1 Introduction

In analytic philosophy, expressivism started out as the idea that some kinds of sentences express mental states other than beliefs. It is most famously applied to moral language, in the hands of Simon Blackburn and Allan Gibbard, but also sometimes to modal, meaning, and mental language.Footnote 1 Often, defenders of this view take it to be local, only applicable to some domains of talk, which are going to be contrasted with language that is in the business of representing the world and expressing beliefs. However, philosophers like Huw Price and Robert Brandom propose to apply this idea globally, to see all domains of language not as representational, but as expressive, in some thinner and broader sense of ‘expressive’ than that of the local expressivists.

Both local and global expressivism are substantial and interesting ideas, having wide ramifications in philosophy of language and metaphysics. They appear to conflict—it seems that there must be a difference between expressivism everywhere and expressivism somewhere—and defenders of both ideas think that they conflict. However, it is a challenge to pin down exactly what they are committed to, and thus exactly where they disagree, for both theses tend to be framed in vague, big-picture ideas, and moreover adopt subtly different vocabularies—especially the two crucial terms here, ‘expressive’ and ‘representational’. As a result, despite their superficial difference and their proponents’ prima facie disagreement, some find it hard to see the difference, or even argue that there really is no difference between them. A telling example is from Matthew Simpson (2020), who argues that the most plausible interpretation of global expressivism also makes it not distinct from local expressivism.

Now here is a problem: are global expressivism and local expressivism really different, and if so, given that the superficial difference between expressivism everywhere and expressivism somewhere is thrown into doubt, what exactly is their difference? This paper offers a resolution of this problem, a definitive answer to the worry that it is hard to see how they disagree, and the worry that they really do not disagree. Through a reframing of the debate, I show not only how exactly they disagree, but also that their disagreement is more profound than it has usually been conceived as.

I begin with the best exposition of local and global expressivism as defended in the literature, with an emphasis on their different conceptual origins (Sect. 2). After giving reasons for suspecting that they do not really disagree, I then reject that idea and pin down their disagreement, first on the concrete level of linguistic content, then on the more abstract level of explanatory order and broader philosophical commitments. This will be the main body of argument where I show how a seemingly minor and subtle disagreement on the surface of the debate opens up to reveal a much deeper discrepancy (Sect. 3). Finally I examine an important consequence of this reframing: the global expressivists’ arguments to convert local expressivists are ultimately unconvincing as they only amount to more exhibitions of their difference. This might be expected because, given that their discrepancy is deeper than is usually recognised, what seems to be common ground turns out to be not common at all (Sect. 4).

2 Expressivism: From Local to Global?

It seems almost trivial to say that global expressivism is a global version of local expressivism. However, it is misleading to think of global expressivism as merely an extension of local expressivism, and of their debate as merely a matter of ‘all or some’, because, as I shall argue, they should be best described as resulting from two different lines of reasoning and rooted in two different conceptual origins. Local expressivism is motivated largely by what is known as ‘the placement problems’ (See Price & Jackson, 1997; Price & Macarthur, 2007; Price, 2011, Ch. 9)—the problem of placing the subject matter of certain domains of discourse in a natural world. Global expressivism, as I will argue, is best described as being motivated by a wholesale rejection of representationalist semantics.Footnote 2 The aim of this section is exegetical, not argumentative; while explaining local and global expressivism in terms of their different background motivations, here I remain neutral on the question of whether these motivations are ultimately coherent or convincing.

2.1 Placement Problems and Local Expressivism

What is moral goodness? Philosophers have been puzzled over the question because it is difficult to find a place for goodness in the natural world. On the one hand, we do talk about it in a way as if moral facts exist. On the other, it seems that we cannot find these facts in the natural world, the world described by our best sciences. We can see that grass is green, but presumably not that murder is wrong. Moreover, natural facts seem to be the wrong kind of facts: moral thoughts are motivating, while typically representational thoughts are not. So here is a problem, at least a prima facie one: how can we reconcile a commitment to naturalism with the fact-stating appearance of moral talks?

