The Elusive Nature of Truth

THE ELUSIVE NATURE OF TRUTH MICHAEL P LYNCH Connecticut College Me gentle Delta beckons from the piam, Then, hzd m shades, eludes her eager swain — POPE Abstract In this essay, I pesent a new argument for the imposszbility of definmg truth by speafyzng the underlyzng structural property ali and only true proposttions have in common The set of consulerations I use to support this clazm take as thar inspiration A lston's recent argument that tt is trn possible to define truth epistemically—in terms of justification or warrant Accordmg to what Alston calls the "mtensional argument", epistenuc def =tons are inconststent with the Tschema or the principie that tt is true that p if, and only tf, p Smce the T-schema has great intuitive appeal, tias is a powerful zndictment of episterruc theones But the basic argument that Alston employs, and the constellatian of considerations whzch pros ecute that argument, work agamst a much broader range of views than he considers W htle this implies that a traditional conceptual analysts of truth rnay be impossible, it opens the door to a piurahst approach to truth Attempts to define truth have never met with spectacular success Plulosophers offer vanous explananons for this fact Trachtion ahsts say that it sunply reveals the difficulty of the problem Work harder, they counsel, and we may yet succeed Defianorasts grumble that any attempt to define truth is nusguided from the start They claun that we can't define truth because there is nothing to define And a durai camp agrees that definttions of truth are unhkels, but stubbornly maintams that there is still something interesting and important about the concept C) Principia, 4(2) (2000), pp 229-55 Published by NEL — Epistemology and Logic Research Group, Federal University of Santa Catanna (UFSC), Brami 230 Michael P Lynch In tlus essay, I present a new argument for the ampossibtlity of defining truth In the dassical sense w e cannot analyze truth ia terras of an underlymg structural property that ali and Only true proposinons share The set of considerations I use to support this claun take as their inspiration W P Alston's recent argument that it is trapos sible to define truth eptstemically—in terms of justification or warrant According to what Aiston calls the "mtensional argument", epistemic definmons are inconsistent with the T-schema, or the principie that tt is true that p tf, and only tf, p 1 Since the T schema has great intuitive appeal, thts is a powerful indictment of epistemic the ones But the basic argument that Alston employs, and the constei lanon of considerattons whtch prosecute that argument, work agamst a much broader range of views than he considers At the very least, they show that a conceptual analysis of truth is even more complex and intractable than previously thought Arguments for the indefinability of truth are not unfamthar Frege, for mstance, once claimed as much 2 And Donald Davidson has de clared that tt is "folly" to define truth 3 So the over ali lesson you draw from this essay will probably depend on your prior stance on the issue An optinustic tradttionahst may well decide that I have gwen a reductio ad absurdum of Alston's inmal argument The pes surnstic deflationtst will take tt as another reason to gwe up on a misguided project A third reaction, neither traditionalist nor defianomst, opens the door to a functionahst understandmg of truth 1 Conceptual Equrvalence and the Tschema If there is one til-111g that philosophers agree on regardmg truth, a is that the followtng schema teus us somethmg important about the concept TS (T-schema) it is true that p it, and only tf, p 4 Specifically, there is general agreement about the necessary truth of tts non pathological, direct instances (what I'll call "T propost nons") 5 A T proposition's component propositions have the same truth-value In every possible world The Eluswe Nature of Truth 231 A familiar explanation for the truth of TS is that "p" and "it is true that p" always express the same proposition, they are synonymous But this deflationary readmg isn't the only one available Wilham Alston takes TS to exemphfy what he calls the "realist concepnon" of truth, accordmg to which a proposinon is true if, and only if, what the proposition is about is as the proposition says that it is (p 1) Ac cordmg to Alston, T propositions are "conceptually, analytically true, true by virtue of the meanings of the terms mvolved, in particular the term `true' " (p 27) Nonetheless, the two stdes of a T proposition are not synonymous, "the proposition that grass is green is true" says something more than the mere assertion that grass is green The first is a statement about a proposinon, the second is about grass (p 47) Alston's posinon illustrates that equwalences hke TS can be In terpreted differently This is not always recognized For instance, one shouldn't confuse TS with Tarski's Convennon T or the related "disquotanonal schema" DS "S" is true if, and only it, S Here, truth is being ascnbed to sentences, not propositions In stances of DS are therefore connngent truths They display an extensional equwalence between certam sentences Each half of the bi conditional supplies necessary and sufficient conditions for the truth of the other half in the actual world Conceptual equwalences, hke T propositions, are necessary truths But not ali necessanly true equwalences are conceptual equwalences For example, there are necessary a posterion equivalences, such as (1) thts is water IFF this is H20 (Kl use "IFF" as a symbol for a necessary equwalence ) A proposition like (1) is necessanly true but—at least on some accounts—not true in virtue of the intensions or concepts involved This shows that modal propernes alone do not distinguish conceptual equivalences from other equwalences Neither do epistemic properties—for example, whether the equwalence or its mstances are knowable a priori Even if ali conceptual equwalences are a pnon, not ali necessary a pnon equwalences are conceptual For mstance, 232 Mtchael P Lynch (2) 5,653+847 = 6,500 IFF 22-3 = 19 is a necessary a priori truth 6 Both of its component propositions have the same truth-value In every world because each is true In every world Yet arguably, (2) is not a conceptual equtvalence While each half of the equivalence may be a conceptual or analync truth it needn't follow that there is a conceptual ar analync connection between the halves Surely it is the conceptual connecnon between the halves which makes an equivalence worth calling a conceptual equivalence I shall call a necessanly true biconditional a conceptual equivalence just when one grasps it a pnon in virtue of a conceptual connection between its component propositions Betng a conceptual equivalence is a matter of degree At one end of the spectrum, under standing the nght-hand side of a conceptual equivalence implies an understanding of the left hand side and vice versa The most obvious example are proposinons of this sort (3) John is a bachelor IFF John is an unmarned adult male Call equivalences hke (3) heavyweight conceptual equwalences Heavyweight conceptual equivalences are necessary truths that one grasps a pnon in virtue of the fact that their component propositions are synonymous, or identical In content There are also weaker con ceptual equwalences—equivalences that are necessary and a priori In virtue of the concepts involved, but lack synonymous component proposittons For instance, it seems necessary and a priori that (4) this object has a shape IFF this