Expressivism and plural truth

Abstract

Contemporary expressivists typically deny that all true judgments must represent reality. Many instead adopt truth minimalism, according to which there is no substantive property of judgments in virtue of which they are true. In this article, I suggest that expressivists would be better suited to adopt truth pluralism, or the view that there is more than one substantive property of judgments in virtue of which judgments are true. My point is not that an expressivism that takes this form is true, but that it more readily accommodates the motivations that typically lead expressivists to their view in the first place.

1 An inconsistent triad

Here are three claims that, when considered independently, can each seem quite plausible:

• (1) Some value judgments are true.

• (2) No value judgment represents reality.

• (3) All true judgments represent reality.

This is a familiar, if inconsistent, triad. Value judgments like harming innocents is wrong seem true, and a venerable tradition has it that true judgments represent how things are. But to many philosophers, it also seems that there is nothing in the world—no properties of facts—that such judgments represent or even could represent.

Philosophers, with rare exceptions, dislike inconsistency, and have devised various ways out of this one. The realist rejects (2); holding that, appearances aside, some value judgments really do represent reality. Error theorists and their cousins the fictionalists reject (1). All value judgments are (literally) false.

The traditional expressivist has always taken a different route. Like the error theorist, she denies (1), but not because value judgments aim to represent how things are but tragically miss the mark. Rather, they are in a different shooting gallery all together. Value judgments do not express representational states of mind. Consequently, the expressivist, as Simon Blackburn has put it, opposes the picture that language or thought is a “seamless web…[that is] the soothing away of distinctions, whether of primary versus, secondary, fact versus value, description versus expression…” (Blackburn 1998, p. 157). This is a key background insight behind expressivism. Call it the diversity of judgment: different kinds of judgments have different functions in our cognitive life. Traditionally, the expressivist says that moral judgments differ from non-moral ones in part because the former do not represent; that is, they do not express representational states of mind.

The hoary bugbear of such accounts is often labeled the Frege–Geach problem, although, in practice, this is used as a name for a family of related objections. One gloss of what they have in common is the thought that expressivists, in their haste to privilege diversity, deny what we might call the universality of reason. Roughly put, the complaint is that value judgments, like any other judgments, are subject to reason. They figure in the premises of (valid) arguments and as the antecedents of conditionals; they are truth-functional. Consequently it is hard to see, on this gloss, how traditional expressivists can be right in thinking that, like other judgments that have these features, value judgments are not true or false.

There are, of course, a multitude of responses to this problem.¹ In this article, however, I am concerned with a different way out of the tri-lemma above. It is a way out that—at least on the surface—seems apt for moral-antirealists, and expressivists in particular. The response in question consists in denying (3), or that all true judgments represent reality.

(3) Is not to be rejected lightly. Underlying (3) is a view about truth:

Truth monism: there is only one substantive property of judgments in virtue of which they are true.²

For present purposes, we can say that a property is substantive just so long as it is not metaphysically transparent. A metaphysically transparent property is such that one can know all the essential facts about the property just by grasping the ordinary folk concept of that property. Being water is therefore a substantive property; being a conjunction is not. One cannot know all the essential facts about being water just from grasping the folk concept of water. But one can grasp all the essential facts about what it is to be a conjunction just by grasping the concept, that is, by grasping that a, b entail a & b and vice versa.³ Understood in this way, traditional theories of truth, such as the correspondence account, are varieties of truth monism. Indeed, the monist aspects of these views is so frequently assumed, and so widely embraced, that it is mostly overlooked. As noted, (3) too presupposes monism. So if we deny (3), we deny truth monism.

In this article, I will discuss two ways of denying (3). The first way is currently dominant, but I will contend it sits ill at ease with the expressivist’s wish to privilege the diversity of judgment. Hence I will defend, or at least begin to develop, the second way, according to which truth is plural. My point is not that an expressivism that takes this form is true, but that it more readily accommodates the motivations that typically lead expressivists to their view in the first place.

2 Quasi-realism

One way to deny (3) is to embrace minimalism about truth. To do so is to hold that the concept of truth is only a logical device for generalization, and denotes, at best, only a metaphysically transparent property of judgments. According to minimalism of a standard variety, our grasp of the concept of truth is exhausted by our grasp of the instances of

TS: <p> is true if and only if p.

