Solving the problem of creeping minimalism

Abstract

In this paper I discuss the so-called problem of creeping minimalism, the problem of distinguishing metaethical expressivism from its rivals once expressivists start accepting minimalist theories about truth, representation, belief, and similar concepts. I argue that Dreier's ‘explanation’ explanation is almost correct, but by critically examining it we not only get a better solution, but also draw out some interesting results about expressivism and non-representationalist theories of meaning more generally.

1. Introduction

Since James Dreier (2004) introduced it, the so-called problem of creeping minimalism has threatened metaethical expressivism. The problem is that expressivism becomes indistinguishable from realism, its arch rival, once expressivists start accepting minimalist views of various concepts like truth, reference, representation, and belief. In this paper I argue that Dreier's own solution to the problem is nearly correct, but that by critically examining it several interesting points emerge. These concern not only metaethical expressivism, but some ideas which are common in the wider debate about theories which fall under labels like expressivism, but also pragmatism and non-representationalism, and have received increasing philosophical interest in recent years.¹

After stating the problem, I explain Dreier's solution, according to which expressivists exclude ethical facts and properties from explaining ethical meaning, while realists do not. I use an objection from Chrisman (2008) to develop three important points:

• (1) The problem should be recast: we should distinguish expressivism from representationalism not realism.

• (2) We should not assume too much about the ontology involved in representation.

• (3) We should focus on explanation in order to solve the problem.

I then offer an alternative solution. While Dreier rightly focuses on explanation, instead of focusing on ethical facts and properties we need merely say that expressivists are distinctive in excluding representation from their explanation of ethical language and thought, as well as things that could plausibly reduce representation. Doing so is enough to protect expressivism from the problem of creeping minimalism.

I then use this solution to illuminate other solutions to the problem, and further questions about expressivism and non-representational theories of meaning more generally. I argue that given my three key points, we can answer Christine Tiefensee's (2016) worrying argument that even the most central philosophical resources of three prominent writers – Chrisman, Michael Williams, and Huw Price – cannot solve the problem. I show that these views survive Tiefensee's objections; they also give us an especially interesting insight into Price's notion of e-representation. These conclusions are valuable not only for understanding metaethical expressivism, but for understanding concepts from the debate about non-representationalist theories of meaning in general, a debate in which Chrisman, Williams, and Price are all engaged. Finally I answer some lurking objections to the solution, concerning the impact of minimalism.

2. The problem

The problem of creeping minimalism is that once expressivists accept minimalism about various notions including truth, reference, belief, and representation, it's hard to distinguish their view from their supposed rival, realism. The problem develops like this. The most distinctive expressivist view is that ethical language and thought differs from other kinds in an interesting way: it is in some way a different kind of thing to other kinds of thought and language. Expressivists have characterised this difference in various ways. They have said that ethical language is not truth-apt or descriptive; it does not express propositions, state facts, or refer to properties; ethical thoughts are not beliefs, or else they are not representational beliefs. This contrasts both with realism, which denies all of these claims, and with what expressivists might think about other kinds of language, such as the language we use to describe ordinary objects in our surroundings.

The problem of creeping minimalism arises because expressivists now accept so-called minimalist views which collapse any distinctions drawn in these terms.² Briefly, such views entail that even by expressivists’ own standards, ethical language and thought does have the features listed in the previous paragraph. Roughly, minimalism about truth, representation, and other similar features entails that if a given sentence or thought has content that can be stated using a ‘that’-clause, then it has the full range of those features. So since expressivists accept that a sentence like ‘stealing is bad’ means that stealing is bad, and that the corresponding thought is the thought that stealing is bad, if they also accept minimalism they will accept that ‘stealing is bad’ has the relevant range of features. The sentence expresses a proposition, namely the proposition that stealing is bad; it describes or represents stealing as bad; it is true iff stealing is bad and is therefore truth-apt; it expresses the belief that stealing is bad; it ascribes the property badness to stealing; if stealing is bad it expresses or states a truth or fact, namely the truth or fact that stealing is bad, and if this is so then we can say the property badness exists.

As Dreier notes (2004, 25), minimalism is attractive to expressivists because it allows expressivists to accommodate ordinary ethical practice, which involves talk about truth, knowledge, description and representation.³ Expressivists think that minimalism lets them straightforwardly explain these practices, once we’ve explained what it is to make an ethical assertion.

The problem of creeping minimalism is that expressivists seem committed to minimalism about all the features which they may previously have denied apply to ethical language and thought. It is therefore difficult to distinguish between expressivism-plus-minimalism and realism. We can see this in action in particular in recent work by Simon Blackburn. On minimalist grounds, Blackburn accepts that there are ethical truths and propositions, that such things represent ethical facts (Blackburn 1998, 79), that such truths and facts can be mind-independent (Blackburn 1998, 311–312), and that ethical thoughts are beliefs (Blackburn 2010, 4). To see how far Blackburn takes this idea, note how in a recent paper, he explicitly accepts the three defining theses of Richard Boyd's moral realism (Boyd 1988), exactly as Boyd states them, saying: ‘I agree to all these claims’ (Blackburn 2015, 843). So, the problem of creeping minimalism goes, what is the difference between Blackburn and Boyd, between expressivism and realism, once expressivism goes minimalist? Not everybody agrees on who is affected by this problem – which philosophical view suffers if the problem remains unsolved. Some think it is a problem for expressivism; others, including Dreier (2004, 31), think it is a problem for metaethicists, of making sense of the different views on offer. Some think that it is a problem for realism; for instance Blackburn sometimes claims that given minimalism, realism loses its content (1998, 294–298). I will remain neutral on this. It's also important to note that there are two distinct aspects to the problem of creeping minimalism. One is distinguishing expressivism from its rivals. The other is making sure we do so in a way that lets expressivism keep its purported epistemological and metaphysical advantages over its rivals (see Chrisman 2008, 347). In this paper I am concerned only with the first element; the other is much more difficult.

