Pascal Engel


Cadernos de Filosofia,  Universidade Novella de Lisboa

2, 1997, 9-30







        It is a commonplace that XXth century analytic philosophy of language has been concerned with an analysis of meaning. Philosophers in this tradition have produced various accounts of this notion in terms of other basic concepts: truth-conditions, speaker’s intentions, use and assertibility-conditions, to name a few. But no paradigm of analysis has really taken precedence over the others. This seems to be due to the fact that meaning is such a complex and multifarious phenomenon. This leads to doubt of the very possibility of a comprehensive theory of meaning for natural languages. A number of philosophers, however, have attempted to formulate various general requirements for the formulation of a theory of this kind, both at the descriptive and at the more fundamental philosophical level. Their move was to try to avoid the general question of what meaning is , which would call for an analysis of this notion in terms of more basic concepts, such as truth conditions, thought or rules. Instead they have asked: what is it, for a speaker, to know a language  and to understand it ? And they have tried to characterize meaning as a certain body of knowledge which every speaker has. They have called this knowledge knowledge of a “theory of meaning” in the following sense: a theory of meaning (a TM, for short) is the kind of knowledge through which a speaker gives meaning to the expressions of his own language. A theory of meaning in the ordinary sense of a philosophical account of this notion would be an account of the principles of a TM for a particular language, of the kind of characterisation of the knowledge of language that one could give.

        Davidson (1984) is perhaps the philosopher who has developped this strategy the most forcefully. He claims that an analysis of meaning must rest upon a proposal about the general form that a theory of meaning should take. Such a theory must account for the speaker’s understanding of his own utterances, and of our understanding (or, as he says, interpreting) the utterances of others. It should, according to him, take the form of a description of the truth-conditions of sentences, and hence be based on a theory of truth for the speaker’s language. This account has attracted much attention and criticism. It is not my aim here to detail this proposal and its difficulties[1]. I want, rather, to point out here that Davidson’s insistence on the understanding of language leads to certain characteristic tensions in his view, which also affect rival views. My objective is to claim that such tensions are not peculiar to Davidson’s view, and that any conception which adopts the basic starting point that a theory of meaning is a theory of understanding encounters or parallel difficulties.



1. Three ways of reading the equivalence


      Let us start from a basic equation: a theory of meaning (TM) is a theory of understanding. This equation has been advanced by Dummett (1975). Basically it says that any sort of description of meaning in a language must be related to the capacity that speakers have of understanding their own language and to have command upon it. It rules out from the start any sort of description of meanings as abstract entities which would play no role in an account of the actual speaker’s competence or in his use of expressions in current linguistic pratice. Or rather it does not rule out such descriptions, but requires that they be related at some point to some features of their competence[2]. This slogan —  “a theory of meaning is a theory of understanding” — seems sound enough. But it can be read in several ways, according to the kind of features of “understanding” one favors.

        The notion of understanding is highly ambiguous, and it can be read at least in three senses. First, understanding what an expression means can be understood as understanding the content  of this expression, something like its linguistic meaning in the language. For instance I understand what “cat” means when I know what this word means in English (that it designates certain animals). Second, understanding the sense of an expression means that I know how to use it (e.g. that I manifest correct grasp of this expression when cats are around and when I am asked to identify them, or asked to feed them, etc.). Third, understanding may mean that I have the proper mental state, or psychological representation, responsible for my grasp of either the content of the word or its proper use. Here “understanding” means a form of psychological competence, located, presumably in the mind or brain of the speaker. Each of these senses gives us a different reading of our slogan. As Barry C. Smith (1992) points out this equivalence can be read, as any equivalence, from left to right, from both sides, and from right to left[3].

        (1) First, when read from right to left, the equivalence says that we grant priority to a theory of meaning for a language. A speaker understands his language when he knows the basic facts of meaning which supply the interpretations of his and other utterances. The idea here is that all there is to know about understanding a language is the specification of the content  of a theory of meaning for that language and of the way such a content is ascribed to a speaker through the practice of interpretation. Facts about the use of sentences and expressions by speakers, will be, on this view incorporated in facts about the method of interpretation of meanings. And facts about the inner cognitive workings of semantical competence will be simply irrelevant.

