1 Introduction

Here’s a familiar philosophical stance. Not all language represents or describes the world. Non-declarative sentences are one example: ‘hello Dom’ doesn’t represent anything. Not all declarative sentences represent either, since we can always transform a non-representational, non-declarative sentence into a declarative one – ‘Dom is hello’ rather than ‘hello Dom’ Dreier (1996). More interestingly, though, some declarative sentences that look, behave, and feel representational may turn out not to be representational after all. The most famous example of this stance concerns moral language and thought. Perhaps a moral sentence, like the English sentence ‘kindness is good’, doesn’t represent or describe some putative aspect of the world. Instead it has a different job: to give voice to a mental state, perhaps a state of approval or disapproval, which doesn’t itself represent the world.

This is actually a little misleading. These days you won’t tend to hear philosophers who defend this stance say that moral thoughts and sentences aren’t representational, full stop. Instead you’ll hear them claim that it is all right to say that moral sentences and thoughts do represent the world, in a certain loose sense. In fact, in the same sense, it’s also okay to say that these things describe the world, that they can be true or false, that we can know and believe them, and so on. It’s this loose sense we’re using when we talk about things like representation and truth in ordinary contexts. However, we should distinguish what it’s all right to say in ordinary contexts from what we should say when we want to give the most illuminating philosophical account of something. And when we want to do that for moral thoughts and sentences, it doesn’t help us to say that they represent or describe things. It’s like saying a recently produced emerald is grue – true, but less informative than saying it is green. So it’s true to say moral thoughts are beliefs, but it’s more informative to say they are attitudes like approval or disapproval. Saying they are attitudes explains so much more about them. It’s a nice question what exactly the loose sense is in which we can say moral sentences represent, and in what way and for what reasons it’s not very illuminating to say that, but we can ignore these gory details for now.

The key idea, then, is that the most illuminating account of moral thoughts doesn’t bother saying that they are representational, but says something else about them. The most illuminating account of moral sentences says that they express those thoughts, and doesn’t say anything about them representing. If this is right, then our crucial task is to give a full account of the mental states moral sentences express, and to say why they get expressed as they do – by declarative sentences which behave much like any others. We also need to say why it’s acceptable, at least in ordinary contexts, to apply concepts like representation, description, truth, and belief, to moral thought and language.

This view is moral expressivism. Philosophers have defended similar views about other kinds of language and thought, like probability, causation, modality, aesthetics, knowledge, and logic. The core idea these views share – that not all language and thought is to be explained by concepts like representation – is sometimes just called expressivism. This is misleading, since the general idea that we can explain some kinds of language and thought without representation is distinct from the specific view that they involve expression of attitudes. Some of these views don’t talk about expression at all. A better name might be non- (or anti-) representationalism, whose accuracy comes at the cost of syllables. Since it’s hard to find good names for theories in this area, and the arguments I’m going to discuss talk of ‘expressivism’, let’s stick with that.Footnote 1

Moral expressivism is in one sense local: it’s just about one kind of language and thought. We might say it’s a local application of expressivism. But it often goes with another view, local expressivism, which is the view that concepts like representation explain some, but not all, kinds of language and thought. Local expressivism is not an application of expressivism but a claim about the potential applications of expressivism: some of them, but only some, succeed. Two well-known local expressivists are also two of the best-known defenders of moral expressivism – Simon Blackburn and Allan Gibbard (see e.g. Blackburn 2013; Gibbard 2015).

Huw Price has famously argued that local expressivism is unstable, and collapses into global expressivism, the view that we can best explain all kinds of language and thought without appealing to concepts like representation.Footnote 2 He has two arguments for this radical claim. One hitches local expressivism to minimalism or deflationism about concepts like representation, which then drags expressivism over all kinds of language and thought. The other exploits the (hoped) success of local applications of expressivism, especially moral expressivism. It says that if any particular application of expressivism succeeds, this shows us that we didn’t need representation to explain any other kinds of language and thought either. Together, these two arguments push local expressivism into global expressivism. Crucially, they only rely on things the local expressivist is already committed to – minimalism in the first case, and the success of at least one expressivist project in the second. So they look hard for the local expressivist to resist. Price’s two arguments are a cornerstone of his view, which has become extremely influential and has prompted a significant literature. So it is certainly worth examining Price’s arguments in detail.Footnote 3

In this paper I show that Price’s arguments fail. The first makes a false claim about the extent of minimalism, which local expressivists have no reason to think spreads as widely as Price needs it to. The second relies on several false claims about how a moral expressivist’s theory generalises; in particular it conflates what a moral expressivist says about moral language and what they say about moral thought, and makes an illegitimate inference about how a local expressivist might explain cases she thinks do not admit of an expressivist theory. While both of Price’s arguments can be fixed by reinterpreting some of their key terms, that makes them miss their target, for after that alteration their conclusions no longer apply to the local expressivist.