3. The ‘explanation’ explanation

In his (2004), Dreier suggests a solution, which he calls the ‘explanation’ explanation (2004, 39). This solution distinguishes expressivism and realism by their stance not on what features ethical language and thought has, but what features explain its content. The ‘explanation’ explanation says that the difference between expressivism and realism concerns what those views say about the proper explanation of certain target statements, which Dreier calls ‘protected normative statements’ (2004, 34), like:

(E) Edith said that abortion is wrong

(J) Judith believes that knowledge is intrinsically good

He argues that expressivism and realism disagree about what constitutes the truth of statements like these – what it is in virtue of which they are true. Realists and expressivists will therefore differ over how to fill in the blank in statements like:

(E*) Its being the case that (E) consists in nothing more than _________

(J*) Its being the case that (J) consists in nothing more than _________

This is the first stage of the ‘explanation’ explanation: expressivists and realists differ over what will (constitutively) explain protected normative statements. Since protected normative statements are just those ascribing normative content to utterances or thought, Dreier's view is that expressivists and realists differ over what constitutively explains the meaning of ethical words – what constitutively explains the fact that words like ‘good’ and ‘wrong’ mean what they do.

Dreier then says how expressivists and realists differ over explanations of ethical meaning. On his view, realists think that what fills in the blank in claims like (E*) and (J*) must involve ethical facts or properties, and expressivists will disagree:

In particular, [says the expressivist, ] to explain what it is to make a moral judgment, we need not mention any normative properties. (Dreier 2004, 39)

On the other hand, a realist thinks that to explain what it is to make a moral judgement, we need to cite ethical facts or properties to which the believer is related in some way. This is the second stage of the ‘explanation’ explanation: expressivists and realists disagree about whether normative facts and properties are needed to explain the fact that terms like ‘good’ and ‘wrong’ have normative content.

In sum: for expressivists, an ethical sentence like ‘abortion is wrong’ describes abortion as wrong, expresses a belief, states a truth, and so on, but it is not because it does any of this that it means what it does. Similarly, a belief that abortion is wrong represents abortion as wrong, and is true iff abortion is wrong, but it is not because of this that it has the content that abortion is wrong. All those features for which expressivists accept minimalism still apply to ethical language and thought, but they play no role in explaining why it means what it does. Realists, on the other hand, will disagree with this: the fact that ethical language and thought means what it does is to be constitutively explained in terms of such things.⁴

This looks promising. Expressivists clearly aim to explain ethical language by saying that it expresses distinctive mental states which can be characterised entirely in terms of the effect they have on their possessor's behavioural and emotional profile, and which can be characterised without appealing to a moral reality to which these states are reactions. Expressivists often claim this as a distinctive advantage, with Blackburn arguing that his view needs ‘no truck with the idea that we somehow respond to an autonomous realm of values: a metaphysical extra that we inexplicably care about on top of voicing and discussing our more humdrum concerns’ (Blackburn 2010, 5). So initially it seems that the lack of appeal to ethical facts and properties is distinctive of expressivism.

Dreier's account is also supported by the literature. To support his specific focus on facts and properties, Dreier draws on Fine (2001), Gibbard (2003), and Price and O’Leary-Hawthorne (1996), but there are others in the debate who support it too. Blackburn argues that it is definitional of what he calls ‘pragmatism’ (which includes expressivism) that its explanation of the relevant language does not cite the ontology associated with it – the ‘referents [of] its terms, or truth-makers [of] its sentences’ (2013, 75). Michael Williams concurs, arguing that an expressivist explanation of ethical language will be ‘ontologically conservative’ (2013, 143), which just means that the explanation will not appeal to ethical facts and properties.⁵ We will hear more from Price and Williams later.

The most important feature of Dreier's solution is that it neatly avoids the threat from minimalism. Expressivists can accept that there are ethical facts and properties on minimalist grounds while denying that such things are part of what explains ethical content. Believing in some facts or properties doesn’t require accepting that they do any particular constitutive explanatory work. This shift to explanatory questions is what stops minimalism undermining Dreier's solution.

However, Chrisman (2008, 347–348) argues that when a belief is false, realists cannot say that what constitutes that belief is a relation between the believer and a fact. If the belief is false, there is no fact for the believer to be related to. Suppose Suzy believes that some given instance of torture T is permissible. Suppose T is in fact not permissible. No one will say that her belief consists in a relation with the fact that T is permissible, since no such fact exists. So realism cannot be identified as Dreier suggests, or else we could only be realists about true beliefs. This would be unacceptable, since a realist will think that the same story should be told for what explains the content of all ethical beliefs.

