Acta Analytica

, Volume 27, Issue 1, pp 13–27

Functionalism about Truth and the Metaphysics of Reduction


DOI: 10.1007/s12136-010-0105-x

Cite this article as:
Horton, M. & Poston, T. Acta Anal (2012) 27: 13. doi:10.1007/s12136-010-0105-x


Functionalism about truth is the view that truth is an explanatorily significant but multiply-realizable property. According to this view the properties that realize truth vary from domain to domain, but the property of truth is a single, higher-order, domain insensitive property. We argue that this view faces a challenge similar to the one that Jaegwon Kim laid out for the multiple realization thesis. The challenge is that the higher-order property of truth is equivalent to an explanatorily idle disjunction of its realization bases. This consequence undermines the alethic functionalists’ non-deflationary ambitions. A plausible response to Kim’s argument fails to carry over to alethic functionalism on account of significant differences between alethic functionalism and psychological functionalism. Lynch’s revised view in his book Truth as One and Many (2009) fails to answer our challenge. The upshot is that, while mental functionalism may survive Kim’s argument, it mortally wounds functionalism about truth.


Functionalism Truth Pluralism Lynch Multiple-realization 

1 Introduction

In a series of papers Michael Lynch has articulated and defended functionalism about truth (“FT” for short) (2001, 2004, 2005, 2008). Many of the motivations for FT are much the same as those for pluralism about truth, the view that there is more than one way for a proposition to be true. However, alethic pluralism faces serious problems and it is as a solution to these problems that Lynch’s view shines. A precondition, though, for Lynch’s project is a commitment to non-deflationism. Once one accepts, in Lynch’s words, “that truth, whatever it is, needs to be substantively explained, not explained away,” then one might be puzzled by the fact that the set of true propositions is a multifarious lot (2004: 385). Just how is it that truths of mathematics, ethics, physics, and so on, are all true in exactly the same way? Consider, for instance, the scope problem. The problem is that when a theory of truth successfully accounts for the truth of propositions in one domain it completely founders when extended to cover truths from another domain. The causal/referential theory works well for propositions concerning mind-independent, middle-sized dry goods, but completely fails when it comes to propositions about abstract objects, that is, objects which do not stand in causal relations to anything. To avoid these problems, Lynch proposes adopting the framework of functionalism, specifically the multiple realizability paradigm, using it to give an account both of the concept of truth and of the property of truth.

In this paper we are not principally concerned with how FT handles the scope problem or other problems; rather, we are concerned with the adoption of the functionalist framework. We will argue that FT faces a similar challenge to the one that Jaegwon Kim laid out for the multiple realization thesis (“MR” for short) in the philosophy of mind (Kim 1992). This is the thesis that creatures with a radically different biochemistry from our own may realize the same psychology that characterizes humans (Kim 1992: 1). Kim argues that MR has the consequence that alleged mental properties, such as pain, are nomically equivalent to disjunctions of their physical realizers. This consequence, together with the claim that such disjunctions are non-nomic, has the further implication that there is no single, species neutral property of pain. As we shall show, Kim’s central argument can be rejiggered to apply to FT. Consequently, FT has the unintended implication that there is no single, explanatorily significant property of truth; rather the property of truth will turn out to be identical to a disjunction of its realization bases the realization bases of which carry the entire explanatory load.1 Just as Kim’s conclusion, in virtue of ruling out nonreductive physicalism, aims to narrow the options in the philosophy of mind to eliminativism, reductivism, or dualism; so our conclusion aims to restrict the positions on truth to deflationism, reductivism, or dualism (primitivism).2

Our paper divides into five sections. In the first section we explain how Lynch attempts to adopt the MR paradigm to explicate the property of truth. In the second section we describe Kim’s challenge to the MR paradigm in the philosophy of mind and adapt this challenge to target FT as well. In the third section of the paper we show how a plausible response to Kim’s challenge on behalf of nonreductive materialism cannot be extended to rescue the functionalist about truth. To tip our hand, we argue that, while there is a unique causal profile possessed by each realization of (e.g.,) pain, there is no unique causal profile associated with each realization of truth. This is because any relevant causal profile associated with a realizer of truth is “realization specific.”3 Up to this point our argument assumes that explanations invoking MR properties are causal explanations. It is typically assumed in the multiple realization literature that multiply realizable properties are causally efficacious. In the fourth section we consider two objections to our thesis and reply to them. In particular we consider the objection that the truth role does specify a unique, but non-causal, profile that each realizer of truth possesses and in virtue of that there is a non-trivial, non-causal explanation involving truth. To anticipate we argue that this response turns the view into a form of deflationism, thereby relinquishing the substantive commitment to realism. In section 6, we interact with Lynch’s new proposal that fruitful but non-causal explanations involving truth can be had. We suggest they cannot. Although he altered his earlier view to avoid certain problems, we show that the problems our paper raises continue to plague his new view.