Where this is understood to imply that, e.g., the proposition that roses are red is true equivalent to the proposition that roses are red in non-opaque contexts. Hence, minimalists deny that there is a substantive property in virtue of which propositions are true (Horwich 1998). And since, arguably, judgments are true just when their propositional contents are true, they are committed to denying what we called truth monism as well.

Many contemporary expressivists embrace minimalism about truth.⁴ By adopting this view, the expressivist can agree with the realist that value judgments can be true. To judge that it is true that murder is wrong is just to judge that murder is wrong by other means. This combination of expressivism and minimalism about propositional truth is what Simon Blackburn now calls “quasi-realism”, although, as we’ll have occasion to remark below, the name has not always been used that way.

Quasi-realism, thus, has some clear virtues. But as a number of philosophers have noted, it also in a precarious position.⁵ Here is one way of putting the point. Expressivism requires a contrast between expressive and non-expressive judgments. Quasi-realists do not have the obvious resources to draw this distinction. Clearly, being minimalists, they cannot do it in terms of truth. Value judgments are as truth-apt as any other judgment, according to the quasi-realist. So unless the quasi-realist can provide some other way to draw the distinction, the view seems unstable. Expressive judgments turn out to have the semantic features had by non-expressive judgments, and so we lose the ability to distinguish between them. We lose the diversity of judgment.

One obvious way to answer the instability charge on behalf of the quasi-realist is to say that while both value and non-value judgments can be true, only non-value judgments express beliefs. In fact, contemporary quasi-realists typically do take moral judgments to express beliefs.⁶ It is worth pausing to see why they are wise to do so.

Suppose that moral propositions are true, I can judge them, but no such judgments express beliefs. Thus, I do not express a belief when I judge that

M: Murder is wrong

Yet according to minimalism M is equivalent to

MT: <Murder is wrong> is true.

So if in judging that M I fail to express a belief, I also fail to express a belief in judging MT. So on the view in question, I can judge that it is true that murder is wrong, but I do not believe it. For if I did believe MT, I would believe M. Yet the view we are entertaining is one according to which moral judgments do not express beliefs. That is;

Where P is a value judgment, if I judge P then it is not the case that I believe P.

So on the view we are considering, there are moral truths, but I can believe none of them; nor can I believe that they are true. That seems unsatisfactory. Just one reason it is unsatisfactory is this. Presumably, in stating her view, the quasi-realist believes that there are moral truths, or at the very least, that there can be (after all, her minimalism ensures this possibility). But it is very difficult to understand why, or even how, you could believe that there are moral truths if, for any particular moral truth, you cannot believe that it is true.⁷

What this shows is that the quasi-realist must also hold, in addition to minimalism about truth, a minimalist view of belief, according to which being in a belief-state does not entail that one is in a representational state. But if quasi-realists are not going to draw the line around the expressive by way of either truth or belief, how are they going to draw it? Another familiar suggestion is in terms of representation: value judgments may be true but they are not representational. They do not describe the world, so to speak. Call this R-expressivism: Some judgments are true but do not represent natural reality and some are true and do represent.

At first blush, this appears to be just the view the quasi-realist desires: truth without representation in some domains, but not in others. But the second-blush shows the opposite. After all, the quasi-realist is a minimalist. So on her view what accounts for the fact that some true judgments represent and other true judgments do not represent?

We know what she cannot say. She cannot say that the difference consists in the fact that some true judgments express beliefs and others express non-beliefs. For we have already seen that the quasi-realist must admit that value judgments express beliefs. Nor can she say that the difference is due to variability in the way that truth depends on representation: i.e., that non-value judgments are true in virtue of representation (where the nature of representation is spelled out in some substantive fashion) and value judgments are not true in virtue of representation. If we are minimalists, there is no property in virtue of which judgments are true. Moreover, it is not clear how else—other than appealing to the dependency of truth on representation—the R-expressivist can hope to explain why some true judgments represent and other true judgments do not represent. So R-expressivism appears to be incompatible with quasi-realism.