        (2) Second, the equivalence can be read in a strict way, by giving no priority on either side. The idea here is that facts about meaning cannot be by themselves facts about understanding: we must have independent grounds for ascribing knowledge of the theory of meaning to speakers. But these grounds cannot be based upon the inner psychological workings of competence. They must rest upon the ways people use  their sentences, and manifest  their knowledge of the TM in various ways.

        (3) Third, the equivalence can be read from right to left, by giving priority to the psychological facts about speaker’s semantic competence. The structure of the TM must bein some sense reflected in the cognitive underpinnings of the mind of the speaker.

        Less abstractly, Davidson is a paradigmatic representative of reading (1), Dummett of reading (2), and Chomsky of reading (3). Let us call them, following Smith (ibidem) respectively the interpretative stance, the descriptive stance, and the psychological stance. Each stance emphasizes the idea that a TM for a speaker’s language is needed, some set of syntactico-semantical rules formulated in a systematic way. But they disagree on the other factors: use and cognitive set up. Against the psychological stance, both the interpretative and the descriptive stance agree that psychological facts are irrelevant for an account of understanding language, mostly on the ground that language is a public phenomenon, which is manifested in the overt and public use of expressions. But they disagree on the way use is incorporated in their account. For Davidson, facts about use and included in facts of interpretation of speakers, whereas for Dummett they must be settled independently. In other terms, we must for Dummett, have grounds, separate from facts of interpretation, for ascribing to the speakers knowledge of meaning. For instance we must say that there are certain conventions of use for words in a population. For all its emphasis on use, Dummett’s view does not amount simply to what is often called a use conception of meaning, and associated to the name of Wittgenstein. For Wittgenstein does not believe that facts about use could be incorporated in any sort of systematictheory : on the contrary he insists on the local, or piecemal, character of meaning “facts”. Indeed, for Wittgenstein, there are no real “facts” about meanings as particular uses, but only a collection of vague and changing criteria which do not lend themselves to any systematic account.[4]




2. Davidson’s interpretative stance


   Having described in an abstract and programmatic manner the various options, I turn now to Davidson’s particular proposal. For Davidson, as it is well known, a TM is a compositional theory, formulated in the Tarskian format of a set of axioms for each kind of expression of a language L, from which one can derive a set of “T-sentences” of the form “ S is true iff p ” (with the necessary adjustments needed for indexical and other sentences the truth of which is relative to a speaker and a time), and providing the truth conditions of the sentences of L. In what sense does such a TM specify the knowledge  that speakers have of their language?

        Not in the sense that speakers have an explicit  knowledge of what the axioms and the theorems of the TM state, nor a grasp of the semantic machinery of a Tarskian truth-theory, with complex logical notions such as sequences and satisfaction of predicates. Not in the sense that speakers could have a tacit  knowledge of such a theory, at some “subintentional” or otherwise unconscious level, deeply rooted in their cognitive modules. What the TM specifies, for Davidson, is the knowledge that an idealised, competent, and rational theorist , equiped with the appropriate theory,would know.  The knowledge that a TM specifies is not the knowledge of the speaker, but the knowledge that the theorist , or the interpreter, should use if he were to interpret the speaker’s linguistic behaviour from the outside, granted that this speaker is a rational animal, obeying certain idealised norms of conduct and of thought. The only facts about use that Davidson allows himelf to consider are behavioral facts about the sentences the speaker produces, and external facts about his environment. In addition Davidson supposes that we have means of recognising in the speaker a certain kind of propositional attitude, which is a kind of belief: belief about the truth  of a number of his sentences, or his holding-true  certain sentences. From this meager input, the interpreter postulates that the sentences that the speaker hold-true have a certain structure, and that he has beliefs which are, by and large true and logically coherent (this is the famous “principle of charity”). From there, he works out his interpretations, linking them to particular circumstances of use, and reaches in the end an interpretation, which is the output of a process of successive revisions, until the “best fit” is attained.

        Hence, for Davidson,  a TM is is a rational reconstruction., from the part of the interpreter, of the speaker’s competence. It is not, therefore, a specification of the knowledge that the speaker has of his language, but a specification of what the interpreter  knows, when he does such ascriptions of meaning with a TM.

        This view has three main consequences, which are endorsed by Davidson, but which immediately raise problems for his account, and which all have to do with the fact that meanings are ascribed from the outside or the “third person” point of view.

