The Use Theory of Meaning (J. Stanley)

MOVED UP FROM MARCH 9. I HAVE MOVED THIS POST UP BECAUSE THE COMMENTS THREAD REMAINS QUITE ACTIVE, AND I HAVE RECEIVED MANY E-MAILS ABOUT IT.

Like any area of philosophy, there are researchers in philosophy of language with very different conceptions of what it is. One project in the philosophy of language centrally involves the thesis of deflationism about the semantic notions of truth and reference. According to the deflationist, truth and reference play no significant explanatory roles. But the great discovery of philosophy of language in the Twentieth Century is that, with the use of these semantic notions, one can give a tractable account of meaning. Indeed, this work has given rise to semantics, an empirical discipline in linguistics with which philosophers of language are now expected to be conversant. But there is a group of philosophers who are skeptical of the presuppositions of all the work on linguistic meaning that has been done in the Twentieth Century. They are trying to come up with an alternative account of meaning, a “use theory of meaning”, one that does not advert to semantic notions such as reference and truth.

Recently, I had cause to look at some of the work in this tradition to evaluate the progress they have been making. Unfortunately, I saw little to give me optimism in the project of rewriting semantics in terms of use. In Paul Horwich’s paper “A Use Theory of Meaning”, I find that the meaning of a word is given by “a set of sentences that are regularly accepted in such-and-such circumstances”. Unfortunately, this doesn’t give me much help if I am interested in, for example, tense and time, and want to see what the most recent theories of meaning say about tense in natural language.

Those philosophers who are outside of philosophy of language and semantics look to philosophy of language for certain kinds of resources. It is a cost to a metaphysical theory that it results in an error theory about certain central regions of discourse. For example, it is a familiar cost to presentism about tense that it seems to falsify many statements that we regard as obviously true (e.g. "England has had several kings named ‘George’"). It is a cost to an epistemological theory if it predicts that (say) knowledge-attributions are context-sensitive, when they are not. It is a cost to the view that a statue and the clay that constitutes it are identical that it invalidates certain apparently valid Leibniz’s Law inferences. In each case, we need a viable philosophy of language and semantics to gauge these costs. The use-theory of meaning will never be developed in a form that can play the role that current semantic theories do in these ongoing philosophical projects.

Some philosophers look at the theory of meaning through the prism of the project of deflationism. As a result, they regard those of us who operate with the notions of reference and truth as shallow un-philosophical technocrats, since we are not busy challenging these presuppositions. On the other hand, we regard their work at best as useless for the philosophical project of understanding the language-world relation, and at worst as a vain attempt to reinvent the wheel.

Comments

[Claire Horisk]

There's been some recent progress on the question of whether, and how, a semantic theorist might get along with a deflationary notion of truth. I'm thinking of my own work, of course, including my co-authored paper with Bar-On and Lycan; but also work by Max Kölbel, Michael Williams, Douglas Patterson, and Paul Pietroski. (Some of my papers are forthcoming or work in progress; email me at Missouri if you would like a pre-publication look-see.)

[Alan Wong]

Prof. Stanley,
Before your post will generates objections, protest and gnashing of teeth, I would like a point of clarification. Do you think use theories of meaning are incompatible with truth conditional theories? I personally do not. As I understand David Lewis (e.g. "Language and Languages"), he shows we do not have to choose between use-conditions and truth-conditions. Obviously we do not have to pick his solutions. If you accept some sort of compatibilist claim, how would you nuance your criticism to target those who insist meaning is only use?

[Kevin]

Even if I accept that currently available use theories of meaning are not much help in gauging the costs of theories from other areas of philosophy, why should I accept that it is impossible for them to play this role? I would like to hear a good reason to believe your claim "The use-theory of meaning will never be developed in a form that can play the role that current semantic theories do in these ongoing philosophical projects."

[Chauncey Maher]

What’s the point of this post? JS does not seem to be informing us about a philosophical project for he offers only a cursory characterization of deflationism and “the use theory of meaning”, pointing to a single author who is not representative of the group. Readers who are interested in being more well-informed on this kind of project should have a look at the work of Robert Brandom and Mark Lance, in which there is a thoroughgoing commitment to semantics, but from a pragmatic point of view.

No, JS is not really informing us about these projects. His primary concern seems to be to say what’s wrong with them. However, since he hasn’t given us much of a characterization of these projects, his criticism is not well-founded. That criticism is captured in this remark: “The use-theory of meaning will never be developed in a form that can play the role that current semantic theories do in these ongoing philosophical projects [concerning tense and knowledge-attributions].” But this is mere assertion given what he actually says. He certainly has not said enough to make plausible the strong modal expression “will never”.

Even if there were premises that supported JS’s claim, it’s not clear that they amount to a criticism. That’s because a use-theory of meaning might aspire to play a different role in the projects to which JS refers. And that’s not obviously a problem; indeed, it might be a good thing.

[Ole Thomassen Hjortland]

Do you intend this criticism to include "meaning as use"-strategies for logical terms, e.g., proof-theoretical semantics?

[Bob Gamboa]

Given the following quotation from your work-in-progress on twentieth-century philosophy of language, your final paragraph in the entry above, JS, reads as excessively pessimistic about the importance of continuing to reflect on usage in the configuration of the total "language-world relation."

"[S]ome of these disputes take place on a meta-level, with advocates of a non-semantic account of the phenomena arguing that a Gricean or quasi-Gricean apparatus does much more explaining than is ordinarily recognized, and advocates of semantic accounts arguing for greater attentiveness to the nuances of natural language meaning and form. But there is an overarching agreement even between most disputants at this meta-level – the overarching agreement is meaning and use should never be conflated, and that any adequate account of meaning fundamentally employs the notions of reference and truth."