Another possibility, suggested by Price (2010) and Price and Macarthur (2007), is to try and sew the seams of language with different, non-representationalist thread. The thought is to deny that any judgment represents in the above substantive sense and hold that while value judgments are different from non-value judgments in other ways. This may ultimately prove fruitful, and I have no general theory that says it cannot be. But I confess I find it close to the philosophical equivalent of destroying the village in order to save it. The key insight in the background of expressivism is what we called the diversity of judgment. But the diversity in question, at least when it comes to the types of judgments that have most exercised expressivists, is drawn precisely in representationalist terms. Consider, by way of illustration, a popular argument in favor of moral expressivism:

Motivation: Necessarily, if S judges that x is right, then S is motivated to some degree to act in accordance with that judgment. Taken by themselves, representational states don’t motivate or generate any motivating states. Therefore, moral judgments aren’t representational.⁸

Whether this is a good argument does not here concern me. I simply note that it is not clear how the argument could be an effective argument in favor of moral expressivism if global anti-representationalism were correct. For the argument’s entire point rests on a difference between states that represent and those that do not.⁹

So global anti-representationalism may yet prove to be a viable theory of content. But it is not clear that it serves to make quasi-realism more stable. Or if it does, it does so at the cost of robbing the view of its distinctive expressivist motivation. What the expressivist wants, or at any rate, should want, is a view which allows us to make sense of our ascriptions of truth and falsity to value judgments but which nonetheless has the resources to accommodate the expressivist’s original insight that no such judgment represents the natural world around us.

3 Constructing truth?

So the quasi-realist has the right idea. The way out of the inconsistent triad is to deny (3). But those who agree—particularly those impressed by the diversity of judgment, and by the original motivations for expressivism—should look elsewhere than current versions of quasi-realism. We should, instead, take our inspiration from an older version:

The problem is not with a subjective source of value in itself, but with people’s inability to come to terms with it, and their consequent need for a picture in which values imprint themselves on a pure passive, receptive witness…To show that these fears have no intellectual justification means developing a concept of moral truth out of the materials to hand: seeing how, given attitudes, given constraints upon them, given a notion of improvement… we can construct a notion of truth (Blackburn 1984, p. 198).

The heart of this marvelous passage is a deep suggestion: the expressivist should “construct” a concept or “notion” of moral truth out of the “materials at hand.” And that means, of course, that we should reject (3)—the idea that moral truths are true in the same way as non-moral judgments (monism). What the expressivist needs to do is say that are true, but they are true in a different way.

As I have emphasized, we cannot reject (3) lightly. Doing so entails a rejection of an entrenched view of truth: monism. We’ve already seen we can do that by accepting minimalism. But the earlier Blackburn seems to suggest a different route. Namely, endorsing:

Truth pluralism: There is more than one substantive property of judgments by virtue of which they are true.

The thought, in brief, is that moral judgments are true, but in a different way than are judgments about tabletops and tugboats.¹⁰

Even in advance of the details, certain advantages of this position are clear. Like the quasi-realist, the pluralist can say that moral judgments are truth-apt. Like the quasi-realist, she is even allowed to adopt a minimalist view of belief that allows for moral beliefs. And like the quasi-realist, she can hold that part of what distinguishes these kind of beliefs from others is an expanded role, one which connects them, for example, more directly to motivation and our sentiments. But unlike the quasi-realist, she can also adopt the traditional expressivist insight that moral judgments do not represent the world. ¹¹ Moreover, she can do this without de-stabilizing her view. Non-moral beliefs and judgments are true by virtue representing the world; moral judgments are true by virtue of something else. The question, as Blackburn, intimates, is what that “something else” happens to be, and how we can construct it.

Prior to getting to that question, some preliminaries. First, note that I put truth pluralism in metaphysical rather than, as Blackburn does, in conceptual or semantic terms. To say that, strictly speaking, there is more than one concept of truth is to say either that (a) there is only one property of truth and we represent it different ways; or (b) there is more than one meaning to the word “true” (and so there are distinct concepts expressed by the truth predicate). The former suggestion is just a variation on monism, one according to which, truth, like gold or water, has a single nature can be conceptualized in different ways. This may have its attractions, but it is not the point at hand. The latter view is distinct from monism, but it has some implausible connotations. It suggests, if not entails, the position that the truth predicate, when assigned to propositions, is ambiguous in the way that “bank” or “step” are ambiguous. This would be unfortunate. For one thing, it would make it difficult to use the truth predicate in the way that minimalists so rightly insist it is used: to blindly generalization (univocally) over an indefinite number of judgments without knowing their content—and the diversity of that content.¹²

Consequently, I favor stating the view in metaphysical terms. Doing so is consistent with the “more than one concept” version of pluralism, but it does not entail it. On this understanding, there is in fact, as the monist believes, only one property of truth. But this property is a functional property—defined by its role within a network of closely related properties described by some familiar truisms. These include the that truth is a property objective judgments have; that it is what valid inferences preserve, it is what makes a proposition correct to believe and a judgment one that we aim at in inquiry. Together, these and other truisms describe a role, or job-description. The pluralism enters by allowing that there is more than one substantive property that can play that role. Thus, there is more than one property in virtue of which a proposition can be true.¹³

