Davidson and Indeterminacy of Truth Conditions
Interview by Richard Marshall.
[Images: Giacometti]
'The main catalysts for the rapid rise of the philosophy of language in the 20th century were the development of modern logic and the so-called “linguistic turn” early in the century. Of course, philosophers have now generally abandoned the idea that philosophical problems are to be answered by looking at language. However, even if “linguistic” philosophy has been demoted from its status as first philosophy, this hasn’t taken the form of an outright denial of the relevance of language to the rest of philosophy.'
'While I do think that Davidson’s indeterminacy thesis can’t avoid collapse into a radical skepticism about meaning, I don’t think that this, by itself, gives us a good reason to abandon his approach. This is because I am not at all convinced that interpretation, even if approached Davidson-style, is really subject to the kinds of underdetermination that he regards as inevitable.'
'The point of a truth theory, formulated in my language, for your language, is not to specify your language, but to provide a model — or representation — of it. As long as there is a determinate language to which my own sentences belong, I have, in constructing a truth theory for your sentences, specified a possible language.'
Arpy Khatchirianworks in the philosophy of language, the philosophy of mind, and related areas in metaphysics and epistemology. Her interests include the epistemology of language, truth, deflationism, and indeterminacy. She also maintains an active interest in the history of analytic philosophy. Here she discusses ordinary language philosophy, its link with philosophy of mind, Donald Davidson and his theory of meaning, his indeterminacy thesis, radical skepticism, why we shouldn't abandon Davidson's approach, why meaning holism and compositionality don't threaten Davidson's approach, puzzles and facts about meaning, and responses to meaning skepticism.
3:AM:What made you become a philosopher?
Arpy Khatchirian:I grew up in Lebanon during the civil war. As a child, my response to the chaos and absurdity surrounding me was to immerse myself in the visual arts. During my teenage years, I became increasingly frustrated with art. I began to doubt my aptitude for it, but also found myself preoccupied with the questions of the nature and point of art as a practice, and of what constitutes a “finished” piece.
After several abrupt moves, my family emigrated to the US — we ended up in New Jersey — just before my last year of high school. During my senior year of high school, I thought, for reasons that are obscure to me now, that I might study architecture. I went to Rutgers University, as it was relatively affordable, but it turned out they had no architecture program! Of course, it didn’t take me long to realize that I felt more at home in abstract disciplines. I ended up majoring in math and philosophy — it was a lucky accident that Rutgers had strong programs in both fields!
Through most of college, I thoroughly enjoyed math and took it more seriously than philosophy. But things changed during my last year. Much like the way puzzlement about what constitutes art had earlier displaced my enjoyment of the activity itself, my increasing preoccupation with the nature and point of mathematical practice took over my engagement with the discipline. By the time I started graduate school at Berkeley, this had morphed into a more general preoccupation with the nature of human practices and capacities.
I am not sure if this answers your question. But, looking back, a common thread running through all this might be that art, math, and philosophy were all ways of responding to a world that I found puzzling: art was a way to escape it, math was a bit of the world that made sense, and philosophy, a naive, yet courageous attempt to make general sense of it all.
[Donald Davidson]
3:AM:You’re working in the field of philosophy of languageand mind, and have a particular interest in Donald Davidson. Before we get into that, could you first sketch for us what you see as the importance of philosophy of language. Back in the fifties it seemed to be the only game in town with Wittgenstein and the ordinary language brigade. So what’s happened since then?
AK:Ordinary language philosophers, for the most part, didn’t offer much by way of a theory of language or of meaning. What they offered was a certain approach to philosophical problems that involves treating them as problems involving language, but their diagnostic and anti-systematic approach to philosophical problems applied as much to problems involving language as to other areas of philosophy. At the same time, ordinary language philosophy’s emphasis on use, nuances of ordinary use, and context sensitivity no doubt influenced subsequent theorizing about language, and in particular, formal semantics.
The main catalysts for the rapid rise of the philosophy of language in the 20th century were the development of modern logic and the so-called “linguistic turn” early in the century. Of course, philosophers have now generally abandoned the idea that philosophical problems are to be answered by looking at language. However, even if “linguistic” philosophy has been demoted from its status as first philosophy, this hasn’t taken the form of an outright denial of the relevance of language to the rest of philosophy. For the most part, it has been replaced by a more nuanced conception of the different ways in which questions about language might interact with object-level philosophical questions. So I do think philosophy of language continues to thrive today, even if it no longer dominates the landscape.