If nothing else, use theorists have helped semanticists to define the parameters of their discipline by forcing them to acknowledge the profound way in which pragmatics are impressed, though NOT (it turns out) indelibly, upon the facade of logico-semantic structures. Doesn't the story go, after all, that it was Grice's interest IN use theories that compelled him to delimit them?

[Jason Stanley]

Chauncey,

You write:

"However, since he hasn’t given us much of a characterization of these projects, his criticism is not well-founded. That criticism is captured in this remark: “The use-theory of meaning will never be developed in a form that can play the role that current semantic theories do in these ongoing philosophical projects [concerning tense and knowledge-attributions].” But this is mere assertion given what he actually says. He certainly has not said enough to make plausible the strong modal expression “will never”."

This concern is also clearly stated by Kevin. My colleague Jerry Fodor likes to say that the problem with philosophy is that there is no such thing as a dead research program. I agree that I haven't given any reason to think that a deflationist use theory of meaning *cannot* be given. But after so many years of waiting, aren't we entitled to what Leiter has called a "pessimistic induction"? How many more decades do deflationist use-theorists get? Remember, this is a project Dummett called for in the mid-1960s -- giving a theory of meaning for empirical language that doesn't appeal to truth-conditions but only to non-semantic notions of use. Isn't it a problem with the discipline of philosophy that we can't look at the progress that truth-conditional theories of meaning have made (especially compared to their competitors) and take it as evidence that truth and reference are notions that have proven their explanatory worth?

Deflationist use-theorists come in many varieties. Both Hartry Field and Brandom are, despite wide philosophical differences, deflationist use-theorists. Neither has come close to developing e.g. a use theory of meaning for subjunctive conditionals to rival the Stalnaker-Lewis line, or a use theory of meaning for quantifiers to rival generalized quantifier theory. But my problem with these theorists is not just their lack of positive success. I find myself completely unmoved by their motivations. They clearly have some metaphysical fear of semantic notions. I guess I've never taken metaphysics seriously enough to allow it to divest me of something so obviously theoretically useful.

[Jason Stanley]

Claire,

I am of course familiar with Paul Pietroski's work. In some of his work, Paul is trying to develop a non-truth conditional theory of meaning. But he is no deflationist, and no use-theorist either. His motivation is the broadly Chomskian one (think *New Horizons*) that the relation between sentences in context and their truth-conditions isn't systematic. Much of my work in the philosophy of language is in the service of challenging this contention of Paul's (that is, I argue that the relation is systematic). But this is a distinct motivation for a non-truth-conditional theory of meaning. Furthermore, the kind of theory of meaning Paul wants isn't something that can help the deflationist use-theorist. Since Paul isn't a deflationist, he feels free to appeal to semantic notions.

Here is a way of thinking about it. Suppose a semantic theory just assigned gappy propositions to sentences in contexts; things that would need to be filled out in various ways to be genuine truth-conditions. The rules for constructing these might very well appeal to semantic notions. For example, Kaplan's character rules are semantic through and through. The game Paul is involved (as I understand him) is arguing that meanings don't get you all the way to truth-conditions; not that they are non-semantic in nature.

[Fritz McDonald]

"I find myself completely unmoved by their motivations. They clearly have some metaphysical fear of semantic notions. I guess I've never taken metaphysics seriously enough to allow it to divest me of something so obviously theoretically useful."

I doubt that fear of metaphysics completely or accurately captures the motivations behind use theories/conceptual role accounts of meaning. Two other motivations that, I think, should be given consideration, are:

(1) CRS is part of a broader functionalist account of the mind. Any arguments intended that support functionalism thus provide support for CRS.
(2) Arguments given in favor of deflationist theories of *truth* (prosententialism for Brandom, disquotationalism for Field, minimalism for Horwich) support a theory of truth inconsistent with a truth-based semantics. The burden of proof is on the theorist of truth-based semantics is to rebut these arguments, and to give a better theory of truth than deflationist.

(1) or (2) may or may not be successful motivations ultimately, but I think these are stronger motivations than "fear of metaphysics," motivations that have to be taken into consideration in assessing the success or failure of CRS/use theories of meaning.

[Jason Stanley]

Fritz,

I don't think we should run together functionalism or conceptual role semantics with *deflationist* use theories (which are my target in this post). I think it's plausible that one must use semantic notions such as reference in giving conceptual roles for various terms. For example, think of the conceptual role of a particular use of "here", given Evans's view in his paper "Understanding Demonstratives". This is a de re mode of presentation, i.e. a way of thinking that can only be described by invoking the referent.

I think arguments in favor of deflationism about truth (that is, your (2)) *are* generally metaphysical in nature (e.g. truth would be a scary, non-naturalistic property).

[Jamie]

Jason,

"Neither [Field, Brandom] has come close to developing e.g. a use theory of meaning for subjunctive conditionals to rival the Stalnaker-Lewis line..."

I thought Dorothy Edgington had done that. But maybe I am misunderstanding the terms of the debate.

[Jason Stanley]

Jamie,

Edgington thinks that indicatives have no truth-conditions, but I hadn't realize she had more than played with the idea that subjunctives have no truth-conditions. I think that Skyrms has given a probabilistic semantics giving assertibility conditions to subjunctive conditionals. But whether this project was successful, and consistent with deflationist truth-conditions, I don't know. One issue here is that subjunctives clearly do embed in the antecedents of other subjunctive conditionals, and I'm not sure how such an account would treat such embeddings (of course, the fact that it's not clear how a probabilistic assertibility-condition account of indicative conditionals would treat embeddings is touted by its proponents as a virtue).

I suspect that most arguments for the thesis that indicative conditionals have no truth-values are inconsistent with deflationism (which, after all, claims that having a truth-value isn't an explanatorily significant property).

