Comments on Donald Davidson: Meaning, Truth, Language, and Reality

Simon Evnine, University of Miami

Donald Davidson is, notoriously, a philosopher who has appeared as many things to many people. He has been interpreted in the light of Husserl, Heidegger, Derrida and the American Pragmatists. The Davidson who appears in the pages of Lepore and Ludwig’s book is none of these exotic things. They offer a straight-ahead, analytic Davidson and their rendition is, I think, pitch perfect. Not only do they get Davidson right in general terms, but on every disputed point of exegesis, their careful reconstructions of Davidson’s views are exactly right. This book is, therefore, a very impressive, and a very useful achievement. It will provide the essential reference point for all future discussions of Davidson’s work on, as their subtitle says, meaning, truth, language and reality. (They touch only lightly on action theory and not at all on anomalous monism or the nature of events.)

The authors’ account of Davidson’s oeuvre falls into three parts which, roughly speaking, follow the chronology of Davidson’s papers, thus giving us an early, middle and late Davidson. The early Davidson, of the late 1960s, set the course for everything that was to come. He was concerned with the nature of meaning, but, as he says, he “became frustrated with the fatuity of the attempts at answers I found in Ogden and Richards, Charles Morris, Skinner and others. So I substituted [for the question “what is meaning?”] another question which I thought might be less intractable: What would it suffice an interpreter to know in order to understand the speaker of an alien language, and how could he come to know it?” Davidson’s answer to the first part of his question, what would it suffice an interpreter to know in order to understand the speaker of an alien language, was a Tarski- style truth theory for the language.

Lepore and Ludwig actually discern two distinct issues Davidson was getting at with his invocation of Tarski. What they call the ‘original project’ was simply to show how one could specify the meanings of a potential infinity of sentences on the basis of something finite. Lepore and Ludwig suggest that if one takes the axioms of a truth-theory, axioms which say things like “’Caesar’ refers to Caesar” and “’is ambitious’ is true of something iff it is ambitious”, as being interpretive – that is, as in some sense saying what the object language terms mean – then the truth theory shows how one can generate interpretive theorems for any object language sentence.

This is not, of course, how Davidson ever put things. The reason is that the proposal for how to specify the meanings of the infinite number of sentences in a language depends on treating

the axioms of the theory as interpretive. This is to rely on some notion of meaningfulness which is taken as primitive. The use of the truth-theory shows merely how to go from the finite base to the meanings of an infinite number of sentences. It does not show what it is for theorems or axioms to be interpretive. The attempt to use a truth-theory to accomplish this latter task is what the authors call the extended project and it is this that occupies most of Davidson’s attention in the philosophy of language. (The original project returns in a companion to this volume, Donald Davidson’s Truth- Theoretic Semantics.)

How can a truth theory fulfill the extended project? The authors show the evolution of Davidson’s thinking on this. At first, he thought that if one confirmed enough theorems, of the form “’s’ is true iff p,” that would be enough to treat those theorems as interpretive – as saying that “’s’ means that p.” But it quickly became apparent, to Davidson and his critics, that this was much too weak. Confirming enough such theorems, and projecting a set of axioms from which those theorems and others could be generated, would indeed yield a well-confirmed truth theory for the object language, a theory that correctly determines the extension of the predicate “is true” over sentences of the object language. It would not show that the theorems of the theory could be taken as saying anything about meaning.