        First, because it is a theory of what the interpreter understands instead of what the speaker understands, it is at least indirect: it does not tell us anything about what the speaker himself  understands. So we cannot be sure that the features of the theory, which are the only one to which we have access, are features of the speaker’s own competence. For instance if the theory is recursive and compositional, there is no garantee that the speaker’s competence is itself recursive and compositional. These features could well be artifacts of the theory, and thus they could have nothing to do with the actual psychological competence of the speaker. Davison would grant this point, but it is problematic if the theory is supposed to be genuinely explanatory of the speaker’s competence.

        Second, because the interpreter works only from the third person point of view, by ascribing meanings and structures to the sentences of the speaker, it does not take into account what the speaker himself, in the first person, considers as the meanings of his words. Here too Davidson would grant this point, and say that this “first-person access” to meanings is just a relic of the psychologistic view of the meanings of words, or of Quine’s “myth of the museum”: if meanings are public, why care about what “in our minds” we take our words to mean? Still, it is problematic, for we have a genuine “first-person authority” over what we mean by our words, and Davidson grants this. But he grants this only for the interpreter. He, Davidson says, has authority over what he means, and from there he works out his interpretations. How can we be sure that this asymetry exists only for the interpreter?

        Third, since understanding a language is a matter of knowing how to interpret it, how can we be sure that the products of our interpretations yield a systematic theory of the language  itself? Interpretations can vary, and there is no garantee that what they reach is a systematic body of knowledge over a single object, the language. Here too, Davidson grants this consequence. He admits that, under the actual conditions of interpretation and of use, “there is no such thing as a language” (Davidson 1986), but only particular encounters, which yield only partial overlaps. A “TM for a language” is only, on this view, a useful idealisation, but it is not something which exists independently of these acts of interpretation. But here again, it seems to deprive of any objectivity the very account of a “theory of meaning for a language” which was aimed at.


3. Meaning and normativity


      These difficulties have led philosophers like Dummett to reject Davidson’s view. As I said above, Dummett (1975) objects to Davidson’s conception of a theory of meaning that it is too modest , because it does not allow really an attribution of knowledge of semantic facts to speakers. Such an ascription of knowledge, according to Dummett, is almost trivial, in the sense that it permits only ascription of the knowledge of T-sentences, without any ascription of what the theory states. It is just as if one said that someone who knows that «“Snarks are boojums” is true iff snarks are boojums» knows that snarks are boojums. But this person may know that the T-sentence is true, without knowing what it states, or what proposition it expresses. In other terms, a TM, in Davidson’s sense, leaves out what it was purported to capture, namely meaning. For Dummett, this is not enough. Ascriptions of knowledge of meaning must be related to facts about use. But what are such facts about use? “Use” is a very ubiquitous and ambiguous notion. But we can formulate Dummett’s requirement in the following way. On the one hand, facts about use must relate meaning to the manifestation of meaning (this is often called the “manifestation requirement”): there must be particular circumstances were an ascription of meaning is deemed to be correct or not. It is not clear whether Dummett conceives of these as behavioral facts, and the point is disputed. But at least he conceives of them as facts about the conditions of assertion  of sentences. The interpretation of this notion is also a moot point, and I do not want here to engage into the complex discussions which it has raised. But the point can be formulated in a way which is more or less independent of the notion of assertibility conditions used by Dummett, and which allows to express a number of his objections.

        Meaning is normative.[5] Facts about use of linguistic expressions are facts about the ways speakers take the applications of these expressions as correct  or incorrect, or as conform to certain rules  of usage. Rules can be obeyed or disobeyed, and there are facts of the matter as to whether people conform to these rules or not. Not at the individual level, for, as Wittgenstein’s remarks about “following a rule” purport to show, one does not obey a rule “privately”, and no state of mind can fix the particular interpretation of a rule. But at the communal or public level, everyone knows, provided he belongs to a community of usage, which rules he follows. The community can sanction incorrect uses, and a speaker can, from the other’s sanction, know when he conforms to these norms of usage. Now such normative facts about meanings are part of what the speakers know when they know they language. They are not, if we accept Wittgenstein’s criticisms of private language, psychological facts, but public facts. But fo all that, they are not inaccessible to speakers; indeed they lie in the open, and are available to anybody. They are an ineliminable part of what a speaker understands when he understands the meaning of his expressions.

