19

Our Grasp of the Concept of Truth by Paul Boghossian

Reflections on Künne Künne’s Modest Account Wolfgang Künne’s Conceptions of Truth (2003) is a magnificent achievement. Wonderfully clear, erudite, compendious, honest and insightful on some very tricky issues – these are some of its many virtues. I have benefited a great deal from studying it. In this short note, I will concentrate on Künne’s own positive proposal about the concept of truth, his modestly named ‘Modest Account’. I will raise some questions about its ultimate viability. The Modest Account is an account of the concept of truth. It takes the form of an explicit definition, given by the following seemingly simple formula: • ∀ (MOD) x (x is true p (x =[p] & p)). ⇔∃ The square brackets function as singular-term forming operators on sentences. The idea is that they take a sentence that expresses a particular proposition and form a singular term that designates that proposition. The first variable ranges over propositions, conceived of as mind- and language-independent entities. Its substituends are singular terms, for example, ‘that Bremen is pretty’, which are taken to designate propositions. What MOD seems to say, then, is that something x is true just in case there is a proposition that p with which it is identical and p. Someone gets to have the concept of truth by grasping and affirming MOD. There are two distinguishable elements here. First, we have the claim that MOD offers a correct explicit definition of the concept of truth. Second, we have the claim that grasp of the concept of truth is to be explained by someone’s grasping and affirming the definition offered by MOD. In other words, Künne is offering us both an account of the concept of truth and an account of our grasp of that concept. All this seems intuitive and plausible. What could possibly go wrong? As Künne is very much aware, there are many challenges. One of them is something that is encountered by virtually any theory of truth, but especially by ones that purport to explain our grasp of the concept of truth, and that is the problem of circularity: such theories tend to assume the very concept they are trying to explain. In what follows, I will look at three possible sources of circularity in Künne’s Modest Account. Interpreting the Modest Account ∃ The first comes up immediately, in trying to make sense of the second, rather strange-seeming quantifier ‘ p’. This cannot be ordinary objectual quantification, because on the one hand ordinary objectual quantification is always quantification into singular term position, but on the other the sentential connective ‘and’ on the right hand side of MOD cannot be followed by a singular term: it requires a sentence. ∃ Künne considers reading ‘ p’ as a substitutional quantifier. He rightly rejects this option because it would (a) lead to the implausible result that one would have to have every concept in order to have the truth concept, and (b) would very likely convict the theory of vicious circularity: it’s very hard to see how to understand substitutional quantification without relying on the notion of truth. Künne’s solution is to suggest that the second quantifier expresses objectual quantification into sentence position. The challenge is to say what this could possibly amount to.∃ Künne provides the following informal gloss: The formula ‘ p (The Pythagorean Theorem = the proposition that p and p)’ expresses a truth iff there is a proposition that satisfies the condition ‘the Pythagorean Theorem = the proposition that p and p’. And a proposition satisfies this condition iff it is identical with the Pythagorean Theorem and true (2003, 263). Now, this characterization seems intelligible enough. But, as Künne is well aware, it won’t do, since it invokes the notion of truth. Does he provide an alternative characterization that doesn’t invoke truth? If he doesn’t we will have no choice but to conclude that the only grip on this quantifier that we have been given presupposes a grasp of the concept of truth and that will be fatal to the Modest Account. Künne doesn’t try to provide a systematic formal account of how this quantifier is to be understood. What he tries to do instead is soften us up for the claim that we understand it perfectly well, independently of the notion of truth, by claiming that we find perfectly good examples of such quantification in ordinary language. For example, he claims that: • (A)For some way things may be said to be, things are that way is the natural language equivalent of: • ∃ ( B) p.p. Since we seem to understand (A) well enough, he argues, that gives us reason to think we understand (B) well enough, despite the lack of a formal account. Unfortunately for Künne, this doesn’t seem right since (A) quantifies over ‘ways’, and that is more like quantifying over properties rather than over propositions (Hofweber 2005 ). Since Künne provides us with no other resources with which to make sense of his admittedly strangeseeming quantifier, we are left without a coherent, non-truth-dependent reading of it. And