3:AM:And how does it link with philosophy of mind?
AK:First, some topics in the philosophy of language — for instance, the nature of linguistic competence — are topics in the philosophy of mind. Second, insofar as our interest is in natural languages, there is bound to be some interplay between questions about the nature of language and questions about the nature of the capacities involved in using language. Of course, philosophers disagree about the extent to which foundational questions in semantics should be seen as enmeshed with questions about the nature of semantic competence. Scott Soames, for instance, thinks it’s a mistake to think of a semantic theory as a theory of semantic competence.
My own interest in language is focused on its interplay with the philosophy of mind and with epistemology. I’m interested in the questions of what it is to understand a language, what it is to be able to communicate with others, and what kinds of knowledge these might involve. And my view is that a promising approach to foundational questions in semantics is one that’s guided by concerns with the nature of semantic competence and the communicability of meaning. My interest is in the nature of our linguistic practices and the capacities underlying them. My questions are not “What is a language? What is meaning?”, but rather, “What are we, as producers and consumers of language, doing here? What does it take for us to be able to do it?”
3:AM:Davidson is a massive figure in all this but maybe a little obscure to those outside. Can you sketch where he comes in? What were the issues and the rival theories he was out to confront with his theory of language?
AK:Davidson wanted to give us a way to think about meaning that was free of certain trappings — three in particular: reductionism, the reification of meanings, and the privacy of meaning.
First, reductionism. Davidson thinks we can shed light on meaning, truth, and other related notions without reducing them to one another or to anything more basic. I do think that Davidson’s anti-reductionist and anti-foundationalist orientation is too often overlooked, which leads to misunderstandings of the nature of his proposals. For example, philosophers have read into his proposal to use truth theories as meaning theories some kind of attempt to reductively explain meaning in terms of truth, or a proposal to replace one notion with the other. Or again, consider his account of radical interpretation: if you assume that Davidson is trying to give a reductive account of meaning, it’s tempting to read into it some form of a “use theory” of meaning.
Second, the reification of meanings. Part of the significance of Davidson’s proposals concerning the form to be taken by a semantic theory is to show how we can do away with meanings as entities — that is, to show that they do no work in explaining the semantic contributions of words to the sentences in which they occur. Instead, one thing Davidson repeatedly insists on is the connection between meaning and understanding: knowing the meaning of a sentence should put one in a position to understand utterances of it. But knowledge of which complex object a theorist assigns to a given sentence as its “meaning” needn’t put one in a position to understand the sentence.
Third, any commitment to the privacy or internality of meaning. For Davidson, communicability is essential to meaning. Whatever it is facts about meaning consist in, they have to be intersubjectively available.
Of course, one should mention the massive influence that Quine had on Davidson. Most importantly, perhaps, Davidson’s anti-Cartesian orientation owes a lot to Quine. But despite Quine’s influence, or perhaps because of it, many central features of Davidson’s approach diverge sharply from Quine’s. To mention a few salient points of contrast: Davidson rejects empiricism, Quine only takes himself to free it of certain dogmas; Quine happily embraces skepticism about intentional discourse, Davidson clearly opposes it; Quine is commonly taken to be a deflationist about truth, Davidson stands firmly opposed to any form of deflationism.
3:AM:So what’s Davidson’s theory of meaning?
AK:This is a difficult question to answer in a straightforward manner, partly because Davidson himself approaches the subject obliquely. Davidson’s approach involves shedding light on the nature of meaning by first asking two questions that might initially not seem to directly concern the nature of meaning.
First, what form should an empirical meaning theory for a given speaker’s language take (where a meaning theory is a theory that, in some sense, tells you what each sentence and meaningful expression of the speaker’s language means)? Of course, to answer this question, one needs to first ask what the point of an empirical meaning theory for a given language is supposed to be.
Second, how could such a theory be confirmed in particular cases? How might a speaker’s ongoing use of words bear on the acceptability of a meaning theory for her language?
What do these questions have to do with the nature of meaning? Well, for Davidson, if you’ve adequately answered the questions of what kind of theory counts as a meaning theory for a particular language, and of how such a theory might be confirmed in so-called “radical interpretation,” then you’ve shed light on the nature of meaning. There is nothing more to be said about what meaning is. And this is in a sense Davidson’s “theory” of meaning.