This second point is broadly consistent with a previous argument I made (Simpson, 2020b) – that the debate between local and global expressivists is merely verbal, based on different understandings of what an expressivist theory is. However, in this paper I want to go further than this. I will argue that it is a mistake to think that global and local expressivism are on equal footing, in light of my conclusions about Price’s arguments. Those arguments do not reveal what they are intended to – that local expressivism is incoherent or unstable. The local expressivist needn’t be worried by them. At best these arguments change the subject, though they make some interesting points along the way. Moreover, Price’s second argument does not naturally suggest the interpretation of global expressivism I previously gave. In any case, examining Price’s arguments in detail reveals several interesting and important points that help us better understand what local expressivism and moral expressivism are committed to. So while my conclusion is broadly consistent with the verdict that the debate is verbal, things look better for the local expressivist than that might suggest. And in any case it is certainly worth spending time scrutinising Price’s arguments in detail, not least because they are meant to be the primary arguments in support of moving from local to global expressivism.

2 Local and global expressivism

It would be nice to start by saying exactly what local and global expressivism are. Unfortunately there are two major barriers to this. The first is that the literature is not entirely clear. The second is that part of this debate concerns how local and global expressivism should be characterised, not how they have in fact been characterised. For this reason it’s best to work with a vague definition and fill it in when needed. So: a local expressivist thinks that there are some terms and thoughts we can explain without using representational concepts. A global expressivist thinks that all terms and thoughts are like this. This definition uses two terms of art: ‘explain’ and ‘representational concepts’. Let’s start with ‘explain’. I take it that expressivists are primarily interested in saying why the thoughts and words from the discourse in question have the meaning or use patterns they do. Why does ‘good’ mean good in English? Why does it get used as it does? What makes it the case that Bertie is thinking stealing is wrong rather than stealing is right, or stealing is illegal? The question of expressivism is then whether we can say why ‘good’ means good, and what it is to have a thought involving good, without using representational concepts, and whether we can do this for every term and thought or only some. This conception of expressivism is now popular, and it is substantive enough to work with.Footnote 4

What are representational concepts? So far they’re just a placeholder for whatever the expressivist wants to avoid in her explanations. But it’s hard to say what place they actually hold, as we can see in the recent literature. Many use a list which usually includes phenomena like truth-conditions, truth, description, belief, reference, fact, predication, and of course representation. Others talk of a family of concepts tied together by definitions. For instance, if we accept that by definition, any sentence with truth-conditions is representational, descriptive, and expresses a belief, then we’re likely to treat all of these concepts as representational, and to include any other concept tied to them by definitions. Some take representational concepts to be those that link a term or thought to the things it is putatively about – their subject matter. Price has suggested that such concepts are those that would be capable of explaining wide ranges of terms and thoughts, for instance reference which would apply quite generally across the category of singular terms. This idea is very important for assessing Price’s arguments, as we’ll see later.Footnote 5

Crucially, on this characterisation neither local nor global expressivism involves straightforwardly denying that representational concepts apply to the language and thought in question. Rather, the claim is that we can explain that language and thought – in the way outlined above – without using representational concepts, whether applying or denying them. Many philosophers are sceptical that this is the right characterisation of expressivism, or that expressivism can really say this coherently. But this is the conception I will use. Price’s arguments, then, aim to show that the view that we can explain some kinds of language and thought without representational concepts really collapses into the view that we can explain all language and thought without those concepts.