We can put the question of how best to spell out pluralism aside for present purposes. But note two side effects of the present formulation. Whether there are “different ways” in which propositions can be true is difficult to assess; it is not even clear what the question means. Whether there is more than one property that plays the truth-role is easier to pin down. Given even a rough job description for truth (a rough description of the role by way of the above truisms, for example) we can ask: does the same property play that role for every kind of judgment? If not, then monism or minimalism seem more plausible; if not, pluralism is vindicated. Second, that the present view clearly distinguishes truth itself—the functional property—and the properties that play the truth-role. Hence, at least at first glance, the view would seem to hand the expressivist a way to avoid denying the universality of reason. Moral judgments, on this view, are true just as any other judgment is true. They can therefore form the antecedents of conditionals, be negated and be included in premises of valid arguments. If truth is a single functional property, then it can be, or so it seems, preserved through any valid inference, no matter whether its content is a mixture of moral and non-moral judgments.

In any event, and however it is best expressed, if some such view as the above could ultimately be made plausible, Blackburn’s earlier insight, that the expressivist needs a distinct notion of moral truth—or as we are now putting it, a distinct property that can play the truth-role—may yet be vindicated.

4 The materials at hand

So we need to “construct” a property that can play the truth-role for moral judgments, where “moral judgments” are understood as an expressivist might understand them—namely as non representational and as internally related to sentiments or attitudes. What are the “materials at hand” by which we can do so? An obvious suggestion is that such a property might be constructed out of a theory of what it is for attitudes and judgments to be warranted or justified.

Perhaps the most entrenched moral epistemologies are those owing their inspiration to Rawls. According to this family of views, moral judgments are warranted to the degree to which the moral framework to which they belong is in a state of wide reflective equilibrium. This is essentially a coherence theory of moral justification: S’s judgment that p is warranted to the degree that it coheres with the rest of S’s moral and non-moral judgments. Variations on this view have garnered wide support, amongst realist and anti-realist alike. Here, for example, is David Brink:

We all have or entertain moral beliefs of various levels of generality. Many of these moral beliefs depend on other moral beliefs. For instance, beliefs about the value of a particular activity depend, among other things, on ideals of the person (i.e., moral beliefs about what kind of persons we ought to be). Moral beliefs also depend on nonmoral beliefs. For instance, beliefs about the moral or political legitimacy of a welfare state depend on nonmoral beliefs about such things as human nature, social theory, and economics. A coherence theory of justification in ethics demands that these and other beliefs be made into a maximally coherent system of beliefs (1984, p. 102)

But while coherence epistemologies are consistent with realism, the view seems more apt for anti-realists, and particularly expressivists, for at least two reasons. First, the metaphysics of expressivism (there are no mind-independent moral facts) make it an uncomfortable fit with any epistemic view that allows that some moral judgments are justified by something independent of other moral judgments. On a coherence theory, in contrast, all justification is inferential, or at the very least, supervenient on solely judgment-to-judgment relations. Second, even where expressivists allow that moral judgments express (non-representational) beliefs, they will hold that these states internally related to attitudes and sentiments. So an expressivist moral epistemology must allow that the relevant epistemic relations can hold between such attitudes. Expressivists have typically argued that our attitudes and beliefs can be assessed in terms of their coherence. Blackburn again:

A system of attitudes and beliefs is open to criticism if it shows any of a number of different kinds of incoherence…If our beliefs are inconsistent…then something is wrong. Similarly, if our attitudes are inconsistent, in that what we recommend as policies or practices cannot all be implemented together, then something is wrong. If our beliefs are somewhat randomly held with different degrees of confidence…then something is wrong. Similarly, if our attitudes show random attachments and rejections, we are not giving some of their objects a fair hearing… (1998, p. 308).

Philosophers often fret over defining coherence. Such fretting is unwarranted unless one makes the mistake of thinking that “coherence” names a single relation. As Blackburn correctly implies, it does not. The notion of coherence picks out a cluster of epistemic desiderata. These will include, but are not limited to, mutually explanatory support, completeness, and consistency. Call these coherence-making features. Such features themselves come in degrees: members of a framework can be more or less consistent, more or less mutually explanatory, etc. A framework of judgments increases in coherence to the degree to which it exemplifies these features, on balance, to a greater degree. ‘‘On balance’’ because the features are not themselves isolated in their coherence increasing power. A framework would not be more coherent on balance, for example, simply by increasing its size (completeness) by including consistent but explanatorily unconnected judgments. Intuitively, by increasing its explanatorily isolated judgments, the coherence of the framework would on balance remain static or decrease.