[Jamie]

Right, I guess neither D-Edge nor B-Sky is a deflationist. It would be interesting to see how deflationism could be made compatible with the No TV theory of conditionals. Someone should do that.

[Claire Horisk]

Oh dear, Jason, I see I left myself open to being misconstrued. I didn't mean to suggest that Paul, or any of the other people I mentioned, were use theorists. (Although at least two of them, Williams and Kölbel, are deflationists.) Nor did I mean to defend use theorists, or deflationists -- that would be atypical behaviour for me. Rather, I was thinking of your remark that "As a result, they regard those of us who operate with the notions of reference and truth as shallow un-philosophical technocrats, since we are not busy challenging these presuppositions." The people I mentioned, including Paul, were intended as examples of people with commitments to semantic theory who are busy examining reference and truth, and their roles in a semantic theory of the kind you and I like.

[mark lance]

Just a couple comments on this rather shallow criticism. First, most generally, deflationism does not preclude anyone from thinking sentences have truth conditions, or even thinking that the meaning of a sentence can be accounted for by giving its truth conditions. What it precludes is that there is a substantive property of truth the having of which is explanatory. But the claim that P has truth conditions Q is a claim that can be perfectly explanatory for the deflationist. After all, think of what this would mean for an inferentialist. I discuss all this in The Significance of Anaphoric Theories of Truth and Reference (reprinted in the new Armour-Garb volume, or on my web site.)

Second, talk of "use-theory" of meaning is very vague and sloppy. There are many views. Brandom's view is that meaning is a matter of inferential norms governing use. My own is a rather large generalization on this.

Third, Brandom offers detailed semantic discussions of quite a few notions. To pick out two that he doesn't and say that this means there is no progress on giving a semantics is rather odd. Anyway, for a discussion of how an inferentialist should account for quantification, see my Quantification, Substitution, and Conceptual Content -- Nous 30:4, pp 481-507. (There are many other approaches. I'm being egocentric in this quick post.) I've also just written a paper on indicatives and subjunctives which can be made available. But the biggest point is that if one thinks that one explains meaning of some locution by a systematic generation of truth conditions, this can just be taken on board by a deflationist anyway.

Finally, and most generally, if anyone is interested, I'm developing a very general framework for semantics that substantially generalizes the Brandomian approach in which both inferential and truth conditional accounts of meaning are demonstrably recoverable as special cases, and which makes room for many independent sorts of content originating in non-declaratival speech acts. The underlying idea is that of a normative transition between social statuses -- eg entitlement to one assertion generating entitlement to another, but generally status s to act a incompatible with status s' to act a'.

Again, I'm not the only person working on things like this. I suggest that people who want to criticize the whole idea of deflationism and "use theories" -- again, a rather unhelpful category -- do a search on "inferentialism" and have a look at some literature first.

Mark Lance

[Jason Stanley]

Mark,

You write:

"First, most generally, deflationism does not preclude anyone from thinking sentences have truth conditions, or even thinking that the meaning of a sentence can be accounted for by giving its truth conditions. What it precludes is that there is a substantive property of truth the having of which is explanatory. But the claim that P has truth conditions Q is a claim that can be perfectly explanatory for the deflationist. After all, think of what this would mean for an inferentialist."

I don't understand this. Is the claim that you can just point to a traditional semantic theory and wave a magic wand and say "the notion of truth and reference therein employed is deflationist truth and reference"? Then why do you need to appeal to an inferentialist semantics at all? Or is your claim the entirely distinct one that one can provide inferentialist characterizations of traditional semantic notions, such as quantification? Those are two very different claims (though I realize they're often made by the same theorist in the same breath).

I don't think either claim is true. In semantics, we appeal to truth and reference in gathering intuitions about our theories. The various theories about a number of constructions, from anaphora to mass terms to plurals, regularly draw on the same resources developed by metaphysicians (possible worlds, situations, properties, events). Though some semanticists would reject this (essentially because they don't want to be bothered by philosophers), they regularly reject theories that seem to involve us in implausible ontologies (we could talk details if you like). For these and other reasons, I think the practice of semantics precludes construing its appeals to semantic value as deflationary in character. As far as the second claim goes (which is actually where your and Brandom's work seems to be directed), I'm happy to see someone seriously attempt it.

[Jason Stanley]

Ole,

I think that inferentialist accounts of meaning are more plausible for some classes of expressions than they are for others. Presumably, the class of expressions for which they are most plausible are logical terms. My concern is whether the strategy employed there can be broadened to empirical discourse generally (not that I would grant that inferentialist accounts of meaning are successful even in the case of logical terms; it's just that there has obviously been a lot of serious work on that question).

[Jason Stanley]

Mark,

Of course (like many people in my generation) I've read *Making it Explicit*. It is an impressive work. But I don't remember a semantic theory emerging from it that I could use the way I use (say) the theory developed in Dowty, Wall, and Peters *Introduction to Montague Semantics*. What I learned from that latter book was to take a fragment of English and derive meaning theorems from it, via clauses that exploited the notions of reference and truth.

To answer the no doubt uninformed skepticism of people like me, I suggest that you guys write up a similar inferentialist introduction to semantics that builds up the inferentialist meanings of sentences from the inferential roles of their parts, one that undergraduates could use and apply in the study of linguistic meaning (I have a bit of a problem figuring out even what the inferential role of "dog" is, but never mind). You seem to suggest that this is simple to do. To answer people like me, you should definitely do it, and then there will be no more cause for complaint.