2 Functionalism about Truth and Multiple Realizations

As explained above FT attempts to account for the intuitions that drive pluralism about truth while avoiding the consequence that there is no single explanatorily significant property of truth. The essence of FT is that truth is a functional property, i.e., truth is the higher-order property of having some property that plays a specific role. This role is specified by compiling core principles concerning truth.4 Lynch then proceeds to use these platitudes to define ‘true’ and its cognates via the Ramsey/Lewis method (Lewis 1970, 1972). The significance of this lies in conceiving of the property of truth as a functional property paralleling attempts to conceive of psychological properties as functional properties.5

How we should understand this property depends on how we specify functional roles, i.e., functional concepts. Functionalists take broadly two approaches to functional roles: a specification of the functional concept according to the best scientific theory or a specification of the concept according to the folk theory. Lynch develops FT along the lines of the latter approach. As Lynch describes a “commonsense” functionalist individuates concepts by their place in a network of interrelated concepts and commonly believed principles the (implicit) knowledge of which constitutes our folk-theory of the topic in question. Included in the network are “platitudes,” which involve the target concepts, those concepts to be specified and clarified by the theory, and other concepts in terms of which the target concepts are to be defined. Such platitudes delineate the functional roles of the items or states in question.

Our folk theory of truth consists of platitudes and principles, a grasp of which is implicit in our everyday goings-on, including in our reasoning and behavior. It consists of two sets of principles. On the one hand, there are those principles specifying the relationship between truth and other closely related alethic properties, for example, ‘true propositions correspond to facts’, ‘the proposition that p is true if and only if p’, and ‘a proposition is true just when its negation is false’ (Lynch 2004: 392). On the other hand, there are those principles specifying the relationship between truth and groups of non-alethic properties. According to Lynch, these may “be analogous to input clauses, such as ‘If a belief is caused by a reliable process, it is probably true’…[and]…‘even if a proposition is justified it may not be true’” (2004: 392, emphasis in original). So these two sets of platitudes and principles can be used to specify the functional role of the concept.

With the two sets of platitudes in hand, Lynch spells out the truth-role using the Ramsey/Lewis method of implicitly defining theoretical terms (2004: 392; 2001: 730f).6,7 Let the T-terms be those we are attempting to define and let the O-terms be the introducing vocabulary. A conjunction is formed of all the platitudes about truth and related concepts, or—to allow for some fallibility8—a disjunction of all conjunctions of most of them. This sentence is our postulate; call it “A”. It has the following form:
$$ {\mathbf{A}}\left( {{{\hbox{T}}_1},{ } \ldots, { }{{\hbox{T}}_{\rm{n}}},{ }{{\hbox{O}}_1},{ } \ldots, { }{{\hbox{O}}_{\rm{n}}}} \right) $$
where T1 to Tn are T-terms and O1 to On are O-terms included in A.9 If A were spelled out then it might look like this: P is true if and only if P and if S knows that P then P is true and if a belief is caused by a reliable process then it is probably true…, or, …), where the T-terms are the instances of ‘true’ and its cognates and the O-terms in this case are terms such as ‘knows’, ‘belief’, and ‘reliable process’.10 Next, the occurrences of ‘true’ are replaced with a variable, prefixing an existential quantifier for that variable. Letting variables of the form “tn” stand in for ‘true’ the resulting modified “Ramsey sentence” of A is roughly the following:
$$ \left( {\hbox{FT}} \right){\hbox{ For every truth bearer}}\,x,x\,{\hbox{is true}} \Leftrightarrow {\exists_{\rm{t1}}}\left[ {{\mathbf{A}}\left( {{{\hbox{t}}_1}, \ldots { }{{\hbox{O}}_1},{ } \ldots, { }{{\hbox{O}}_{\rm{n}}}} \right){ }\& x\,{\hbox{has }}{{\hbox{t}}_1}} \right] $$

Paraphrasing, this says that x (a proposition, say) is true just when there is a certain alethic property t1 that is related to the non-alethic properties as specified in Aand x has t1 (Lynch 2001: 732f; 2004: 393). When one and only one property plays the role specified in A for a discourse, that property is the “unique realizer” of the truth role for the propositions of that discourse. By functionally specifying the role of truth, (FT) gives us the conditions under which a proposition is true, namely, the proposition has a property playing the role specified in A.