that is my first worry about circularity. Understanding ‘proposition’ A second circularity worry derives from the Modest Account’s reliance on the notion of a proposition. Künne, as usual, is well aware of this worry. But he considers it in a version that fails to give it its greatest bite. The worry he considers derives from Dummett. Here is how Künne formulates Dummett’s argument: (P1) The modest account presupposes a grasp of the concept of a proposition. (P2) Propositions are sentence-meanings. (P3) The notion of sentence-meaning cannot be explained independently of the notion of truth. (C) Hence the modest account is circular (2003, 368). Künne, following others, (Soames, for example) gives some good arguments why (P2) should be rejected, why we should not always identify propositions with sentence-meanings. We can have two expressions in a language that are synonyms – they have the same meaning. ‘Serpent’ and ‘snake’ are arguably two such expressions. A child, however, can wonder whether all serpents are snakes but not wonder whether all snakes are snakes. That seems to show that propositions are distinct from sentence meanings. But the crucial issue is not whether propositions are sentence-meanings. The crucial issue can be stated directly, without talking about language: (The Question) Can we understand the notion of a proposition independently of the notion of truth, so that we can teach someone the notion of truth for the first time by leaning on his or her antecedent grasp of the notion of a proposition? Künne grants that propositions are essentially truth-evaluable, but he denies that this enforces a negative answer to the Question: Propositions are essentially truth-evaluable, but this does not imply that the concept of a proposition is to be explained in terms of truth. (Triangles are essentially figures whose internal angles add up to 180°, but we can say what a triangle is without invoking the notion of a sum of angles.) If the concept of a proposition is explained in the way I suggested in Chapter 5.1.1, without invoking either the notion of meaning or the notion of truth, then there is room for combining a modest account of truth with a potentially illuminating theory of expressing a truth (Ib., 373). The comparison to the case of the concept triangle is apt. It goes without saying that it can be a necessary truth that all F’s are G’s without the concept G being presupposed in the understanding of F (think about water and H2O). But as the triangle case illustrates: All F’s are G’s can be not only necessary, but also a priori as well, without its being the case that the concept of G is presupposed in our grasp of the concept of F. It can come as an a priori discovery that all F’s are G’s. There appears to be room in logical space, then, for the Modest Account. The question is whether that room is occupied by something plausible in the present instance. Well, how could we understand the notion of a proposition independently of the notion of truth? Künne follows others in supposing that we could understand it through our grasp of the locutions through which we attribute propositional attitudes, an approach that Stephen Schiffer has done much to develop. Although at several points Künne says that his approach is very close to Schiffer’s, I think it is distinct in certain important respects. Künne (Ib., 251) suggests that we could grasp the concept of a proposition by learning to accept, as a conceptual matter of course, any inference from an instance of the schema: • (I)A V’s that p for example: • 1Pauli believes that tachyons exist to: • (II)That p is the content of A’s Vn (where Vn is a verbal noun corresponding to the verb in (I). for example: • 2That tachyons exist is the content of Pauli’s beliefAnd, finally, to an instance of: • (III)The proposition that p is the content of A’s Vn i.e.: • 3The proposition that tachyons exist is the content of Pauli’s belief. However, since many propositions will not be the actual content of anyone’s propositional attitudes, Künne adds the caveat that something which could be thought or said in some mode or other may never in fact be thought or said in any mode. You grasp the notion of a proposition, according to Künne, by being prepared to regard such inferences as analytic and by endorsing the caveat. Künne’s idea, of course, isn’t to make empirical remarks about how ordinary people actually come to understand the notions of proposition and truth, but to show that they could have come to understand them in the indicated way. If that were so, we could hope to explain our grasp of the notion of a proposition on the basis of our prior grasp of these sorts of mental state attribution and the analytic inferences involving them; and we could then hope to explain our grasp of the notion of truth on the basis of our prior grasp of the notion of a proposition, just as the Modest Account would have us believe. What are the ways in which this strategy could misfire for Künne’s purposes? There are two central ways. First, it could misfire if the proffered story