper authority.

So, if speakers hold (1) true and make an immediate inference to (3) (or generally countenance such an inference), this is also a kind of evidence for regarding (A) as a non-logical axiom. Indirectly, it is also evidence for distinguishing our interpretations of (1) and (2), especially since we find no immediate inference from (2) to (3).

All of this adds up to empirical evidence for different interpretations of the terms ‘dismissed’ and ‘asked to leave’. Given the kinds of evidence sketched here, there are grounds for attributing different satisfaction conditions to ‘x is dismissed’ and ‘x is asked to leave’, even if we observe no situation where native speakers hold (1) true and (2) false or vice versa. (Our informants might, for instance be mostly observing from the distance and not hear the actual words used.) However, once we gain a broader picture of the native language-theory, e.g., the connections between ‘dismissed’, ‘permis-sion’ and ‘authority’, then we might also gain further insight into the way in which the observable situation is theory-laden for the native observer. That is, for instance, an alien investigator of our English-speaking community might come to notice verbal or non-verbal signs or symbols of authority which will allow a finer discrimination between the situations in which (1) and (2) differ in attributed truth-value. With the native language-theory in hand, we can also expect to come to see the world in the way the natives do.

5. Quinean ObjectionsThere are, of course, important objections which might be made to the kind of proposal sketched above. For example, there are sure to be Quinean objections arising out of that much discussed cluster of views associated with the indeterminacy of translation, the inscrutability of reference and ontological relativity. These views have been under discussion for more than twenty-five years now, dating from the publication of Word and Object. The great problem, following so much detailed discussion, is to still see the general outline of the forest after so much detailed study of so many trees.

Davidson, I think, has already made substantial inroads on this problem, in spite of the fact that he often seems to sympathize with Quine’s thesis of the indeterminacy of translation. If we accept Davidson’s view that we can reasonably describe an alien language by relating it to standard quantification

20

theory, that this is no imposition or revision but rather captures something objectively present, then indeterminacy of translation has already been stripped of a very large part of its significance. This much translation is only accomplished in Chapter Two of Word and Object by means of the metho-dology of analytic hypotheses, and is thereby afflicted, on Quine’s view, by the indeterminacy of translation.

Davidson remarks, in “Belief and the Basis of Meaning,” that “on my approach, the degree of indeterminacy will...be less that Quine con-templates.” The reasons for this are “partly because I advocate adoption of the principle of charity on an across-the-board basis, and partly because the uniqueness of quantificational structure is apparently assured if convention T is satisfied.”34

Davidson appears to take the teeth out of the indeterminacy thesis. “If there is indeterminacy,” he says, “it is because when all the evidence is in, alternative ways of stating the facts remain open.”35 Quine himself, in the review article cited above, approves of McDowell’s characterization of the indeterminacy thesis “as a version...of the strong verificationist objection to realism in a theory of meaning..”36 For Quine, that is, there are no facts to the matter if we impute quantificational structure when translating. For example,in Word and Object, he objects to a test for synonymy of two predicates ‘F’ and ‘G’ on the basis of whether the native is willing to assent to the standing sentence ‘All F’s are G’s and vice versa’ “following any stimulation that might be imposed at t.” Quine’s objection is that where languages other than our own are involved, co-extensiveness of terms is not a manifestly clearer notion than synonymy or translation itself; it is no clearer than the considerations, whatever they are...that make for the contextual translation of the identity predicate, the copula, and related particles.”37 In short, the objection is that to test for synonymy of terms we must assume the translation of the logical apparatus, but since the latter is subject to indeterminacy oftranslation, so is the former. Davidson’s position, on the contrary, is that since we can discover the native equivalents of our standard quantified logic, we should also be able to provide correlations of non-logical vocabulary in the finite base of our truth-theory. We should, that its, be able to find reasonable interpretations of native terms. That we can translate native terms, it should be noted, is something assumed in my proposal above regarding the aim of interpreting theories.

34. Davidson 1974, “Belief and the Basis of Meaning,” p. 153.35. Ibid., p. 154.36. Quine 1977, “Review of Evans and McDowell,” p. 228.37. Quine 1960, Word and Object, p. 54.