I don't deny (and don't deny in my post) that there is work within the inferentialist tradition that seriously respects the data and tries to make progress within the constraints of a non-truth-conditional approach (I'm thinking e.g. of Chris Gauker's recent *Conditionals in Context*). The point of my post was two-fold. First, the work of redescribing truth-conditional semantics in non-truth-and-reference terms is still in the promissory note phase, and people shouldn't pretend otherwise. Secondly, there should be some healthy skepticism about the project of setting aside all the progress we've made in semantics in an attempt to rewrite semantic theories in inferentialist terms. At best, we'd just be doing a lot of work that has already been done (and this is at best, since I doubt it's even possible).

[Ned Block]

Jason complains that “deflationist use theories” are no help in solving certain problems, e.g. about tense,. But his criticisms have nothing to do with deflationism. The use theories he is talking about are inflationist. In particular, Horwich is an INFLATIONIST about meaning—his use theory is an attempt to reduce meaning in non-semantic non-intentional terms. Meaning, so construed, has a substantive potentially explanatory inflationist nature. So to characterize his view as a deflationist use theory is misleading. Horwich combines an inflationist account of meaning with a deflationist account of truth and reference. Whereas Jason’s favored strategy is to explain meaning in terms of truth and reference, Horwich prefers to go the other way around: to explain meaning in terms of use (non-semantically and non-intentionally described) and take advantage of what he takes to be an analytic truth: if x means DOG (if x is used in the characteristic manner of 'dog'), then x is true of all and only the dogs.

There are 2 quite different semantic projects, what is sometimes called the theory of meaning vs theory of meanings. The use theory is of the former kind: a theory of the metaphysical nature of meanings. Theories of the latter kind, the kind that interests Jason have more to do with relations among meanings than with what meaning is, metaphysically speaking. To object to a metaphysical account of meaning on the ground that it doesn’t help with questions about the relations among meanings is like objecting to mind-body dualism on the ground that it doesn’t help with answering the question of whether emotions always involve sensations. Dualism is a theory of the general metaphysical nature of mind, not a theory about the relations among mental states. Similarly, metaphysical theories about the nature of meaning are not intended to answer questions—not all by themselves-- about the relations among meanings.

An inflationist use theory like Horwich’s can be combined with an inflationist view of truth and reference and a view of the relations among meanings, so nothing in use theories is incompatible with Jason’s project.

Jason says “I suggest that you guys write up a similar inferentialist introduction to semantics that builds up the inferentialist meanings of sentences from the inferential roles of their parts, one that undergraduates could use and apply in the study of linguistic meaning.” But inferentialist versions of use theories take sentences as the metaphysically basic items of language since sentences are what we infer from and to, not words. The inferential roles of words are abstractions from the inferential roles of sentences. But sentences can be basic from the metaphysical point of view without being basic from a psychological point of view or from the point of view of a theory of meanings.

[Jason Stanley]

Welcome to the debate Ned! I'm a bit taken aback by your comment, though. Obviously, by "deflationist use theorist" I meant a *deflationist about truth and reference* who is therefore driven to give a substantive theory of meaning solely in terms of use and deflationary truth (as of course Horwich was). All of the commentators, and all of the many people who have been e-mailing me outside the comment threads, took me as meaning this. Nobody was misled by my post in the way you suggest.

[Jason Stanley]

Ned,

It's a reflex with you to accuse everyone you disagree with of confusing the distinction between semantics and meta-semantics! Much as I like that distinction (recall that you hammered it repeatedly into me in graduate school), I simply don't see how it is applicable to the issue we're discussing. A semantic theory specifies the semantic types of lexical items, and explains how the semantic content of complex expressions comes from the semantic contents of their parts. A meta-semantic theory is an account of what makes it the case that a lexical item has the content it does. There, now we're clear about the distinction. It's irrelevant for the issue at hand.

An inflationary semantic theory is one that specifies the semantic content of lexical items by reference to inflationary semantic notions such as reference and truth. An inflationary semantic theory has as consequences theorems that are both contingent and non-trivial, such as:

(1) "Dogs bark" is true at world w if and only if every dog in every world in which there are only biologically normal dogs barks in barking situations.

Or perhaps:

(2) "John is a tall fifth-grader" is true if and only if tall(John)=d, where d is a degree of height that is above the average for fifth-graders.

Deflationary semantic theories cannot straightforwardly appeal to contingent, informative statements of truth-conditions such as those given in (1) and (2). These are not analytic truths. But it is these theorems that have proven so useful in various projects, both within and outside philosophy (e.g. metaphysicians need semantics because they want to know some of the costs of various theories, e.g. whether those theories will result in error theories of some stretch of discourse).

There are two strategies available to the deflationist. The first strategy is to reject all such theorems, and accept only purely deflationist statements of reference and truth-conditions. The second is to accept such theorems, but argue that what links the right-hand sides of them to the left-hand sides of them is an account that can be given purely in terms of use and deflationary reference and truth.

In my post, I point out that there is all this valuable work that has been done on the semantics of various constructions. This work proceeds by assigning references to parts of sentences, and building the truth-conditions of the whole sentence from the reference assignments to the parts. Use-facts may enter into this project, but only in helping to specify the relation between lexical items and the things and properties in the world to which they refer. A compositional semantic theory does the rest.

A deflationist use-theorist, by contrast, if they want to exploit the valuable analyses that philosophers of language have produced over the years, must give an explanation entirely in terms of use plus deflationary truth of the relation between the object-language sentence and the meta-language sentence that gives its truth-conditions. Furthermore, the explanation in terms of use cannot appeal to inflationary semantic notions. This seems to me to be an extremely difficult task, and I'm glad I'm not engaged in it. Be that as it may, curious to see how much progress they had made on it, I went looking around in some literature. The post is the result of the lack of progress in what I discovered.