Notice, however, that this analysis has yet to divulge the nature of truth. (FT) has only given us a general specification of truth-conditions for propositions. We now face a choice, says Lynch: either we identify truth with the properties satisfying (FT) or we take truth to be the ‘role’ property (2004: 394; 2001: 734f). Lynch argues that the first option brings with it several problems. For instance, it makes truth-talk ambiguous. For it may be the case that the “near-perfect” or “unique realizer” of the truth-role for one discourse could differ from that of another discourse.11 If ‘true’ in one context picked out a causal/referential property and in another context picked out supercoherence then ‘true’ will behave like other ambiguous expressions. There will be the one expression ‘true’ expressing different properties in different contexts.12 This is just one of the alleged problems besetting pluralism about truth, according to Lynch. In order to avoid the problems with pluralism, Lynch prefers to take truth itself to be the role property. Truth is the higher-order property of having a property that plays the truth-role.13 This has the virtue of allowing a plurality of realizer properties within a commitment to monism, all the while avoiding the threat of ambiguity.

Does Lynch think that this higher-order property is explanatorily significant? He had better think so. If functionalism about truth is to compete with deflationism then the higher-order property of truth must be explanatorily significant. After all, a commitment to realism about the property of truth is undermined if all the relevant explanation can be carried out by appeal to only lower-level realizer properties. This is important because it is consistent with deflationalism that the concept of truth is a functional, but empty, explanatorily idle concept. Compare the situation with non-reductive physicalism. Psychological concepts are functional concepts. These concepts do enter into explanations of behavior and hence we have good reason to think that there are corresponding functional properties. To the extent that one should accept a form of realism about higher-order properties it seems crucial that the properties are explanatorily significant.

Lynch does explicitly state that FT is consistent with truth being an explanatorily significant property.14 He writes:

Some have argued that the truth of a belief is part of the causal explanation for the success of action based in part on that belief. Thus, one might think that the truth of my belief that there is coffee in my mug table [sic] helps to explain why I am successful in drinking coffee from my mug. Nothing about FTP prevents us from appealing to truth in this way (2004: 395).

Further, he writes: “…just because truth is a role property does not mean it cannot do explanatory work” (2005: 40). In sum, not only is it incumbent for a form of realism about functional properties that they carry an explanatory burden, Lynch recognizes that FT is at least consistent with such a view.

3 Higher-order Properties and the Multiple Realization Thesis

The MR thesis holds that mental properties can be realized by different physical realization bases. Pain, for instance, may be realized by a Martian brain, a mollusk brain, or mammalian brain. As Kim explains, the main point of MR is that “mental properties do not have nomically coextensive physical properties, when the latter are appropriately individuated” (1992: 6). The attraction of the MR thesis in the philosophy of mind is that it provides a way of understanding how psychological properties can be distinct from physical properties and yet—in a way—nothing more than physical properties. In short, it allows one to be both an explanatory realist about the mental and a good card-carrying physicalist.

As Kim explains, though, a challenge for the advocates of MR is to explain the integrity of mental properties. Since the advocates of MR hold that mental properties may have different realization bases, the challenge is to maintain an explanatory realism about mental properties in light of the nomic equivalence between the alleged mental property and the disjunction of its realizers. Kim presses this challenge by an analogy with jade. We used to think of jade as a single natural kind and accordingly the property being jade as a single explanatorily significant property. Now we know that jade is not a natural kind: every instance of jade is either an instance of jadeite or nephrite. The challenge for advocates of MR is to explain why pain doesn’t suffer the same fate as jade. This challenge is all the more pressing for advocates of MR who also accept the physical realization thesis. This thesis has two components: (1) pain (e.g.,) occurs in a system when and only when appropriate physical conditions are present in the system and (2) the significant properties of mental states, in particular nomic relationships among them, are due to, and explainable in terms of, the properties and causal-nomic connections among their physical “substrates” (Kim 1992: 14).

To press his argument further, Kim defines the property N to be the disjunction of Nh, Nr, and Nm (the Ns are the neurological realization bases for pain in humans, reptiles, and Martians). On the assumption that these bases exhaust the ways pain can be realized, the advocate of MR needs to explain why pain is not identical to this disjunction of realizers. For given (2), the property N will have all the same causal-explanatory powers as the property pain, and these powers will be due to and explainable in terms of their realizers.