Searching for what one would have to know of a truth theory, besides its extensional adequacy, if it were to be taken as interpretive, brought Davidson to his middle period, roughly, the 1970s, the era of radical interpretation. Davidson’s commitment to the idea that meaning is essentially a public phenomenon, what the authors call the primacy of the third person point of view, led Davidson in the direction of trying to show how publicly available evidence, behavioral evidence, could be used to confirm a truth theory. Confirming a theory on the basis of such evidence alone is the project of radical interpretation. But how exactly is this supposed to work? The authors reveal a certain confusion in Davidson’s views on this. The question is whether confirmability by the radical interpreter is a restriction on how a truth theory must be confirmed if it is to be interpretive, or whether what must be confirmed is that the theory can be confirmed by a radical interpreter. Once this question is put, it seems clear that neither answer can be correct. If, from the point of view of a radical interpreter, one confirms that a given truth theory is indeed correct for a given language, then all one knows is that that truth theory is true, i.e. extensionally adequate. But as we indicated above, mere extensional adequacy is insufficient to guarantee that a truth theory so confirmed is interpretive. On the other hand, why should a radical interpreter’s (or anyone’s) showing that a truth theory can be confirmed by a radical interpreter add anything to showing that the truth theory is extensionally adequate that is relevant to its being interpretive? Ludwig and Lepore’s response is that what Davidson should have said is that a radical interpreter must be able to confirm a truth theory and confirm that it is interpretive, thus picking up on the idea they introduced in redescribing Davidson’s original project of providing a compositional theory of meaning.

Presumably, at least one reason that Davidson did not put matters thus is that it appears to beg the question. If a radical interpreter can confirm that a truth theory is not just extensionally adequate but also interpretive, is she not already employing a semantic concept of the kind that is supposed to be the outcome of radical interpretation, rather than its enabling condition? Ludwig and Lepore argue that their suggestion does not beg the question, though. Semantic concepts may not enter into the description of the evidence the radical interpreter uses to confirm the truth theory (and confirm that it is interpretive); they may, however, enter into a statement of what the radical interpreter is trying to confirm on the basis of that evidence. I do wonder, though, what is involved in showing that a truth theory is interpretive. The authors seem to reject the view that showing that the truth theory can be confirmed from the point of view of a radical interpreter suffices to show that it is interpretive. I am not sure, however, what criterion of adequacy they substitute.

The method of radical interpretation is described in fascinating detail. Exactly what form the evidence takes and how one gets from that evidence to a truth theory, via the principle of charity are carefully discussed and laid out perspicuously. Ultimately, however, the authors argue that Davidson fails in his attempt to show that a radical interpreter could confirm that a given truth theory was interpretive of a speaker. Thus, the extended project attributed to the early Davidson must also fail. Davidson does not succeed in showing what it is for language to be meaningful.

The third part of the book, dealing with the late Davidson (from about 1980 onwards) deals with consequences that Davidson thought followed from his views about radical interpretation, and also with some arguments that might be thought to rebut the authors’ conclusion that radical interpretation is impossible. The consequences include rejection of alternative conceptual schemes, an account of a person’s knowledge of her own mental states, a rejection of skepticism and the necessity of language for thought. All of these famous positions associated with Davidson are also judged to fail, since they all, in one way or another, presuppose the possibility of radical interpretation. Davidson’s well-known triangulation argument, which might be hoped to take care of the authors’ arguments against the possibility of radical interpretation, also does not succeed.

The heart of the book, then, is the authors’ argument that radical interpretation is not possible. Their reason for thinking this is that the whole idea of using radical interpretation, in which a person tries to arrive at a theory of what another means and believes on the basis only of publicly available (i.e. behavioral) evidence, depends on seeing the theoretical concepts of an interpretation theory (the concepts of meaning, belief, etc.) as what the authors call “purely theoretical.” That is,