about how we might be in a position to get the notion of a proposition were wrong or inadequate. Second, it could misfire (even if that story were right about one’s acquisition of the notion of a proposition) if one could show that it relied on his antecedent grasp of the notion of truth. I think there is something to worry about on each score. Let me start with the first. We know where we need to end up. The concept of a proposition has to turn out to be a certain kind of concept – the concept of a mind-independent, language-independent, entity that is designated by a singular term. This entity has to be essentially truth-evaluable, but that information won’t be included in mere grasp of the concept. On this view, then, our concept of a proposition embeds both a metaphysical thesis and a semantic thesis; and both are essential to the Modest Account. Call this conjunction the Required View of propositions. I have two worries about how we get to the Required View on the basis of Künne’s story. To see what they are, note first that our job had better be done by the time we get to (II), since the word ‘proposition’ in (III) is just an adornment: Künne considers (III) to be pleonastic on (II) in the way that: The number seven is prime is pleonastic on Seven is prime. So what we need to assess is just whether it’s true that we could be said to have grasped the notion of a proposition, as that’s spelled out by the Required View, simply by our being prepared to infer, as a conceptual matter of course, to an instance of (II) from an instance of (I). One problem with this claim is that it’s not really clear that the locution ‘the content of’ is an idiom of ordinary language: few non-philosophers would be found using it in the course of routine talk about their own or other people’s mental states. Or, to the extent to which they would be found using it, it seems to me that they would be equally willing to infer from: • (i)Pauli is experiencing painto • (ii)Pain is the content of Pauli’s experience. But no one would think that someone’s willingness to infer from (i) to (ii) gave any support either to the claim that ‘pain’ is a singular term that denotes an individual pain, or that we are now committed to thinking of pains as entities, let alone ones that could be discovered to be essentially truth-evaluable. So where did that information come from in the case of propositions? I think that Künne’s (I) – (II) inferences are analytic. But I think they serve to provide an implicit definition of a technical notion –‘the content of’– rather than serving as our means of grasping the notion of a proposition. Perhaps Künne could instead appeal to some of the inference patterns that Schiffer appeals to, which are different from Künne’s (I) to (III) patterns. Schiffer arrives at what is basically the Required View by arguing that it is the best explanation for the analytic validity of such inferences as: Harold believes that there is life on Venus and so does Fiona. So, there is something that they both believe – to wit, that there is life on Venus. But this claim seems vulnerable as well, for it seems that we could equally point to an inference of the following kind: Harold experiences pain and so does Fiona. So, there is something that they both experience – to wit, pain. Once again, though, no one would think that this provided any support for a view about pain that would correspond to what the Required View says about propositions. Furthermore, Mark Sainsbury (2005 ) has shown that there are some inference patterns that support the metaphysical thesis but not the semantical one, and vice versa. Consider: Harold goes to market and so does Fiona. So, there is something that they both do – to wit, go to market. This may be some evidence that acting is a relation between an agent and an action; but few would be tempted by the view that ‘goes to market’ is a singular term that refers to an individual action. On the other hand, we have the inference: Harold weighs 150 pounds and so does Fiona. So, there is something that they both weigh – to wit, 150 pounds. It might be fine to say that ‘150 pounds’ refers to 150 pounds; but it would be wrong to say that weight is a relation between an object and 150 pounds. If we are to use such inference patterns to support the Required View we need to formulate a principle that tells us why some of them support it and others don’t. But we have to hand no such principle. Furthermore, as many philosophers – Sainsbury (Ib.), Moltmann (2003 ) and others – have pointed out, there are cases in which substituting ‘the proposition that p’ for the allegedly coreferential ‘that p’ does not seem to preserve truth or even sense: Pauli fears that tachyons exist is clearly not equivalent to Pauli fears the proposition that tachyons exist. And while it makes to sense to suppose that Pauli hopes that tachyons exist it seem to make no sense to suppose that Pauli hopes the proposition that tachyons exist. Künne says that proper attention to the distinction between a proposition’s being the content of a