        But Davidson’s kind of view seems to be unable to account for this normativity of meaning. Rules of meaning, or semantic rules are, on his view, reduced to a descriptive feature of the structure of a TM. Now this seems to be unfair to Davidson’s view. For does he not insist on the fact that an interpreter must, to set up his interpretative procedure, rely on certain very general norms of rationality, such as the fact that the speaker or thinker is in general truthful, reliable, coherent and justified in his beliefs? Aren’t these maxims of rationality in interpretation what accounts for the normativity of meaning? Isn’t this very feature what explains Davidson’s rejection of any sort of reduction of semantic facts to physical or natural facts and the ground of his “anomalous monism”? Although this is certainly Davidson’s reason for this rejection, the answer is that they do not account for “the normativity of meaning”, at least on Dummett’s view. For Davidson’s “standards of rationality”, although they certainly imply that there is something normative in meaning, only imply that this normativity is a feature of the theory  of interpretation, not of the speaker’s knowledge of it. The kind of normativity that Davidson has in mind when he uses this word is not the normativity which attaches to particular rules of usage , but a much “higher profile” kind of normativity, which goes with the theory of interpretation itself.[6] It is one thing to say that certain sorts of “theories” that we use for interpreting the behaviour of agents, such as the theory of decision (which Davidson incorporates into his theory of interpretation) are “normative”, in the sense that they rest upon idealised standards,  or idealised descriptions, not in the sense that they incorporate specific norms which would go together with kinds of words or expressions. In Davidson’s sense, a norm is an idealised description. It is not a particular rule  of usage. Indeed, when he deals with language use, Davidson actually denies  that there is anything intrinsically  “normative” in meaning: interpreters do not need to know any sort of rules or conventions which speakers would obey. Sofor him if there is something normative in meaning, this something is not a feature of meaning itself, it is a feature of its description, and a feature of the theory of interpretation. The only aim of interpretation is getting people right. The notions of convention, rule, practice, norm, according to him, is of no use in the practice of interpreting another person’s words. It’s only and idle wheel. As he says famously, “just as Lear  gains power through the absence of Cordelia, I think treatments of language prosper when they avoid uncritical evocation of convention, linguistic rule or language games. [7]

        On Dummett’s view, on the contrary, such notions as rules or conventions attached to particular expressions and sentences are essential to a description of semantic competence. What a speaker knows, when he know how to speak a language are certain patterns of correct use of his words. These patterns are relatively fixed, and their set constitutes what it is to be a language.[8] Meaning, therefore, is normative in the sense that speakers know ways of evaluating the applications of their expressions. This knowledge belongs to what Dummett call the “assertibility conditions” of sentences, in terms of which he proposes to formulate a theory of meaning.

        If we focus in this way on the normative character of meaning, we have a better grasp of Dummett’s emphasis on use, and of his descriptive stance. Indeed it should rather be called a normative stance. We can also understand better why Dummett, like Davidson, eschews any attempt to reduce meaning to a set of semantic rules which would be “tacitly” known by speakers, in the Chomskyan or cognitivist fashion. For such “rules” are only “rules” by courtesy: they are not rules which the speaker himself recognizes. A rule, to be a rule, must be articulable by the speaker at the conscious level, and it makes not sense to say that a rule is folowed tacitly or unconsciously (unless one reduces a rule to a regularity). This is the gist of Wittgenstein’s reflexions on rule-following.[9]

        The emphasis stressed on the normativity of meaning, as an instrinsic  feature of the rules of usage that the speakers attach to their expressions, and sanctionned by a community of users, is but another way of insisting on the fact that the speaker’s semantic competence must not exceed what he is able to grasp, and what he can recognize, from inside  his practice, as right or wrong about his use. Dummett’s (or for that matter, the Wittgenstenian) objection to Davidson’s account is that the construal of semantic understanding  as an extrinsic  feature of the speaker’s practice (a feature of the theory of interpretation) generates “an intolerable divide between the concepts of meaning and understanding”[10].