their significance must be taken to be exhausted by what they imply for the domain of behavioral evidence that determines when they are applied or not. Advocacy of the pure theoreticalness of the concepts of meaning and belief counts as commitment to the adequacy of the third person point of view. The reason, then, why the authors find the possibility of radical interpretation problematic is that they think that the third person point of view is not adequate. In particular, the fact that people can know about their own meanings and beliefs without the kind of behavioral evidence which is supposed to exhaust the content of these concepts from the third person point of view, shows that there is more to them than their implications for behavior. Davidson struggled to give some account of a person’s knowledge of her own mental states and meanings that was consistent with his overall third-person approach. Few have found Davidson’s attempted solutions to this problem at all convincing. What is novel in Ludwig and Lepore’s critical treatment of this part of Davidson’s philosophy is the way in which they connect Davidson’ failure here to a failure in his whole philosophical project. The theory of radical interpretation is seen as part of a grand rejection of a Cartesian tradition in philosophy. But this rejection stumbles over a modest Cartesianism that just won’t go away: people do know their own meanings and beliefs without evidence and with greater surety than that with which they can know the meanings and beliefs of others. Unless one can finesse that fact away, no account of meaning and belief which attempts to exhaust their significance in behavioral evidence, as radical interpretation does, can possibly succeed.

The asymmetry of first- and third-person attributions of meaning and belief has proved a thorn in the side of every theory that has ventured into that territory. I cannot help feeling that Davidson was on the right lines here, at least to this minimal extent: a solution should come from some kind of deflationism about the first-person point of view. I think that my instinct on this puts me at odds with the authors who, contrary to prevailing philosophical currents, seem to hint at a sneaking admiration for Cartesianism. About how deflationism about the first person should go, and whether it will ultimately be compatible with other Davidsonian commitments, I have no idea.

The general reason the authors have for thinking that radical interpretation must fail manifests itself at a greater level of detail in their treatment of indeterminacy and its relation to the mechanics of radical interpreation. Lepore and Ludwig argue that the evidence available in radical interpretation underdetermines an interpretive truth theory. That is, working from the admissible evidence alone, one is able to confirm a wide range of distinct truth theories equally well. Davidson, of course, has expressed allegiance to the notion of indeterminacy of meaning. It might, therefore, be thought that the underdetermination that the authors detect is simply an instance of indeterminacy and therefore not harmful to the Davidsonian project. Ludwig and Lepore’s case against the possibility of radical interpretation depends on characterizing indeterminacy and distinguishing it from the underdetermination of truth theory by evidence which they find. Indeterminacy is a special case of underdetermination. It occurs when all the relevant evidence fails to favor one of a number of theories and the concepts in those theories are purely theoretical, in the sense we specified above. In that case, the authors say, the several theories will “express the same facts.” In mere underdetermination the various theories will not express the same facts.

Since, as we have just seen, the authors disagree with Davidson’s view that the concepts of an interpretation theory, the concepts of belief and meaning, are purely theoretical, it is clear that they will take any underdetermination allowed by the method of radical interpretation to be of the non-indeterminacy variety. Supposing that radical interpretation does allow mere under- determination, its use will lead to a number of different theories which do not “express the same facts.” Each of these theories will have the form of a truth theory for the language under investigation. If someone knows that the totality of evidence is compatible with distinct truth theories for the language then, even if it is permissible arbitrarily to adopt one of these theories as a truth theory for the language, it could not be taken as an interpretive truth theory. Hence, if mere underdetermination is possible in radical interpretation, it cannot succeed.

How is underdetermination possible in radical interpretation and why must it be of the non- indeterminacy variety? Here is a crude reproduction of the authors’ long and complex argument. The radical interpreter starts with behavioral evidence and on that basis attempts to confirm a set of L-sentences for a speaker. L-sentences say, roughly, under what conditions the object of interpretation holds true a given (occasion) sentence. When we have a set of L-sentences, we use the principle of charity to infer from them a set of TF-sentences, which say under what conditions those occasion sentences are true. From the TF-sentences, we project a set of axioms that articulates their compositional structure, and the set of TF-sentences plus those axioms constitute a truth theory for the language we are interpreting. We can then take this truth theory as interpretive, if we know how it was arrived at (and perhaps some other things, such as that its concepts are purely theoretical – I’m not sure about this point, as I mentioned above).