This is the guiding question for local expressivists like Blackburn and Gibbard. To answer it, they took inspiration from old moral non-cognitivists like Ayer (2001/1936), but added an important twist to their idea. The old idea is that moral statements do not genuinely represent the world, but instead express certain non-cognitive mental states (Ayer’s notorious equation: saying ‘murder is wrong’ is like saying ‘Boo Murder’). The twist is that expressivists recognise the need to respect the fact-stating appearance of moral assertions. They seek to make sense of the fact-stating appearance of moral discourse and talk about moral truths and facts as the realists do, without committing themselves to a metaphysical realism about morality. While noting that ‘murder is wrong’ and ‘grass is green’ have the same syntactic structure, they also emphasise that these sentences have very different underlying functions. To borrow the later Wittgenstein’s (1953, §12) metaphor, for expressivists, looking at language is like looking into the cabin of a locomotive: the handles all look more or less alike, but they do very different things. Both Blackburn (1993) and Gibbard (1990, 2003) have worked at length to theorise about the expressive function of moral talk, but details are not relevant here.

Given the need to reconcile functional pluralism with the uniform fact-stating appearance of assertoric sentences, expressivists, both local and global, generally come to adopt some kind of semantic deflationism.Footnote 3I will take Paul Horwich’s (1990) theory as an example and provide a quick gloss. According to this minimalism, truth, considered as a property of propositions, is not a substantive property. This means that there can be no general formula of the following form: ‘For all propositions x, x is true iff x is F’ (ibid., p. 120–2), which explains the facts about truth. Furthermore, the following trivial schema contains all there is to know about truth: ‘ < p > is true iff p.’ (ibid., p. 35). This account, on the one hand, makes truth non-substantial and non-explanatory: truth does not explain meaning, but follows from it. When a proposition < p > is true, it is true because p, not the other way around. Conversely, when we know a worldly fact such as p, we know that the corresponding proposition, < p > , is true. The T-schema tells us which sentences are true on top of a non-truth-conditional account of meaning. On the other hand, it leaves space for deflationists to explain why we come to use the truth predicate (despite its insubstantiality) by appealing to its use: it allows us to make generalisations where we previously cannot. Without the truth predicate, I cannot express < everything Sally said was true > without a potentially infinite conjunction (Sally said a and a, Sally said b and b, etc.).

Thus construed, minimalism is very friendly to expressivism. It helps expressivism explain the fact-stating appearance of, say, moral talk, without contradicting the expressivist and functional pluralist spirit.Given minimalism about propositions, we can make sense of talking about moral propositions; combined with minimalism about truth and properties, expressivists could then make sense of talking about moral truths and properties. Since we know how to use moral sentences, we also automatically know how to predicate truth of them. Meanwhile, this thin, deflationary sense of semantic concepts gives us the space for substantial, functional, and piecemeal explanations of the different domains of our language. Since minimalism stripped concepts like truth of their explanatory role, that role has to be played by more pragmatic, use-based stories.

In sum, local expressivism is motivated by placement problems, and solves them by arguing that languages have functions other than representing. In saying this, Blackburn and Gibbard also concede that one of the functions of language is representing. This will remain an important point of dispute between local and global expressivists.

2.2 Global Expressivism as Anti-representationalism

Now, if we simply take this line of thought and globalise it, we run into an obvious problem. The problem is this: for local expressivists, we tackle these ‘problematic’ discourses with placement problems by telling (functional, genealogical) stories of why we go in for them. Crucially, in telling these stories, we do not use the elements of these discourses, but merely mention them. In the moral case, very crudely, the story might tell us that agreeing on our praises and reprimands towards others’ behaviour is indispensable for large-scale cooperation, an important survival advantage; it does not use moral vocabulary in telling the story, but merely mentions people’s moral talk. But stories like this necessarily have to assume a background: a natural environment that contains various middle-sized objects, including human beings themselves. Now if we expand our expressivism and propose that we should give such stories for our ordinary descriptive talk (e.g. of trees and stones) as well, we find that they are impossible to tell. It is impossible to explain why we go in for tree talk without assuming the existence of trees and our causal relations with them. As Blackburn puts it: ‘the terms in which [our] situation can be the most barely presented must occupy a special role in any of our explanatory endeavours.’ (2010a, p. 261) This is Kraut’s (1990) ‘no-exit’ problem: our explanations have to assume something and start somewhere, and those are precisely what we cannot give an expressivist treatment of.

This point is certainly right, but running into such an obvious absurdity suggests that this is not the right understanding of global expressivism. I suggest that global expressivism is best seen as motivated by a rather different line of reasoning, though it does reach something like a globalised version of the latter in the end. Global expressivism should, first and foremost, be seen as a form of global anti-representationalism. It begins with the question: what is wrong in seeing our languages and thoughts as representational? It answers: there are two bad ideas associated with it.Footnote 4 First: the world has a pre-existing structure that our representational practices (thoughts and languages) seek to latch on to.Footnote 5 Second: representational relations—word-world relations—are largely atomistic and independent from one another.Footnote 6 When asked: what should we do with it then? It answers: we should replace it globally with something else; that ‘something else’, ultimately, comes to some kind of expressivist treatment of our language in general.