There are two strategies the deflationist can employ. First, she can reject all the informative, explanatory T-sentences. Secondly, she could try to appropriate them, by appeal to deflationary truth plus use. Here is the kind of thing that a deflationary use-theorist would have to do. She would e.g. have to take Lewis's account of counterfactuals, and argue that the relation between the object-language counterfactual on the left-hand side and the meta-language statement of its truth-conditions on the right-hand side could be explained entirely in terms of use-facts plus deflationary truth. For example, as my colleague Barry Loewer suggested to me, she could argue that Lewis's semantics (or something like it) codifies the vaid inferences we make using counterfactuals (and she would have to also argue that the notion of validity does not mask appeal to inflationary truth). She would e.g. have to argue that the distinction between no-truth-value views of indicative conditionals and other views is not one that is fundamentally explained in terms of the property of having a truth-value (though Richard Holton has made some progress here, as has Field). In general, she would have to provide some strategy of showing that for every informative, contingent statement of truth-conditions, there is a way of explaining the relationship between the object-language sentence and the meta-language sentence in terms of use facts plus deflationary truth (and she'd have to argue that the use facts didn't involve semantic properties).

As many people, both in the comments thread and in e-mails to me, have pointed out, I haven't given an argument that this project can't be done. But the fact that so little progress has been made on it --the fact that it's even difficult to see *how* it could be done, for many cases-- suggests that it is rather difficult. And if it can't be done, then one thing the success of such semantic explanations will have shown is the explanatory value of the inflationary notions of reference and truth.

[Richard Heck]

I take Jason's question to have been along the following lines.

Expressions have semantic properties, which we can state using the notions of truth and reference. We may well then want to ask in virtue of what expressions have these properties, and one might propose that the answer is: In virtue of how they are used. Maybe that's right, and maybe it's wrong, but it's an intelligible view. And conceptual role semantics, as I understand it, is precisely such a view.

Perhaps one thinks that the semantic properties of expressions should not be stated using the notions of truth and reference. Dummett proposed, for example, that we should instead use some notion like verification. One advantage to this proposal, as Dummett sees it, is that it makes the meta-semantic question in virtue of what expressions have these semantic properties relatively easy to answer, since what counts as a verification of a given claim can (plausibly) be read off use. This too is an intelligible view. The problem is that it has proven very difficult to work out the details. It has, in particular, proven difficult even to identify a notion of verification that is appropriate for empirical statements (a point first emphasized, I think, by Putnam). Not that people haven't tried---Wright has tried quite hard---but there haven't been many successes.

There are other possibilities, too. As mentioned above, Paul Pietroski has been arguing that truth and reference aren't the right notions. But Paul has views about what the right notions are.

What is common to these views proposals is that they take compositionality seriously. It is NOT an intelligible view that every expression has its semantic properties in virtue of how it is used. There are simply too many expressions that never will be used but that have determinate semantic properties. So, to borrow a point from Higginbotham, if one is going to give any account of why such expressions mean what they do, one will have to specify the semantic properties of sub-sentential expressions and tell a story about how these determine the semantic properties of larger expressions.

And so, I take it, the question Jason means to be raising for "deflationist use theories" is: What are these properties, and how do they compose? It's no good simply to say that, whatever they are, they are abstractions from the use of sentences. One would like to know first what the semantic properties of sentences are, and then one would like to know what we're supposed to abstract and how we're supposed to do it: What, for example, are the relevant properties of quantifiers? How are they determined by the uses of sentences in which they occur? How do they then determine the appropriate use of sentences that have not been and never will be used? So far as I know, there aren't even any bad theories of this kind that are compatible with "deflationist use theories". There are just gestures and promissory notes.

Similar problems arise even before we reach the subsentential level. How are the 'use properties' of subjunctive conditionals determined by the 'use properties' of their parts? One can't just say that each subjunctive conditional means what it does in virtue of its being used as it is. There are too many subjunctive conditionals that have perfectly determinate meanings but which will never be used. Here again, I don't even know of any bad theories of this phenomenon that are compatible with "deflationary use theory". As noted above, Edginton's work does not fit the mold here: As I understand her view, the assertibility conditions of subjunctive conditionals are determined by the semantic properties---in particular, by the truth-conditions---of their parts. Whether the conditionals themselves have truth-conditions isn't to the point.

To my mind, however, the most serious issue here concerns the "autonomy of meaning". To borrow an example from Chomsky, it just does not seem right to say that it is because of how it is used that "John is too fast to catch" means that John is too fast for anyone to catch him, rather than that John is too fast for him to catch anything. It is, of course, true that competent speakers know how the sentence ought to be used. But that is not isolated knowledge about this single sentence: It is systematic and productive, as Chomsky makes very clear in his discussion. What one wants is a story about how speakers know that this sentence is to be used in the way it is. Part of the story is that speakers (tacitly) know what its syntactic structure is. But that is not enough. After all, speakers also know a lot about the sentence's phonology, but that information isn't what explains their knowledge of how it is to be used. Why not? Because there aren't enough (if any) systematic connections between phonology and semantics. The rest of the story is thus that speakers (tacitly) know how the syntactic structure of the sentence constrains its possible meanings. And so, in short, the sentence is used how it is because of what speakers know about its meaning, not conversely.

[mark lance]

Hey folks:
First, I sent a response yesterday to some of this which wasn't posted. Not sure why that is, and I won't try to reproduce most of it now. [Obnoxious comment deleted by editor]. He seems to have rather gotten more of the point now, and acknowledges that rejecting the idea that truth is a substantive property doesn't commit one to saying that a sentence with the word 'true' in it is not explanatory.

But this whole discussion suffers from a bit of a problem. It began with Horwich as the example and the claim that his account doesn't get very far. I have nothing to say on that. But then it has morphed into a discussion of all sorts of folks under this label "deflationary-use". Brandom is explicitly included in this and by implication anyone who does semantics in anything like a brandomian way.

