Representation, Deflationism, and the Question of Realism

2.1. Interlude: Explanationism and Metasemantic Expressivism

It is worthwhile to note that explanationism, in any of its versions, dovetails with recent attempts to redefine expressivism as a metasemantic view (e.g., Silk 2013; Ridge 2014; Chrisman 2016).

These new versions of expressivism rely on the distinction between first-order semantics and metasemantics. A semantic theory for a given language is a formal model that tells us how to assign semantic values to linguistic items, and how these values combine to give rise to semantic contents for entire sentences. Metasemantics examines what it is in virtue of which linguistic items have the semantic values that they do.¹⁰ Expressivism was traditionally expected to provide a psychologistic semantics that would match normative expressions with desire-like attitudes, and which would compete with a truth-conditional semantics for normative terms. Metasemantic expressivists do not attempt to provide such a semantic model. They aim to provide instead an expressivist account of why normative language fits the description provided by truth-conditional semantics, which competes with a representationalist metasemantics for normative discourse.¹¹

Expressivism thus redefined can avoid the well-known difficulties faced by expressivist semantic theories in accounting for the content of normative terms in logically complex sentences—the cluster of issues known as the Frege-Geach problem.¹² If expressivism is compatible with an orthodox truth-conditional semantics for normative terms, including truth-conditional accounts of validity and logical connectives, then Frege-Geach issues disappear, or can be addressed using tools available to cognitivists as well.

Moreover, metasemantic expressivism can fit within a particular truth-conditional semantic model that has become paradigmatic in linguistics and the philosophy of language, namely a Kratzer-style contextualism for deontic modals (Kratzer 2012), which can be expanded to deontic and evaluative terms more generally.

Chrisman (2016: 175–76), for example, defines expressivism as a view about what we mean when we attribute context-sensitive meanings to normative claims: truth conditions for a sentence articulate the thought that someone who asserts that sentence canonically ought to have, and the thought to which we are committed in making normative assertions has a directive functional role, rather than a descriptive one.

Ridge’s (2014) view is even more deeply linked with contextualism: he proposes that desire-like attitudes help determine the semantic content of deontic and evaluative terms in normative contexts of use. For example, “You ought to donate a large part of your income to charity”, when used in a normative sense, means that any acceptable standard of practical reasoning would recommend that you donate a large part of your income to charity, and to judge that a standard is acceptable is to endorse a normative perspective that does not rule out that standard. More generally, on this picture, any normative claim expresses (1) a normative perspective, understood as a non-cognitive practical stance, and (2) a belief to the effect that the given object of evaluation is ranked in a certain way by standards not ruled out by the agent’s normative perspective.¹³

However it is implemented, the metasemantic conception of expressivism seems to support the explanationist response to the problem of creeping minimalism, and in particular Simpson’s representational explanationism. If quasi-realism is taken to involve an expressivist metasemantics, then it looks like we can draw a clear divide between this view and genuine realism: while the two views might agree in their attributions of truth-conditional meanings to normative terms, they disagree about how to explain semantic content. Realists explain meaning by invoking representational relations between normative language and the world, while expressivists appeal instead to desire-like attitudes in explaining why normative claims have the meanings that they do. For instance, an expressivist like Ridge will not appeal to semantic notions like reference or representation when explaining how the standards parameter of normative claims is fixed, identifying instead the desire-like mental state that constitutes judging a standard to be acceptable. In contrast, we should expect a realist to invoke representational relations in explaining this element of normative meaning—say, by claiming that “acceptable” refers to the property of acceptability.

To be clear, I am not attributing any specific version of metasemantic expressivism to quasi-realists like Blackburn and Gibbard, though both have played down their commitment to semantic expressivism in recent works. I should also note that actual proponents of metasemantic expressivism do not embrace the broad deflationism that defines Blackburn’s and Gibbard’s quasi-realism, so they do not face the problem of creeping minimalism as typically stated. Chrisman (2016) disavows the project of making good on realist claims about objectivity in an expressivist framework, while Ridge (2014) pursues a version of quasi-realism that does not rely on deflationism about truth and other related notions as an essential component. But it is possible to be a quasi-realist who adopts both metasemantic expressivism and an expansive deflationism. (I for one am attracted to the quasi-realist project thus understood.) And the metasemantic conception of expressivism would then seem to support explanationism as an account of what separates quasi-realism from realism. This might be taken as further evidence that explanationism is the right answer to the problem of creeping minimalism.