3 The external argument

Price gives two arguments that local expressivism collapses into global expressivism. He calls his first argument ‘external’, since it exploits local expressivism’s link with a separate view – minimalism. According to Price, minimalism about some group of concepts says that they ‘play no significant theoretical role in a mature theory of language and thought’ (2015a p.138). It follows that if minimalism about representational concepts is true, those concepts play no role in explaining language and thought in general, and so global expressivism is true. But, Price argues, anyone supporting local applications of expressivism should accept minimalism, and indeed many expressivists do accept it. So local expressivists are committed to global expressivism. We can set out Price’s argument like so:

  1. (E1)

    Minimalism about all representational concepts entails that all terms and thoughts can be explained without representational concepts.

  2. (E2)

    Local expressivists are committed to minimalism about all representational concepts.

  3. (E3)

    So local expressivists are committed to global expressivism: all terms and thoughts can be explained without representational concepts.

(E1) is meant to be true by definition, while (E2) is harder-won. The inference to (E3) is justified by the definition of local and global expressivism as the view that some or all terms and thoughts can be explained without representational concepts.Footnote 6

Let’s accept Price’s definition of minimalism as the denial of explanatory role, and ask whether local expressivists are committed to minimalism about all representational concepts. I think they are not. For local expressivists count as representational the concept of causal tracking, but they are not committed to accepting minimalism about tracking. So there is a straightforward counterexample to (E2). The idea is that certain uses of a term or tokens of a concept are caused by the presence of something in the speaker’s environment. Just as a thermometer causally tracks the ambient temperature, because its reading of n degrees is caused by the temperature’s being n degrees, so uses of a word like ‘tree’, or tokens of the concept tree, are meant to be caused by the presence of trees. The latter case isn’t really so simple, but the basic idea is the same.

Without knowing anything more about minimalism than that it excludes the concept in question from explanations of thought and language, it’s already clear that nobody is committed to minimalism about tracking. Tracking has its explanatory role built in. If a term tracks something, this entails that (certain) uses of that term are caused by the presence of that thing in the local environment. So to say that ‘tree’ tracks trees just is to suggest a causal explanation of those uses of that word. Since it’s plausible that tracking is an essential part of the explanation of some terms – like natural kind terms – minimalism about tracking must be false. And anyway, nothing in local expressivism suggests that it supports minimalism about tracking.

But is tracking really a ‘representational concept’, one that’s banned from an expressivist theory? Price’s external argument will only fail if it is, since his claim is that expressivists are committed to minimalism about representational concepts – it’s irrelevant whether they deny minimalism about anything else. The trouble is that it’s clear that local expressivists think that tracking is a representational concept. Localists like Blackburn and Gibbard have argued against global expressivism precisely because some terms are tracking terms.Footnote 7 Moreover, if we accept the idea mentioned above, that representational concepts include those that involve a relation between words and thoughts on the one hand and their subject matter on the other, tracking will count as representational. So premise (E2) is false: the external argument fails.

There is a way to fix this argument: reject this view of representational concepts. I have suggested that Price takes representational concepts to be those that are suitably general, capable of both applying to and helping explain whole kinds of terms (Simpson, 2020b, pp.152–155).Footnote 8 Truth is general in this sense, because it is meant to apply in some way to all sentences of the right kind, at least in that they generally have truth-conditions. Reference is general because it’s meant to apply to singular terms generally, since they generally have reference conditions. Clearly causal tracking is not like this, because there’s no reason to think that all predicates (or all singular terms) track things. Price himself doesn’t exclude tracking from his explanations: it falls under what he calls ‘e-representation’ which sometimes has an explanatory role (2013 pp.36ff.) Another global expressivist, Michael Williams, happily appeals to tracking to explain colours terms (2013 p.140). So global expressivists clearly don’t think tracking must be avoided.

This protects premise (E2) by removing tracking from its scope. But it destabilises the argument. For now what the conclusion (E3) says local expressivists must accept is not what local expressivists like Gibbard and Blackburn actually reject. It is the view that we can explain all terms and thoughts without general concepts like truth and reference. Importantly, this does not entail that we can explain all terms and thoughts without relations like tracking. It just says that any concept general enough does not play an explanatory role. Crucially this leaves it open that tracking, which isn’t general in this way, might explain some terms – and that’s just what the local expressivists think, and why they think global expressivism is false! So Price can preserve the premise (E2), but this makes the conclusion irrelevant to local expressivists.