Undertermination may creep in at various points. For example, when projecting the axioms of a truth theory from the confirmed TF-sentences, there may be several ways to do that. So long as the concepts of reference and satisfaction are taken as purely theoretical, however, such underdetermination will be harmless indeterminacy. The problem comes, according to Ludwig and Lepore, over the possibility of many, non-equivalent sets of L-sentences an interpreter can infer from the behavioral evidence. (They show how if there is one set, there will be many.) There will thus be non-equivalent TF-sentences and non-equivalent truth theories. Our knowledge of the

existence of (or the possibility of the existence of) such alternative truth theories in turn will prevent us from taking any of them as interpretive.

Why cannot this be treated as a case of benign indeterminacy? Ludwig and Lepore write:

It is not clear from the point of view of the radical interpreter that he can treat different theories he could confirm as stating the same fact, which is required in order to treat the resulting range of theories as an expression of… indeterminacy… rather than simply underdetermination [of the non-indeterminacy variety]… From the point of view of the interpreter, the different L-sentences he can confirm are non-equivalent…[This will] generate corresponding TF-sentences which from the interpreter’s perspective must be treated as assigning different interpretations to the speaker’s sentences. Thus, the interpreter must regard the different theories he can confirm as strictly incompatible with one another (not just apparently or intuitively)… In other words, it is incoherent for the interpreter to regard the different theories which he could confirm as both true. (239)

Their point is that the interpreterknows that different starting points (different sets of L-sentences) will lead to different TF-sentences. These TF-sentences are what will be used to interpret the sentences of the interpretee. But they use, on their right-hand sides, sentences of the interpreter’s language which the interpreter herself knows to mean different things. How, then, can the interpreter take the different interpretation theories to express the same facts? The argument, therefore, hinges on the interpreter being able to know something about her own meanings that she cannot find in others on the basis of radical interpretation.

There is, actually, one way in which an interpreter could choose indifferently among theories she knew to be non-equivalent in interpreting something: if she took her own conceptual repertoire (or language) to be richer than the interpretee’s. This is what happens in the application of intentional concepts to non-human animals. Although I know that “my feeder is home” and “S.E. is home” mean different things, the difference is irrelevant when interpreting my cat. But in interpreting other humans, this cannot work. The authors point out that the interpreter may be faced with others who speak the same language she does. In that case, the interpreter cannot take her own language to be richer than her interpretee’s. Another reason, not given by Lepore and Ludwig, but consonant with their outlook, is that an interpreter must, on Davidson’s view, take the objects of intepretation to be able to interpret back. If a condition of interpretation were that the interpreter must take her own language to be richer than the interpretee’s, this would lead to a

conundrum: A’s taking distinctions in her language to be irrelevant in interpreting B, while also holding that B is taking distinctions in her language to be irrelevant in interpreting A.

As I stated above, the authors take themselves to be showing that radical interpretation entails a kind of underdetermination that is not of the benign indeterminacy variety. But Davidson describes indeterminacy as allowing for “equally acceptable alternative theories which differ in assigning clearly non-synonymous sentences of [the interpreter] as translations of [the interpretee’s] same utterance.” The authors, therefore, might better have put their point by saying that, contrary to Davidson’s claims, indeterminacy is not benign after all. And the reason it is not benign has to do with the fact it will represent as irrelevant differences that an interpreter will know are relevant, in application to herself.

The book, after 420 dense, difficult and fascinating pages, ends abruptly, in this sense. Having argued that the project of radical interpretation must fail, and with it a variety of arguments Davidson pressed on its basis, the authors do not take any final stock of Davidson’s contribution to philosophy. If the authors are right, almost everything Davidson wrote on meaning, truth, language and reality is flawed. (The same is true, of course, of all interesting philosophers.) What, then, should we make of Davidson? What are the lessons to be learned? How does Davidson’s project fit into the bigger picture of 20th century philosophy? It is, perhaps, understandable that, having stared so hard and so perceptively for so long at the details of Davidson’s oeuvre, the authors should have found it difficult to refocus their gaze on such Brobdingnagian questions. But Lepore and Ludwig’s already excellent book might have been even better had they done so.