With these notions in hand, we can say what it would be for a moral framework to improve in coherence. Framework F is more coherent at t2 than at t1, when on balance, it has at t2 either more of the coherence-making features or some of those features to a greater degree. So if completeness, consistency and explanatory connectedness are coherence-making features, adding a consistent and explanatorily connected judgment to the system will increase that system’s coherence along those dimensions. Consequently we can say that P coheres with moral framework F if, and only if, the result of including P in F would, on balance, make F more coherent.

So we have the materials. How do we use them to build a property that could plausibly play the truth-role for moral judgments? Here is the early Blackburn:

The simplest suggestion is that we define a ‘best possible set of attitudes’ thought of as a limiting set which would result from taking all possible opportunities for improvement of attitudes (1984, p. 198).

The idea in short is that the property that plays the truth-role for moral judgment is a type of idealized coherence. But how exactly are we to define this property? There are at least three variations possible.

First, we could say that moral judgments have the property that plays the truth-role for morality just when they are members of a maximally coherent framework, where a maximally coherent framework is one to which no further improvements can be made. But this idea is refuted by the simple observation that there are no such frameworks; hence the theory implies there are no true moral judgments.

Second, we could go subjunctive: moral judgments have the property that plays the truth-role for moral judgments just when they would be members of a maximally coherent moral framework, were one to exist (cf. Dorsey 2006). This is not hostage to the existence of coherent moral frameworks. Nonetheless, it does require us to understand what they would be like were they to exist. Otherwise we would not understand what it would be for a moral judgment to be true. And yet understanding that in which an ideally coherent moral framework consists seems a difficult task, simply because, in addition to understanding what it would be for a framework to improve, we would also have to understand what it means for a framework to be unimprovable.

A third suggestion would be to borrow a page from Crispin Wright. Wright has influentially suggested that those attracted to epistemic theories of truth should cast their views not in terms of what cannot be improved, but in terms of constant improvement. Wright calls this property (misleadingly in my view) superassertibility (Wright 1992). It might be better called superwarrant; and where warrant consists in coherence, supercoherence. We can define supercoherence as:

SUP: The moral judgment P supercoheres with F if and only if P coheres with F at some stage of inquiry and would continue to do so through every successive moral and non-moral improvement to F.

SUP does not demand that we understand an ideally coherent system. Rather, we need only understand two things: what it is for a judgment to cohere to a framework (see above) and what it would be, relative to any arbitrary stage of inquiry, for that framework to improve in coherence (see above). Thus, our third suggestion then comes to this:

Supercoherence: A moral judgment P is true in virtue of being supercoherent with F.

Thus, where the correspondence theorist takes it that moral judgments are true in virtue of their relation to facts (or propertied objects) the constructivist takes it that moral judgments are true in virtue of their relation to other judgments. But note: Supercoherence, at best, is only a theory of a property that plays the truth-role for a specific kind of judgment. It is not a theory of truth simpliciter. That gives it several advantages.

Coherence theories of truth simpliciter are typically charged with being too permissive or liberal. Any judgment that has the relevant coherence properties (is or would be a member, for example, of a maximally coherent system) will be true. Applied to moral judgments, this would appear to open the door to barbaric moral judgments being true.

Despite the obvious difficulty in achieving supercoherence, a similar charge can be made here. Our present view of what plays the truth-role for moral judgments implies that a moral judgment is true when it would be part of a durable system of moral and non-moral judgments. Crucially, there is no requirement that the non-moral judgments themselves be true. And that might well be seen as a problem. In other words, if certain false non-moral beliefs are forever held fast, and enough other adjustments are made to the system to compensate for holding them fast, even the craziest moral views might turn out to be supercoherent, and thus even the craziest moral views might be true.

Our expressivist cum supercoherentist can mitigate this worry by adding some constraints (Lynch 2009; Capps et al. 2010). Thus, they might claim that the correct theory of moral truth is not SUP but

Concordance: The moral judgment P is true in virtue of being concordant with F, where F is concordant just when (a) P supercoheres with F; (b) P supercoheres with the non-moral truths; and (c) the proposition P supercoheres with F is not true in virtue of being supercoherent with F.