Now back to the two questions. Here, in a nutshell, is Davidson’s answer to the first question, about the form an empirical meaning theory for a given speaker’s language should take. For Davidson, a meaning theory for a language L is to take the form of a recursive truth theory for L, a theory modeled after Tarski’s truth definitions for particular languages. This theory is to recursively generate, for each sentence of L, a theorem of the form:
s is true in L if and only if p
where what replaces ‘s’ is a name, or structural description, of the relevant sentence of L, what replaces ‘p’ is a sentence in the language of the theory that is true if and only if the sentence of L mentioned on the left is true, and ‘if and only if’ is the material bi-conditional. (Of course, there’s a question about what further requirements — in addition to truth — there might be on such theorems, if the theory entailing them is to serve as a meaning theory. And this has been the subject of much lively discussion.)
Davidson’s answer to the second question is his famous, or perhaps infamous, account of radical interpretation. As I see it, there are three central features to it. First, radical interpretation of a speaker is interpretation “from scratch,” that is, without drawing on any prior knowledge of either the meaning of any of the speaker’s words and utterances, or the content of his psychological states. This doesn’t mean that the interpreter is not presupposing that the speaker’s words are meaningful or that the speaker has a rich mental life, but only that the interpreter does not take herself to know what any of the speaker’s words mean and what his thoughts are. So the interpreter must start with some facts about the speaker that can be gleaned from his behavior without drawing on such knowledge. Davidson’s proposal here is to with facts about the conditions under which the speaker holds various sentences to be true. Second, radical interpretation is governed by the so-called principle of charity, which requires us to aim for a theory that, as Davidson puts it, “finds [the speaker] consistent, a believer of truth, and a lover of the good (all by our own lights, it goes without saying)." In particular, one essential aspect of this principle involves taking the speaker’s utterances, in basic perceptual cases, to be about and true of the very causes — salient to both speaker and interpreter — of these utterances. Third, the constraints on interpretation are supposed to apply holistically, to one’s overall interpretation theory for the speaker, which includes, in addition to a meaning theory, a theory of the speaker’s beliefs and other psychological states.
3:AM:And what’s his indeterminacy thesis regarding language and what’s the problem?
AK:Davidson’s indeterminacy thesis says that there are different equally acceptable theories of interpretation one could construct for another speaker, with no fact of the matter which one is correct. This is supposed to arise from the underdetermination of interpretation theories by the constraints governing interpretation, along with the idea that there are no facts about meaning beyond what an interpreter could discover on the basis of publicly available evidence.
To understand why the labels ‘underdetermination’ and ‘indeterminacy’ are apt here, it’s important to note the kinds of differences Davidson thinks can obtain between two equally acceptable interpretation theories. One type of difference is to be found at the level of reference: Davidson thinks that we could construct two (or more) equally acceptable truth theories for a given speaker that diverge in their assignments of reference to the speaker’s terms — that assign what we think of as different objects to the same referring expressions — but result in the same assignments of truth-values to sentences. This is of course Quine’s “inscrutability of reference,” and Davidson accepts it for essentially Quine’s reasons. Whether such indeterminacy exists, and to what extent it would be problematic, are good questions.
But my own focus has been on another kind of indeterminacy Davidson accepts, found at the level of whole sentences: what we might call “the indeterminacy of truth-conditions.” Davidson thinks there are bound to be different equally acceptable truth theories for a speaker that entail divergent assignments of truth-conditions to some of her sentences: one and the same sentence, as uttered by a speaker on a particular occasion, could come out true on one acceptable way of interpreting her and false on another. This type of indeterminacy is supposed to arise because of the holistic nature of interpretation, in particular, the so-called inextricability of meaning and belief. An interpreter relies on considerations of overall plausibility to disentangle the contributions of belief and meaning in explaining the speaker’s holding-true of a sentence. This is supposed to generate cases where there’s no telling whether an apparent disagreement with another speaker counts as a genuine difference of opinion, because there are ways for us to reinterpret the other’s words that would remove the appearance of disagreement. Since, for Davidson, facts about meaning do not outstrip the evidence available to an interpreter, the upshot is that there are equally acceptable ways of interpreting the speaker such that one and the same sentence comes out true on one of these ways and false on another.