Finally, note that even without tracking as a specific counterexample, there’s simply no reason to think that minimalism does apply to all representational concepts. For Price only makes the case that it applies to these general things like truth and reference. It leaves it completely open that narrower concepts that the local expressivist might think are representational are completely out of the reach of minimalism. Tracking is one example that shows this, but even without that there’s no reason to accept (E1) and (E2) together on the basis of Price’s claims. There’s a more general moral here: while minimalism famously ‘creeps’ over many concepts, there’s no reason to think it will creep over all relevant ones.

4 The internal argument

Let’s now consider the ‘internal’ argument. This draws on the notion of a successful expressivist project, which explains its target language without appealing to representational concepts. Among other things, a successful expressivism must explain why those representational concepts still do apply, or at least appear to apply, to the relevant language and thought. After all it’s fine to say, for instance, that by thinking of someone as evil I’m representing them as evil, and when I voice that view I’m describing them as evil. So a good expressivist theory needs to explain why it’s fine to talk in terms of representation and description in this context, though those concepts will play no role in the underlying explanation of what’s going on. Such a theory also needs to say why moral language comes in declarative, assertive form, why we call moral sentences true and false, and so on Price (2015a, p.140). Let’s say the successful expressivist will explain the representational appearance of the target area without representational concepts,Footnote 9

Price argues that if a local application like moral expressivism succeeds, this is good reason to think that similar theories can be given everywhere, even for cases localists think are only explicable using representation. It will turn out that everything can be explained without representational concepts after all, so global expressivism is true (Price, 2015a, pp.140-1). The local success of expressivism is proof of concept for an aggressive global expansion. Moreover, Price argues, if a localist says that representational concepts are needed to explain certain kinds of language, this just shows she hasn’t successfully explained the representational appearance of her original target. For in such a case the representational features we needed must be ‘associated with some characteristic of use that the [expressivist explanation] cannot emulate’ (2015ap.141).

So we can set Price’s internal argument out like so:

  1. (I1)

    If moral expressivism is true, we can explain moral language and thought without using representational concepts.

  2. (I2)

    If we can explain moral language and thought without using representational concepts, we can explain all language and thought without using representational concepts.

  3. (I3)

    So if moral expressivism is true, so is global expressivism: we can explain all language and thought without using representational concepts.

Note the focus on moral expressivism here: this is because Price appeals to its details, and because it is the most developed form of expressivism defended by Blackburn, Price’s main target.

Premise (I1) is true by definition; (I2) is doing the heavy lifting. But why think it’s true? Why think that just because we can explain moral language and thought without representation, we can do the same everywhere? Price points to the details of the moral expressivist’s theory. The expressivist’s explanation of the representational appearance of ethical language ‘will offer some function, or ‘point,’ for the practice of expressing moral judgements (say) in declarative form, and ascribing truth and falsity to the resulting claims’ 2015a p.140). The expressivist will say why humans would have language which expresses the attitudes that constitute moral thoughts: it helps us disagree, debate, and co-ordinate action. We don’t need to say that we use moral sentences because there are moral truths they describe. Instead all the work is done by saying what attitudes moral sentences express and why those attitudes would get expressed by representational-seeming sentences.

Price claims that this explanation carries over to apparently bona fide representational cases. The expressivist said that the reason we have representational-seeming moral sentences is that they express attitudes it’s useful to be able to share. But if that explanation works for ethical attitudes, which supposedly contrast with beliefs, then it should work ‘in what seems a much easier case: that of the expression of the behavioural dispositions we call beliefs’ (2015a p. 140). Why would beliefs be expressed by declarative, representational-seeming sentences? For the same reason attitudes would be – so we can disagree, debate, and align our beliefs. As Price says, (2015a p. 140) ‘isn’t it plausible that it is even more useful for social creatures to have the linguistic machinery to align those sorts of psychological states?’ But if that’s so, then our non-representational explanation of why moral attitudes are expressed as they are will also apply (with some changes in detail) to beliefs. And that is proof of concept for an expressivist explanation of the entire range of sentences that we might have thought can only be explained using representation.