The addition of clause (b) means that, according to the present theory, in order for one’s moral judgment P to be true, it must not only supercohere with one’s other non-moral judgments, but those judgments must be true—in whatever way non-moral judgments are true. This is a move made possible by the fact that the position in question is a form of pluralism about truth. Not every judgment on this view need be true by coherence. For some judgments, what plays the truth-role may be a representational or “correspondence” property. Thus, some moral judgments that might supercohere with one’s moral and non-moral judgments may not be supercoherent, simply because they are inconsistent with the empirical facts (that is, the judgments that accurately represent some non-moral reality).

The addition of (c) tightens a different sort of slack in the view. It means that it is not sufficient for a judgment supercohering with F for that framework to include the judgment it itself supercoheres with F. This is of obvious importance; without it, a framework’s supercoherence would, intuitively, be too easy a matter. What we want instead, and concordance gives us, is more objectivity: Judgments of supercoherence are not true by virtue of supercoherence.

According to our functionalist version of truth pluralism, a property realizes truth by playing the truth-role. And a property plays the truth-role when it fits the job-description given by our truisms about truth. Concordance would seem to satisfy this constraint. First, given the principle that

P if and only if P is concordant

(CM) is consistent with the T-schema. Of course, the principle has an idealist flair: applied to morality, and read in a left-to-right direction, it implies that, formally speaking, moral reality depends on the coherence properties of our moral judgments. But of course this is just what we would expect from any anti-realism about morality. Yet objectivity, and a rather radical degree of judgment independence is still obtained. For concordance is distinct not only from mere warrant, it is distinct even from superwarrant:a judgment can be superwarranted by the evidence at some stage of inquiry but not concordant. Thus, the possibility for moral error is ample: just because I and everyone else I know justifiably believe some proposition does not mean that it is concordant. Likewise concordance is a stable notion: if a proposition is concordant, then it is concordant at any stage of inquiry. And finally concordance has the relevant normative import: it could hardly fail to be correct to believe what is concordant, and it seems plausible, given the Rawlsian account of moral theorizing, that we aim at concordant judgments when we engage in moral inquiry. In short, it appears to give the expressivist exactly what he wants:

…an anti-realist story [that] can make sense of several ideas: that truth is the aim of judgment, that our disciplines make us better able to appreciate it, that it is, however, independent of us, and that we are fallible in our grasp of it…(Blackburn 1998, p. 4).

Indeed; all this, and the ability to keep the traditional expressivist insight: that value judgments do not represent the world around us.

Of course, none of this shows that concordance actually does realize the truth-role for moral judgments. That depends on whether it can be shown that the domain of moral judgment is epistemically “constrained”; (see Wright 1992, 2001); that is, that where P is a moral judgment:

P only if the judgment that P is capable of warrant.

And of course, this would have to be argued.¹⁴ As noted above, I am principally concerned to lay a particular possibility for expressivists, not argue that either expressivism under this possibility is true.

5 Return of the bugbear?

So truth pluralism appears to bring some real advantages to expressivism. So, why, one might wonder, has Blackburn in particular abandoned it for a comparatively ascetic minimalism?

One speculation (and it is only that) is that Blackburn has always heard the siren call of a more global pragmatic quietism. And deflationary viewpoints clearly whistle that tune better than pluralism. But they do so, I have argued, at the cost of losing the diversity of judgment so essential to expressivism.

In any event, and independently of Blackburn’s own motivations for abandoning his more pluralist sympathies, there is a reason to be worried about whether truth pluralism is as helpful to the expressivist as it first may appear. The worry is worth taking seriously; but it can be mollified.

The issue, in a sentence, is that pluralism unpacked in the way that we have done it here, seems to bring back the issue underlying the Frege–Geach problem: namely, the worry that expressivists must deny the universality of reason. This is because truth pluralism would seem to bring along in its wake a form of logical pluralism. Logical pluralism is the denial of

Logical monism: there is one and only one consequence relation.

One way to deny logical monism would be hold that certain logics operate over some domains of inquiry (say morality) and not over others.¹⁵ But now a familiar problem reappears: how will the expressivist assess the validity of mixed-inferences like the following?

MIX: Torture is wrong or pigs fly. Pigs don’t fly, so torture is wrong.

The conclusion is normative; one of the premises is not. If what manifests truth for moral judgments is distinct from what manifests truth for non-moral judgments, then how do we understand the validity of this inference?