This suggestion is best justified by taking a quick detour back to Richard Rorty and Wilfrid Sellars, the origins of two chief anti-representationalist ideas which heavily influenced contemporary global expressivists like Price, Brandom and Michael Williams, and on which they have all worked at length to develop and refine. This detour is crucial not only for understanding global expressivism, but for illustrating their disagreement in the next section.

Rorty has attacked the first ‘bad idea’ with several arguments, and finally settled on the line that seeing the world as having a ready-made structure which we could only represent more or less accurately is to recognise an external authority before which we have to humble ourselves. His ‘anti-authoritarian’ argument in (1999), as I reconstruct it, is roughly this:

P1::

It is in the nature of representational practices that representations are responsible to the represented (to get it right, to represent it accurately), that representations take the represented as the authority.

P2::

Our linguistic/cognitive practices, if seen as representational, are (generally) directed towards things external to us.

P3::

We are only responsible to our fellow human beings, not anything external to us.Footnote 7 (Brandom, 2020) calls this, approvingly, ‘social pragmatism’.)

C::

Our linguistic/cognitive practices are not representational.

Both Price and Brandom explicitly endorse this idea and acknowledge Rorty’s influence on them, although their approaches are less ‘slash-burn-uproot-sow-with-salt’ than Rorty’s (Price 2013, p. 193). Both agree with Rorty that the representationalism characterised above is a deeply problematic picture, but neither wants to altogether abandon the notion of representation. Most notably, Price wants to separate two senses of representation: one he calls e-representation, which is the environment-tracking paradigm of representation, representation in the sense that thermometers represent temperature. Another he calls i-representation, which is the internal, functional notion of representation: ‘something counts as a representation in virtue of its position or role in some cognitive or inferential architecture.’ (2013, p. 36) On his account all meaningful sentences are i-representational (that is, expressive in his global expressivism), and they are so in virtue of word-word rather than word-world relations, but only some are e-representational. He says, then, that the important lesson is not to conflate the two, which is the big mistake of what he calls ‘big-R Representationalism’; as long as we do not conflate them, these two senses of representation are harmless. Brandom similarly seeks to recover a notion of representation ‘that is freed of the burdens and consequences [Rorty] saw as inevitably encumbering it’ (2013, p. 91). Price notes, and Brandom concurs, that they remain ‘card-carrying Rortyian anti-representationalist[s]’, and that the recommended philosophical shift to global expressivism ‘does involve a global rejection of a certain representationalist conception and hence, to that extent, amounts to a defence of Rorty’s anti-representationalism.’ (2013, p. 193).

Sellars famously rejects the second idea. His central insight is that meaning and understanding are constitutively connected, and both are holistic, related to what he famously calls ‘the space of reasons’. To understand a proposition, for Sellars, requires that one could use it to infer, to reason, and to justify, i.e. to know how it appropriately relates to other propositions (1956). A helpful example from Brandom concerns parrots: a parrot might be trained to have a reliable disposition to utter ‘this is a carrot’ whenever it sees a carrot. Nonetheless, intuitively, it doesn’t understand the meaning of that utterance at all because it doesn’t know how to use it: it cannot justify the utterance, cannot provide reasons for it, or infer anything from it. If this is right, then to know the meaning of a term must involve knowing the meaning of lots of other terms, and thus atomistic representations cannot account for our linguistic practices.

Again, both Price and Brandom are fully on board with this idea. Brandom has done more than anyone else to refine and develop Sellars’s normative inferentialism (1994). Price also doesn’t hesitate to admit their conceptual affinities. He argues (2013, 2016) that Sellars’s distinction between truth as S-assertibility (correct semantic assertibility) and picturing is essentially the same distinction as his between i- and e-representation, and then says: ‘I may be somewhere between Rorty and Brandom—though if so, I think, then not very far from either… [but] I may be closer to Sellars than to either.’ (2013, p. 193–4) He also takes the Sellars-Brandom inferentialism as a ‘pre-eminent candidate’ for spelling out his placeholder notion of i-representation: ‘my default model for a conception of its [i-representation’s] natural home is provided by inferentialism.’ (2013, p. 40).