4. A dilemma for the meaning-theorist


     Now how can we appraise the argument from the normativity of meaning? Given its obvious links with the three difficulties that I raised about Davidson’s view, I take it seriously. But if we formulate it as a requirement that we must reintroduce the notions of rule, of practice, or of norms of usage within the project of a theory of meaning, it is in fact quite destructive of this very project.

        It is not quite clear, first, that Dummett himself can integrate this requirement within his own descriptive stance on a theory of meaning. For he conceives of this project as requiring that we specify the meanings of the expressions of a language without using the notion of meaning , i.e without presupposing the speaker’s prior command of any language. Remember that Davidson’s conception is “modest”, since it presupposes that the interpreter has a command of his own language when he interprets others. In this sense, he does not construct his theory of meaning “from the outside”. On the contrary, the requirement of “full-bloodedness”, as Dummett’s calls it, prescribes that the theorist of meaning takes an “outside” perspective on the speakers. It amounts to the idea that in order to be able to account for a speaker’s grasp of meanings through an account of his linguistic abilities. To account for what a speaker understands, we must, for Dummett, state what he is able to do, to describe certain skills. How are we to construe this claim? It seems that, like any skill or ability, ability to master a form of expression must be the outcome of a tacit  or implicit  knowledge, which is not articulated at the propositional level. Dummett, however, does not want to account for this form of tacit knowledge in the psychological, Chomskyan fashion, as a knowledge of tacit rules governing the expressions and included somewhere in the speaker’s cognitive capacity. Sometimes, Dummett speaks as if this tacit knowledge could be described in a behavioural  fashion, as is we could associate to each kind of expression a certain kind of behaviour. But this raises familiar problems: there is no set of behaviour(s) which can be associated to a set of meanings, or of beliefs, of a speaker. Dummett rejects holism  in the theory of meaning, and he criticizes precisely Davidson for being committed to a form of holism in his theory of interpretation. But holism seems to be unavoidable at least at the methodological level.[11]

        If he wants to stick to the normativity argument, and if he does not want to espouse a psychological view of tacit knowledge, the partisan of the descriptive stance must admit that there is no more to rules of usage than what speakers can themselves know, quite openly, and not through some form of implicit knowledge. He must grant the Wittgensteinian point that rules are “there”, laid in view, and there is no “deeper” account of them either in psychological or in behavioural terms. But if he grants this, the partisan of the descriptive stance must also grant that a systematic  description of meaning in the form of a “theory” is threatened. He allows that speakers have knowledge of their language which does not reach deeper that the judgments that they make about the correctness of certain patterns of usage, but that when we reach these judgments, we reach “bedrock”.[12] Such a move would amount to the view that what we could do is only a descrption , from the inside of a community’s practice, of the rules of usage. This quite Wittgensteinian move would be, as it were, purely descriptivist.

        But it seems then that we face a dilemma. If, on the one hand, we insist on the fact that for a TM to be a genuine theory of understanding, it must not be divorced from what a speaker himself (and not an interpreter) must actually know and master, such a knowledge must be knowledge of the rules that are in order within the practice of a community which speaks this language. But these rules cannot be associated in any systematic way with the speaker’s competence, and do not lend themselves to a systematic account. So the argument from normativity prevents us from giving a TM. On the other hand, if we accounted for semantic competence in terms of a tacit knowledge of semantic rules instanciated in the psychology of the speaker, the normativity of linguistic usage would be lost. Rules would become merely causalregularities , and normative force would be lost. So there is a tension between the requirement of descriptivity (of use) and the requirement of systematicity (of a TM).


5. Another way of dealing with normativity


    The tension, and the potential dilemma, can indeed be removed if we reject he argument from normativity the outset. Indeed this is what Davidson does. But it is also what the psychological stances does. According to Davidson, the normative features of meaning are but epiphenomenal features that we attach to meanings when we know our language. But an account of understanding language does not need them. According to the partisan of the psychological stance, knowledge of language is the product of an explanatory psychological capacity which allows us to recognize, tacitly, certain features of syntax and semantics, which are internal to the speaker’s psychology. This is indeed in line with Chomsky’s attitude towards the notion of a language itself. For him, what the description of semantic competence deals with is not an external  language, in the ordinary sense of a set of rules mastered by the speakers of French, English, Portugese, etc., but what he calls an I-language, an internal language, or the idiolect of a speaker (Chomsky 1986)[13]. But on this view, normativity of linguistic usage would also be an epiphenomenon.