This is a problem. Richard Heck is right that it is bizarre to say that use makes certain things correct. It is also just wrong to say that meaning is use. Much closer is the claim that language amounts to a system of norms and to make a claim about meaning is to make a claim -- within language -- about how a word ought to be used. (Not a moral ought, but a linguistic one.) I've got a book with John Hawthorne on that topic, but many others have written similar things. Brandom clearly identifies meaning with inferential norms, not inferential behavior.

Given this, this whole idea of using the truth conditions as a deflationist is really completely simple. A is true iff B. This sentence endorses the propriety of a material inference from A to B and B to A. Now someone who takes on the sort of meta-semantic position that Ned Block mentions is also committed to reducing this fact to something about use. Maybe Horwich is. Brandom isn't. And in no case does deflationism about truth commit you to this. The anaphoric version of deflationism tells us precisely what this sentence says. (Again, have a look at my paper for the argument which involves no appeal to magic.)

Think about it this way. Hardly any two substantivalists about truth agree on what it is. Is it a causal relation to the world, picturing, functional isomorphism, something epistemic? Seems that the success of truth conditional theories of meaning -- as Block rightly points out -- depend in no way at all on the resolution of this. So the work can't be done by the substantive content of "is true". What is doing the work -- as Davidson pointed out ages ago, and as I am trying to explain here (and which I take to be the main missed point of Block's post) -- is the way that the truth theory establishes connections between the truth of A and the truth of B, which is to say between A and B.

Mark Lance

[Douglas Patterson]

A brief comment in this, my first philosophy blog post, since my name has already come up here. Mark Lance claims that deflationism is perfectly compatible with thinking that "the meaning of a sentence can be accounted for by giving its truth conditions". I think this claim rests on a confusion, at least if we take "accounted for" as gesturing at serious explanation, as I explain in a piece that came out in Phil Studies last summer (124:3).

The basic point there is that Lance is perfectly right in thinking that if one can give an account of meaning (I introduce some qualifications on this notion in the paper) for the "true"-free fragment of a language, it can be extended to the full language on the principle that "s is true" always shares its meaning with sentence s itself. This, however, I argue, is something that everyone should accept. The controversial claim is rather that the account of meaning for a language can be expressed in a language that has only a deflationary truth-locution. This is the controversial view, since it forces us to understand any attributions of truth to the object language in terms of some antecedent notion of meaning, since our account of our own truth attributions is that they're equivalent in meaning to certain other sentences, something we can't sensibly say without having some other understanding of what attributions of meaning attribute. About this interesting form of deflationism I think Jason is right: known attemtps to explain meaning in other terms haven't worked out, and the failure is robust enough to give us good reason to expect that they won't work out in the future. This isn't an argument against all possible such attempts, of course, and I don't claim here or in the paper that it is, though I do share Jason's pessimism on the point.

[Christopher Gauker]

I am grateful for Jason's nod to my book on conditionals, but I would like to say that I am not an inferentialist. We need to be able to give an account of the validity of inferences in semantic terms. And we have to ground our semantic terminology in a theory of linguistic communication. If we want to eschew real reference relations, as I do, then we cannot define logical validity as preservation of truth on an interpretation. (That's not just obvious, but I think it's true.) Fortunately, there are other ways. In particular, we can define logical validity as preservation of what I have called assertibility in a context. A context is a certain sort of structure built up from atomic sentences (and maybe some other things), and we have to be able to say what it takes for one of those structures to be the one that pertains to a given conversation; but we can do that, I claim, without appeal to reference relations.

[Jason Stanley]

Mark,

I'm sorry, but I never received the comment to which you allude, and there is no record of it arriving on Typepad.

Doug Patterson has basically said what I would say in response. I also don't think the "missed point" of Ned's post was that you could reap all of the benefits of truth-conditional semantics with a deflationary truth-predicate. I don't think that was a point he was making at all, and I have no reason to think he believes that very controversial claim (though he very well may). Also, I don't believe that "the anaphoric version of deflationism tells us precisely [what a T-sentence says.]" There are a host of problems for that claim (among them that on an anaphoric version of deflationism, we not being *told* much of anything by a T-sentence).

I also think the success of truth-conditional theories have very much to do with inflationary semantic notions, most saliently notions such as reference and satisfaction. The role of semantic theories in assessing costs in metaphysics, and the interplay between metaphysics and semantics more generally, is I think only explicable on the view that a semantic theory gives us the metaphysical commitments of the truth of a sentence (what we've discovered essentially is that thinking about metaphysics is a darn good way to be a semanticist). I think this role of semantics shows Davidson to be precisely wrong when he says that as far as the truth-predicate in semantics is concerned, even if the emotivist is right, 'Bardot is good' is true if and only if Bardot is good.

This post was rushed, because I've just arrived in my hotel room and am rushing out the door again...

[Douglas Patterson]

Lance Writes:

"Think about it this way. Hardly any two substantivalists about truth agree on what it is. Is it a causal relation to the world, picturing, functional isomorphism, something epistemic? Seems that the success of truth conditional theories of meaning -- as Block rightly points out -- depend in no way at all on the resolution of this. So the work can't be done by the substantive content of "is true". What is doing the work -- as Davidson pointed out ages ago, and as I am trying to explain here (and which I take to be the main missed point of Block's post) -- is the way that the truth theory establishes connections between the truth of A and the truth of B, which is to say between A and B."

The first problem here is that the fifth sentence doesn't follow from the ones that precede it. Suppose that truth conditional theories of meaning explain something worth explaining. Now we ask ourselves: is it in virtue of the substantive content of "is true" that they do so? Grant that we currently find ourselves at a loss when it comes to saying what this substantive content is (picturing, isomorphism, whatever). Does the fact that we can't do this (not now, and maybe not later) show that the content isn't doing any work? No. The fact that we're not in possession of an obviously good reductive account of the content that does the work doesn't show that there isn't content there doing the work. Something might show that, but this doesn't.