On the face of it, this result seems incoherent: two theories end up being equally acceptable, even though they entail incompatible results as to the truth-value of a given sentence. But Davidson reminds us that the truth of a sentence is relative to a language and describes the alleged indeterminacy as an indeterminacy in the language we can take the speaker to be speaking. If this is right, the indeterminacy here does not involve describing the same sentence as both true and false. Instead, it involves describing the sentence as true in one language and false in another. So all we have to make sense of is the idea that a speaker can be taken to be speaking different languages relative to different, equally acceptable theories of interpretation.
Now, I don’t think this way of trying to make sense of truth-conditional indeterminacy really works, at least not without undermining some of the most distinctive — and in my view attractive — features of Davidson’s approach. As I already mentioned, Davidson is no eliminativist about meaning and intentional discourse. Quite the contrary: he wants to show how we could take intentional discourse seriously while doing away with, on the one hand, meanings and other intentional objects, and on the other, a commitment to reducibility. So his take on the significance of indeterminacy — or, of the kinds of indeterminacy he countenances — contrasts sharply with Kripke’s and Quine’s “no fact of the matter” views of meaning.
For Kripke, accepting the skeptic’s contention that there is no fact of the matter whether one meant one thing or another by the symbol ‘+’ involves facing the prospect that, as he puts it, “the entire idea of meaning vanishes into thin air.” Similarly, for Quine, the indeterminacy of translation establishes “the baselessness of intentional idioms.” By contrast, Davidson construes indeterminacy as involving relativity to an overall theory or scheme of interpretation, but then likens this relativity to the harmless relativity in the measurement of magnitudes to scales of measurement: if I first say that a table is 1 meter long and then that it is 3.28 feet long, I haven’t changed my mind about its length. Likewise, the different ways of assigning truth-conditions to a speaker’s sentences are supposed to simply amount to different ways of keeping track of “the same facts.”
For Davidson’s solution to work, we have to make sense of two ideas: first, that we can equally correctly attribute two different languages to a speaker relative to two different acceptable overall theories of interpretation for the speaker. Second, that when we do so, the different ways of assigning reference and truth conditions to the speaker’s expressions are different ways of capturing the same facts.
My claim is that we can’t make sense of the first idea, let alone the second. Here’s why. If we accept Davidson’s conclusion that there is no determinate language to which anyone’s words belong on a particular occasion, then in particular, there is no determinate language to which an interpreter’s words belong on a particular occasion of interpretation. But if there is no determinate language to which the interpreter’s words belong, then when the interpreter uses these words to construct a “theory of truth” for the sentences of another speaker, she has not thereby specified a particular language. Let me put this in a slightly different way: unless my own words possess certain semantic features in an absolute sense, rather than relative to a theory of interpretation, then I cannot, in using these words, assign semantic features to another speaker’s words. So, whatever it is that I have done in constructing a truth theory or an overall theory of interpretation, it does not involve having successfully pinned down a particular language. But if a particular language has not been pinned down, the speaker’s sentences and words have not been assigned any specific roles. We therefore lose our grip on what it is that distinguishes Davidsonian indeterminacy from the damaging kinds found in Kripke and Quine.
3:AM:So if it leads to a radical skepticism about meaning why not just go to another candidate theory?
AK:Well, while I do think that Davidson’s indeterminacy thesis can’t avoid collapse into a radical skepticism about meaning, I don’t think that this, by itself, gives us a good reason to abandon his approach. This is because I am not at all convinced that interpretation, even if approached Davidson-style, is really subject to the kinds of underdetermination that he regards as inevitable. In other words, the problem that Davidson is attempting to solve might not really arise in the first place.
It’s telling that Davidson doesn’t provide many examples of the alleged indeterminacy of truth-conditions. And the handful of examples he does provide are severely underdescribed. One might think this is because the details are pretty easy to fill in. But I actually think it’s because convincing examples are hard to come by.
So my take is that it is not at all clear that Davidson’s approach really forces us to accept indeterminacy. But if it did, yes, this would undermine its plausibility.
3:AM:You try and find a defence of the theory but it ends up with no one able to know what anybody else means doesn’t it? Can you take us through your argument here?