        But taking the psychological stance is at odds with the phenomenology of linguistic understanding, just as the notion of tacit knowledge of semantic rules is at odds with the customary notion of knowledge. Can speaking a language really be described in norm-free terms?

      It seems to me that we cannot, and that on this point the argument from normativity is correct. But it makes the norms mysterious if it just registers the fact that certain applications of words are correct and other not, and the fact that a community sanctions certain uses, and not others. In that respect I agree with the Chomskyans and with Davidson that their is something obscurantist in the uncritical evocation of rules and practices. How then could be remove the dilemma?

     It seems to me that we can, if we do not divorce the various stances that we distinguished: interpretative, descriptive and explanatory. There is no reason why certain patterns of use of words that we recognize as correct should not be explained, at least partly, through the exercise of a psychological, tacit, competence in the way it is envisaged by the psychological stance. If we find that a word applies, that a sentence is correct, it may well be that it is because  we have certain tacit knowledge, deeply rooted in our psychology, which allows us to recognize these features, just as other “modular” structures in our visual system aloow us to recognize certain visual scenes. Now the problem with this causal “because” is, as the critics of the notion of tacit knowledge of rules point out, it is not substituable readily to a  normative “because”. Speakers have the feeling that they obey usage not because of certain causal patterns in them, but for certain reasons. But the fact that reasons explanations are usally divorced from causal explanations does not imply that there is no connexion between the two. To take a closely related case, consider perceptual knowledge. To have a true perceptual belief, and for such beliefs to be knowledge, it must be justified in some way. And the justification does not simply amount  to the existence of a reliable causal mecanism which produces the belief. But this causal element can be part of the justification. Indeed it may be held that a justified perceptual belief is a belief produced by a causal mecanism which is known to be reliable. [14] I would claim the same for knowledge of language: to know what an expression means, one has to know that the kind of processes by which we grasp this mean are reliable. This is certainly not enough to give us "knowledge of meaning" in the intuitive sense, but it is certainly part of it. Even if we grant this, it will be said, does it account for the normativity of meaning?

    No, it does not, since normativity is a feature which belongs to the first-personal level of knowledge of meaning. I must have access to the meaning of my words in order to be able to appraise their use. In fact, as I have already pointed out above, Davidson’s conception of interpretation incorporates this feature. Davidson grants that there is a first-personal access of the interpreter to the meanings of the words of his language (or of his idiolect), and that it is where we start from. But here somebody who is impressed by Wittgenstein’s rule-following considerations will inevitably raise the question: how do I know that what seems to me to be right is  actually right? What reason is there for basing any account of the meaning of another person’s words on what I  take my own  words to mean? Davidson answers that this first-personal knowledge of meaning is only the starting point of the interpretation procedure. It is essential that, at a later stage, the interpreter should compare his own conception of the meaning of his words to the conception that he hypothesizes in others. He might even be led to correct his own account.  A theory of meaning is the product of a process of comparison between what the interpreter takes his own words to mean, what he takes another person’s  words to mean, and the external features of an environnement. This triangle of elements will give the proper fit. In other terms,the interpreter will triangulate. [15] Thus we have in fact, within Davidson’s own conception, a clue for an account of the origin of the normative character that we attach to the meanings of our words. This origin, I suggest, is located in the first-person authority that we have about our own judgments about meanings. Davidson, however, does not say much about this first-person authority.[16]