object has a size Understanding that something has a size is tmpossible without under standing (perhaps implicitly) that it also has a shape—even if it is an indefirute shape Yet to talk about shape and size is not to talk about the same thing So (4) is a weaker conceptual equivalence than (3) Call (4) a muldlewetght conceptual equwalence (4) is clearly weaker than (3) but not so weak as (5) An ommscient being knows that snow is white white 7 IFF snow is The Elustve Nature of Truth 233 The concept of an omniscient being requires that it have knowl edge of any state of affairs However, understandmg that snow is white does not regime that one understands anythmg (even implic ttly) about omniscience In the sense in which (4) is weaker than (3), (5) is weaker than (4) Of course, we may have trouble distinguish tng very weak conceptual equwalences of thts sort from equwalences whose necessity denves from other (1 e nonconceptual) sources But let us wawe such concerns for the moment and say that (5) is a lightweight conceptual equivalence Biconchtionals are transitwe Thus, two material equwalences that share a component proposition can entali a third ap -= q) &(q m r) D (p m r)] But intuttively, a conceptual equivalence cannot pass on more conceptual weight than it has Compare thts to eptstemic justification epistemically speakmg, it is permissible to inter a weakly justified behef from one that is strongly justifted, but not vice versa A belief cannot make another behef more justifted or probable than it is itself Thus a cham of such mferences is only as strong as its weak est member Analogously, one conceptual equwalence cannot make another weightier than it itself An equwalence cannot be shown to be heavy or middle weight by a chain of equwalences one of whose members is lightweight 8 To return to the T-schema, the weight class of T proposinons is a matter of contention Deflationists typically see them as heavyweight conceptual equivalences like (3) This the deflationist's chief evidence for beheving that truth talk is "content-redundant" Al ston, on the other hand, belteves that T-propositions are lightweights akm to (5) On his view, the two halves of a T-proposition dif fer in content, but one can grasp that grass is green and "yet lack the concept of truth" (p 48) The more usual view amongst nondeflationists, however, and in my opimon the most plausible posinon, is that T proposinons are middleweight equwalences hke (4) As Cnspm Wnght has remarked, ti is a platitude that "to assert is to present as true" (1992, p 34) The concept of truth may not be required to understand any particular proposition, but it is required to grasp the concepts of assertion and something being the case For how could one know the dtfference between something being the case and its not being the case, or between assertion and dental, wah 234 Mwhael P Lynch out understanding the difference between truth and falsay ? As I see a, an implica understandmg of the concept of truth is required to understand any proposition This essay is concerned with attempts to define our concept of truth, but the word "defimtion" is notonously difficult to define One often hears it said that in the halcyon days of yore, our predeces sors attempted to conceptually analyze or define concepts in a very strong sense—by providmg heavyweight conceptual equivalences for the concepts in question I have my doubts about this tall-tale Even without much reflection, a seems highly doubtful that one could provide that sort of analysis for any philosophically interesting concept In any event, now days philosophers are less interested in reducnon and more mterested in defininons or analsses which draw illuminating connections between concepts Of course, such connections are still conceptual and are typically represented by biconditionals—at least they are when the subject is truth, as any survey of the recent literature will attest Therefore, in this essay, I shall take instances of the following to be conceptual definitions or analsses of truth (TD) It is true that p IFF the proposition that p is X, where "X" stands for some property, relational or otherwise, and where the "IFF" denotes a less than heavyweight conceptual equivalence 9 2. Alston's Intensional Argument For Alston, to grasp the concept of truth is to understand that "the content of a proposition determines a (necessanly) necessary and sufficient condition" for its truth (p 27) In this sense, Alston's account is muurnaltst It is also rectlast in that Alston takes the T schema to imply that truth is a germine non-epistemic property of proposi tions, he rejects defiationism Thus, Alston nghtly categorizes lus view as a mtnimal realism (p 37) Alston's positive account of truth deserves senous attention, certainly more than 1 am able to give it here 10 But his minimal realism is not my topic I want to discuss the implications of what Alston calls his zntensumal argument agamst epistemic accounts of truth The Eluswe Nature of Truth 235 As we shall see, the "mtensional argument" actually denotes a family of consideranons aimed at undermining a particular assump non that lies behmd epistemic analyses of truth I should note at the outset I am more mterested In the argument itself than in Alsto man interpretatton Thus my way of presenting these considerations differs markedly from Alston's at certain points The "episternic accounts" Alston has In mtnd attempt a concep tual definition or analysis of truth m epistemic terms Such views assert a conceptual equwalence between a proposition's being true and as having art epistemic property of some sort Alston takes the strongest epistemic theory of truth to be Putnam's internai reahst view as a was In 1981—roughly the icica that truth is Justifiability in ideal condinons (p 195) Putnam has subsequently pointed out that he never took the theory as a "reductwe" analysis of truth—one wluch defined truth in terms of more basic epistemic notions (1989, p 115) Instead, he clauns the two concepts are traerconnected (ibid ) This imphes that understanding the one concept requires an understanding of the other, which in turn suggests that Putnam's takes instances of the following as middleweight conceptual truths (E) It is true that p IFF the proposinon that p would be ideally Justified (that is, p would be Justified under ideal epistemic cacumstances) Whether or not this is correct as an interpretation of Putnam, (E) is a reasonable representative of eptstemic accounts of truth Alston's basic argument against (E) is simple The potra of the T schema is that a proposmon specifies the necessary and sufficient condinons under whtch a is true But (E) implicitly dentes thts The content of the proposition (obtaming) isn't sufficient or perhaps even necessary according to (E)—a requires something more The proposition in question must have a certain epistemic status Snow's being whtte is not sufficient In order for it to be true that snow is whae, what is necessary and suffictent for the truth of the proposition is that the proposition--or a behef with that proposition as content—be Jusn fied m ideal epistemic circumstances When first encountenng thts point, rnany people react with exasperation Surely, that can't be right, they say—there is more to be 236 Michael P Lynch said Indeed, as Alston admits, til% first remark only establishes a "prima facie case for incompanbility of the T-schema with an epistemic conception of truth" (p 211) The full nature of the inten sional argument only becomes apparent when one considers the