Chrisman then argues that Dreier can only avoid the false belief issue by appealing to representation: the realist will say that what constitutes Suzy's false belief is that she represents T as permissible. This is possible even though her belief is false, since thinkers can represent the world falsely. Expressivists can then deny that representation is part of what constitutes Suzy's belief, and we get our distinction.

However, Chrisman argues that this relies on a ‘distinction between representational and nonrepresentational mental states’ (2008, 348) which collapses given minimalism. The idea is that minimalism, in the same way as I described in Section 2, simply entails that ethical beliefs are representational, even by expressivists’ own standards. It therefore collapses the distinction as drawn above.

4. Learning from the ‘explanation’ explanation

I think that Chrisman's objections both fail. Dreier's account does not collapse for the reason Chrisman offers, and a representation-based alternative does not collapse under minimalism. However, I think Chrisman's discussion raises some important and interesting issues concerning the problem of creeping minimalism.

I’m going to use my discussion of the above debate to argue for three claims. First, we have been conceiving of the problem of creeping minimalism in the wrong way. We should recast it, and then examine Dreier's view in light of this. Second, Dreier's solution focuses too much on the ontology of representation, and the best solution to the problem of creeping minimalism should not do so. Third, Dreier is right to focus on explanation, and doing so lets us refute Chrisman's second objection. After establishing these three points, I will argue for a solution which improves on Dreier’s.

4.1. Three lessons

The first point is that we should recast the problem of creeping minimalism. We should aim to distinguish expressivism not from realism, but from representationalism. Representationalism says that ethical thought and language is best explained in terms of representation, that ethical thoughts are best explained as beliefs, and so on. Representationalism is not sufficient for realism. While realists typically accept representationalism, so do many non-realists. For example, error theorists are not realists, in the sense that they think there are no ethical truths or facts.⁶ But they typically are representationalists: they think ethical thoughts are best explained as representational beliefs. Therefore, distinguishing expressivism from realism is not the right route: error theory differs from expressivism in exactly the same relevant respect as realism does, over whether ethical thought is belief-like and representational.⁷

This leads neatly on to the second point. We should not focus too much on the ontology invoked by representational accounts. Error theory gives us a nice route in to this point. Consider again Chrisman's first objection to Dreier: the belief that p cannot be constituted by a relation to the fact that p where it's false that p. A very natural reply to this is to point out that the realist will of course not think it is so constituted, but will instead say that Suzy's belief that T is permissible is composed of concepts, including the concept of permissibility. This concept represents things as permissible, and does so in virtue of a relation to the property of permissibility. This route saves Dreier from Chrisman's objection: it retains the explanatory role of ethical properties rather than ethical facts.

However, it is at the very least controversial whether all accounts of representation will take this form, of appealing to a relation with properties. Start with error theory. The property view I just outlined would commit error theorists to relations between thinkers and uninstantiated properties, perhaps necessarily uninstantiated properties. If representation means a relation with a property, since error theorists think such properties are not instantiated, they can at best say we bear relations to properties which exist but are uninstantiated. Not all error theorists would want to accept this.⁸ So this is a drawback for Dreier's view, though it is not necessarily decisive.

However, error theory is just one tricky case: there is a much more general point here, namely that not all accounts of representation take the property or fact view Dreier discusses. There are plenty of representationalist views that do not. Consider a propositionalist view, which takes a belief that p to be explained in terms of the believer bearing the belief relation to the proposition 〈p〉. Some take propositions to be composed of senses or concepts, which don’t require corresponding properties or entities. Such a view could treat my belief that something is F as explained by my bearing the belief relation to a particular proposition, yet it would not require my representing something as F to involve a relation between me and F-ness. Contrast this view with the expressivist, who doesn’t think we have to appeal to a belief relation between me and the proposition 〈stealing is bad〉 in order to explain my belief.

Consider also views which say that we can believe things about non-existent objects. Le Verrier's beliefs about Vulcan are still representational in a way which expressivists think his beliefs about right and wrong are not. But it would be a stretch to think that we should account for his beliefs in terms of a relation with an existing object, since Vulcan doesn’t exist. The same idea goes for uninstantiated properties. We all want to account for systematically false beliefs, for instance about magic or witches. Not everyone will want to accept uninstantiated properties being magical and being a witch to help us explain such beliefs. Finally, consider adverbialism about perception (Ducasse 1942; Chisholm 1957). Such a view says that to see something red is to see red-ly, not to bear a relation with redness. Yet this view is still clearly representationalist: this view will accept that seeing red is best explained as representing something as red.

So this is our second point: we should not commit the representationalist to a specific ontological view in order to account for representation. For any proposed ontology, there are plenty of representationalists that will reject it. And we should not rule out any particular representationalist views just to solve the problem of creeping minimalism, a problem, as Dreier says, in ‘metametaethics’ (Dreier 2004, 31)! This point has been obscured because of the focus on distinguishing expressivism from realism specifically, rather than from representationalism.