However, I will argue that, if we adopt a deflationary approach to the relevant metaphysical and semantic notions, there need be no explanatory divide between expressivism and anything recognizable as a plausible notion of normative realism. Again, my arguments will focus on Simpson’s representational explanationism, but they can be extended to other versions of explanationism as well.

3.1. The Argument from Deflationism

Let me begin with the first horn of the dilemma. I will assume a deflationary theory of truth, facts, and other related notions. Here are the main tenets of this theory, first with respect to truth and truth-aptness:¹⁴

• For a sentence or mental state to be truth-apt, it is enough for it to meet certain minimal conditions concerning its structure and its interaction with other parts of language and thought. (For instance, according to the version of deflationism known as disciplined syntacticism, any sentence that has a declarative form and is governed by sufficient standards of warrant is thereby truth-apt.¹⁵ But the details may vary between different deflationist theories.)

• The notion of truth has an expressive role, allowing us to endorse first-order claims, and to abbreviate and generalize when talking about claims that we endorse or reject.

• The meaning of ‘true’ is exhausted by biconditionals of the form “ ‘p’ is true if and only if p”.

• The notion of truth plays no substantive explanatory role in the theory of meaning.

Importantly, this deflationary package does not entail the rejection of truth-conditional semantics. Deflationists can accept that the meanings of sentences are given by truth conditions. What they claim is that attributions of truth conditions are not grounded in some metaphysical property or relation between language and the world, called truth. Instead, deflationists take the function of truth talk to be expressive rather than explanatory, even in such semantic contexts.¹⁶

This deflationary picture can be extended to other metaphysical notions in terms of which realism is typically stated—for example, fact and property—as well as other semantic notions with which truth is connected in representationalist theories of meaning, such as reference, satisfaction, or representation itself (Dreier 2004: 29–30; Taylor 2019: 13–16).

I will focus on deflationism about representation in what follows, primarily because this notion plays a key role in Simpson’s explanationism. In particular, I will rely on the following deflationary theses about representation:

1. For a sentence (mental state) to be representational, it is enough for it to have a declarative form (be expressible through a declarative sentence) and be governed by sufficient standards of warrant.

2. The notion of representation, understood as a general semantic notion, plays no substantive explanatory role in the theory of meaning.

Note the qualification in thesis (2). What deflationism rejects is any explanation of meaning in terms of general semantic notions like truth, reference, or representation, understood as metaphysical relations between language/thought and the world, and applicable in principle to a wide range of domains of discourse.¹⁷ But it allows that there are representational relations between certain mental or linguistic items and the natural world, namely causal tracking relations that can be fully described in naturalistic terms. Indeed, deflationists can rely on such tracking relations in local explanations of meaning, for example, for our terms for mid-sized dry goods, while denying that these relations provide a reduction base for semantic notions of reference or representation that can play a broader explanatory role in the general theory of meaning.¹⁸ Moreover, according to deflationism, the ordinary notion of representation involved in such claims as “‘$2+2=4$’ represents a fact” or “‘Genocide is wrong’ represents genocide as being wrong” cannot be reduced to tracking relations, and thinking of representation as a metaphysical relation between language/thought and the world will not shed light on what the word “represents” means in such contexts either. Deflationists adopt instead a sideways approach to this ordinary concept of representation: they identify its expressive function as a device for endorsement and generalization, akin to the expressive function of “true” and “fact.”

Now, this is only a quick overview of deflationism about representation, and I will not try to provide an argument for the view here. But I will assume this deflationist view in what follows, not only because I find it plausible and Simpson relies on it in his explanationist proposal, but also because quasi-realists have come to adopt deflationism about representation,¹⁹ and this expansive deflationism has become the framework for the debate about how to distinguish quasi-realism from realism: it is when we extend deflationism from truth and facts to reference, properties, or representation that the question of how to keep quasi-realism and realism apart seems to become intractable.