natural objections to this minai. point The objections that Alston dis cusses himself he treats as distinct In fact, they are ali vanations of a single strategy, one which usefully bnngs to light a particular as sumption imphcit m any attempt to analyze truth The assumption is that there is a bridge-pnnciple with connects the T-schema with the purported definition In quesnon Specifically, it seems that we could make TS consistent with the epistemic conception of truth by denv mg that concepnon from TS together with the following premise (E2) p if, and only if, the proposition that p would be ideally jusn fied If one grants the transinvity of the biconditional, the forrn of the argument is straightforward The clara is that one can grant the conceptual truth of TS while holdmg (E) as the correct definition of truth because TS and (E2) together imply (E) The success of the point depends on how we interpret (E2) 11 There seem to be four possible interpretations Interpretation A is that (E2) is a non-conceptual truth of some sort There are two ways in which a proposition can be non conceptually true it can either be a contingently true proposition, or it can be necessanly true but not in virtue of the concepts involved The first, and less plausible, of these alternanves is the one that Alston considers, but the stronger version tares no better The more plausible version is that (E2)— like (1) above—is a metaphysical dali-ri The epistemic theorist is explairung what it is for a certain state of affairs (snow's being white) to obtam It is a necessary truth (a synthetic a priori truth perhaps) that when p obtams it is ideally justifiable and vice versa But so in terpreted, (E2) not only smacks of idealism, the conclusion that the derivation supposedly guarantees becomes unwarranted A realist can grant the mference and yet deny that tias tact implies anything about the concept of truth Specifically, one can grant (E2) without thereby beheving that our concept of truth is definable as idealized justifiability For suppose it could be the case that p just when the The Eluswe Nature of Truth 237 proposition that p is ideally justifiable—this could merely be a fact about the limas of the unwerse or the extent of our mmds So interpreted, (E2) need not entall any particular concepnon of truth at ali Interpretation B of (E2) involves what Alston calls a reinterpre tation of content (p 214) A defender of the epistemic concepnon of truth might try to get around the tntensional argument by claiming that any instance of the right hand side of the T schema, such as snow is white "states" that that proposawn would be justified under ideal epistemic conditions In other words, (E2) is a heavyweight conceptual equw alence—even if (E) itself is not This strategy has the ment of direct ness, and it certainly allows the eptstemic view to use the T-schema But it faces not only the problems of the more modest interpretations of (E2) (see below), it imphes a rather nasty regress To take (E2) as a heavyweight equwalence amounts to saymg that any proposinon is a proposinon about the epistemic status of some proposition, namely itself As Alston points out, this imphes that it is impossible to spec ify which proposinon it is that I am talking about (p 216) For if every proposition is a proposition about the eptsternic status of ttself, then the proposition that p is the proposition that the propostuon that p would be &any jusgfied It follows in turn that the proposition that the proposaton that p would be tdeally justtfted is identical to the proposmon that the proposition that the proposztion that p would be zdeally pistified would be tdeally justtfied And so on The content of our statements is unspecifiable, and so mterpretanon B ends at incoher ence 12 13 One might wonder if the epistemic theonst couldn't deflect this argument by pointing out that a similar regress obtams In the case ot "true" itself That is, if one takes the T-schema as a heavyweight conceptual equwalence, then asserting that p is equwalent to asserting the proposition that p is true And to assert the proposition that p is true is equwalent to asserting that the proposition that the proposttion that p is true is true, and so on But, someone might argue, this regress clearly isn't vicious, since we take it that the truth values of ali these proposmons are determmed simultaneously—the vanous 238 Michael P Lynch "true's" cancel out, so to speak 14 But, the reply to Alston might go, tf they cancel out ia this case, surely the eptstenuc theonst can make the same claim on behalf of her view Several points are relevant in reply First, recall that Alston does not take TS to be a heavyvveight conceptual equwaknce And netther does the astute eptstemic theonst Philosophers who regard "it is true that p" and "p" as strongly content-equwalent mean to say, In effect, that "it is true that" is a mere grammancal operator It adds nothing to the content of p itself The point is that truth-talk is redundant But this can't be the attitude of the epistemic theonst toward either the TS or (E) Her pomt is not that truth-talk is redun dant, it is that "p would be tdeally justitied" reveals the real content of "p" In some profound and complicated sense ot "real content" The eptstemic theonst, in offenng a conceptual definition, takes herself to be explaming our concept of truth, not elimmating it Thus there is an Important dtfterence between the regress mvolving "true" and the regress involving "ideal justifiability" The final two strategies argue that (E2) is a lighter weight conceptual equwalence of some sort The more plausible alternatwe is C, which takes (E2) as a middleweight conceptual equwalence As I argued above, TS and (E) are themselves best interpreted in tlus way, and it seems intuitive that the eptstetruc theonst would take both (E) and (E2) to have the same status as TS A typical way that realists respond to this sort of suggestion is by counter-example Thus, Alston notes that "it does not ~late [the concept ot p's being ideally justifiable] to suppose that in some cases a behef that p is ideally justifiable without it being the case that p or vice versa" (p 214) For instance, it isn't incoherent to suppose that the number of stars in the unwerse at this moment is odd But that is a claun which surely couldn't be justftted by creatures hke us even in ideal epistemic conditions This is In marked contrast to TS, In that it is a clear vtolation of our concept of truth to suppose that it could be true that p wahout its being the case that p Of course, it is always possible for the epistemic theonst to Insist on the metaphystcal necessay of (E2) Indeed, such principies may well be the consequence of an ontology that has it that "the mind and the world tomtly make up the mind and world" (Putnam, 1981) The Elustve Nature of Truth 239 That is, deep rrietaphysical facts about reference or the reach of hu man expenence could entail that necessanly, a state of affairs can obtam when and only when a certain proposmon would be justified In ideal epistemic circumstances If so, then propositions such as, e g, the number of stars in the uruverse at this mament is odd may turn out to be neither true nor false and therefore "incoherent" after ali Yet even tf these metaphystcal clauns are correct, that fact alone cannot prove that (E2) is a conceptual equivalence as opposed to a surd metaphysical fact on the order of interpretation A above In order to show that (E2) is a middleweight conceptual equtvalence, one must argue addinonally that