As such, it might seem that the best route is to say this: expressivists deny that ethical language is best explained in terms of its being representational, while representationalists affirm that it is. We then decline to say anything more specific about what representation is. Before exploring this idea and suggesting an even better alternative, recall Chrisman's second objection: isn’t this proposal hopeless, since expressivists won’t deny that ethical language is representational, and indeed can’t deny it if minimalism is true?

No. The whole point of Dreier's account is that what matters is not what features expressivists think ethical language and thought has, but which of those features explain the fact that ethical language mean what it does.⁹ It is not what features words like ‘good’ and ‘permissible’ have, but which of their features explain the fact that they mean good and permissible. Minimalism does not imply that the content of ethical language and thought is to be constitutively explained in terms of representation. It only implies that it is representational, in the sense that ethical language and thought can be said to represent the world. Though minimalism implies that ethical thought and language is representational, it does not imply that this fact helps to explain the protected statements Dreier identifies. Its being representational does not help to explain its meaning. So Chrisman's objection completely misses Dreier's crucial manoeuvre: the shift to explanation.

4.2. A better account

So, we should be distinguishing expressivism from representationalism rather than realism, and we should remember that the key way to do so is to read the two views in explanatory terms. Representationalists think that ethical language and thought needs to be treated as representational to be properly explained; expressivists do not. However, we should also not assume too much about the ontological commitments of a representationalist view. What, then, should we say about creeping minimalism?

The first thing to do is to make a defensive point. By adopting Dreier's explanatory focus, expressivists can defend themselves from the threat of minimalism. Minimalism was a threat because it seemed to undermine the expressivist's negative view: she can no longer say that ethical language is not representational. This threat has disappeared now: even if she has to say ethical language is representational in that it represents the world, she can deny that we should explain it in terms of representation. This shift to explanation neatly resolves the problem of creeping minimalism: minimalism no longer collapses the expressivist's view into representationalism.

However, at this point we might ask more about what representationalism comes to. So far I’ve characterised it as the view that representation is required to properly explain ethical language and thought. But this is only a first step. For starters, how do we classify accounts which don’t use words like ‘representation’ (or ‘belief’ and ‘description’) in their theory? For instance, a reductionist view, perhaps one which reduces an ethical belief to a causal tracking state, would not mention representation in its explanation of ethical thought, yet would still clearly count as representational. More generally, on the above view, expressivism only seems to oppose those accounts which use representation as a primitive – as an unexplained explainer.¹⁰

The best response to this objection is that we need simply understand the expressivist's ban on explanatory roles for representation and belief as including anything which might plausibly serve as a reduction base for those things.¹¹ So the reductionist view above rests on the idea that causal tracking is representation; her explanatory chain goes:

w means good ← w represents things as good ← w tracks goodness

(where w is just a word.) The expressivist will not accept this. Representation is itself trivially entailed by w's meaning good, which is directly explained in terms of attitudes:

w represents things as good ← w means good ← basic sentences involving w express intentions, plans, attitudes &

This captures the idea that representationalists are reducing representation, while expressivists are not putting it in any explanatory capacity at all.

So, to conclude: Dreier's account is right in its basic approach, but it faces two issues. It wrongly tries to distinguish expressivism from realism rather than from representationalism, and it attributes too much of an ontological commitment to representationalist views, perhaps because its focus is more on realism.¹²

Instead we should step back, and argue that while expressivists must accept that ethical thoughts are beliefs and that they represent the world, as do ethical assertions, they need not say this in their explanation of what gives ethical thought and language its content. Expressivists say that the content of ethical thoughts and beliefs is not to be explained in terms of representation or any plausible reduction base for it. This is completely compatible with minimalism.

5. Other accounts

In my view, this alteration of Dreier's original solution distinguishes expressivism not merely from realists but representationalists in general, and does not collapse given minimalism. However, before we look at some potential issues with the new solution, it is worth applying some points from Section 4 to other work on the problem of creeping minimalism. I will examine Christine Tiefensee's arguments against three solutions to the problem. One of these solutions is Chrisman’s. The other two use two concepts recently developed by two prominent non-representationalists: Michael Williams's notion of an ‘EMU’ (an explanation of meaning in terms of use) and Huw Price's notion of ‘e-representation’. Tiefensee argues that none of these solutions works.

Tiefensee's arguments are worth discussing for two reasons. First, her pessimistic conclusion is cause for concern: if even the sophisticated non-representationalist machinery recently set out by Chrisman, Williams and Price cannot solve the problem of creeping minimalism, we might well worry that no solution is likely to emerge soon. Second, it turns out that every key point in Tiefensee's critique can be answered by using two of the ideas I defended above: (1) by not focusing on the ontology of representation and belief, and (2) by recognising the importance of explanation.

5.1. Chrisman's inferentialism

With this in mind let's look at Chrisman's idea. On the basis of his criticism of Dreier, Chrisman rejects representation-based approaches to the problem of creeping minimalism. Instead he thinks we should replace representation with inference as the central tool for understanding expressivism (Chrisman 2008, 335). On an inferentialist view, the meaning of a sentence is constituted by its inferential role, which Chrisman takes to consist in two things: (1) what circumstances license asserting the sentence and (2) what further assertions and actions are licensed by asserting the sentence (2008, 350).