4.1 Explaining language and explaining thought

This argument for premise (I2) is too quick. To see this we should distinguish explaining moral sentences from explaining moral thoughts. Price is drawing attention to the expressivist’s favoured explanation of moral sentences. This explanation is psychologistic: it explains the features of moral sentences in terms of their function of expressing attitudes. Among these features is their representational appearance. The expressivist thinks we can say why moral sentences are declarative, truth-apt, and so on, just by appealing to facts about the nature of expression, the point of expressing attitudes, the benefits of aligning attitudes, and so on. We needn’t say anything about what the sentences describe or represent. Price thinks this explanation will generalise to other sentences: if attitude expression is beneficial, belief expression will be too.

We could deny that the expressivist’s psychologistic explanation will generalise to other sentences. I’m not going to do that.Footnote 10 Instead I want to raise a different problem: even if the psychologistic explanation does generalise, this doesn’t imply global expressivism. It shows us that explaining all these sentences doesn’t require representational concepts, but crucially leaves it open that we need those concepts to explain the mental states they express. Consider the belief that grass is green. At most, Price establishes that given an account of that belief, we can explain the sentence ‘grass is green’ just in terms of expression, without any representational concepts. But if we need representational concepts to explain that belief, then our theory isn’t expressivist. And Price’s argument hasn’t established that we don’t need representational concepts to explain thoughts in general. The argument assumes that moral thoughts can be explained without representational concepts, but it hasn’t established that this is true in general. Further examination reveals several reasons to doubt that we can do without representational concepts altogether.

Price talks about belief here, and one might argue that he therefore undermines his own point: he’s using a representational concept to characterise certain thoughts. But that would be too quick – there’s no reason to take Price as using this term in a serious way. Perhaps beliefs themselves can be accounted for without using representational concepts. Price’s description of beliefs as ‘behavioural dispositions’ suggests such a theory. A view of beliefs as dispositions to act doesn’t obviously invoke concepts like representation; in fact, historically it opposes those theories of belief with representation at their heart. Moreover, note that the local expressivist will not want to criticise Price on this issue, for two reasons. First, expressivists may be committed to minimalism about belief, which (given Price’s understanding of minimalism) means it cannot help explain anything at all. So they cannot agree with the apparent self-contradictory point that some mental states must be explained as beliefs. Second, we might think that regardless of minimalism, expressivists shouldn’t use belief as an explanatory primitive. They should say more about the nature of belief, and doing so opens up room for a global expressivist view again.

The relevant question is therefore whether a general non-representational theory of belief is possible. It may well be. But two important points arise. First, this possibility is not guaranteed by the success of moral expressivism. The expressivist story is not proof of concept that we can explain all kinds of belief-like thought without representation, even if we can explain its linguistic expression without representation. The possibility of a general non-representational account of belief is independent of whether moral thoughts are attitudes. So it’s just wrong to say that moral expressivism’s explanation of moral thought and language generalises. At best, Price has only shown that its explanation of moral language generalises.

Second, there is a clear way in which explanations of at least some beliefs will appeal to what local expressivists typically take to be representational concepts. Earlier we saw that local expressivists like Blackburn and Gibbard count tracking as representational. But if, as seems likely, there are some words that can only be explained by tracking relations between them and their subject matter, that’s because (given the expressivist’s psychologistic model) the concepts those words express track their subject matter. And so beliefs involving those concepts will be explained in terms of tracking. If tracking is a representational concept, then the explanation of those beliefs isn’t expressivist, and global expressivism is false. So the success of moral expressivism doesn’t guarantee global expressivism at all: it allows that we need representational concepts at least to explain some mental states. Note that this point goes beyond tracking too: the important point is that there is room for explanations of thought to require representational concepts even if explanations of language do not. Nothing in Price’s argument rules this out.

4.2 Are local expressivists missing something?

Price has a more general, abstract argument that can be used to reply to this point. He argues that if the moral expressivist does need representational concepts in order to explain some language or thought, this shows that their original explanation missed something:

But then the claim that the easy cases are genuinely descriptive—that is, have some more substantial kind of semantic property—seems problematic in one of two ways. Either it is an idle cog, not needed to explain the relevant aspects of the use of the statements in question; or, if it is associated with some characteristic of use that the merely quasi kind of truth cannot emulate, then it shows that [expressivism] is a sleight of hand—it fails to deliver the goods, just where it really matters. If [expressivism] is really successful by its own standards, in other words, then it inevitably escapes from the box, and becomes a view with global application. Price (2015a)[pp.140-1]