AK:I attempt a reading of Davidson’s indeterminacy thesis on which it avoids the outcome that there is no determinate language to which any of our utterances belong, since it’s this that forces us into an unpalatable form of skepticism about meaning. But as I go on to explain, this strategy doesn’t ultimately work.
To avoid the outcome that there is no determinate language used by anyone, we need to be able to say something like this: that even if you and I can’t take each other to be speaking a determinate language, we can each see ourselves as speaking a determinate language.
So how might we do this? We could try to do it by changing our conception of the job of a truth theory for a speaker, formulated by and in the language of another speaker. On the reading I consider, the point of a truth theory, formulated in my language, for your language, is not to specify your language, but to provide a model — or representation — of it. As long as there is a determinate language to which my own sentences belong, I have, in constructing a truth theory for your sentences, specified a possible language. You in turn might very well be speaking a certain language, one that you can specify by constructing a theory of truth for your own sentences in your own language. But the correctness of my theory of truth for your language doesn’t require that the language specified by my theory be identical to the language specified by your theory. Rather, the correctness of my truth theory for your sentences simply involves its serving as an adequate model of your language.
One apparent advantage of this construal is that we can happily accept the idea that there might be more than one adequate model of something. But what does it take for the truth theory to constitute an acceptable model of the speaker’s language? The answer would have to be: its playing a role in an acceptable overall theory of interpretation for the speaker. So this reading tries to avoid incoherence by confining the indeterminacy to the interpersonal level: the level of one individual’s interpretation of another’s words. The indeterminacy does not concern the semantic features of a speaker’s own words, as understood by the speaker herself.
But clearly, the resulting picture of communication is deeply unattractive, especially by Davidson’s own lights: the best each of us can try to achieve, in trying to communicate, is to construct internal (i.e., not intersubjectively available) representations of the languages of others! This outcome is not any more intelligible that one that simply denies the possibility of communication.
In other words, I really don’t see how Davidson’s thesis about the indeterminacy of truth-conditions can ultimately avoid collapsing into skepticism about meaning. Saving meaning by making it incommunicable is just not saving it at all.
3:AM:Would you say that the problems faced by Davidson’s theory are problems faced by any meaning holism type theories, and is compositionality a general problem for all of them?
AK:I think that any plausible conception of meaning needs to make sense of successful communication. It needs to explain how two individuals can understand a word, sentence or utterance in the same way, and know that they share this understanding. Similarly, any plausible account of concepts needs to exhibit a concept or thought as graspable by more than one individual and on more than one occasion. It also needs to explain how two individuals can know that they are applying the same concept or “thinking” the same thought.
In particular, this is true of holistic theories of meaning and content, and some have argued that holism threatens the possibility of shared meanings and concepts. The idea here is that if what I mean by a given word depends on what I mean by every other word in my language, and the same is true of you, then you and I can never mean the same thing.
Now, I don’t think that meaning holism is generally problematic, though particular versions of it might be. For instance, suppose you identify the meaning of a word in a speaker’s idiolect with the set of beliefs associated with it, that is, the beliefs the speaker would express by using the word in various sentences. Then it’s hard to see how you can avoid the result that meanings are fleeting and unshareable.
But the kind of holism we find in Davidson doesn’t, it seems to me, have such unpalatable consequences. For one, Davidson doesn’t identify the meaning of a word with any kind of thing: not a set of beliefs, not a set of inferences. He simply doesn’t assign meanings to words. What’s holistic in Davidson’s picture is interpretation: how one person is to determine what another’s utterances and beliefs are about. And the fact that we do so on the basis of holistic constraints doesn’t mean that we can’t be successful — that we can’t make sense of one another as having shared beliefs and therefore shared concepts, or as using words “in the same way.”
To me, the worry that holding meaning hostage to holistic constraints governing interpretation might threaten the possibility of shared or determinate meanings is similar to the worry one might initially feel when faced with the Wittgensteinian point that meaning and communication depend on agreement in judgments. The worry here is that taking the possibility of meaning to depend on agreement in basic judgments makes it too precarious — something that might easily be lost. Similarly, one might worry that if interpretation depends on an interpreter’s finding herself in sufficient agreement with the speaker on basic matters, or on her finding a sufficient degree of rationality in the speaker’s attitudes, then this makes interpretability and therefore the possibility of meaning itself too fragile. In both cases, my reaction is: yes, success here is precarious, in the sense that it depends on contingent facts about us and our circumstances, but isn’t that to be expected? The alternative, that somehow our linguistic acts are bound to be successful, seems unreasonable to me.