My conclusion does not guarantee, however, that it will always be possible to rule out all alternatives to a favored interpretation of a native language-theory. Perhaps it will always be possible to formulate alternative interpretations of a native language-theory which cannot be ruled out given the evidence available. Still, it does not follow from this alone that there is no fact to the matter in settling on one particular interpretation.

The “empirical slack” of theories strongly suggests that where we have a theory in the home language, we can always invent new theories designed to cover the same evidence and the same domain as the established or favored theory. Thus, once having translated a theory expressed in an alien idiom into our own, we could then proceed to invent a slightly different theory covering the same domain—thus producing an alternative translation. But, if the significance of such facts regarding the empirical slack of theories does nor force an ontological agnosticism regarding the referents of our own best theo-ries, (a kind of position the orthodox Quinean avoids), then it seems reasonable to hold that ontological relativity has no greater significance as applied to translation. This is to say that one must choose between arguing from the inscrutability of reference to the indeterminacy of translation or accepting the paradigms of the Tarskian theory of reference. Because, as Quine puts it, radical translation begins at home.

One comes to suspect that alternative language-theories in the home idiom, whether formulated directly as accounts of natural phenomena, or proposed as interpretations of an alien system, will bear one of two relations to each other. They will be strong alternatives, such that we suspect that some evidence might decide between them, or they will be weak alternatives, i.e., plausibly regarded as mere notational variations. I do not suggest that this distinction between theories which are strong or weak alternatives to each other will always be clear-cut and straight-forward. Indeed, this distinction can be no more clear-cut than the prospects of finding new evidence as required by a theory. Sometimes we just have to wait and see. But sometimes there will be very reasonable grounds for believing that no evidence will decide between two theories: in such cases it is reasonable to speak of notational variation. Any notational variation in our own idiom will be as good as any other for the purpose of interpreting an alien language-theory.

Quine makes good use of the concepts of truth and reference. “Within our own evolving doctrine,” he says, “we can judge truth as earnestly and absolutely as can be.”38 But, if we judge of truth “as absolutely as can be,” and this is truth as elucidated within the Tarskian theory of referential notions, then this is surely to say that where we have good reasons to believe

38. Quine 1960, Word and Object, p. 25.

22

that a sentence is true, e.g., ‘There are men,’ then we have the same good reasons to judge of the reference of our words, i.e., that ‘men’ denotes or is satisfied by men. If we take truth and reference seriously within our own evolving viewpoint, then whatever there is to the thesis of the inscrutability of reference does not, obviously, forbid earnest judgments of truth and reference. The kernel of truth in the doctrine seems to come down to the fact that we can only make judgments of reference, or assign reference to terms, given the perspective of a theory we take to be true. But however this may be, as I have argued elsewhere,39 it is surely a pointless exercise to merely imagine possible reinterpretations of our theory, or distinctions we do not make, and on such grounds remain agnostic concerning the reference of our words. Davidson’s program, since it is an application of the Tarskian theory of reference to inter-linguistic interpretation, attempts to carry this point over into the case where we interpret the words of others. “Translation,” Quine says, “arrests the free-floating reference of the alien terms only relative to the free-floating reference of our own terms, by linking the two.”40 But, if we can and must “judge truth as earnestly and absolutely as can be,” and reference and truth go together, as in Tarski’s work, then “the free-floating reference of the alien terms,” is sufficiently arrested by relating them to our own. At least the inscrutability of reference does not suffice to demonstrate the contrary.

ReferencesCallaway, H. G. 1979, “Reference, Variables and the Empty Universe,”

Logique et Analyse 85-86. pp. 85-98.1981, “Semantic Theory and Language: A Perspective,” Philosophical

Topics, Summer, pp. 61-70.1985, “Meaning without Analyticity,” Logique et Analyse, 109, pp. 41-60.Chomsky, Noam, 1977, Language and Responsibility, Pantheon Books, New

York.Chomsky, Noam, 1980, Rules and Representations, Columbia University

Press, New York..Davidson, Donald, 1967, “Truth and Meaning,” Synthese 17, 304-23;

reprinted in Davidson 1984, pp. 17-36. 1970, “Semantics for Natural Language,” reprinted in Davidson 1984, pp.