Chrisman argues that realists think ethical terms play a role in theoretical reasoning, whose premises give ‘evidential support’ to the conclusion which if true ‘will usually constitute theoretical knowledge about the world’ (2008, 350). Expressivists, on the other hand, will play a role in practical reasoning, whose premises ‘provide practical support for the conclusion’ which if true ‘can constitute practical knowledge about how to interact with the world as we know it to be’ (2008, 349–350, original emphasis). As such, realists and expressivists differ in that the former take ethical commitments to be theoretical, while the latter take them to be practical.¹³

However, Christine Tiefensee (2016) forcefully criticises Chrisman's account, arguing that it is just as vulnerable to creeping minimalism as other approaches. She focuses her initial objection on the distinction between practical and theoretical commitments. Consider the claim that ethical statements express theoretical commitments. According to Chrisman, this means that they express commitments which if true will constitute theoretical knowledge about the world. But Tiefensee argues that given the right minimalist theories, the expressivist will accept that ethical commitments are theoretical: given minimalism about ‘true’, and ‘world’, she says, there's no reason for expressivism to reject that true ethical commitments can constitute ethical knowledge about the world (Tiefensee 2016, 2443).

I think Tiefensee is absolutely right that expressivists can accept a minimalist version of the claim that ethical commitments are theoretical commitments. However, as with Dreier, Chrisman can reply by turning to explanation. He shouldn’t say that for expressivists, ethical commitments are not theoretical. He should say that for expressivists, the fact that they are theoretical has no role in explaining them. A sentence like ‘torture is wrong’ has its meaning because it is practical, not because it is theoretical. It means what it does because it gives practical support for conclusions based on it as a premise. But it is still theoretical: it still gives evidential support to the conclusion, and when true it constitutes knowledge about the world. It's just that this fact plays no explanatory role whatsoever. As with Chrisman's objection to Dreier, Tiefensee's objection can be avoided with more of a focus on explanation. This isn’t to say his solution is correct, just that it can avoid Tiefensee's objection.

5.2. The EMU

Michael Williams (2013) claims that the notion of an EMU – an Explanation of Meaning in terms of Use – is a good way of understanding expressivism about ethical language as well as distinguishing between non-representationalism and representationalism more generally. According to him, an EMU has three components, which I will characterise exactly as Tiefensee does (Tiefensee 2016, 2448)

(I–T) A material-inferential (intra-linguistic) component, comprising the inferential patterns in which a concept stands, thus determining its conceptual meaning.

(E–T) An epistemological component, specifying the epistemological circumstances of competent language use.

(F–T) A functional component, detailing what the concept is used for.

(I–T) is meant to specify the inferential role of a term. For instance, part of the (I–T) clause for a colour-word like ‘red’ will specify inferential connections between that word and other colour terms: from ‘x is red’ you can infer ‘x is coloured’, from ‘x is scarlet’ you can infer ‘x is red’, and so on. (E–T) is meant to specify the circumstances under which a speaker is licensed to make assertions using the term; for instance part of (E–T) for ‘red’ will specify that a speaker can use ‘red’ only to apply to objects which are clearly red. (F–T) is meant to tell us why we have a term of which (I–T) and (E–T) are true (Williams 2013, 135). Williams says: ‘The F-clause appeals to use as expressive function: what a word is used to do, what it is useful for’ (Williams 2013, 135).

Tiefensee then suggests that we could try to distinguish realism and expressivism by saying that a realist EMU will differ from an expressivist one. She sets out two EMUs which follow the above pattern, and points out where they match and where they differ. For our purposes, the crucial point is that Tiefensee thinks the difference between expressivism and realism will come down to what these views say about tracking. Tiefensee considers the following claim: (Tiefensee 2016, 2449, my emphasis):

• (1) In a reporting use, tokens of ‘x is good’ express reliable discriminative reactions to an environmental circumstance. Their role is to keep track of goodness, in this way functioning as language entry transitions.

What this roughly means is that ethical statements express beliefs which track ethical properties, perhaps in the same way that we think our language about our environment tracks the objects in it.

Tiefensee argues that expressivists can accept a suitably minimalist reading of (1). On this reading the idea that ethical statements ‘track the moral truth’ is just the idea that some ethical statements are true, and some people are good enough that they assert more or less only the moral truths, and that their moral statements are counterfactually responsive to the moral truth so that, for example, if x had been good, they would have said that x was good, and if it had not been, they would not have said so. On this minimalist reading of tracking, an ethical truth-tracker is the person who gets the ethical truths right, and who wouldn’t easily have got them wrong. Since we’re assuming that expressivists can make sense of ‘getting the ethical truths right’, and can evaluate counterfactuals involving ethical statements, they will not deny that ethical statements track the truth in this sense.

So, Tiefensee concludes, focusing on the notion of tracking as Williams does will not distinguish expressivism and realism, because expressivists will end up asserting the same things as the realist. And though Tiefensee does not explicitly say so in this passage, this is because of a minimalist reading of ‘truth-tracking’, which Tiefensee calls ‘i-tracking’ due to its link with Price's ‘i-representation’, a concept we’ll discuss shortly.