There are actually two questions that might be raised here. The first is: can the truth pluralist allow that there is a single property preserved in such mixed valid inferences? The answer is arguably yes if we interpret truth pluralism in the functionalist manner that I have suggested (Lynch 2009). The functionalist holds that there is a single property of truth—it is just multiply realized. Thus, if an inference is valid—mixed or not—it preserves a single property, truth. Truth, as it were, is preserved across distinct realizations.

So truth pluralists, at least in their functionalist guise, have no problem with mixed inferences, at least if we interpret that problem to be one about property preservation. The real problem arises if we concede that a given realization of truth brings with it a non-classical logic. And arguably, concordance does. To see this, consider whether it is likely that there are any moral judgments such that neither they nor their negations would be concordant. Arguably, there are such judgments. Consider, for example,

(S): Sophie’s choice is morally right.

Sophie’s choice, in Styron’s novel of that name, is forced to choose which one of her two small children to give away to the Nazi camp guard to be killed. (S) claims that her actual choice—to give up the younger of the two children—is the morally right one. But this is questionable, and not because the other decision would be right. The more natural thought—indeed, a central thought behind the novel—is that that the issue is so morally explosive as to make neither choice right nor wrong. And one could imagine, with little difficulty, that this would be the case no matter how information improves. If so, neither (S) nor its negation would be concordant. And if concordance plays the truth-role for moral judgments, then (S) would be neither true nor false, and hence neither (S) nor not (S). So here we have a violation of both bivalence and the Law of the Excluded Middle. And that constitutes a violation of classical logic.

Whether one agrees with this specific example, the larger point is that it is possible that the moral domain compels us (by way of a priori considerations like those above) to hold that there are some moral judgments for which we lack warrant for either for them or their negations. Conjoined with the supposition that what realizes the truth-role for truth in the moral domain is itself defined in terms of warrant, we arrive at the conclusion that something weaker than classical logic is operative in the domain.

It is here that we see a problem arising that is parallel to the Frege–Geach problem. For now consider an inference that employs (S):

NIX: If it is not the case that Sophie’s choice is morally right, then grass is not green. But grass is green; so Sophie’s choice is morally right.

Is this inference valid or invalid? The argument is valid by the lights of classical logic. If so, then we have a valid logic that fails to preserve truth, since under our present assumptions (S)—the conclusion—is neither true nor false. On the other hand, the argument is invalid on a number of weaker non-classical logics, including any that (like an intuitionist logic) does not sanction double-negation elimination.

To re-cap. Expressivism, I have argued, is best interpreted as being committed to truth pluralism. But truth pluralism, together with some reasonable assumptions about the moral domain, implies logical pluralism. Yet logical pluralism throws us right back into the problem of the universality of reason. One might think we have gained no ground.

There are reasons, however, to believe that this worry is not as serious as it looks. The logical pluralist can modify her view in a way that appears to meet the universality challenge. Actually, there are several ways to do so (Lynch 2009). One way is to adopt a simple “modesty” principle. Such a principle is motivated by the thought that we should not count any inference as valid that would require a judgment be true which our theory of truth, together with a priori considerations about the content of that judgment, tells us is not true. That is, reasoning across judgments of different kinds should always be done cautiously in the following sense. We should limit the number of logical truths we endorse, so as to respect those domains that, by virtue of the property that plays the truth-role in that domain, enforce less logical truths than others. One way to do so, for example, would be to endorse:

Modesty: Where an inference contains judgments with distinct kinds of content, the default governing logic is comprised by the intersection of the logics that govern the relevant contents.

Modesty cautions logical conservatism. Assuming that the logic governing moral judgments is intuitionist, for example, NIX will be counted as invalid, since the intersection of the weaker intuitionist and a classical logic will be intuitionist. More importantly, such a principle would give the pluralist a systematic way of incorporating and allowing for the universality of reason.

To re-cap. The universality of reason is a worry only for our truth pluralist only if concordance brings in its wake a non-classical account of logical consequence. Whether it does or does not is a substantive matter for meta-ethical inquiry, and hinges, as the example of Sophie’s choice would indicate, on how one evaluates moral dilemmas. Nonetheless, what the present section has endeavored to show is that even if concordance does imply a form of logical pluralism, all is not lost for the expressivist/pluralist.