We can adapt Kim’s strategy to apply to alethic functionalism. We can introduce property T as the disjunction of all the properties that play the truth role, i.e., \( {{\hbox{T}}_{\rm{r}}} \vee {{\hbox{T}}_{\rm{a}}} \vee {{\hbox{T}}_{\rm{c}}} \). T has all the same causal-cum-explanatory powers as truth in virtue of the causal-cum-explanatory powers of its realization bases. So alethic functionalism faces a similar challenge to the one Kim puts to advocates of MR in the philosophy of mind. Although Kim puts his point in terms of natural kinds and natural laws, the fundamental concern is one of explanation. Hence the specific problem for the alethic functionalist is to maintain that truth is an explanatory significant property, given that it has the same causal-cum-explanatory powers as property T. The worry here isn’t that property T fails to be a genuine property; rather the worry is that, given alethic functionalism, the property of truth is identical to property T. Because T has no unique truth-related explanatory powers, everything we want to explain about truth or in terms of truth can be explained in terms of the disjuncts of T.

To make the problem explicit, functionalists about truth face the following argument.
  • P1: The truth property, as characterized by FT, is identical to the property T.

  • P2: Property T has no unique truth-related explanatory powers.


  • C: The truth property, as characterized by FT, has no unique explanatory powers.

P2 is true because property T has just the truth-related explanatory powers that are had by its disjuncts. P1 is true because necessary coextensiveness + same causal-cum-explanatory powers suffices for property identity.15 Thus the argument for P1 need not assume the questionable premise that necessary coextensiveness itself is sufficient for property identity. Lynch himself attempts to escape the worry that “FTP [alethic functionalism] implies truth is a disjunctive property and disjunctive properties are somehow suspect” by denying that questionable premise (2005: 42n9).16 This move, however, is inadequate to dispel our argument since we explicitly indicate the problem is not with disjunctive properties themselves. Rather, to repeat, the problem is that the higher-order property of truth is identical in explanatory power to the property T. Since property T inherits all its explanatory powers from its disjuncts, T itself is explanatorily idle.

4 Unique Causal Profiles and the MR paradigm

There is a plausible reply to Kim’s argument on behalf of nonreductive materialism. This reply is fashioned by L. Antony and J. Levine, L. Clapp, and Antony (Antony and Levine: 1997: 90; Clapp 2001: 132; Antony 2003: 18f). The reply goes like this: pain is not identical to the disjunction of its realizers because each realizer will have extra causal powers on account of the heterogeneous nature of its realizers. Thus the union of the causal powers of each realization will have more causal powers than pain. Rather than identifying pain with the union of the causal powers of its realizers, pain is identical to the intersection of causal powers of its realizers. And in the case of genuine MR properties this is a substantial set. Another way to put this reply is that genuinely MR properties are identical to a distinctive causal profile that is possessed by each realization of the MR property.

A nice aspect of this reply is that it makes sense of the standard response to the disjunctive strategy, i.e., the strategy of reducing mental properties to the disjunction of realizers. As Fodor has stressed the disjunction of realization bases does not itself constitute a bona fide natural kind (1997). What each realization has in common is a distinctive causal profile. This causal profile is not uniquely possessed by any first order property. It would be wrong, therefore, to identify the MR property with the disjunction of realization bases because that would leave out the crucial thing: the distinctive causal profile.

Does a similar response work on behalf of alethic functionalism? Initially one may think that this kind of response would work for any nonreductive realism that adopts the MR paradigm and hence it would work for alethic functionalism as well. But this thought fails. As the Antony et al., reply shows, a genuine MR property must possess a distinctive causal profile that is realization independent. The problem, though, is that there is no realization independent causal profile possessed by each alleged realizer of the higher-order truth property. This claim follows from reflection on the scope problem. Let us explain.

Recall that one of the central motivations for alethic functionalism—if not the central motivation—is to resolve the scope problem. The scope problem is that a first-order account of truth, i.e., an account that identifies truth with some single, first-order property, works at best for a limited number of domains. The causal-referential account works well for medium sized dry goods but completely founders if extended to abstracta. In addressing the scope problem, Lynch proposes to identify the property of truth with the higher-order property of having some property that plays the truth role, where that role will be played by different properties in different domains. It’s crucial that the specification of the truth role is realization independent. That is, whether or not a realizer plays the truth role does not depend on the specific nature of the realizer. However, the scope problem suggests that the specification of the truth role is realization dependent. Consider that the scope problem rides on two claims: (a) that a first order account of truth works well in a limited domain and (b) that a first order account completely fails in other domains. But these claims imply that the causal-cum-explanatory powers of a realizer of truth are realization specific.17 To see this think about the realizer of truth with the domain that works well for the causal-referential story. Here this realizer has its set of causal-cum-explanatory powers, which are naturally related to the causal-referential story. But, given the complete failure of this story to cover truths in ethics and mathematics, the realizers of the truth-role in those other domains will not have the same powers as the realizer in the domain that works well for the causal-referential story. In consequence, reflection on the scope problem indicates that the realizers of the truth role are realization dependent.

5 Objections & Replies

We consider two objections to our thesis and reply to them.