grasping that any proposition is the case (or understanding that any state of affairs obtams) necessanly presupposes grasping that the proposition in quesnon is ideally justified Contrapositively, it would have to be shown that if one cadt understand how a proposinon (e g one about the chstant past) would be jusnfied even rn ideal epistermc circumstances, one can't be said to understand that propostnon But tt is precisely this that the above metaphysical arguments do not show I can grant that (E2) is nec essanly true and yet sun maintain that In a straightforward, =mal sense ot "understand" that I understand that the number of stars zn the universe right now is odd wahout having the faintest inclination to thmk that this proposition would be ideall justified or unjustified In potra of fact, many folks entertain, doubt and wonder about ali sorts ot propositions about, e g God, without having any sort irnplicit un derstandmg of whether they would be justified in ideal epistemic circumstances The truth is that most people wouldn't recogruze ideal eptstemic ctrcumstances af they carne up and Int them on the head Thus, even it the usual counter examples to (E2) don't prove that a is not necessanly true, they do illustrate that as necessity is not conceptual in character Furthermore, to say that (E2) is a middleweight conceptual equtv alence is to imply that grasping the concept of ideal epistemic cir cumstances is implicitly necessary for grasping any proposttion The range of proposaions that we can understand, and the range of con cepts we use in understanding them is extremely wide, and due to the ever-changing human situation, indefinitely open-ended Yet (E2) meant to apply equally well to any proposinon It is theretore un Mtchael P Lynch 240 surpnstng that general equivalences about understanding along the Imes of (E2) are not particularly convincing Interpretation D, thankfully, needs bule discussion It takes (E2) as a lightweight conceptual truth This would mean that there is a conceptual connecnon between the halves ot (E2) without an un derstanding of snow's being white implying even an imphcit under standing of ideal eptstemic circumstances But as I noted above, a conceptual equivalence cannot pass on more conceptual weight than it has Therefore, if (E2) is lightweight then (E) must be as well But most epistemic theonsts, I think, would balk at denymg that even an implicit grasp of "rational acceptability" or "justifiability" is required for an understanding of truth The point of epistemic theones, after ali, is typically that the concept of truth whtch we use In our datly practice imphcitiy mvolves an appeal to what would be rational or justffied to accept So from the eptstemtc theonses pcnnt of view there seems to be 'tule mouvanon for takmg (E2) and hence (E) as lightweight equivalences—other than the mere dem-e to avoid the present argument 3 The Global Intensional Argument Our Alstoman argument has considerable force—even more force than Alston hunself intended If the intensional argument is sound, then so Is the global intensional argument If epistenuc definitions are set aside, there is only one senous con tender for a truth-definition the correspondence account Proving that the correspondence theory is no better off when it comes to the intensional argument than epistenuc accounts will therefore suffice to prove the more general clatm The essennal core of any correspondence theory of truth is that a proposition or statement is true when a corresponds, fits, or matches realtty Theones of this type are distmgutshed from each other along three pnncipal limes, namely their respective views about the nature ot the truth bearers, the truth makers and the relanon of truth self 15 Thankfully, the success or tailure of the intensional argument does not rest with the specific content of the theory mvolved Therefore, 1 shall take as my target a fairly unassummg statement of the The Elustve Nature of Truth 241 view (C) It is true that p IFF the proposition that p corresponds to a fact One might tiunk that (C) is a tad bit too unassuming, since many assume that it is a plantude that a proposition is true when it fits the facts (e g Wnght, 1992) That is, deflanorusts typically take it that they can grant principies like (C) but still deny that truth is a substantive or authentic property (Horwich, 1990) 'The tnck is sim ply to read "corresponds to the facts" in a metaphysically innocuous way—as not commating one to any substantial metaphysical rela tionshap between propositions and mind mdependent things called "facts" So if the correspondence theonst is gomg to be seen as as sertmg somethmg distmetwe, she can't simply assert (C) as a mere banahty The correspondence theonst wants more than the right to use the words "correspondence to fact" Therefore, let us take it that the "correspondence" relationship mennoned in (C) is a meta phystcally substannal, complex relationship which obtams between a mmd-mdependent fact and a proposition—without specifying (so as to be neutral between compenng views) the exact na ture of this complex relation (whether it is a structural relationship between objects and parts of the proposinon for mstance) Furthermore, "fact" can be taken simply as a placeholder for whatever parts or aspects of the world are on the other side of the correspondence relation with the proposition—that is, as neutral between compenng accounts ot the nature of facts Finally, and as we did with (E), let's take our repre sentative correspondence theory as ottenng a less than heavyweight conceptual analysis of some sort Yet even mterpreted tn thts way, the correspondence theory wouid seem to be incompatible with our schema For the T schema does not say anythmg about "correspondence", ali it says is that if snow is white then it must be true that snow is white, and that if it is true, then snow is white This is ali that is needed, there is nothing said about a propos-n-1°n correspondmg to reality Again, the obvious strategy is to show that the correspondence definition of truth is conststent with the T schema because that defiration is entatied by TS and the following principie 242 Michael P Lynch (C2) p if, and only if, the proposition that p corresponds to a fact Again, interpretation A regards (C2) as an non conceptual claim, according to which the state of affairs of snow's being white can only obtam if a proposmon corresponds to reality That is, the state of affairs of snow's bemg white depends for its actuahty upon there being a proposinon, which corresponds to reality, that snow is white But so understood, version A again does not warrant the desired conclusion For why would this surd metaphysical fact---if (C2) were a facthave anything more to do with our concept of truth than any other fact ? Suppose God teus us that snow (actually) is white only if the proposition that snow is white corresponds to, e g a fact, and so on for every other proposinon Why would this fact necessanly entali that the concept that we humble humans presently employ when we predicate "true” to a behef or statement is the one marked out by the correspondence schema? Things are no better with version B either The correspondence theonst could also reinterpret the content of ali of our statements That is, the correspondence theonst could take it that the propos' non thát grass is green is identical to the proposmon that the proposition that grass is green corresponds to sorne fact And so on The mfinite regress is exactly the same heie as it was for the episternic theonst