55-64.1973, “Radical Interpretation,” Dialectica 27, 313-28; reprinted in

Davidson,1984, pp. 125-140.

39. Cf. Callaway 1985, “Meaning without Analyticity,” pp. 8ff.40. Quine 1981, “Things and their Place in Theories,” p.20.


1974, “Belief and the Basis of Meaning,” Synthese 27, pp. 309-23; reprinted in Davidson 1984, pp. 141-154.

1974, “On the Very Idea of a Conceptual System,” Proceedings and Addresses of the American Philosophical Association, 47; reprinted in Davidson 1984, pp. 183-98.

1976, “Reply to Foster,” in Evans and McDowell (eds) 1976, pp. 33-41; reprinted in Davidson 1984, pp. 171-80.

1977, “The Method of Truth in Metaphysics,” reprinted in Davidson 1984, pp. 199-214.

1980, “Toward a Unified Theory of Meaning and Action,” Grazer Philosophische Studien 2, pp. 1-12.

1984, Inquiries into Truth and Interpretation, Oxford University Press, Oxford.

1985, “Incoherence and Irrationality,” Dialectica 39, pp. 345-54. Dummett, M., 1974, 1976, “What is a Theory of Meaning?” Part I. In

Guttenplan 1974; Part II in Evans and McDowell 1976.Evans G. and J. McDowell, 1976, Truth and Meaning, Essays in Semantics,

Clarendon Press, Oxford. Fodor, J. A., 1979, The Language of Thought, Harvard University Press,

Cambridge. Follesdal, D., 1973, “Indeterminacy of Translation and Underdetermination

of Theory,” Dialectica 27, pp. 289-301. Foster, J. A., 1976, “Meaning and Truth Theory,” in Evans and McDowell

(eds) 1976, pp. 1-32.Frege, G., 1977, Logical Investigations, Yale University Press, new Haven.Guttenplan, S. (ed), 1975, Mind and Language, Clarendon Press, Oxford. Gochet, P., 1986, Ascent to Truth, A Critical Examination of Quine’s

Philosophy, Philosophia Verlag, Munich. Hintikka, J., 1969, “Semantics for Propositional Attitudes,” in Models for

Modalities, D. Reidel, Dordrecht.Putnam, H., 1975, “The Meaning of ‘Meaning’,” in Putnam 1975, Mind,

Language and Reality, Philosophical Papers, Vol. II, pp. 215-271. Quine, W. V., 1951, “Two Dogmas of Empiricism,” reprinted in Quine

1953, pp. 20-46.1953, From a Logical Point of View, Harvard University Press,

Cambridge. 1953, “Notes on the Theory of Reference,” in Quine 1953, pp. 130-38. 1960, Word and Object, MIT Press, Cambridge. 1974, “On the Very Idea of a Third Dogma,” in Quine 1981, pp. 38-42. 1977, “Review of Evans and McDowell,” Journal of Philosophy 74, pp.

225-41.

24

1981, Theories and Things, Harvard University Press, Cambridge.“Things and their Place in Theories,” in Quine 1981, Theories and Things,

pp. 1-23. Root, M. and J. Wallace, 1982, “Meaning and Interpretation,” Notre Dame

Journal of Formal Logic 23, pp. 157-73.Schlick, M., 1936, “Meaning and Verification,” Philosophical Review, 45.

Excerpted in Lehrer and Lehrer 1970, Theory of Meaning, Prentice-Hall, Englewood Cliffs, N.J., pp. 98-112.

Wittgenstein, L., 1922, Tractatus Logico-Philosophicus, Routledge and Keagan Paul, London.

Date post:	25-Nov-2023
Category:	Documents
Upload:	temple
View:	1 times
Download:	0 times

SEMANTIC COMPETENCE AND TRUTH-CONDITIONAL SEMANTICS

Documents