5.1 Objection 1

It is misguided to saddle FT with the requirement that the higher-order property of truth have a unique causal profile; all that is required to unify the various realizers of truth is their satisfaction of the job description.18 In other words, what all realizers of truth have in common is just that they all satisfy the open sentence ‘x plays the truth role.’ This amounts to saying that what all realizers of truth have in common is that the predicates expressing them can replace ‘is true’ in the T-schema without loss of truth-value. On this response, what Lynch has done is to replace one description, namely, the T-schema, with a more elaborate description, namely, the modified Ramsey sentence containing the platitudes about truth. But this view is just a drawn-out form of deflationism.19 Instead of the standard deflationist line about truth—there’s nothing more to truth than satisfying the T-schema—this view says that there’s nothing more to truth than satisfying the truth role.20 As we have argued there’s no unique explanatory property that satisfies this role. Moreover, it’s consistent with deflationism that the property of truth is a higher-order, explanatorily idle property. Of course, because explanation is our guide to reality many deflationists prefer the more standard line that all there is to truth is satisfying the T-schema. But nonetheless there’s nothing about simple deflationism that prevents an expansive ontology of explanatorily idle properties. Consequently, to move away from the focus on robust explanation and causal profiles escapes our argument only to adopt a nominal variant of deflationism.

Why think this consequence is troublesome for functionalism? After all, might not the functionalist claim that she is already half-way to deflationism by denying the first-order accounts of truth and that this modified form of deflationism is really her considered view? In reply, though, if all there is to truth is satisfying the truth-role then we fail to see how there can be substantive explanations that involve truth. Consider the analogy with positions in philosophy of mind. Functionalism, one might say, is halfway to an eliminativist position by arguing that token identity theories are not true. Mental properties can't be identified with their lower-level realizers. So, one might think, on functionalism all there is to a mental property is it’s satisfying a certain job description. But this analogy utter fails. The functionalist holds that there are genuine functional properties that stand in causal relations to other properties. As we have argued the alethic functionalist cannot adopt the same response as the mental functionalism. Thus the alethic functionalist must maintain that truth is more than merely satisfying a certain complex job description. In fact, Lynch recognizes this and considers substantive explanations involving the notion of truth. One of the arguments Lynch rehearses is the success argument: true beliefs are more likely to lead to successful action. On Lynch’s account this means that beliefs having the property of truth help to explain successful action (2009: 121). But if all there is to truth is satisfying an elaborate description it’s unclear why that would explain successful action. Thus it's incumbent on the functionalist to offer more than a modified deflationary account. In fact, Lynch attempts to do just this in his recent book. We interact with his proposal below.

5.2 Objection 2

A second objection might claim, following Fodor, that disjunctive properties per se are not the problem (Fodor 1997). The fact that the realization base of truth is disjunctive is not the issue; the problem is that the disjunction of truth’s realizers is “metaphysically closed” instead of “metaphysically open,” where a disjunction of realizers is metaphysically open just in case there exist possible but non-actual realizers for the property in question and a disjunction of realizers is metaphysically closed just in case it is not metaphysically open. We can secure the autonomy, that is, the non-reducibility, of truth by arguing that there exist possible worlds with realizers for truth absent in the actual world. To our knowledge Lynch has never made such an argument.21 But we can head off this response by considering a distinction offered by Antony and Levine between “realization-specific” and “realization-independent” generalizations (1997). An example of a realization-specific generalization is (a) “Dormative substances cause dry mouth.” The truth of this generalization depends on the particular “realization” in question for, as they note, antihistamines do but barbiturates do not cause dry mouth. As an example of a realization-independent generalization consider (b) “Dormative substances, if ingested before driving, cause traffic accidents.” The second generalization, on the other hand, is true no matter the particular realization. Anything that induces sleep, if ingested before driving, causes traffic accidents. Related to FT, in order to show that the disjunction of truth’s realizers is metaphysically open one must provide realization-independent generalizations of the truth role. But deflationism aside, this is precisely what the scope problem shows can’t be done. 22 Hence FT is in a tough spot: either provide instances of realization-independent generalizations involving ‘true’ and its cognates or do not. The latter option leaves the functionalist about truth facing our primary objection. The former option leaves the functionalist about truth facing our objection that whatever these realization-independent generalizations are they are explanatorily weak and consistent with deflationalism.