In particular, this expansive deflationism undermines the contrast between representational and non-representational mental states, on which expressivism has traditionally relied. If deflationists are right, then we do not need to appeal to representation in order to explain how a thought gets to be representational. Representationality is instead grounded in structural features of mental states and how they interact with other mental states, and these features can be found in any domain of thought where we make assertions. Moreover, those who take deflationism seriously should not only claim that all mental states that meet the minimal structural conditions are representational, but also that there is no meaningful distinction between thoughts that are representational only in a “minimal” sense and those that are “robustly” representational: any mental state that meets the structural conditions is thereby fully and genuinely representational.²⁰

For all I have said so far, explanationism still seems to stand as a response to the problem of creeping minimalism. After all, its core idea is that quasi-realism and realism do not diverge in the features they ascribe to normative language and thought, such as truth-aptness or representationality, but in how they explain these features. In particular, Simpson can still claim that, even though expressivists might agree with realists that normative judgments are representational, they reject an explanatory role for representation in the theory of normative meaning and in an account of why normative judgments have this feature. Indeed, Simpson’s proposal seems to become even more plausible if we explicitly include deflationism about representation in the quasi-realist package. Take again (RE):

(RE) Representational relations explain attributions of meaning to sentences in D.

Both expressivism and deflationism reject (RE) with respect to the normative domain, and this seems to be the sense in which quasi-realism is opposed to realism.

Here is now the first argument against (RE) as a criterion for realism.

If we understand (RE) as referring to representation as a general semantic notion, then expressivism is indeed in conflict with (RE): for example, Ridge’s explanation of how the standards parameter of normative claims is fixed relies on normative perspectives, understood as desire-like mental states, rather than on representational relations between normative terms and facts about acceptable standards.²¹ Moreover, combining expressivism with deflationism reinforces this conflict, given that deflationism rejects an explanatory role for representation in the theory of meaning more generally.

However, this is not a good way to understand the divide between quasi-realism and realism, precisely because it because it relies on attributing to realism explanations of meaning that are incompatible with deflationism. Here is what I take to be a constraint on any good response to the problem of creeping minimalism: it should provide a criterion for realism that is compatible with a deflationary framework. (RE) understood in terms of general semantic notions does not meet this constraint.

To be clear, I am not claiming that all normative realists should be happy with this constraint on criteria for realism. Many self-professed realists do accept inflationary explanations of meaning in terms of notions like truth and reference, and might even take such explanations to be constitutive of their realism. But the question in which we are interested is whether we can draw an explanatory divide between quasi-realism and realism in a deflationary framework. That is, we are looking for a criterion for realism that would allow someone to be a quasi-realist about normativity and a genuine realist about other domains of discourse, as quasi-realists like Blackburn and Gibbard take themselves to be.

(RE) understood in terms of general semantic notions fails to deliver such a criterion, precisely due to the wide reach of deflationism. Deflationism applies to all domains of discourse, not only to normative discourse: it denies that representation and similar semantic notions explain meaning even for paradigmatically descriptive regions of discourse, such as our claims about mid-sized dry goods. This means that the deflationists’ rejection of (RE) does not allow us to draw an explanatory contrast between domains of discourse where realism holds true, for example, our claims about tables and chairs, and areas amenable to a quasi-realist account, like normative discourse. From a deflationary standpoint, realism defined in terms of (RE) disappears as a theoretical option in all areas of discourse, so acceptance of (RE) cannot constitute a plausible criterion for normative realism.²²

3.2. Tracking Explanations of Meaning and the Missing Realists

A key premise in my argument above was the claim that quasi-realists must reject an explanatory role for representational relations in the theory of meaning in all domains of discourse. Someone might object, however, that this ignores a type of representationalist explanation of semantic content that is not undermined by deflationism, namely a causal explanation that appeals to tracking relations between language and the world to explain the emergence and use of certain bits of language. Therefore, we can still articulate a divide between quasi-realism and genuine realism by taking (RE) to refer to such explanations. Or so the argument would go.

Tracking explanations of meaning can take different forms: for example, causal theories of reference for proper names or natural kind terms, or teleosemantic theories of the sort offered by Millikan (1984) or Dretske (1986; 1988). But we do not need to engage with these details here. What matters for our purposes is that it is exceedingly plausible that some such story is true for some parts of language, and deflationary expressivists should have no problem acknowledging this. Indeed, the endorsement of tracking explanations of meaning for our terms for mid-sized dry goods and the like is a staple of expressivist and pragmatist literature. See, for instance, Price’s (2013: 36) notion of e-representation, which applies to mental and linguistic items whose function is to “co-vary with some (typically) external environmental condition”, or Williams (2013), whose explanations of meaning in terms of use for ordinary descriptive terms like “red” appeal to reliable discriminative reactions to environmental circumstances.