5.3. E-representation

Tiefensee argues that as a result of this, the only way to get a distinction between expressivism and realism is to understand the notion of tracking in (1) as ‘e-tracking’, which we can make sense of in terms of Huw Price's distinction between what he calls ‘e-representation’ and ‘i-representation’. Roughly, e-representation is a causal relation between terms and their environment, while i-representation is a feature emerging from relations between terms, including inferential and causal relations (see Price 2013, 36 for a fuller explanation). As such, a term e-represents if it tracks things in world. A term i-represents if it has the inferential role required to give it assertoric content. We can then read the notion of ‘e-tracking’ in terms of e-representation, taking the two to be more or less the same idea.

Crucially, a term can i-represent without e-representing: a term can be used in assertions, without having the job of tracking the world. Expressivists think that ethical terms are i-representational but not e-representational. We might distinguish them on this basis from realists, who think that ethical terms are both i- and e-representational. This is just the same as saying that expressivists deny, but realists accept, that ethical terms e-track the world

Tiefensee argues that we need to explain e-representation in more detail, but that the only substantial way of doing so faces a dilemma: it yields either a characterisation of realism which expressivists can accept (thus failing to yield a distinction) or else one which realists won’t themselves accept.¹⁴ Tiefensee argues that on the best way of cashing out the notion of e-representation (and thereby, e-tracking) ethical language is e-representational just in case it has two features: (i) ‘reports of the presence of goodness must be caused by goodness’ and (ii) such reports ‘must be default justified’, i.e. ‘do not require inferential support from other premises’ (Tiefensee 2016, 2454). However, many realists will wish to deny (ii). Moreover, there are two readings of (i), one of which realists will reject, the other of which expressivists will accept. Either way, we don’t get a distinction between expressivism and realism.

Nothing in my argument will depend on whether ethical sentences are ‘default justified’ so instead I shall focus on claim (i), that reports of the presence of goodness must be caused by goodness. Tiefensee begins by pointing out that non-naturalist realists, who believe in the causal inefficacy of ethical facts and properties, will reject (i). I think this is sufficient reason to reject this reading, since it fails to account for one of the two major forms of realism. However she goes on to make a further point. She argues that there are two readings of (i), the strong reading and the weak reading. On the strong reading, (i) says that ‘moral properties are themselves causally efficacious and that we perceive them through a special, sensory moral faculty’ (Tiefensee 2016, 2454). While expressivists will indeed deny that moral properties are like this, and therefore will deny (i) on its strong reading, Tiefensee points out that many realists will also reject (i) due to scepticism about the notion of a special sensory moral faculty. The strong reading, then, fails to be acceptable to realists, and so cannot be a necessary component of realism.

On the other hand, when read weakly, (i) seems more moderate: it says that goodness can cause reports about goodness by being identical to a natural property or properties which do all the causal work. So ‘whenever we detect a causally efficacious natural property that constitutes a moral property & we also perceive the corresponding moral property’ (Tiefensee 2016, 2455). This seems more acceptable to realists. However, Tiefensee argues, it will also be acceptable to expressivists, for the following reason.

The weak reading of (i), says Tiefensee, depends on the idea that moral properties are, ‘in some sense or other, nothing over and above natural properties such as causing pain’ (Tiefensee 2016, 2455). Expressivists, she says, can make that claim too, though for them it will be understood as the expression of an ethical claim, about what wrongness consists in (e.g. causing pain). But if this is so, she argues, then since wrongness just is some natural property, then if those natural properties cause our ethical utterances, then it follows that ethical properties cause our ethical utterances. And so expressivists will be happy to accept (i) on the weak reading.

Tiefensee concludes that e-representation, as cashed out via (i) and (ii), cannot give us a reasonable distinction between realism and expressivism: either it will not be a distinction at all or else will yield a characterisation of realism which realists themselves will reject. It's worth noting at this point that Tiefensee also thinks of e-representation as being a mark of ontological commitment: those bits of language which are e-representational are the ones which carry some genuine ontological commitment, and as such are the ones for which realism is appropriate. She therefore adopts a pessimistic stance about the notion of ontological commitment here: no such notion has been cashed out which distinguishes expressivism from realism and which realists will accept.

To sum up, Tiefensee argues that neither Williams's EMUs nor Price's distinction between i-representation and e-representation can solve the problem of creeping minimalism. EMUs cannot help by themselves because the best EMU-based strategy uses a notion of tracking that is acceptable to expressivists when understood minimalistically. When cashed out in terms of e-representation, either expressivists will still accept that ethical language is e-representational, or else realists will deny that it is.

5.4. Replying to Tiefensee

As I mentioned earlier, Tiefensee's arguments are important because they target resources – EMUs and e-representation – given a central role in recent work by prominent non-representationalists. These arguments should worry us quite independently of the problem of creeping minimalism for ethics: they threaten some of the core components of contemporary non-representationalism. While I think Tiefensee's argument fails as a whole, discussing it does reveal some important lessons.

First, we must recognise that it's because Williams's EMUs are taken to be explanations, that Tiefensee's criticism of EMU-based strategies fails. Tiefensee claims that expressivists might be able to accept the realist's claim about tracking (1), which is part of the realist's EMU for ‘good’. But if Williams's EMUs really are explanations, then the fact that expressivists can accept that ‘good’ can express reliable discriminative reactions is not enough to undermine this particular EMU strategy. More is required: expressivists need to think that (1) is part of what explains the fact that ‘good’ means good. They might accept (1) as true, but say that it doesn’t belong in the EMU for ‘good’, because it doesn’t have any explanatory role. In other words, even if you are good at tracking the ethical truth, you don’t have the concept good in virtue of this fact.