6 Truth and Multiple Manifestation

Finally, we must consider Lynch’s most recent formulations of alethic functionalism, and, specifically his attempt to provide robust, non-causal explanations involving truth. According to Lynch, “the property of truth is the property that has the truish features essentially or plays the truth-role as such” (2008: 127, emphasis in original). The truish features are just those features specified by the “core truisms” about truth.23 We illuminate truth by specifying its essential features, which are captured by the central platitudes. More recently, adopting the framework of manifestation, Lynch writes, “Necessarily, P has the property that manifests truth for propositions of D if, and only if, it is a priori that, when had by atomic propositions of domain D, the truish features are a proper subset of M’s features,” where ‘P’ is a propositional variable, ‘D’ is a domain variable, and ‘M’ is a variable ranging over possible manifesting properties, or realizers, of truth (2009: 77, emphasis in original).24 On this new framework, a property such as correspondence will play the truth-role or manifest truth when and only when it is a priori that the truish features are a proper subset of the features of correspondence. One alleged advantage of treating realization as the manifestation relation is that on the latter construal realizers stand to the property realized much like (but not the same way as) determinates stand to determinables, and as Yablo argues, thus conceived a determinate does not compete for “causal honors” with its determinable (1995: 487; 1992).25

A key commonality between this newer formulation and the older one is that both the “realizer” property and the property of truth are present in any given case but are ontologically distinct (Lynch 2009: 75n4). Lynch stresses this, saying, “…this approach also allows the functionalist to claim that truth is, as it were, immanent in ontologically distinct properties” (2009: 74, emphasis in original). The challenge from above was that it followed from the earlier functionalist view of truth together with a few plausible assumptions that truth just was the disjunction of its realizers and, hence, anything explainable by the former is explainable by the latter. Here a similar question arises. In any given case where an appeal to truth is required for an explanation we ask whether the property carrying the explanatory load is the property manifesting truth, correspondence, say, or whether it is truth.26 Another way of putting the matter, involving a slight modification of the terminology from the discussion of Antony and Levine above, is to ask whether the generalizations offered involving ‘true’ and its cognates are manifestation-independent or manifestation-specific, where a generalization is manifestation-independent when and only when the truth of the generalization does not depend on any particular manifestation. It is essential for Lynch’s new view that truth, rather than the disjunction of its realizers, is explanatorily significant. Lynch recognizes that burden and presents several considerations to demonstrate the explanatory significance of truth, as he conceives it. We argue that his account leaves much to be desired. Hence, our primary thesis remains in play: accounts of truth must opt for deflationism, reductionism or primitivism.

Let us begin by considering Lynch’s appeal to the success argument. The success argument aims to show that truth is explanatorily significant because it plays an important role in elucidating successful action. Lynch summarizes the success argument as follows:

I would be less likely to succeed in fulfilling my desires if I lacked true beliefs about how to get what I want, and more likely to succeed if I possess true beliefs about how to get what I want. Beliefs which are true are those that have the property of truth. Hence having beliefs with that property helps to explain why I get what I want when I do. (2009: 121)

Lynch’s goal in commenting on the success argument is not to defend the argument but rather to show that the success argument favors his functionalist view of truth over other accounts. If the success argument does favor Lynch’s functionalist view that would provide a powerful consideration that truth, as he conceives of it, is explanatorily significant.

Lynch begins by rejecting the requirement that the explanations involving truth be causal (2009: 121–6). He argues that we can still get counterfactual dependencies--the hallmark of good explanations--without support from causal relations. What non-causal explanatory work might an appeal to truth accomplish? For one thing, Lynch claims, an explanation appealing to the property of truth “does more explanatory work” than similar explanations appealing to one of truth’s manifestations. The problem with appealing to one of truth’s manifestations is that it involves appeal to a “particular metaphysical theory of my belief’s structural properties,” while, presumably, in the case of an appeal to truth no such appeal is made (Lynch 2009: 125). This appeal is left out of explanations involving truth itself. Hence, his argument continues, explanations involving truth get by with a “minimum of irrelevant information” (2009: 125). Appeals to truth to explain success involve less irrelevant information than the same explanation involving appeals to one of truth’s manifesting properties.

We question this argument. It may be true that appeals to truth involve less information than the same explanation involving some particular first-order account of truth. But Lynch hasn’t shown that that information is irrelevant. It would seem to be irrelevant only if one had already accepted the functionalist view and so thought that the particular realizer of the truth role was an unnecessary metaphysical detail for the task of explanation. Furthermore, even on Lynch’s account, an appeal to truth in explanation involves a complex functional property. It’s hard to see how this isn’t a kind of metaphysical detail the folk lack. Even more, Lynch’s account requires a domain marker that is tagged to each proposition. This complicates explanations involving truth. Consequently, we fail to understand how the fact that appeals to truth involve less information privileges functionalism over some first-order account.