Indeed, Blackburn and Gibbard often invoke such explanations of semantic content when articulating the sense in which normative discourse is not representational. Gibbard (2015: 213), for example, takes environmental tracking to be a paradigm of representation, which can be used to articulate a bifurcation thesis between genuinely descriptive language and those areas of language which only mimic genuine description, like normative discourse. Moreover, Simpson (2018: 525–26) appeals precisely to tracking explanations of meaning to argue that deflationism does not undermine his version of explanationism: expressivists can still say that normative discourse is non-representational in an interesting sense, he argues, given that they take its content not to be amenable to tracking explanations, while accepting that such explanations are available in other areas of discourse.²³

This brings us to the second horn of my dilemma for representational explanationism.

I agree that there is a narrow notion of representation, understood in terms of tracking relations, which survives the expansion of deflationism and allows us to claim that normative discourse is non-representational in a meaningful sense. Again, deflationism rejects an explanatory role for representation understood as a general semantic notion, but is compatible with local explanations of meaning that appeal to tracking relations.

However, such a naturalistic notion of representation cannot be used to answer the question of realism in metaethics. This is because standard forms of non-naturalist realism, like the views defended by Enoch (2011) or FitzPatrick (2011), reject tracking explanations of meaning as well, given that they take normative facts and properties to be causally inert. Moreover, even certain versions of naturalist realism, like Jackson’s (1998) analytic descriptivism, might not be covered by (RE) thus understood, as they do not seem to rely on tracking explanations of meaning either: on Jackson’s view, the semantic content of moral terms is fixed by our folk theory of morality (or, more precisely, by an idealized version of said folk theory), rather than by causal relations between our language and the world.²⁴ But all these realist views should be covered by any plausible criterion for normative realism. Therefore, the appeal to tracking explanations of meaning cannot vindicate (RE) as an account of the divide between quasi-realism and realism either.

At this point, Simpson might object that his goal is to articulate what separates deflationary expressivism from representationalism, rather than to define realism in a way that keeps it distinct from quasi-realism. But insofar as representationalism is meant to include all forms of realism, as it should, the objection I raise here still applies: there is little use for a notion of representationalism in metaethics that leaves out standard forms of non-naturalist realism and perhaps even some versions of naturalist realism.

To sum up the argument so far: if (RE) is understood in terms of general semantic notions, then it is incompatible with deflationism and therefore cannot serve as a good response to the problem of creeping minimalism; if (RE) is understood in terms of a naturalistic notion of tracking relations, then is it is too narrow as a criterion for realism.

3.3. A Disjunctive Explanatory Criterion for Realism?

Someone might argue that the two horns of my dilemma do not exhaust the options for explanationism. We need not seek a unique explanatory commitment that is characteristic of all forms of realism, the argument would go, whether this is a commitment to tracking explanations or to explanations of meaning in terms of general semantic notions. Perhaps the right way to understand (RE) is in a disjunctive way: a view is committed to (RE) if it endorses either tracking explanations of meaning or other kinds of metasemantic explanations that involve representational relations. This might deliver a criterion for realism that covers all forms of normative realism.

Now, remember the key constraint on any good criterion for realism in the context of this debate: it should be compatible with a deflationary framework. In particular, it should not rely on ascribing a substantive role to general semantic notions of truth, reference, or representation in realist explanations of meaning.

Here is a disjunctive reading of (RE) that explicitly runs afoul of this constraint: a view complies with (RE) if it subscribes either to tracking explanations of meaning or to explanations of meaning in terms of general semantic notions. (RE) thus understood may well capture all versions of normative realism, and moreover, it allows quasi-realists to distinguish between regions of discourse: they can accept (RE) in some areas, in virtue of endorsing tracking explanations of meaning in those domains, while rejecting (RE) for normative discourse because they reject both inflationary representationalist explanations and tracking explanations of meaning for normative terms.

However, this reading of (RE) does not provide a good response to the problem of creeping minimalism, because it invokes an explanatory commitment that is in conflict with deflationism. Again, many self-professed realists will see no problem here, but for quasi-realists and others who accept deflationism, this will not do: if we take deflationism seriously, we should treat inflationary representationalism as a mistaken view about the nature of truth and representation, rather than as the ingredient of a substantive form of normative realism that we should reject. (Compare with the question of realism in other domains: someone can be both a deflationist about truth, facts, etc. and a scientific realist in any sense that matters, at least by their own lights.)