6 Conclusion

Near the end of his discussion of expressivism and truth in Spreading the Word, Blackburn notes that there is “no causal story…to justify ourselves as good signalers of virtues, duties, obligations and goods” (1984, p. 257). But that, he says, does not mean we cannot talk of moral truth. Then he asks:

Does this make moral commitments true in the same sense as others, or only in a different sense? I do not greatly commend the question (1984, p. 257).

As should be clear by now, I do commend it. Blackburn’s pessimissim here is motivated, he claims, by the fact that he does not see anyway to force the issue of whether to treat a property like supercoherence or concordance as “a kind of real truth” or as a “legitimate, but imaginary, focus upon which the progress of opinion is sited”. I am more optimistic. The real question is not “is concordance a real kind of truth” but “does concordance realize the truth-role” for moral judgment. This is a question, I have suggested, we can sink our teeth into. And if it turns out that we are able to answer it in the affirmative, we would be well on our way to showing how those sympathetic to expressivism can retain both the universality of reason and the diversity of judgment.¹⁶

Appendix

Above, I argued that the quasi-realism realism is a poor fit for what I called R-expressivism, because R-expressivists presumably hold that truth depends on representation in some domains and not in others. And this is incompatible with minimalism.

There is another argument available by which to make this point. Whether you judge it sound depends on whether you accept certain other premises. Since these premises are controversial, I think it is best left as an afterthought to the article. I present the argument here in abbreviated, informal, form.

R-expressivists deny that the following holds across the board:

REPRESENT: The belief that grass is green is true because it represents that p and p.

Here beliefs are understood as mental items. “Representation” here is understood to be some sort of metaphysically thick relation: one that is identical to, or supervenient on, a natural relation. The “because” signals a metaphysical grounding relation (you can, if you prefer philosophers’ English, read it as “in virtue of”). Someone who endorses REPRESENT across the board is endorsing that the truth of every belief depends on two factors: (a) the fact that it represents, in a non-trivial way, what it does; (b) and the way the world is. To deny REPRESENT is to hold that there are at least some beliefs whose truth does not depend on their representing some state of affairs. And that’s just the quasi-realist view we are entertaining: ethical beliefs can be true or false without representing. But of course it also contradicts minimalism, for it entails that some true beliefs are true by virtue of possessing some substantive property.

So our quasi-realist endorses REPRESENT only locally. So she takes some instances of the schema to be true. One such instance she will presumably take to be true is:

• 1. The belief that grass is green is true “because” it represents that grass is green and grass is green.

But our quasi-realist, I argued in the article, was a minimalist about belief as well as truth. Consequently, it seems plausible that she is committed to minimalism about belief-truth, that is, she will endorse the view that the following instances exhaust our concept of belief-truth:

• 2. The belief that grass is green is true iff grass is green.

If we grant this, then from 1 and 2 we can conclude via substitution:

• 3. Grass is green “because” the belief that grass is green represents that grass is green and grass is green.

This is a surprising result on its face. The latter half of 3 is a conjunction. Consequently, 3 claims that grass’s being green depends on there being (a) beliefs, and (b) beliefs that represent certain states of affairs. This is presumably not a welcome result for the quasi-realist. She claims that REPRESENT holds in some domains. But it is in those domains we want to be realist. The argument indicates however, that in the domains that REPRESENT holds, states of affairs represented by beliefs in those domains obtain in virtue of there being beliefs that represent those states of affairs. This gets the mind/world fit exactly backward from what one would expect: we get antirealism in those domains where realism was desired.

There are several ways to respond naturally. One might first deny the argument is valid because propositions of the form p “because” q create opaque contexts. If propositions of the form p because q are opaque—where “because” again indicates a metaphysical grounding relation—then one cannot expect to substitute co-referring expressions and preserve truth. So the minimalist can reject the move from 2 to 3. The question is how this move is motivated. Why should we think that propositions of that form are opaque? The following examples do not seem to be:

• Grass is green “because” the Pope says it is.

• Grass is green “because” that’s how the universe is.

In these examples, “The proposition that grass is green is true” can substituted for “grass is green” without loss of truth-value. Likewise, with “The belief that grass is green is true”. Consequently there seems no reason—other than to avoid the argument—to say that propositions of this form are opaque.

Alternatively, one might deny (1). This is tantamount to holding that REPRESENT is never true. Perhaps this is correct; but as I noted above, it seems a bit of a drastic move for an expressivist to adopt.

Finally, one might deny (2). This is to deny belief-truth minimalism. Not surprisingly, this is the route I would suggest. Minimalism is no friend of expressivism.