Tiefensee approaches this point herself in the final footnote in her paper, claiming that ‘a stronger focus on function might do the trick’, though she is sceptical of this (Tiefensee 2016, n. 26, 2458). However, if we read ‘function’ just in the same explanatory terms as Dreier introduces and I have endorsed, then the ‘stronger focus on function’ Tiefensee mentions can be understood as putting more emphasis on what explains ethical meaning. And this is precisely the route I want to follow. On this view, the idea Tiefensee discusses becomes the claim that expressivists say that ethical terms ‘establish language entry transitions’ (roughly that they track ethical facts), but that this fact does not explain their meaning. So the EMU-based solution doesn’t collapse given minimalism after all

While Tiefensee's argument against EMUs fails, I think Tiefensee is more or less right about e-representation. Price's notion of e-representation can be used as a mark of neither realism nor representationalism. There are two reasons for this. First, whether or not a given kind of language is e-representational depends on whether it has an appropriate causal connection with the kinds of objects, properties and facts that it is meant to represent. But this cannot be a mark of realism or representationalism in general because there will be views which accept representationalism (and perhaps realism too) yet deny that the relevant facts and properties are causally active, as Tiefensee rightly points out. For instance an ethical non-naturalist might think that ethical language can only be explained in terms of representing the world, but that the properties and facts it represents are not causally linked to it.¹⁵

This is one reason we can’t rest the general distinction between expressivism (or nonrepresentational pragmatism more generally) and representationalism on e-representation. The second reason is that e-representation does not have explanation built in. In other words, a piece of language may well be e-representational, but this does not entail that its meaning or use is explained by its being e-representational. (This is why expressivists could accept that ethical language was e-representational in Tiefensee's weaker sense.) Since I have pinned the distinction between representationalism and expressivism on the issue of what explains the content of ethical language, e-representation is not by itself going to play a crucial role – we need to include explanatory considerations too.¹⁶ So we shouldn’t think that e-representation is a mark of the difference between representationalism and expressivism in ethics. More generally, therefore, we shouldn’t think that it's a mark of representationalism versus non-representationalism either.

So while Tiefensee's pessimistic argument fails as a whole, it reveals some important general points about the debate about expressivism and representationalism. Moreover, we’ve seen that the main moves in this debate can all be better understood and answered when we appreciate that explanation is the key to distinguishing expressivism. Dreier and Chrisman's points are best understood in these terms, and understanding Williams's EMUs in this way helps us see how we might use them to answer Tiefensee's objections.

6. Minimalism's revenge

To finish I want to discuss two spanners minimalism may yet throw in the works. First, the solution I’ve defended seems to make minimalism incompatible with representationalism, a consequence we should avoid if possible. Second, minimalism seems to prevent expressivists from distinguishing their opponents’ views from their own. I will answer these in turn.¹⁷

6.1. Incompatibility?

The first worry is that minimalism about representation is incompatible with representationalism. This is because minimalism rules out explanations in terms of representation – and such explanations are definitive of representationalism. This is a common point in the literature. In a recent paper Camil Golub rejects some proposed solutions to the problem of creeping minimalism for this reason (Golub 2017, 1392f; 1406). Many writers think this point is important for the wider debate about non-representational theories of meaning.¹⁸ While a full answer requires more space than I have, I will sketch out what I take to be the best response.

It's not clear why we should care about this supposed incompatibility. We might think that expressivists can live with it: if rival views are incompatible with minimalism, so much the worse for them. However this would make expressivists, who accept minimalism, unable to be representationalists about areas of language outside ethics, for instance about the language of middle-sized dry goods. At least some expressivists want to be representationalists in these cases, so they should try to answer this problem.¹⁹

The short answer is this: minimalism about representation rules out some but not all of the kinds of explanation which count as representationalist. Recall that in Section 4 I pointed out that representationalism includes at least two kinds of explanation: (i) those using representation as a primitive, and (ii) those using a plausible reduction base for representation. Minimalism only rules out the first of these, not the second.

While there is not enough space to fully explain why, minimalism about representation blocks the use of representation as an explanatory primitive. In particular minimalism entails that any claim of the form

x means y because x represents the world in a certain way

is false. This is because representation is a trivial unexplanatory consequence of meaning, not the other way round. So minimalism rules out explanations of kind (i). Similarly, minimalism about belief rules out explanations using belief as a primitive.²⁰

Minimalism also entails that representation is not reducible. Horwich takes minimalism to involve the view that the property in question is not constituted by any other property. For him this means that there is no property F such that (a) all and only representational terms have F and (b) (a) explains the facts about representation (Horwich 1998b, 104–107). Since reductionism would give us such a property, it is false if minimalism is true. As such, any reductionist view of representation is incompatible with minimalism: minimalism rules out explanations of kind (ii) above.²¹

However, this is no threat to the solution I’m defending. Expressivists aren’t primitivists or reductionists about representation. They do not think that we can explain any kinds of term using representation as a primitive – they usually opt for a causal relation like causal tracking to explain middle-sized dry good terms (see Blackburn 2013; Gibbard 2015). Nor do they want to reduce representation: they don’t think representation can be reduced to any one property, partly because they think that ethical terms and middle-sized dry goods terms both represent the world but don’t do so for the same reasons. So they will not care about this incompatibility. Primitivists and reductionists won’t care either: so much the worse for minimalism, they will say.