propositional attitude and its being the object of such an attitude helps with these sorts of problem. I think he means that if we put our propositional attitude attributions into the content format (‘That p is the content of A’s Vn’), substitutions of ‘the proposition that p’ for ‘that p’ don’t result either in changes of sense or in nonsense. That may be true. But the Required View would appear to license these substitutions quite generally, and not just for occurrences within the preferred format. I wasn’t sure how Künne’s (2003 observation was supposed to help with that more general problem. Belief and truth , 258–260) All this arises by way of questioning whether we could be said to understand propositions solely on the basis of the sorts of inferences that Künne and Schiffer present. I now want to turn to posing a different question, and for that purpose I will grant the Required View. My question will be whether, even if the Required View were true, we could be said to understand ‘propositional attitude’ talk without reliance on a prior grasp of the notion of truth. Could we, for example, come to understand belief talk without reliance on an antecedent grasp of the notion of truth? One might think that, in picking belief to press this question, we have stacked the deck against the Modest Account. But (a) it’s arguable that the concepts of the other propositional attitudes depend asymmetrically on our grasp of the notion of belief, so that this case is central (for considerations in support of this, see Boghossian 2008 ); and (b) it would introduce a significant element of ad hocness into Künne’s account if he had to pick and choose which propositional attitudes could serve as the basis of our understanding of propositions. It would certainly be odd to claim that we could get the notion of a proposition on the basis of desire and hope talk but not on the basis of belief talk. In the case of belief, I am inclined to think, along with many others, not only is it plausible that it is essential to beliefs that they can be right or wrong, correct or incorrect, but that this normative fact forms a very basic part of the concept of belief. You don’t understand belief unless you understand that (a) it can go right or wrong, and (b) that it goes right by being true and wrong by being false. In other words, to have the concept you have to grasp what we might call the: (Norm on Belief): A belief goes right by being true and wrong by being false. Some writers, like David Velleman (2000 ), have sometimes argued for a stronger claim: that there is an important sense in which belief aims at the truth and that this fact helps explain why we have the Norm on Belief. I don’t know if this stronger claim can be made out, but I don’t need to rely on it here. However, even if only the weaker thesis were true, it would make trouble for the Modest Account by implying that one could only grasp the notion of a proposition through its role in belief talk by already having the notion of truth. Künne doesn’t address this issue in the book, but he does make some comments about it in a reply to one of his critics (Sundholm): The principle that an act of judging, or a belief, is correct just in case its content is true is not a norm: it just sets up an analytic connection between the concepts of correctness and of truth (2008, 392). He goes on to suggest that the only real norm on belief is one not in terms of truth but in terms of epistemic justification. The issues here are complex and I don’t have the space to go into them in anything like the detail they deserve (again, see Boghossian 2008 ). Künne is assuming that the only kind of principle that deserves to be called a norm would have an ‘ought’ in it. I think that’s too restrictive. ‘Correct’ and ‘right’ often act as normative terms, and so does ‘reason’. One can have a reason to phi even if it’s not the case that, all things considered, one ought to phi. In addition, even if one stuck with ‘ought’ norms, it’s not clear that one should say that it’s not a norm on belief that one ought to believe that p only if p is true, that the only norms on belief are in terms of epistemic justification. Surely, the truth-norm underwrites the justification norm in much the way in which the norm ‘Buy Low, Sell High’ underwrites the various more directly followable rules that traders on the stock market actually employ: in both cases, the former norm explains why the latter are to be followed. Propositions and truth So this is a problem about getting a truth-neutral concept of proposition out of our grasp of belief talk. Well, perhaps that’s a reason for privileging the other propositional attitudes in our account of our grasp of the notion of a proposition. It seems to me, however, that there is an even deeper problem here, one that would show up no matter which propositional attitude we were to pick. And that is that I think the notion of truth is implicated in the very concept of a proposition. We have already noted that everyone (at least everyone party to this debate) agrees that propositions