A sketch of these ideas suffices to show that global expressivism has a motivation very different from local expressivism.Footnote 8 Local expressivists start with the placement problem: if we see moral and modal discourses as representing (as ordinary descriptive discourses), then we run into ontological and explanatory problems; so we should see them as doing something else. Global expressivists, however, start with a wholesale rejection of representationalist semantics and a replacement with something else. They therefore don’t need an ‘exit’—an external point of view from which they could account for the problematic discourse without using it—because they are not after the goal of solving the placement problems in all cases, or, because they are not simply globalising what local expressivists are trying to do locally. They propose to look at the entirety of language from a different angle, to give a comprehensive non-representational theory of meaning. In short, they have a different criterion of success from that of local expressivists.

In short, global expressivists are not trying to accomplish globally what local expressivists want to accomplish locally. This interpretation is justified not only by the conceptual genealogy of global expressivism offered above, but also by its capacity to provide us with a framework that allows us to best distinguish between local and global expressivism and make sense of their disputes. I now turn to illustrate the fruitfulness of this idea.

3 Do They Really Disagree?

3.1 Why Suspect that They Don’t Disagree?

Quick answer: because they seem to commit to very similar things. They both agree that the fact-stating appearance of assertions is to be respected, and thus adopt deflationism. They—especially Blackburn and Price—both emphasise that language has different underlying functions. It might be said that the difference is whether representation is included in this pluralist picture—whether representation is among the many functions of our language. Blackburn certainly says yes, but instead of plainly saying no (like Rorty would), Price’s distinction between i- and e-representation allows him to say: yes, certainly there are parts of our language that e-represent the world, and e-representation is indeed one function among the many, in his functional pluralist picture. Now again he and Blackburn start sounding similar: they now could even agree that representation, at least in one sense, is an important part of our language.

But there is the lingering worry that their difference must be obvious, or even cheap: one wants to apply expressivism everywhere, and the other only somewhere; one wants to say that no discourse is representational, and the other that some discourses are representational. However, the problem is that they understand the two crucial terms—‘expressive’ and ‘representational’—subtly differently. As is indicated in Sect. 2 (and will be developed below), local expressivists tend to understand ‘expressive’ psychologically (expressing mental states other than beliefs), while global expressivists understand it inferentially (expressing inferential commitments); while global expressivists reject the notion of representation embraced by local expressivists, their notion of e-representation seems to draw the line precisely where local expressivists draw it. Hence the suspicion that, essentially, they just use different words to draw the same distinction.

Matthew Simpson (2020), for example, argues that the only interpretation of global expressivism which makes it plausible also makes it identical to local expressivism, rendering the debate empty. To briefly summarise his argument: global expressivism is taken to be the claim that no term ‘robustly’ represents. There are broadly three ways of spelling out what a ‘robust’ representation is. (1) A term robustly represents if it bears some kind of co-variation with some aspect of the external environment. This cannot be right, as Price acknowledges that there are e-representational terms, as do other global expressivists (this is too common sense to deny). (2) A term robustly represents if the best explanation of its meaning and use cites its subject matter. For Simpson, this cannot be the right reading either: it would be implausible to say that the causal relation between the object tree and the term ‘tree’ has no bearing at all on its meaning. He further cites passages from Price and Williams to show that this is not the interpretation they intended either. (3) A term robustly represents if the best explanation of its meaning and use requires ascriptions to it of a general relation between language and the world (and for some terms, the best explanation of their meaning and use doesn’t require mentioning their subject matter). However, on this third interpretation, globalism falls back to localism: this claim is perfectly acceptable to local expressivists, and in fact accepted by Blackburn and Gibbard. Therefore, any plausible interpretation of global expressivism will show that it does not really disagree with local expressivism.

3.2 Disagreement over Content

Now I go on to show their disagreement, first on their different theories of linguistic content. The difference is, briefly, that global expressivists like Price, Brandom and Williams think that the content of all terms are inferentialFootnote 9 (i.e. it always and primarily depends on the content of other terms we use) while local expressivists like Blackburn and Gibbard think that the content of some terms are representational (i.e. it sometimes primarily depends on its relationship with the environment rather than its relationship with other terms). I shall first explore the three global expressivists’ view of content, focusing on Brandom—because he has the most worked-out, systematic account of meaning—and then contrast it with the view of Gibbard and Blackburn.

Briefly, Brandom’s overarching project is to build a theory of semantics upon a theory of pragmatics. On the fundamental, pragmatic level he describes ‘a game of giving and asking for reasons’ (1994), a picture of our social practice that involves rational creatures counting commitments and entitlements for themselves and others. The semantics that naturally falls out of this picture is what he calls ‘normative inferentialism’. Inferentialism, in the sense that the content of an utterance is constituted by its inferential role in the web of language (by what counts as evidence/justification for it and what legitimately follows from it); normative, because the content is not constituted by what inference anyone actually makes or is causally disposed to make, but by what inference is appropriate to make given the normative social practice.