As is the consequence if every proposinon is about the correspon dence of some proposinon to fact, then there will be no way to specify which proposition is being discussed Interpretanon C, however, may seem more plausible in the present case Under das interpretation, (C2) is a middleweight conceptual equivalence As I noted above m the case of the epistemic theonst, the defender of the correspondence account might see thts as the most natural reading, since as we've argued above, TS is best inter preted as a middleweight conceptual equivalence itself To say that (C2) has this status is to take it that an understand ing of correspondence relationships is imphcaly necessary for grasp mg any proposinon On such a view, understanding correspondence relationships is a conceptual presupposition of understanding period Remember that the con-espondence theonst, in advocating both (C) and (C2), is not simply pushing plantudes She takes herself to be of The Elusure Nature of Truth 243 fenng a substantive conceptual analysis of truth Hence the concept of correspondence betng assumed here must be robust and complex in nature This is perhaps unremarkable as long as we are only deal mg with (C) itselt In any event, whether (C) is a plausible conceptual truth m as own nght is not the present concern The quesnon is the status of (C2) Once we recall that the correspondence being spoken of is a metaphysically complex relationship, how plausible is it that there is even an imphat conceptual connecnon between understanding simpliciter and understanding the philosophical notion of a "correspondence" relanon between a mind mdependent fact and a proposmon7 Consider, for mstance, the proposition that Smith is selfish h seems quite plausible that one could understand this proposition perfectly well, and even take it to be true, without havmg any sort of grasp, implicit or otherwtse, of a metaphystcally thtck relatton ship between that thought and an objective entity, the fact of Smith's selhshness As we noted when dtscussing (E2) above, the range of propositions we assert In everyday hfe is incredibly wide Perhaps it is true that we must have the concept of a complex correspondence relationship between proposition and fact in order to understand some propositions But having a concept of correspondence can hardly be a presupposition of understanding across the board To thmk other wise would place an undue burden on those who know much about the real world but httle about philosophy So we arnve at interpretation D, where we take (C2) as a hght weight conceptual equwalence On this reading, (C2) is conceptually necessary even though one can understand any proposition without knowmg (even implicitly and indirectly) anything about cor respondence Again (because a conceptual equwalence cannot pass on more conceptual weight than tt has itselt) this makes (C) a light weIght conceptual truth as well As was the case when we considered tias alternative under the auspices of the eptstemic theory, one wonders what the motwation would be for such a position & yen that TS is a middleweight con ceptual equwalence, why thmk that (C2) and (C) have lightweight status 7 But perhaps the correspondence theonst can supply an answer Perhaps (C2) is a conceptual equwalence not because of di rect connections between its component proposinons and their con- Machael P Lynch 244 cepts, but indirectly, because (C2) is ttself denvable from some further equivalences One suggestion, for example, would be to derive (C2) trom the following F schema (FS) h is a fact that p IFF p (C3) It is a fact that p fact IFF the proposition that p corresponds to a (C2) p IFF the proposition that p corresponds to a fact Having deduced (C2) trom FS and (C3), we can then (together with TS) infer (C) as a lightweight conceptual equivalence Yet httle real ground has been gamed by this maneuver For we are now faced with deciding how to interpret these addinonal premises Again, either could be read in a deflationars way—as non-metaphystcal plantudes But thts won't do for the correspondence theonst, who sees facts as real entales out in the world That ande, let us grant that the FS \vill have the same conceptual status (the same It wetght class") as TS—whatever that turns out to be The real issue is (C3) Once we read (C3) In the mtended metaphysically senous sense, it turns out to be as poor a candidate for a conceptual equiv alence as (C2) Why should we thmk that In order for anyone (not just philosophers) to even understcmd what a fact is that she must understand—in every case—complex metaphysical relationships be tween tacts (seen as real things out in the world) and propositions (seen, e g as abstract objects) ? Without an additional argument, this seems to be an unwarranted assumption So we are left with taking (C3) as a lightweight conceptual equivalence Which is exactly where we were with (C2) to begin with There is not more mon vanon for holding (C3) as lightweight conceptual equivalence than (C2) Further, the correspondence theonst must face a sertous question we have so far avoided As we asked when we were concerned with their epistemic counterparts, what makes (C2) or (C3) lightweight conceptual equivalences as opposed to straight non conceptual equiv alences of either the "mathemancal" kind of example (3) or the "metaphysical" land employed in interpretation A above 7 For familiar Quinean reasons, being unable to make out sharp border between The Elustve Nature of Truth 245 non conceptual and conceptual truths is perhaps what we should ex pect But that fact will hardly help the traditional correspondence theonst The quesnon we are interested in is precisely whether a is possible to gwe even a "weak" conceptual defimtion of truth in terms of correspondence The correspondence theonst, in so far as she beheves that such an analysis is possible, is committed to beheving that one can sensibly call some equwalences conceptual and others nonconceptual, and furthermore, that we can do so with certainty In this very case For instance, even if we grant the truth of (C3), In other words, why should we take it that it is true in virtue of its component concepts (e g the concept of a fact) as opposed to the way the world is ? One way to make the point, of course, would be to appeal to still further equwalences, and attempt to show that what latle conceptual strength (C3) has denves from them But the prospects of finding such equwalences seem dim to say the least, even if we wawe the obvious fact that the same questions could be raised again at that point The upshot is that, as we saw with (E2), taking (C2) as a lightweight conceptual equwalence is poorly monvated or adhoc But ex,en tf we put the questzon of mottvatzon aszcle, (C2) seems to me a poor candidate for being a conceptual equwalence of any sort At a mmimum, to take an instance of (C2) as a conceptual equwalence is to hold that we grasp it In virtue of a conceptual connection which obtams between its component propositions Yet (C2) is not meant as a sunple defiationary plantude Whatever theory of correspondence we have m mind, the alleged conceptual connection in question is gomg to be of a very fine gramed and specific nature This means that it should be a priori that snow cannot be white unless a particular proposition is In a particular metaphysically substannal and complex relationship to reality But a hardly seems a conceptual truth that snow can be as a is only if there are relations of correspondence between it and propositions The world does not require—by