Davidson as I understand him sees things in just these terms: we ought to conclude from the failure of known reductive accounts that truth is indefinable. (I don't agree there, but let it pass.) This doesn't mean that "is true" doesn't have substantive content, it just means that the content doesn't admit of a good reductive analysis. I had thought it was rather clear in "The Structure and Content of Truth" that his view was that "is true" has to have such content for a truth conditional theory to do the work he wants it to do.

Thus I don't really see what Lance is on about in the last sentence I've quoted; neither the connection between the truth of A and the truth of B nor that between A and B is the issue, and I can't see that either ever was for Davidson. What Lance writes only grammatically makes sense if we read "A" and "B" as names of sentences or objectual variables that range over them, not when we read them as schematic sentence letters, which builds a concern with relations between sentences into the very way he's putting things. The issue is, rather, the connection between the truth of "A" and it's being the case that A (read the letter substitutionally there, and the quotes as corner-quotes). A truth-conditional theory states the conditions under which sentences are true; it doesn't just map sentences onto other sentences.

[Paul Horwich]

Jason’s critique of the deflationist use-theoretic view of meaning goes as follows:

1. Forty years of work in truth-theoretic semantics (TTS) have yielded a host of valuable results and explanations;
2. This work presupposes inflationism with respect to truth and reference;
3. Therefore, it constitutes a powerful case against deflationism;
4. Especially so, since no use-theoretic semantics has yielded any such results, despite decades of effort;
5. Therefore, we have good reason to reject deflationism and to abandon any hope of a deflationary use-theory of meaning.

Here’s why I think this is all wrong:

(1*) The accomplishments of TTS amount to the discovery of ways of deriving truth conditions of sentences (of an increasing variety of kinds) from assignments of referents to certain words and from principles concerning other words – connectives – which specify how the referents of the expressions connected by them determine the referents of the connected expressions. No doubt ‘progress’ has been made over the years in this endeavour. No doubt a great deal of ingenuity and intelligence has been displayed. But the scientific value of such derivations is questionable. For if linguistics is to be a serious science, it must stand alongside psychology, neurology, biology; chemistry, etc. in helping to provide causal explanations of concrete events of linguistic activity. And it is quite unclear how the derivations of TTS could do that.

(2*) The derivations themselves – the deductions of truth conditions from reference conditions – are perfectly consistent with deflationary construals of the semantic notions (as noted by Alan Wong). A deflationist can perfectly well agree (following Frege) that the truth of a sentence is determined by the referents of its parts, and can set himself the project of figuring out what the referents of various kinds of part would have to be. What he does not accept is that reference is a natural relation, susceptible to unified non-semantic reductive analysis. He thinks rather that each word’s reference is engendered by whichever non-semantic property explains its overall use -- i.e. explains the vast array of acceptance/rejection facts concerning the sentences that contain the word. So, if ‘U(w)’ is the non-semantic property (perhaps itself a core use-property) that explains the overall use of “dog” then it’s a necessary, a posteriori fact that
U(w)  w is true of all and only the dogs.
Moreover, he thinks that a science of semantics must invoke these underlying use-explaining properties if it is to account causally for concrete phenomena of linguistic activity, and hence be part of the holistic scientific enterprise. My own conjecture is that these meaning/reference engendering properties are themselves use-properties – more specifically, explanatorily basic properties of the form (very roughly speaking), ‘such and such specified sentences containing w are accepted’. (See my Reflections on Meaning , Ch 2, for details).
Note that this deflationary perspective on truth and reference, and on their utility in the science of language, is motivated by an honest empirical attempt to come up with the basic principles governing our deployment of those notions. If there is anything like ‘fear of metaphysics’ here, it is the assumption that we are aiming for a linguistics that is scientific.

(3*) Clearly, then, progress within TTS does not in fact constitute any case at all against deflationism and against the use-theoretic view of meaning. Contrary to JS, it is not central to deflationism that truth be, in no sense, ‘explanatory’. In fact, that the truth of a sentence may be ‘explained’ (in some sense) in terms of the referents of its parts, is a case in point. Moreover, not only has Jason made no case against deflationism, but (as Fritz MacDonald observes) he has not even attempted to rebut the various arguments in favour of it.

(4*) The deflationist use-theoretic view (a) that what really matters empirically are the various non-semantic grounds of reference-relations, and (b) that these grounds are to be discovered in the way just outlined, has the potential to make semantics truly scientific and explanatory.
Granted there has been very little detailed work within this framework – there are. hardly any plausible results as to what the basic use- explainers for specific words (e.g. “dog”) actually are. And given the holistic nature of this enterprise (stemming from the fact that, since each sentence contains several words, the explanation of its acceptance would need to invoke the meaning-constituting use-property of each of them) the difficulties are formidable.
But where there has been substantial progress (to which I hope to have contributed in my Meaning 1998 and Reflections on Meaning 2005) is in elaborating and defending the basic approach. If I can be forgiven an immodest comparison: to complain at this stage at the lack of technical results within the framework is about as reasonable as it would have been to dismiss Davidson’s “Truth and Meaning” in the early 70s..
Let me clarify what sorts of result should, and should not, now be expected. We should be aiming to explain linguistic activity -- in particular, given our focus on semantics, the acceptance and rejection of sentences. We should NOT be trying to show how the ‘basic uses of sentences’ are determined by the basic uses of the words in them; for no notion of ‘the basic use of a sentence’ has been defined or should be deployed. Rather the meaning of a sentence is assumed to be constituted by the fact that its words are combined (functionally) as they are and have the basic use-properties they do. Thus there is no challenge to avoid making assumptions that violate compositionality, as there is in TTS. (Here I think I am agreeing with Chauncey Maher and disagreeing with Richard Heck). The challenge rather is to find basic ‘laws of use’ for words that will, when taken together, and when combined with general principles and circumstantial factors, explain the acceptance of the sentences containing them.