The ‘idle cog’ horn of this dilemma is that we don’t need representational concepts to explain any thought or language – i.e. that global expressivism is true. We’re disregarding this option at this point in the dialectic, because we’re assuming my earlier claim that we need representational concepts to explain some thoughts. But then given the other horn of the dilemma, Price argues that the representational concepts we appeal to must be associated with some features of genuinely representational language and thought, and the expressivist can’t explain why moral language and thought has those features – it cannot ‘emulate’ them. This means that moral expressivism has failed to explain the representational appearance of moral language. Since we’re all assuming it doesn’t fail, it follows that there can’t be any relevant features of other kinds of language that expressivists cannot explain. So global expressivism must still be true. Note that this doesn’t just focus on language – it applies to thought too, so my point from the previous subsection is irrelevant.

This argument also fails. In a very abstract way, its key point is this. If global expressivism is false, there is some kind of language and thought (perhaps scientific language and thought) that has some feature F we can only explain using representational concepts. But then we also cannot explain why moral language has F without using representation, so moral expressivism has failed. This abstract formulation lets us see two mistakes. First, just because we cannot explain why (say) scientific thought and language has F without appealing to representation doesn’t mean that moral expressivism even has F in the first place. For instance perhaps the ‘direction of fit’ of scientific thoughts must be explained in terms of representation, but that doesn’t mean that moral thoughts have that same direction of fit. So even if we couldn’t explain the former feature without representation, that would be irrelevant. Second, even if we cannot explain why scientific thought and language has F without appealing to representation doesn’t mean that we must use representation to explain why moral thought and language has F. It might have F for different reasons. For instance the belief-like phenomenology of scientific thought might be explained in terms of its subject matter, while the belief-like phenomenology of moral thought might be explained by the nature of the relevant attitudes. Nothing in Price’s argument blocks this possibility, because nothing in the argument says that if two different things have F, this must be for the same reason. In fact, it’s a core expressivist idea that two kinds of language and thought with different underlying functions and behaviour may share representational appearance. So there’s no abstract route from moral to global expressivism here. Price’s internal argument fails.

As before, Price could emphasise the conception of representational concepts as those that would explain a suitably wide class of terms and thoughts. The internal argument then goes through easily. For if moral expressivism is right, then at least one family of predicates – moral ones – can be explained without appealing to concepts like reference, representation, predicate satisfaction, and so on. This immediately entails that no single representational concept explains all predicates: moral predicates are an immediate counterexample. Since on this reading global expressivism just is the view that no concept explains all terms of a given kind, we’ve established global expressivism. (I2) is now plausible: if moral expressivists can avoid using representational concepts in their explanation of moral thought and language, it immediately follows that no such concepts explain all predicates (and their mental correlates). However, now the argument interprets expressivism in a way that local expressivists reject. The local expressivist wants to avoid Price’s general relations, but she also wants to avoid explaining terms and thoughts using relations like tracking, which put those things in a relation with their subject matter. These are not general in Price’s sense: for instance, there’s no reason to think that tracking is so general that it even applies to all predicates (or singular terms), let alone explains their meaning. So if Price takes this alternative route, his argument works but misses its target, for (I3) is now true but irrelevant to the local expressivist.

5 Consequences

So, Price’s two arguments that local expressivism collapses into global expressivism both fail. The first ‘external’ argument fails because local expressivists aren’t committed to minimalism about all representational concepts – concepts like tracking count as representational but are beyond minimalism’s reach. The second ‘internal’ argument fails because the possibility of an expressivist account of moral language and thought doesn’t show that we can explain any other kind of language or thought without representational concepts (like tracking). The false premises of each argument can be made true if the notion of a representational concept is understood in a different way than local expressivists understand it, specifically if it is understood as including only suitably general concepts. However, in this case the arguments’ conclusions, while now plausible, no longer apply to local expressivists. The proposition the phrase ‘global expressivism’ now expresses in (E1) and (E3) is consistent with the proposition local expressivists take ‘local expressivism’ to express. So Price’s arguments either fail, or miss their target.