definition'—that there be correspondence relations at ali I have found that some philosophers who are inchned to favor these sorts ot consideranons when they are directed agamst epistemic theones react with somethmg akm to outrage when they are apphed to the con-espondence account Two points need underlirang First, 1 am not disputing (C)'s intuitwe plausibility, nor indeed (except in 246 Mtchael P Lynch directly) have I been concerned with (C) at ali The target has been (C2)—and again, not with regard to its truth, but with regard to its status as an (alleged) conceptual equivalence Second, there is a tendency among some to slide back and forth between seemg the correspondence vtew as an mnocent platitude and regarding it as a chstinctive philosophical theory of truth opposed to other such the ones Unnoticed, such shding can make it seem msane to doubt the conceptual necessity of (C2) Yet once we become aware of such waf fling, and regard (C2) in its proper light, there is a senous quesnon as to what sort of conceptual equivalence it could be 16 4 Implications There are many reasons, over and above those just presented, for thinking that neither the correspondence theory nor the epistemic theory of truth is plausible So it is not surprising that these defininons of truth do not succeed The surpnsing fact is that they ultimately fail for structurally similar reasons In order to be consistent with the T schema, both accounts require it to be conceptual truth that a state of affairs obtams when and only when a proposition has a certam robust property This means that there must be a conceptual connecnon between understanding any proposition and understand ing that a has that property Yet on examination we fmd that any Interpretanon of (E2) or (C2) is either too weak or too strong to do the job Interpreted strongly, neither pnnciple apphes to every proposinon, mterpreted weakly, they become indistinguishable from stnctly metaphysical truths Therefore neither (E2) nor (C2) can transfer the proper conceptual weight to either (C) or (E) Since the global intensional argument works against reahst ac counts of truth, and Alston. is a realist, one might condude that the monster has eaten its maker This is too hasty Alston does hold that our concept of truth imphes that truth is a property of proposi tions But he stops short of presenting a definition of truth in terras of that property To grasp that T propositions are necessanly, conceptually true, just is to grasp the concept of truth on Alston's view The T schema, understood as a light or nuddleweight conceptual equiv alence, gives us a recipe of sorts for construcnng T-propositions To The Elustve Nature of Truth 247 understand this recipe is to have the concept, no definition is needed What the global intensional argument does reveal is that any Po sition, including Alston's, which takes truth as a property must also take truth as conceptually primitive in a certam sense For in order to be a reahst and avoid the global mtensional argument, one must (at a minimum) hold that (a) our concept of truth is a concept of a property of propositions, but (b) that concept is not definable as the concept of an underlying robust property ali and only true propos' tions share To deny (a) would be embrace defiationism, to deny (b) would mean facmg the global intensional argument Truth turns out to be a basic concept As a consequence, our discussion mas seem to encourage deflationism As Paul Horwich remarks, the defiationary perspective in general is that the search for an analysis [of the torra "13' is true IFF 'p' has property 'E") is misguided, that our concept is exhausted by the uncontroversial schema, and that there is no reason at ali to expect that truth has any sort of underlyzng nature (1995, p 358) If the global intensional argument is sound, then not only is the attempt to define truth nusguided ot not needed, it is impossible For accordmg to the argument, the concept ot proposmonal truth (the temi "true") cannot be defined as a certain tspe of property If one is already tempted by deflationism, the mference from "the temi 'trile' cannot be defined as a particular property of propositions" to "there is no property of truth" \Anil no doubt prove irresistible We seem stuck between saying that truth has an mdefinable nature and holding that truth has no nature at ali This is not a comfort able position We need a third alternative In the remaming pages, I'll try to provide a quick sketch of what I thmk that alternative could be The global intensional argument shows that we cannot analyze our concept of truth by defining it in terms of a single underlymg property shared by ali true proposttions This does not entall that truth has no nature, but it does suggest that there is no single nature ot truth Thus a third alternative is that truth has more than one na ture Crispin Wright has called this position alethic plurahsm (1992, 2001) 248 Michael P Lynch If the only essennal propernes of a truth prechcate are formal— a matter of its use complymg wtth certam yery general amorns (plantudes) —then such predicates may or may not, In different areas of discourse, have a varymg substance (1992, p 23) According to Wright, our concept of truth may be quite =mal, but this leaves open the possibility that the underlying nature of truth may take on distinct forms m Ifferent domams of discourse Sum.larly, Terence Horgan has suggested (1986, 1991, 1996) that while truth m every discourse is "correct assertibility", what correct assertibility amounts to may differ dependmg on the context On a plu ralist account, principies like (E) or (C) may fit some discourses but not others Thus, the correspondence theory may apply to propos' tions about the middle-sized dry goods of the physical world but not to proposmons of an intuitively more contextuai sort Accordingly, a principie hke (E) could be applied to e g moral truth or the truth of propositions about psychological atntudes Recall the proposition that Smith is selfish We take such propositions to be capable of being true but the fact of Smith's selfishness seems quite dtfferent In kind than the fact that there is a book on the desk I can't knock up agamst Smith's selfishness in the same way I can bump into the desk and the book Accorchngly, perhaps Smith is selfish just when the behef that he is selfish is a member of an internally coherent system of propositions about human acnon in general and Smith's actions in particular The truth of Smith's selfishness may be radically humandependent in a way that the truth that there is a book on my desk is not The pluralist's point, which our global mstensional argument seems to support, is that while (E) and (C) work well m certam do mains, they fali when apphed globally A particularly clear way of understanding how truth might be plu ral In nature, and one that I've defended at length elsewhere (see Lynch 2000, 2001) takes truth as a funcnonal concept On this account, we don't interpret the T schema as something that needs to be explamed by some other more fundamental equivalence, but as revealing a fundamental aspect of what true propositions do, their funcnonal role We can understand it as tellmg us that our concept of truth is the concept of whatever property a proposinon has when The Elustve Nature of Truth 249 the world is at that proposinon says that it is Roughly speaking, "saymg it like it is" is part of the funcnonal role of true propositions and proposmons that do so (and also fulfill vanous other conditions) have the property