Lynch employs another strategy apropos the success argument to illustrate the superiority of his functionalist view. Lynch claims that explanations involving the property truth conceived in terms of his functionalist account carry more relevant information than explanations involving some traditional, first order account of truth. Lynch discusses this in the case of ‘Bridget’. He writes,

The explanation in terms of her [Bridget] belief’s causal mapping tells us why in the actual world she gets what she wants. But the explanation in terms of the truth of her belief tells us that in any relevantly similar situation, as in the actual situation, Bridget’s having a true belief about how to get what she wants will be a good explanation for why she gets what she wants.” (2009: 125)

In other words, appealing to the truth of her belief “gives us modal information that an explanation just in terms of the underlying manifesting property does not” (Lynch 2009: 125, emphasis in original).

We agree with Lynch that good explanations involving truth support counterfactuals. But we fail to see how this feature of good explanations favors Lynch’s functionalist view. The main problem with Lynch’s Bridget case is that we are never told the conditions under which a situation counts as “relevantly similar.” This is crucial for evaluating Lynch’s proposal. In the scenario Lynch provides, Bridget’s belief is true because the property of causal mapping manifests truth for the domain of her belief.27 But if we are to consider relevantly similar situations we’ll consider situations in which Bridget’s belief still has that property. Changing the domain type for Bridget’s belief (on Lynch’s account the type of realizer property for truth) changes a very relevant detail. For instance, explanations of successful behavior will say, in this case, that Bridget successfully drank water because she had a true belief of the appropriate domain and one is more likely to get what one wants when one has true beliefs of the appropriate domain. It isn’t the case that, as Lynch says, “other true beliefs would have brought about success had the world been different than it is…” unless those other beliefs are relativized to the appropriate domain (2009: 126). The problem here is that the domain quantifier fixes the range of alethic properties and closely related properties. Changing the domain is analogous to changing the underlying biochemistry for beings in pain, for example. Just as there is no guarantee that changing the biochemistry leaves the same psychology, so there is no principled rationale for thinking that changing the domain type leaves all the same closely related alethic properties. Thus, Lynch’s functionalist account doesn’t guarantee that appeals to truth give us additional, relevant modal information. The upshot of this is that Lynch’s key claim that explanations involving truth support counterfactuals fails to support his functionalist view.

The final consideration Lynch offers to demonstrate that his functionalist account of truth makes truth explanatorily significant is that his view has a virtue which deflationism lacks, namely, that it ranges over “worlds and situations that I might not even understand, and hence would not recognize were one to include them in some long list of conditionals” (2009: 126). Deflationists claim that the predicate ‘is true’ exists primarily to provide us with certain expressive resources, resources which are required to make, for example, blind-ascriptions or to express infinite conjunctions or disjunctions. A deflationist could then say that the success argument is just shorthand for the following:

(W) If Bridget believes that using the water-fountain will get her what she wants and using the water fountain will get her what she wants, then she will be more likely to get what she wants, and if she believes that talking to Patty will get her what she wants and talking to patty will get her what she wants, then she will be more likely to get what she wants, and so on. (Lynch 2009: 122)

The “and so on” indicates that (W) is a possibly infinite conjunction. A deflationist would claim that without ‘is true’ we could not express what the infinite list expresses. But whereas the deflationist would have to include in their lists situations “that I might not even understand, and hence would not recognize,” Lynch argues, the alethic functionalist, by appealing to the truth of one’s beliefs, covers these situations without having to specify them.

This argument, however, fails to consider versions of deflationism that do not restrict application of the truth-predicate to sentences as understood by a speaker. A deflationist who countenances propositions is free to use a variable ranging over propositions, including propositions potentially incapable of being expressed by any natural language sentence, to formulate a claim with the same expressive power as Lynch’s. For example, a deflationist could assert the following: For all propositions p, if I believe that p and p, then I will be more likely to get what I want. We do not need to claim that this is correct; we only need to claim that it is an alternative version of deflationism that possesses the same alleged virtue Lynch claims of his position.28

To sum up, we’ve considered Lynch’s reasons for thinking that the functionalist account of truth makes truth an explanatorily significant property. We’ve argued that Lynch’s reasons for thinking so are unconvincing. His appeal to the functionalist property of truth as involving less irrelevant information and also involving more relevant information fails to favor his account over some first-order account of truth. Also his attempt to stave off deflationism fails to consider the full resources of deflationism. Our main thesis remains: functionalism does not offer a viable account to the nature of truth. Theories of truth must choose between a form of reductionism, deflationism or primitivism.


Our argument does not address the question whether the concept of truth is a functional concept. Rather, as with Kim’s argument, our argument targets the claim that the property of truth is a multiply realizable property.


Davidson (2001), for example, defends a primitivist account of truth.