        Let us try to say a little bit more, although sketchily. When I judge that a certain word, say “cat”, has a certain meaning, say that it applies to these feline animals around there, I have certain belief. Now this belief may be wrong: I may misapply he word, and confuse cats with, say foxes, or they might have been robots sent by Martians, unknown to me. This is quite familiar from Twin-Earth kinds of arguments, or from externalist arguments such as Burge’s.[17] So I may be wrong about the content  of my belief. But in general (except perhaps in cases of cognitive disorders) I cannot be wrong about the fact that I have this belief. My self-ascribed belief that I believe that these are cats  is always warranted, an immune to error. As Davidson (1988) and Burge recognize, this is perfectly compatible with externalism about content.[18] I suggest that this authority that we have over our own self-asscribed beliefs is the origin of the authority that we have over our own judgments about meanings. We may be wrong, and the community, or our subsequent exchanges with our fellow speakers might prove us wrong, about the nature of these meanings, but it cannot prove us wrong about the fact that we have such judgments. This feature is in fact a feature of our beliefs in general, and we could put it in terms of the well known “Moore’s paradox”: “ There are cats, but I do not believe that there are cats” is not only an impossibility about our assertions, but also an impossibility about our thoughts.  A speaker would who think such thoughts would be in an incoherent state of mind, or he would not master the very conditions of his having beliefs.[19] To believe that p  is to believe that  p is true, and to believe that one believes that p.  This constitutive or conceptual link between believe, believing true, and believing that one does believe, is integral to belief. Now consider Davidson’s proposal about theories of meaning. He claims that interpretation gets started from sentences held-true  by speakers, and held-true by interpreters. The basic propositional attitude of “holding-true” certain sentences of one’s own language is the condition of interpretation. Because it is authoritative, its authority is conferred to judgments about meaning, or about self-knowledge of meaning. That we are always correct about such attributions is the origin of the fact that we take them to be normative. Indeed the community imposes it norms on meaning. But on Davidson’s account of interpretation, we could not recognise these norms unless we could not recognize the basic normative character of our own self-ascriptions. So we have here, in the end, the connexion between meaning and normativity. It does not lie in the rational norms, such as the principle of charity, that are presupposed by the interpretation procedure, but it lies in this first-personal feature of our own thoughts. In this way, we can reduce the gap that seemed to exist in the first place between Davidson’s third-personal point of view on meaning and the first-person point of view required by the notion of understanding a language.

        This is only half of the story. For remember that one of the main objections raised by Dummett about Davidson’s kind of conception of meaning-theories (TM) was that they are overtly realistic, in the sense that they construe meaning in terms of “transcendent truth-conditions” which might be inacessible to speakers. On Dummett’s view, if such truth conditions are inacessible to speakers, we cannot say that they understand them. But this anti-realist threat does not exist, if we reflect on the fact that the interpreter, on Davidson’s view, starts up with the truth conditions that he  understands. The truth-conditions do not reach farther that those that he himself is able to grasp. Still, it may be claimed, the concept of truth that Davidson’s kind of TM is a realist one in Dummett’s sense. The problem is not that it is realist, but that it is deflationist. On Davidson’s view, a sentence is not true because it is “true to the facts”, or because of the existence of a particular relation of correspondance with an independent reality. There is no more to truth than the Tarskian equivalence “ S is true iff p  ”.[20] But this seems to deprive us from recognizing one important feature of truth: that truth is not simply a feature of our asserting certain sentences as true, or taken-them to be true, or holding true these sentences. Truth is also the aim of our assertions, and in this sense, it is a norm of our assertions that we aim at truth. I have also suggested that it is a norm of our beliefs that we aim at truth.[21] This feature should be incorporated in Davidson’s interpretative conception of meaning. I am not sure that Davidson would agree. But what I suggest is that if we revise Davidson’s theory of interpretation in order to take these normative features into account, we shall reduce the tensions which exist between the various stances on a theory of meaning that we have distinguished.






Bilgrami, A. 1992Belief and Meaning, Blackwell, Oxford

Burge, T. 1979 “Individualism and the Mental”, Midwest Studies in

           Philosophy,  4

            1988 “Individualism and Self-knowledge”, Journal of Philosophy,

           85, 649-663

Davidson, D. 1984 Inquiries into Truth and Interpretation, Oxford,

             Oxford university Press

             “First-person authority”, Dialectica, 38, 101-110

              1986 "A nice derangement of epitaphs", in Le Pore, ed.

              Meaning and Interpretation, Essays on Davidson, Blackwell,


             1987, “Knowing one’s own mind”, Proceedings of the American

              Philosophical Association, 60, 441-458

             1991 “Three varieties of knowledge”, in A.J. Ayer,

              Memorial Essays, ed. A. P. Griffiths, Cambridge, Cambridge

             University Press

          “The Folly of Trying to Define Truth”, Journal of Philosophy,

            XCII, 6, June 1996.