of truth A funcnon is a type ofjob And a funcnonal concept is the concept of that job or functional role Thus bemg a mousetrap is a funcnonal property in this sense, and our concept of a mousetrap is the concept of a device that does the job of catchmg mice But of course, this single job can be done or realized In quite different ways To define this job, we wnte a job-descnption specifying how that job relates to others in the immediate economic vionity We define the job in terms of as place in a larger network of jobs, ali of which are understood in relation to each other Accordmg to what is often called "commonsense" functionalism, mental concepts such as behef and desire can be understood In precisely this way Such concepts are "package deal" concepts In that they are not mdividuated one by one but by their place In the network of implica and exphat psychological platitudes that make up our common-sense psychology The totality of platitudes makes up the job descnptions for these mental states In my view, we also have a folk theory of truth, or a network of platitudes, pnnciples, concepts and generalizations a grasp of which con stitutes having a sense of the true and the real These plantudes and principies needn't be explicitly beheved of course Like the analogous psychological pnnciples, it is enough that a grasp of them is imphcit in our reasoning and behavior Some of these platitudes will concern the interconnections between alethic concepts, including e g "true propositions correspond to facts", "facts are what make propo sitions true" and "the proposition that p is true if and only if p" and lt a proposinon is true just when its neganon is false" Some relate truth to other sorts of concepts Some of these, such as "If a behef is caused by a rehable process, it is probably true", will be analogous to input clauses they take us from concepts on the outside or near the edge of our alethic network to an apphcanon of "true" Others (the output clauses) will take us in the reverse direction, for example "If a proposmon is true, you ought to beheve it" While sun others ("a proposition can be jusnfied but not true and true but not justified") may not fali determinately into one group or another Unhke our folk 250 Michael P Lynch psychological theory, our folk theory of truth is not prunanly a causal theory In the case of human psychology, most of the platitudes urdi be causal tn nature (e g pam causes worry) But not ali will be Others, like "toothache is a type of pam", will be quasi logical With regard to our folk theory of truth, this order is reversed one expects that most of the pnnciples will be quast logical, although there is nothmg to rule out the possibility that some may ais° be causal So truth funcuonahsm does not define truth as an underlying property shared by ali true propositions It defines our concept of truth hohsncally—by as role In this network formed by the conjunc non of the common-sense plantudes invoiving truth In short, to be true is to play the truth role Much more needs to be said m order to expiam the functionalist theory But even our short descnption points to a way in which that theory's pluralist credennals allows it to bypass the threat of the global intensional argument The hallmark of a funcuon is that it can be muinply realized without our concept of the function chang ing in any essennal way Yet we can allow that the role does take on a more or less robust character from domam to domam We may take it that the piantudes compnsing our folk theory of truth are ranked, and it is the higher-ranked plantudes, such as the T schema, that compnse the essential truth role Nonetheless, we allow on this view that In some domains certain additional plantudes could be added to our folk theory of truth, and thus the concept of truth in those domams takes on a more robust character The truth role, therefore, has a nurnmal core that is subject to more or less robust ennchment m different domains In our discussion ot the global intensional ar gument, we found that both the correspondence and epistemic ac counts foundered on findmg a conceptual bridge to the T-schema Netther (C2) nor (E2) is a plaustble candidate for the status of conceptual equivalence when apphed across the board to every propost non Yet unlike more tradinonal theones of truth, the functionahst theory—by allowing for diffenng reahzations In dtffenng domainsis consistent with this conclusion While it is implausible that either principie is a necessary presupposition of understanding In general, both principies may be conceptual presupposinons of certam types of understanding There may be certain norms or principies that op The Elustve Nature of Truth 251 erate over some types of propositions such that the grasping them presupposes grasping that they correspond to reality (In the case of proposmons of some domains) or are ideally jusnfiable (as may be the case with propositions of other domams) If so, then we might allow that relattve to these specific domams, (E2) or (C2) act as middleweight conceptual equivalences They do so by helpmg to constitute the set of plantudes that, relative to that domam, mark out the truth-role Thus a functionalist account ot truth allows that the nature of truth may vary by allowing more than one property to play the role marked out by the concept But it also allows for the possibility that our min imahst, functional concept of truth may be expanded and ennched in distinct domains Numerous questions about any pluralist etfort remam, but I think pluralism about truth, especially in its functionalist form, is promis ing At the very least it suggests a new explanation for our apparent inability to define truth as a single underlymg property that ali true propositions share Perhaps the nature of truth has eluded us, not because it has no nature, but because it has more than one 17 References Alston, W P 1979 "Yes, Virginia, 'There is a Real World " Proceedings of the Amencan Philosophical Association 52 779-808 — 1996 A Rectlist Conception of Truth Ithaca Cornell University Press Blackburn, S 1984 Spreadzng the W ord Oxford Clarendon Press David, M 1994 Corres pondence and Disquotation Oxford Oxford University Press Davidson, D 1996 "The Folly of Trying to Define Truth " Time Journal of Philosophy 93 6 Dummett, M 1973 Frege Phdosophy of Language Cambndge Harvard University Press Fine, A 1989 ` Truth-Mongenng Less is True " Canadzan Journal of Philos ophy 611-16 Frege, G 1967 "The Thought A Logic-il Inquiry " trans By A M and M Quinton, In P F Strawson (ed ) Philosophical Logic Oxford Oxford .. University Press Horg in, T 1986 "Psychologism, Semanncs and Ontology " Nous 20 21– 31 252 Mtchael P Lynch 1991. "Metaphysical Realisin and Psychologistic Semantics " Erkenntras 34 297-322 1996 "The Penls of Epistemic Reductionism " Phz/osophy and Phe nomenological Research 56(4) 801-7 Horwich, P 1990 Truth Oxford Basil Blackwell — 1995 "Meanmg, Use and Truth " Mind 104 355-68 Kalderon, M E 1997 "The Transparency of Truth " Mmd 106 423 Kirkham, R 1992 Theorzes of Truth A entical Introduction Cambndge MIT Press Kovach, A 1997 "Why We Should Lose Our Natural Ontological Attitudes " Southern Journal of Philosophy 35 57-74 Lynch, M 1997 "Mnumal Realism or Realistic Minimalism " Philosophical Quarterly 12(189) 512-8 1998 Truth zn Context Cambndge MIT Press 2000 "Alethic P/uralism and the Functionalist Theory of Truth " A cta A nalytica 15(24) 195-214 — (ed ) 2001 The Nature of Truth Classic and Contemporary Perspec aves Cambndge MIT Press — 2001 "A