The fact that Price’s arguments fail on the first reading but can be fixed is consistent with my previous argument that the debate between local and global expressivists is merely verbal. To some extent, this paper has reached that conclusion from a different direction, by looking directly at what are intended to be the two crucial arguments that push local expressivists into global expressivism, rather than aiming (as I previously did) to find an interpretation of global expressivism on which it might be true. It is tempting, therefore, to see the above as just confirming the verbal debate verdict, and so to conclude that local and global expressivism are on equal footing. After all, this is a natural conclusion when a debate turns out to be verbal.

However, I think this is the wrong diagnosis of the debate, for the following reasons. First, Price’s arguments are intended to reveal some kind of instability or incoherence in local expressivism. In a recent paper, Price says of the verbal debate idea: ‘In one sense this conclusion is congenial to me, for I don’t want there to be a coherent alternative to [global expressivism] in this neighbourhood.’ Price (2019, n.31 p.148).Footnote 11 Price’s arguments are meant to show that local expressivism is incoherent, and that his global expressivism is the result of making it coherent. However, what I’ve shown in this paper is that the local expressivist can completely reject this claim. For Price argued that the local expressivists wanted to have their cake and eat it: to accept minimalism but distinguish representational and non-representational language in a way minimalism forbids, and to deliver the representational appearance of moral language and thought while tacitly admitting to be unable to explain certain elements of that appearance. But neither of these claims is true. The distinction the local expressivist wants is still there, and is not threatened by minimalism at all. And just because some kinds of language might need to be explained using representation (in some sense), that doesn’t mean moral language, or any other language, needs to be explained in that way too. The incoherence in local expressivism that Price’s arguments suggest is just not there. Price’s arguments are an attempt to shift local expressivism, a well established view, somewhere new, on the grounds that the old view is incoherent. They simply fail to do that. The most they achieve is to point out some other facts, concerning ‘general’ representational concepts.

Second, it’s worth noting that while the generality interpretation of Price’s global expressivism is well supported by some things that Price says, it does not very naturally apply to Price’s internal argument. For given that interpretation, the argument goes through far more easily than is suggested by the level of detail Price goes into in defending it. For it really doesn’t matter how the moral expressivist’s explanation works in detail. All that matters is that the success of any local application of expressivism just is an immediate counterexample to the thesis that we can explain thought and language by appealing to general relations between them and the world. This conclusion doesn’t need any of the work that Price puts into establishing it. It’s enough just to say that if moral language isn’t representational, then there aren’t any general representational relations in Price’s sense. This is far easier to establish than the details of Price’s argument, which as we’ve seen don’t succeed on their own terms anyway. So while the internal argument can be interpreted in a way that makes it succeed, this is not a very natural interpretation. The more natural interpretation takes the detail seriously – but then we find that the argument doesn’t succeed, because it makes false claims about how the moral expressivist’s theory might generalise. This is important, because it supports the idea that local expressivism is coherent and stable after all.

Finally, even if the best interpretation of this debate shows it to be verbal, it is certainly still worth seeing why Price’s arguments fail when read in the local expressivist’s own terms. In the internal argument, Price makes several claims about how easily the moral expressivist’s explanations generalise. These turn out to be false when we look at them closely: the mere possibility of the success of moral expressivism doesn’t guarantee – or even suggest – that its psychologistic explanation of moral sentences can generalise, nor that if it can, that all mental states can be explained without appeal to representation and the like. These are flaws in Price’s arguments, but they are interesting to note and important for a proper understanding of what moral and local expressivists are up to.

This is not to say that nothing interesting comes from the points Price makes. For instance there is a flourishing debate about the impact of minimalism about truth on mainstream explanations of meaning, which often proceeds in truth-conditional terms.Footnote 12 It is just less clear what this has to do with local expressivism. For whatever the outcome of this debate, there’s no suggestion that any of the distinctions the local expressivist might want to draw – between terms that express attitudes and terms that track features of the environment, for instance – would be threatened at all. Even if minimalism stops us characterising ordinary, traditionally descriptive language and thought in terms of truth and reference, there’s no reason to think this would mean it had to be characterised in a way that made it indistinguishable from other kinds of language. Perhaps instead the most interesting questions to ask under the ‘global’ heading are other, narrower ones. For instance, whether or not Price is right, this leaves untouched the other major debate about ‘globalising’ expressivism – the one most forcefully put forward by Mark Schroeder (2008) about the need for, but potential impossibility of, a global version of the moral expressivist’s semantics. In any case, Price’s arguments give local expressivists nothing to worry about.