Now we zoom in on the focal point of contention and consider terms like ‘red’. Obviously, the application of term stands in some inferential relations to that of other terms: the inference from ‘x is crimson’ to ‘x is red’ is always good, so does the inference from ‘x is red’ to ‘x is coloured’. However, it seems that the content of ‘red’ can’t simply be fixed by these inferences; a paradigmatic way of justifying our application of ‘red’ is pointing to some red object in the environment. This is precisely how our complex web of meaning gets latched on to the world and avoids spinning in the void.Footnote 10 Brandom, following Sellars, accommodates this intuition by noting that there are ‘language-entry’ and ‘language-exit’ rules: there are some inferences, in the broad sense, that could start from non-inferential circumstances or end with non-inferential behaviours. Given our norms, the inference from the presence of a red thing to the judgement ‘this is red’, and that from judging ‘stealing is wrong’ to preventing people from stealing, are also constitutive of the meaning of ‘red’ and ‘wrong’. The former constitutes an ‘entry’ into the world of inferences, i.e. the ‘space of reasons’, and the latter constitutes an ‘exit’ out of it.

However, saying that we are related to the environment doesn’t mean that global expressivists agree with local expressivists. Almost everybody agrees that we are related to the environment. The crucial question is what kind of relation it is. For global expressivists, the crucial answer is that the relation is causal rather than semantic. Causal, in that we do have some reliable dispositions to react to the environment in certain ways, in virtue of our sensory receptors that are causally influenced by external stimuli; not semantic, in that this relation, in itself, cannot tell us what any of our terms mean (recall the parrot example on p. 7). ‘“Red” does not refer to red things by virtue of this causal relation’, as Williams puts it. He goes on to say: the ‘presence of [observation] terms is a condition on meaningfulness for all terms, even though mere causal relatedness to environmental circumstances does not fix the conceptual content of any.’ (2013, p. 142, emphasis original) This means that our causal relation to the environment could only be part of the story of meaning, and it plays a part only because it plays an inferential, justificatory role in our linguistic practices. The content of all terms, therefore, is primarily inferential, and the content of some terms also includes a causal word-world relation in its inferential role, but such relations are only secondarily significant, only significant via its place in the inferential web. The mistake of representationalists, as per global expressivists, is not only that they take word-world relations to be sufficient in accounting for linguistic meaning (for some or all terms), but also that they take such relations to have primary importance in determining meaning.

This, as Price notices, fits nicely with his distinction between i- and e-representation, which he admits to be more like placeholders rather than well-developed notions. He wants i-representation to be the general case, covering all assertoric language; inferentialism contends that all assertions have a normative inferential role. He also wants some parts of language to be e-representational, to be causally related to the environment, but, importantly, even for these ordinary descriptive terms/assertions, e-representationality and i-representationality come apart. The picture above, again, nicely cashes out this claim: even for terms like ‘red’, it is e-representational in virtue of its causal relation to the environment, and i-representational in virtue of its overall inferential role in our linguistic practice. The former captures its specific function, the latter captures its being able to figure in assertions.

Now we contrast this picture with Gibbard and Blackburn. Gibbard (2015), while acknowledging his affinities with Price, points out the difference sharply: for him, it is wrong to start with i-representation as the general case, and see e-representation as merely one function among the many. Rather, we should start with the ‘paradigmatic’ cases of representation (that of the middle-sized environmental objects), where ‘the standard representationalist story in logic books is right’ (ibid., p. 214), and where, for him, i-representationality and e-representationality coincide (or, to use his terms, the relation of denotation and that of tracking are the same). Furthermore, it is the e-representationality of these cases that explains how i-representational features arise in other domains of discourses. Note, again, that he not only argues that in the ‘paradigmatic’ cases the semantic notion of representation gives us the whole picture of meaning, but he also argues that e-representational features have the explanatory primacy over i-representational ones. It’s not that e-representational features play a role in meaning fixing because they have a justificatory and inferential role; instead, it’s that i-representational features fix meaning in moral and modal discourses because they are derived from e-representational features of some other terms. As he puts it himself: ‘on [Price’s] picture, the general case is one of i-representation, where genuine explanations [of content] cite only features that are internal to thinking and language. My own picture is significantly different: I picture a paradigm of representation where genuine e-representation explains i-representational features’ (ibid., p. 216).