(5*) I have to conclude that Jason’s dismissal of the deflationary use-theoretic approach does not address what is to be said in favour of that perspective, and does not confront the weaknesses within his truth-theoretic semantics.

[Jason Stanley]

Paul,

This is very helpful. Though I was already aware that this was your position, this helps me to clarify some points of disagreement. Let me begin by questioning (1*); in particular, let me begin by explaining how TTS plays a causal explanatory role in the explanation of linguistic activity.

The TTS explanation of facts about linguistic activity is as follows. People have tacit knowledge of the referents of words, that is they know contingent a posteriori facts of the form "Bill Clinton" refers to Bill Clinton. They know the syntactic structure of sentences, and they know the composition rules corresponding to the syntactic structures. This knowledge allows them to grasp the truth-conditions of sentences they hear, and express propositions they want to express. Truth-conditional semantics provides an elegant explanation of how our knowledge of language allows us to gather information about the world, and communicate it as well. That is, together with Grice's work on pragmatics and refinements thereof, we have a remarkably well-functioning account of linguistic activity.

You write "if linguistics is to be a serious science, it must stand alongside psychology, neurology, biology; chemistry, etc. in helping to provide causal explanations of concrete events of linguistic activity. And it is quite unclear how the derivations of TTS could do that." Here is how the derivations of TTS do that. John utters "There is milk in the fridge". Mary hears John, and using her tacit knowledge of the reference of the words in his sentence, and their modes of combination, and her trust in John, acquires the belief that there is milk in the fridge. Semantic theory is part a perfectly scientific causal explanation of the process from Mary hearing a certain acoustic signal to her acquiring a belief about the world; the derivations of TTS play a crucial explanatory role in this process. They explain how Mary so quickly acquires a belief involving the fridge and milk from a mere sound.

This remarkably well-functioning account presupposes a certain relation between linguistic items and the world, the relation of reference. It seems very difficult to give a naturalistic reduction of the reference relation. But I would think its value in yielding such a well-functioning causal explanatory account of linguistic activity (and the present lack of a serious competitor) should lead us to think that the reference relation is a genuine relation. If we are unsuccessful in reducing reference to something else (whether it be use or nomic relations), its remarkable success in explaining how we can hear a sound we have never heard before and thereby acquire a belief about the world should nevertheless lead us, on grounds of methodological naturalism, to recognize its reality.

O.k., my next question concerns (2*). In particular, I'm curious about your views about the 'no conflict' thesis between deflationism about truth and reference, and TTS. There are a number of facts about the theorems of TTS that seem incompatible with the thesis that the word "true" used in them is deflationist (e.g. that these theorems are contingent, a posteriori truths, that there is indexicality and vagueness in the language, etc.). But I'm aware that you have detailed responses to these worries (though surely you don't think that these worries are completely groundless, as Mark Lance seems to think, otherwise you wouldn't have written a book responding to them). So put these worries aside (though for the record I think they are very serious, though I recognize that you and Field have said much of interest in response).

For the deflationist to adopt TTS in the way you suggest, doesn't she have to provide use-theoretic explanations of all the reference clauses in a semantic theory? Furthermore, to make it the case that her theory is deflationist about reference, these use-theoretic explanations cannot amount to a systematic naturalistic use-theoretic account of the reference relation (since then reference would be a naturalistic relation, and we wouldn't be deflationists). So you need to provide a use-theoretic explanation of each truth of the form "dogs" is true of dogs, without it being the case that there is one general kind of pattern of explanation that will work across the board? Isn't that what is required to adopt the successes of TTS? This is exactly the kind of project I was hoping to see some progress on, and my post stated that I hadn't found any. If we do in fact never make much progress on this, can't we conclude from the elegant causal account of linguistic activity that appeals to TTS that reference is a genuine relation?

[Jason Stanley]

Paul writes:

"Clearly, then, progress within TTS does not in fact constitute any case at all against deflationism and against the use-theoretic view of meaning....Moreover, not only has Jason made no case against deflationism."

But by Paul's own admission, adopting TTS requires the deflationist to provide a use-theoretic account of various reference facts (or at least making it somewhat viable that this can be done). Given the lack of progress on this project, which is, after all, precisely what I was remarking on in my original post, progress within TTS does in fact provide a case against deflationism, rather than "clearly not" constituting "any case at all" against deflationism.

[Paul Horwich]

Jason,

You suggest that TTS helps provide causal explanations in virtue of the fact that people tacitly know the referents of words, which brings about tacit knowledge of the truth conditions of sentences.

But there is a problem as to how this 'bringing about' takes place, given that the knowledge is not explicit and so the mechanism cannot be inference. And there is similarly a problem of how tacit knowledge of truth conditions can get put to work.

The use-theorist offers a strategy for answering the question, "What is it for someone to implicitly know the referent of a word?", and suggests -- because the answers differ so greatly from word to word -- that the causal stories behing linguistic activity must be told at this underlying level.

As for the relation between deflationism and TTS, it seems to me that the conflict between them emerges when TTS is taken (as it usually is) to include the thesis that TTS premises and conclusions are causally efficacious (which would mean that reference would have to be a fairly unified natural relation).

I should clarify that I am not recommending that use-theorists 'take over' TTS derivations. For I think that these have no causal-explanatory utlity. I'm suggesting, rather, that scientific understanding requires us to get down to the highly-variable use-theoretic underpinnings of reference-facts.