3:AM:Where do we go to avoid any troubling skepticism about meaning in contemporary philosophy of language? Do words have determinate meanings, are there facts of the matter about what words mean?
AK:There are facts of the matter about what our words mean, at least in the following sense: there are true claims to be made by one speaker concerning what another speaker’s sentence means, or the conditions under which it’s true, on a particular occasion of utterance. There are also facts of the matter about whether or not two speakers understand a sentence in the same way, and whether or not two utterances of the same sentence, or different sentences, are to be understood in the same way.
But if this is right, and if we want to hold on to the idea that a speaker understands a sentence at least partly on the basis of understanding its meaningful parts and the significance of their modes of combination, then we should also hold that there are facts about meaning — or at least understanding — to be captured at the sub-sentential level. If understanding a sentence involves understanding a language, then there are facts about the contributions made by different bits of a language to determining the meaning of each sentence in the language.
If meaning puzzles us, it’s because we demand so much of it. Here are some contrasting pairs of requirements often imposed on meaning: reducibility and normativity; compositionality and context-sensitivity; publicity and first-person authority; meaning as determining truth-conditions while also being determined by use; as explaining the expressive function of language while also explaining its fact-stating role. Skepticism about meaning is the result of the notion being pulled apart by one or more of these contrasting pairs — for instance, the thought that if meaning is determined by use, then meaning can’t determine truth-conditions.
As for responses to skepticism, there seem to be three possibilities:
(1) The first kind of response involves showing, contra the skeptic, that the relevant constraints on meaning are indeed jointly satisfiable: for instance, by giving a naturalistic account of meaning together with reading the normativity requirement in a way that allows meaning facts to count as normative.
(2) The second kind of response involves agreeing with the skeptic that the constraints are not jointly satisfiable but rejecting one or more of these constraints. One example of this is the argument, against Kripke’s skeptic, that meaning facts are not essentially normative. Another case involves rejecting the requirement that meaning determine truth-conditions, by means of an appeal to some form of deflationism about truth and reference. This seems to pave the way for a use theory of meaning — Horwich’s account of meaning in terms of basic acceptance properties is a good example here — that’s unconstrained by the requirement that facts about meaning themselves determine facts about reference and truth-conditions.
(3) The third kind of response, and the kind of response Kripke offers the skeptic, is a “skeptical solution.” It involves accepting the skeptic’s conclusion that there are no facts about meaning, but showing how we can live with it — that is, showing how our ordinary attributions of meaning are perfectly in order even if there are no meaning facts for them to capture.
Now, I don’t see how we could happily carry on with our linguistic practices if we really adopted the skeptic’s attitude towards these practices. So (3), the skeptical solution, is not really an option for me.
Where then can we go to avoid skepticism? Should we go with (1) or with (2)? Should we try to show that the various constraints on meaning are, despite appearances, jointly satisfiable, or should we reject one or more of them? To make a good judgment here, we’d need to be confident about which alleged constraints are legitimate and what their significance is. But I am not at all sure we’re there yet.
As I see it, our best bet at the moment is to focus on some of the basic questions about meaning and language: Is meaning essentially normative, and if so, in what sense? Does language have an essential fact-stating function? Are the notions of truth and reference essential to our understanding of meaning? Making sense of meaning will involve getting clear on what we demand, or should demand, of this notion, even as we figure out how these demands can together be satisfied. My own view is that this will involve the recognition that there is a non-negotiable connection between meaning and truth, where our fundamental notion of truth is an interpersonal one, applicable across speakers and languages.
3:AM:And finally, are there five books you could recommend to readers here at 3:AM to take us further into your philosophical world?
AK:
Huw Price, Facts and the Function of Truth. Unfortunately, I think this book is out of print. It’s a challenging and important book (which is not to say that I agree with its central claims).
Hartry Field, Truth and the Absence of Fact
Quine, Word and Object
Wittgenstein, Philosophical Investigations
Kripke, Wittgenstein on Rules and Private Language.
ABOUT THE INTERVIEWER
Richard Marshallis still biding his time.
Buy his new book hereor his first book hereto keep him biding!