Functionalist Theory of Truth " in Lynch, 2001 Pitcher, G 1964 Truth Oxford Oxford University Press Putnam, H 1981 Reason, Truth and History Cambndge Cambndge University Press 1989 Representatzon and Reality Cambndge MIT Press — Williams, C J F 1976 W hat is Truth 7 Cambndge Cambndge University Press Wright, C 1992 Truth and Objectivity Cambndge Harvard University Press 1995 "Truth In Ethics " Ratio (New Senes) 8(3) 210-26 — 2001 "Minimahsm, Detlationism, Pragmatism, Pluralism" in Lynch, 2001 751-89 Keywords Truth, realism, annrealism, Alston, correspondence, pluralism Michael P Lynch Department of Phtlosophy Connecticut College New London, Connecticut USA mplyn@conncoll edu The Eluswe Nature of Truth 253 Notes 1 In his 1996, p 208ff Unless otherwise noted, ali parenthencal references are to this text Z Frege, 1967 Frege's argument is a distinct pomt from Alston's, despite the fact that both arguments do share a common premise—namely, the Tschema Frege's poli-R revolves around what we might call the transparency of truth, see Dummett 1981, ch 13 and Blackburn, 1984, Sc 7 2 The best recent discussion of Frege's poli-a is Kalderon, 1997 3 Davidson has been committed to tias position for some time I refer here to his 1996 4 1 will take this formulanon of TS as equivalent to "the proposinon that p is true if, and only if, p" ir is a matter of contention whether one can meanmgfully quantify over equivalences like TS In this paper, I will just assume that one can do so by employmg subsntutional quannfication But I don't think that much hangs on this assumption, since I thmk that the problems that generate the intensional argument (in either its local or global forms) would remam even if we could only assert T-propositions mdividually 5 Pathological instances generate the liar or related paradoxes Indirect instances would mvolve so-called bhnd ascnptions of truth, such as "everything John said was true" 6 It may also be knowable a postenon, of course By calhng a truth "a pnon" I only mean that it is knowable mdependent of expenence 7 One might Object that (5) must assume that an omniscient being exists For those with such quibbles, there is (5*) An ommscient being woukl know that p 1FF p Of course, as I say in the text, vanous considerations of this sort may cause one to suspect that with regard to any very weak equivalence of this sort, we aren't really dealmg with a conceptual necessity here at ali 'This is an important question, but 1 am here only pointing out that it looks like lightweight conceptual equivalences are possible 8 Many thanks to Robert Barnard for suggesnng the labels ` heavyweight" etc , as w ell as the analogy with episterrac jusrification 9 Of course, this isn't the only use of the word "definition", and thus other types of dei-mations of truth may escape the argument But I think it is clear that traditional attempts to say what truth is have either explicitly or implicitly been couched in terms of such biconditionals Some readers have suggested to me that we do not need to appeal to necessary equtvalences, however Perhaps instead of TS, for instance, ali one really needs are Instances of DS understood as contingent a pnon truths Perhaps, but note that we are inchned to take " `snow is white' is true iff snow is white" as a 254 Mtchael P Lynch pnon only if we take "snow is white" to mean that snow is whne Only interpreted sentences—sentences that can be taken as expressing a proposmon, are true or false Thus, in order to take an instance of DS as an a pnon truth, we must be considenng its component sentences not as contentless scnbbles but as meaning something And given that "snow is white" means what is does, then tt is necessardy tnie that "snow is white" is true just when snow is white 1° For a more detaded summary and analysis of Alston's position, see Lynch (1997) 11 Of course, to demonstrate the consistency of TS and (E) one needn't run the denvation in just this way Any of the three equivalences could be the conclusion I simply put the pomt this way because it seems the most intuitive None of the arguments m this section, or the next, rest on the order of the equivalences 12 'mis particular poliu (in a somewhat different form) can also be found in Alston's APA Presidennal Address, 1979, p 795 Arthur Fine later made a similar argument m his 1989 For an illummating discussion of the latter, see Adam Kovach, 1997, p 60-61 13 It is w orth notmg that as Alston makes olear, this infirme regress needn't threaten Dummett's venficationist theory of meaning and content, provida-1g that theory is interpreted in a particular way (Alston, 1996, p 220) If, for example, we take the theory to be that the content of an assertion is identical to a statement of that assertion's venfication conditions (e g the proposinon that roses are red is the proposition that roses look a certain way to observers in normal conditions) then reference to the ongmal proposinon is avoided m the analysis of the proposition and the regress is avoided (See, e g Dummett, 1973, p 586 for a statement of his view that is similar to this) 14 This is the so-called transparency property of truth, which lelds to Frege's argument mennoned above See Kalderon, 1997 15 For discussions of different types of correspondence theones, see Pitcher 1964, pp 9-11, and Kirkham, 1992, pp 119-49 For an excellent discussion of a particular type of correspondence theory (one which takes the truthbearers as sentences) see David, 1994 16 It is worth bnefly considenng w hether the global intensional argument works against one last theory—namely, Tarski's semantic theory of truth Just what that theory is, (and whether or not it should be considered an example of a correspondence account) is disputed Furthermore, Tarsla was concerned with linguistic, or sentential truth, while we are necessanly concerned vvith proposmonal truth But Joel Fnedman has suggested to me Th£ Elustve Nature of Truth 255 that we might cut through these problems by simply considenng whether the argument applies to the followmg Tarskian definition (F) the proposition that p is true IFF every sentence that expresses the proposinon that p is sausfied by every mfimte sequence of objects Here, we assume a powerful enough meta-theory and that "sansfaction" is defined iecursively in the usual way Perhaps (F) could be made consistent with the T-schema if we add (F2) p IFF every sentence that expresses the proposition that p is sansfied by every mfinite sequence of objects But (F2), I suggest, is no better a candidate for the status of conceptual truth of any sort than were (E2) or (C2) Its two component propositions are certainly not synonymous, and it is difficult to see how understandmg that snow is white could somehow imply that one understands anything about sequences, infinite or otherwise 17 Earher versions of this paper were read at the Pacific Division of the Amencan Philosophical Association and The Wheaton Conference on Truth and Realism I thank the audiences at those readmgs for helpful discussion and especially my commentators, Zlatan Dammanovic and Richard Fundi respecuvely Atm Plantinga, Joel Fnedman, David Anderson, Michael Rea and Adam Kovach deserve special mention An earher version was also read and discussed by a readmg group composed of members of the University of Memphis Philosophy Department, includmg Jennifer Case, Bob Barnard, Terry Horgan, David Shoemaker, John Tienson and Mark Timmons Paul Bloomfield also provtded helpful commentary Finally, special thanks to William Alston for helpful cnticism and support