Can we articulate a different disjunctive criterion for realism, which does not rely on explanations of meaning incompatible with deflationism? For instance, someone might suggest that the explanatory commitments of non-naturalist realism, and perhaps those of analytic naturalism as well, can be captured in terms of a non-causal notion of tracking or representation that is neutral with respect to the question of deflationism in the same way that the causal notion of tracking seems to be neutral on this issue. This might deliver the second disjunct that we need in a successful explanatory criterion for realism.²⁵

I believe this proposal does not work either. In what follows, I will briefly discuss two options for capturing non-naturalist realism in explanatory terms, but the lesson is more general: there is no way of making good on the metasemantic commitments of non-naturalism without relying on explanatory theses that are in conflict with deflationism. I am less certain what to say about analytic naturalism, but ultimately this does not matter much for my case against explanationism, given that any plausible criterion for realism should cover standard versions of non-naturalist realism.

The two ideas I want to discuss are two responses to the problem of semantic access for non-naturalist realism: how is it that our normative terms refer to certain non-natural properties? The first proposal is reference magnetism: on this picture, normative properties are reference magnets for normative terms in virtue of being metaphysically elite.²⁶ Can this alleged metasemantic mechanism be the non-naturalists’ version of tracking relations and provide a second disjunct in an explanatory criterion for realism?

Without weighing on the merits of this proposal, what matters for our current purposes is that the reference magnetism view relies on metaphysical and semantic notions in central explanatory roles, particularly the notions of property and reference, and for this reason cannot be used as part of an explanatory criterion for realism in a deflationary framework.²⁷

Enoch (2011) proposes a different answer to the problem of semantic access, partly based on the conceptual role semantics proposed by Wedgwood (2007): the hypothesis that our normative terms refer to certain normative properties—for example, “wrong” refers to wrongness—best makes sense of the inferential rules governing our use of normative terms, like the norm that someone who believes that a certain action is wrong should form the intention to avoid that type of action. Moreover, Enoch argues, we can explain the correlation between our relevant inferential practices and the normative properties by appealing to a pre-established harmony story: evolutionary factors pushed us toward certain patterns of thought and action, such as valuing reciprocity and acting accordingly, which (we can independently judge) happened to track normative facts.²⁸

I believe Enoch’s proposal does not fit in a deflationary framework either. It still relies on an inflationary notion of reference, that is, on reference understood as a substantive metaphysical relation between normative terms and non-natural properties, which can play a broader explanatory role in the general theory of meaning.²⁹ To be clear, this is an inflationary picture that quasi-realists reject, and many realists accept. But this disagreement between deflationists and inflationary representationalists cannot provide us with a criterion for realism in the context of this debate, which assumes deflationism as a framework.

You can see where this is going. Any explanatory criterion for realism that will cover standard forms of non-naturalist realism will have to invoke inflationary semantic and/or metaphysical notions, but precisely for this reason no such criterion can be a good response to the problem of creeping minimalism. In other words, precisely because they reject an inflationary representationalist metasemantics, deflationary expressivists should think of their view as compatible with realism pure and simple.³⁰

One final issue: how does Jackson’s (1998) analytical descriptivism fit into this discussion? I previously noted that this version of naturalist realism also seemed to be left out by an explanatory criterion for realism in terms of causal tracking relations, and any good criterion for realism should arguably cover this view as well. As a reminder, for Jackson, the semantic content of moral terms is fixed by an idealized version of the folk theory of morality. This metasemantic picture would presumably extend to normative terms more generally. Is there an explanatory conflict between this view and expressivism, and if so, would this help vindicate Simpson’s explanationism?

Now, Jackson explicitly relies on metaphysically robust notions of property and reference (e.g., 1998: 15–16, 117–18), so his view would seem to fall under the first horn of my dilemma: it is in conflict with expressivism, but its explanatory commitments are not compatible with deflationism, and for that reason cannot help answer the problem of creeping minimalism.

However, I am not sure that analytical descriptivism as a general method must rely on inflationary metaphysical and semantic notions. Deflationists might be able to accept some version of this method (roughly speaking, using platitudes of moral folk theory and reflective equilibrium to assign extensions to normative terms) by reinterpreting it as an exercise in normative theory that aims to answer questions such as what makes an action right or wrong—though, to be sure, this wouldn’t be exactly Jackson’s project, and even calling it a metasemantic view might be a stretch. But this view would also seem to be compatible with expressivism, so focusing on it would not help vindicate explanationism.

In any case, the status of analytical descriptivism in this debate makes little difference to my case against explanationism, which primarily relies on the idea that we cannot capture the metasemantic commitments of robust non-naturalist realism in a way that is compatible with a deflationary framework.