Moreover, minimalism about representation is compatible with some kinds of representationalist explanations. While minimalists cannot think that representation itself can do explanatory work, they can say that a plausible reduction base for representation can do such work, so long as they do not say that representation reduces to that base. For example, an explanation in terms of a causal tracking relation C which denies that representation reduces to C would be representationalist, since C is a plausible reduction base for representation. Yet since this view does not actually say that representation reduces to C, it is compatible with minimalism.²²

More generally, explanations appealing to things which are plausible reduction bases for representation will be unaffected by minimalism about representation so long as they don’t take representation as reducible to such things. So expressivists can accept minimalism about representation, yet argue that some terms, for instance, bear causal relations with the world in virtue of which they mean what they do, so long as they don’t think that representation is or is reducible to those relations. Minimalism rules out some representationalist explanations, but not all of them.

What if minimalism creeps all the way to these reduction bases, for instance causal tracking? I see no reason to think it will. Consider C, the tracking relation. This doesn’t bear the close conceptual relations to truth and meaning that belief, representation, and description do, in virtue of which minimalism crept to them. Nor is it part of ordinary practice to say that ethical terms causally track the world. So there's no pressure on expressivists to accept minimalism here. And so the threat of incompatibility is answered.

6.2. Unintelligibility?

The second point is that if expressivists only have a minimal notion of representation to hand, they will not be able to distinguish their own view from representationalism. This is because they will not be able to characterise representationalism, which requires a non-minimal notion. This concern rests on the idea that minimalism yields a minimal notion of representation; this idea is common in the literature (see e.g. Williams 2013, 131; Macarthur and Price 2007, 140; Dreier 2004).

We can understand minimalists as defining a minimal notion of representation, by offering definitions like

‘F’ represents _m x iff x is F.

or perhaps, if we’re willing to quantify over properties

‘F’ represents _m things as having the property X iff X is F-ness.

Minimalists think that representation, in its ordinary sense, just is representation_M: ordinary usage of ‘represents’ is best explained in terms of these schemas.

However this doesn’t mean that minimalists cannot understand non-minimal notions of representation which are defined differently, for instance a causal tracking notion defined like so:

‘F’ represents _c things as having the property X iff uses of ‘F’ are caused by property X

Minimalists can understand representation _c. They can understand what causal tracking is. They may even think it's sufficient for representation in general: all terms which represent _c do represent. They just think representation _c isn’t representation in its ordinary sense, and that it doesn’t explain our ordinary use of words like ‘represents’.

Nor do minimalists need to hear their rivals’ theories as using a minimal notion. A minimalist can understand a causal tracking theorist when she says: ‘ethical terms represent _c the world’. She's saying that ethical terms causally track the world. The minimalist doesn’t need to interpret this claim by replacing ‘represent _c’ with ‘represent _m’.²³ The minimalist just thinks the tracking claim is false. So there is no threat of unintelligibility, so long as we properly understand what it means for minimalists to ‘only have’ a minimal notion of representation.

So neither incompatibility nor unintelligibility threatens the solution I’m defending. Minimalism about representation neither blocks representationalism tout court, nor makes it unintelligible from the expressivists’ point of view.

7. Conclusion

To conclude: Dreier's ‘explanation’ explanation distinguishes expressivism and realism by their explanations of the meaning of ethical language and thought. Prompted by Chrisman's ultimately unsuccessful criticism of this account, I argued for key three points. First, we should distinguish expressivism not from realism, but from representationalism. Second, we should not assume too much about the ontology involved in accounts which explain in terms of representation; Dreier's account made such assumptions because of his focus on realism not representationalism. Third, we should focus on explanation, as Dreier does – Chrisman's second criticism of Dreier missed this fact.

As such, I argued that expressivism is distinct from representationalism because it does not use representation, or something that plausibly could reduce representation, in its explanation of ethical thought and language. This protects the expressivist from minimalism, since she can happily accept minimalism without saying that ethical language has to be explained in terms of representation. This gives us a solution to the problem of creeping minimalism.

After doing this I showed how the key points I defended help us resolve several issues raised in the recent literature: Chrisman's criticism of Dreier, and Tiefensee's criticisms of Chrisman's inferentialism, as well as Williams's ‘EMUs’ and Price's ‘e-representation’. I argued that Chrisman himself should simply focus more on explanation, that we should recognise that Williams's EMUs are explanations, and that Price's e-representation cannot be a mark of either realism or representationalism. Finally, I pointed out how on this solution, minimalism remains compatible with representationalism, and representationalism can still be intelligible to expressivists, despite worries to the contrary.

As such, a critical examination of Dreier's solution to the problem of creeping minimalism gives us not only a better solution to the problem, but some interesting results concerning philosophical concepts from the wider debate surrounding non-representationalist theories of meaning.

Funding

This work was supported by a PhD studentship from the Arts and Humanities Research Council.