are essentially truth-evaluable. What the Minimal and Modest theories must deny is that, despite being essential to propositions, truth-evaluability needn’t be something that you have to know up front about a proposition in order to grasp the concept. It’s something that you can come to discover about them, after you have grasped what they are, just like you come to discover that the internal angles of any triangle sum to 180 degrees, after you have grasped what triangles are. Now, I think it has to be admitted that the non-circularity and therefore non-vacuity of both the Minimal and the Modest theories is hanging by a very fine thread. It’s conceded that truth-evaluability is essential to propositions; it’s conceded that it’s a priori that they are truth-evaluable. In the face of all this, though, it’s maintained that knowledge that propositions are essentially truth-evaluable is not strictly needed for grasp of the concept of a proposition. Well, how do we decide whether F’s being G is just a necessary a priori consequence of F, or whether it’s constitutive of grasp of the concept of being F? These sorts of questions are always very hard, of course. But a reasonable suggestion is this: if F’s being G were constitutive of grasp of the concept of F’s, we would be very surprised if, even knowing not very much at all about F’s, someone gave us reason to think that not all F’s are G. Whereas if it’s just necessary and a priori that all F’s are G, and we didn’t know that much about F, we might well take the claim that it’s not G in stride.1 For example, prior to seeing a proof that the interior angles of a triangle must sum up to 180 degrees, it would not have offended against my concept of a triangle if you had told me that they summed up to 160 degrees. This may be mathematically impossible. But it would not have shown up in my mere grasp of the concept. It’s arguable, however, that the relation between propositions and truth is not like that, that I know up front, merely on the basis of understanding: that snow is white that it has conditions of truth, indeed, that it’s true iff snow is white. Both Horwich’s Minimalist view and Künne’s Modest view are committed to holding that introducing the notion of propositional truth into one’s language is a highly optional matter. All of our talk of the propositional attitudes can be firmly in place without our needing to bother about truth at all. On Horwich’s view, truth is something we need to do introduce just in case we need to be able to blindly assent to a proposition that we cannot display or express. (How such a device can function to abbreviate an infinite conjunction that we could not have grasped in the first place is another story.) Künne doesn’t commit himself to such a rationale. But I think he still goes wrong in holding that one could understand the notion of a proposition without understanding the notion of truth. What would be wrong with a Primitivist account of truth? The general tendency of these considerations is that attempting to explain our grasp of the notion of truth by relying on the notion of a proposition is likely to move in a very small circle. We grasp the concepts of the propositional attitudes and of truth as a package. While this is strictly just a view about grasp of the concept of truth rather than a view about the concept of truth itself, I am inclined to think that it comports best with a Primitivist account of the concept of truth, according to which truth is a simple unanalyzable quality that satisfies the Denominalization Schema: The proposition that p is true iff p. This is a position that was held by Frege, Moore, the early Russell and Ernest Sosa. What would be so terrible if we had to accept such a Primitivist account? Why do we need a definition along the lines of the Modest Account? What we are often told is that we need a definition because otherwise the concept of truth would be left mysterious. There are at least two problems with this rationale for seeking a substantive account of the concept of truth. First, when I think about those concepts whose intelligibility worries me and for which it would be good to have an account before we could be fully confident that they made sense, truth is just not one of them. ‘Ought’is one of them; but not truth. Of course, there are the semantic paradoxes. But none of these sorts of account of truth even try to help with those, as Künne would be the first to acknowledge. Second, I don’t see how much demystification we can be said to gain from a theory that relies on the notion of a proposition. I mean: the notion of a proposition really is a mysterious notion. An abstract, mind- and language-independent object that has conditions of truth and falsity but which could not literally be said to be claiming or saying anything. That really does call for explanation. And the great theorists of the proposition – Frege and Russell – worried a great deal about what it could be that ‘held’a proposition together, so that it could, unlike a mere ordered pair for example, be true or false.*