Davies, M. 1981 Meaning, quantification and necessity , Routledge,


Dummett, M.(1959) “Truth”, reprinted in his Truth and other enigmas,

             Duckworth, London, 1978

           1975  "What is a theory of meaning (1)", in S. Guttenplan,

       Mind and Language, Oxford, Oxford University Press 1975

Engel, P. 1994 Davidson et la philosophie du langage, Paris, PUF 199

              1995 “Une théorie naturaliste de l’esprit évacue-t-elle la

                subjectivité?,Cahiers de philosophie de l’Université de Caen,

              28, 113-127

               1996 Philosophie et psychologie, Paris Gallimard 1996

              1997 La vérité, réflexions sur quelques truismes, Paris, Hatier

Goldman, A. 1986, Epistemology and Cognition, Harvard, Harvard

             University Press

Kripke, S. 1981, Wittgenstein on rules and private language, Blackwell,


Mc Dowell, J. 1987 “In defense of modesty”, in Taylor 1987, 59-80

Montague, R. 1974 Formal philosophy, Yale University Press

Shoemaker, S. 1996 The first-person perspective, Cambridge, Cambridge

                University Press

Smith, B. C. 1992 "Understanding language", Proceedings of the

           Aristotelian society ,  109-141

 Taylor, B. ed 1987 Michael Dummett, contributions to

                          philosophy,Kluwer, Dordrecht     

Wright, C. 1986 Realism, Meaning and Truth,  Blackwell, Oxford

                1992 Truth and Objectivity, Harvard, Harvard University



*  This paper is the product, in a compressed form, of three lectures given in April 1997 at the Universitade Nova de Lisboa, in the Department of philosophy of language, at the invitation of Antonio Marques and Joao Saagua. I thank them for their invitation, their hospitality, and their comments, as well as the participants in these occasions, among whom Desiderio Murcho.

[1] I have done that in Engel (1994). the contrast between a “theory of meaning” in the philosophical sense and a TM (or a “meaning-theory”) is well put by M. Davies 1981, ch.1.

[2] For the opposite view, see for instance Montague (1974)

[3]B. Smith’s important paper provided the stimulus for the present one. See also Engel 1994, ch.7

[4] Some writers, like Wright (1986) have attempted to formulate, on Wittgensteinian grounds, “criterial semantics”; but these attempts are not congenial to Wittgenstein’s asystematic  conception of meaning.

[5] This point is well put in Wright, 1986, preface

[6]The notion of a “higher profile” here is borrowed from Bilgrami (1992) who set up very well the contrast between these two sorts of conceptions of normativity.

[7] cf. in particular “Communication and convention” in Davidson 1984, and Davidson 1986.

[8] see in particular Dummett’s (1986) criticism of Davidson 1986

[9] cf. Kripke 1981

[10] The phrase is not Dummett’s, but Wright’s (1987, p.210), commeting about Forster’s characterisation of Davidson’s project.

[11] I cannot dwell upon this here. See Engel 1994, ch.5.

[12] This is, in substance, Mc Dowell’s reaction (1987) to Dummett’s view. See his exchange with Dummett in Taylor 1987, and the account of it by Smith (1992), p. 127-129.

[13] Actually, as it has often been noticed, Davidson’s interpretative view insists on the idiolect of the speaker too (see especially Davidson 1986).

[14] This version of reliabilism seems to me to be that defended by Goldman (1986). It does not eleminate normative elements. cf. Engel 1996, ch. 5, where I also claim that there is not necessarily an untolerable divide between normative explanations in terms of reasons and causal explanations in terms of psychological facts.

[15] See in particular Davidson 1991. In his recent Jean Nicod lectures, given in Paris in 1995, Davidson has insisted on this triangulation as the basis of “objectivity”.

[16] But see Davidson 1984a. He reduces first-person authority to the very feature of the interpretation procedure, which starts from the knowledge of meaning that the interpreter has.

[17] See for instance Burge 1979

[18]  See also Engel 1997

[19] see Shoemaker 1996 about variations on this theme.

[20] See Davidson 1984, and Davidson 1996

[21] Dummett stress this point in in his classic paper “Truth”(1959), as well as Wright (1992). I have argued that a deflationist position on truth suffers from this ignorance of the normative nature of the concept of truth in Egel 1997. See also Engel 1994, p144-146, 219-220.