This terminology comes from Antony and Levine (1997). We discuss their view in the objections and replies below.


For the most recent discussion of the platitudes, or “truisms,” and their role in specifying a theory of truth, see Lynch (2009: 70–3).


Lynch explicitly acknowledges the parallel: “By acknowledging that truth is always and everywhere the property of playing the truth role, functionalism is not abandoning pluralism. Consider the case of a mental state like pain. According to functionalist accounts that take pain as a (higher-order) role property, there is indeed a sense in which the nature of pain is uniform across species. This is because, by and large, the pain role is uniform across species. But explaining that role does not explain what pain is in a more fundamental sense. It does not explain how that function is performed in a particular organism. For that, we must look to the details of the organism's neuronal structure: we must look for the lower-level property that realizes the pain role. In the same way, the functional role of truth does explain how that role is filled in a particular discourse. For the underlying nature of truth, we must look to the details of the type of thought in question” (2001: 735).


For a criticism of his platitude-based strategy, see C. Wright (2005). Lynch (2005) responds.


This paragraph paraphrases Lynch (2001: 732–5) and (2004:393–5).


On the concession for fallibility see Lewis (1972).


Lewis calls this the “postulate” of our theory (1972: 253). ‘A’ stands for our sentence, the theoretically significant terms of which are listed parenthetically to the right.


To follow Lewis explicitly one would need to change ‘true’ and its cognates into names.


Lynch introduces these terms in discussing the weight principles of one discourse receive relative to other principles in the alethic network. Discourses without at least one near perfect realizer of truth or with more than one are unalethic, that is, the propositions in them are neither true nor false (Lynch 2001: 739f; 2009: 77).


It might turn out to be so on the assumption that a difference in extension entails a difference in concepts.


Recently, Lynch has rejected this specification of the property of truth (Lynch 2009: 66 and 66n18). We consider the newest formulation below. As will become apparent, our reasons for rejecting the older formulation differ from his.


Lynch (2009) argues that truth figures in non-causal explanations. We consider this argument below in Section 6.


Lynch (2009: 75) accepts a view of property identity whereby two properties are distinct just in case they are individuated by “non-identical sets of essential features and relations.” This leaves it open whether the essential features are causal.


Lynch (2005: 42n9) grants that his view implies that truth is necessarily co-extensive with some disjunctive property while denying the stronger identity claim.


We address non-causal explanatory features below.


We are grateful to Adam Podlaskowski for pressing this objection.


Strictly speaking, the view might turn into a form of “minimalism,” akin to Crispin Wright’s (1992). Wright, however, does not think there is a unique property meeting the general criteria for a predicate’s counting as a truth predicate. Because Lynch eschews pluralism, we do not discuss this option.


Lynch explicitly accepts the challenge of providing robust explanations involving truth (Lynch 2009). This demonstrates that specifying platitudes or truisms is not enough. One must also show that something plays the role as specified. We turn to this below.


Lynch does claim that in any possible world where ‘truth’ refers at all that it refers to the property with the features specified by the platitudes, i.e., to the property with the “truish features” (2009: 78). It is not claimed that all realizers, or manifesting properties, of truth, are present in every possible world.


We turn to Lynch’s discussion of the success argument below, which aims to tackle this concern.


Lynch (2009: 70) provides a list.


Note the domain variable. On Lynch’s account each proposition is tagged to a domain.


Lynch recognizes there are differences between the two views, however. For one, truth, when present, manifests itself, but a determinable does not manifest itself. Second, determinants “differ from one another along some linear ordering.” And third, “determinants of a determinable mutually detest one another” (2009: 75, 75n4).


It might be suggested that reading ‘or’ in an inclusive sense avoids this choice. Yet doing so threatens to overdetermine the causal role, for then in each case there would be two properties playing the role, truth and the manifesting property. Why posit two when one will suffice? Thanks to Cory Wright for pointing us to this possibility.


The full scenario is described in Lynch (2009: 122).


For an account of “pure disquotational truth,” a version of deflationism according to which ‘is true’ properly applies to sentences as understood by a speaker, see Field (1986). For a version of deflationism countenancing propositions see Horwich (1998).



For comments on earlier drafts, we are grateful to Clare Batty, Kamper Floyd, Eric Loomis, Michael Lynch, Adam Podlaskowski, Chase Wrenn, Cory Wright, and the audience at the 2008 Alabama Philosophical Society Meeting.

Copyright information

© Springer Science+Business Media B.V. 2010

Authors and Affiliations

  1. 1.Department of PhilosophyUniversity of KentuckyLexingtonUSA
  2. 2.Department of PhilosophyUniversity of South AlabamaMobileUSA

Personalised recommendations