Blackburn is more subtle to interpret here. He does not say that there are ‘paradigmatic’ representations, because he does not conceive of other kinds of discourses (moral, modal…) as representational at all. His functional pluralism is ‘messier’ than Price’s, as he construes it. Price wants an inferential role account for i-representation, the general case for assertoric languages. Contrarily, Blackburn wants nothing more than the bare minimum at this general level. For him, functional pluralism requires that the content of different kinds of discourses is given differently; for our ordinary descriptive and scientific discourses, the e-representational features are sufficient to give us a substantial representational semantic relation. The following passages are indicative:

Why aren’t these relations, which Field calls ‘indication relations’ [i.e. e-representational relations], themselves sufficient to give birth… to the very notion of a truth-condition that deflationism exists to deny? … our conception of ourselves as responding to things and facts about things seems a promising starting point for a theory of representation… (2012, p. 200)

[The explanation of our use of ordinary descriptive terms] obviously, draws on the referential resources of the object language and, according to the account in front of us, amounts to a victory for representationalism over pragmatism. (2013, p. 79, emphasis added)

To say that we mirror their [tables’ and chairs’] doings now becomes a way of summarising a whole host of facts about our sensitivities that come along with first positing them… A mirror is quick to reflect the surrounding scene; I am not quite so quick but I do such a good job that comparing myself to a mirror becomes almost irresistible. (ibid., p. 82–3)Footnote 11


So, pace global expressivists, Blackburn believes that we could have a representational semantic theory based on the complex causal relations connecting us and the environment. Where global expressivists think that causes are merely causes and never give us substantial semantic relations, Blackburn thinks that the causal relations already give us enough material to build up substantial semantic relations. Where global expressivists think that causal relations only give us entry into ‘the space of reasons’, Blackburn thinks that ‘the space of reasons’ is much more closely entangled with ‘the space of causes’. Where global expressivists think that the meaning of all terms are primarily about their inferential roles, Blackburn thinks that the meaning of some terms are primarily about the external environment and our causal relations with it.

3.3 Disagreement over Explanatory Order

Now we know they disagree over theories of content. But this is a disagreement on the surface; to understand it better we need to consult a bigger picture. Why do global expressivists believe that content is always constituted by inferential roles inside our linguistic practices, while local expressivists believe that it is sometimes external to them? As Blackburn (2013) notes, they want the opposite explanatory orders in some cases (this is what my repetitive uses of ‘primary’ were highlighting).

Blackburn finds in Rorty the sharpest contrast with himself. For Rorty (see Sect. 2.2), the primary aim of our linguistic practices is always to foster ‘agreement with our peers’, metaphorically put, not with the environment.Footnote 12 Since he rejects the idea that nature has a pre-given structure for us to accurately capture or represent, we could use any vocabulary to conceptualise it, and the only criterion for our vocabularies is whether they allow us to successfully cope with the environment and with one another. The main global expressivists are on board with that idea. Brandom is straightforwardly and self-avowedly a social pragmatist and a follower of Rorty on these matters (see his 1999). Price, in identifying content with i-representation, also agrees that our explanation of language should always start from social practices, from intersubjective agreements. This is demonstrated in his (1990), (1998) and (2003), where he talks about the function of assertion and truth as specifying a norm, a norm where no-fault disagreements are minimised, where we need to reason and argue with one another to exchange information and reach agreements whenever we could, which is very useful for cooperation and survival. ‘Truth is the grit that makes our individual opinions engage with one another’, as he puts it (2003, p. 169). Then his reasoning goes like this: meaning is a way of summarising our use; we use assertions (roughly) to minimise no-fault disagreement with our peers; so meaning must be constructed (roughly) from agreement with our peers. Afterwards we might say that in different kinds of discourses we want to reach agreements for different reasons, and that for some terms we want to reach agreements because they causally co-vary with the environment. But that only comes later into the picture.

The explanatory primacy of agreeing with our peers is nicely captured by the theoretical primacy of normative inferential roles, which essentially tell us how to use our language according to social norms. And Blackburn’s disagreement with this theoretical primacy of inferential roles stems from his disagreement with the explanatory order endorsed by Rorty, Brandom and Price, as he makes clear in his (2013). His objection is that for him, in some cases—the e-representational cases—we primarily want to agree with the environment, and agreeing with our peers is merely a by-product of that. Scientists do talk in cafes to reach agreements, but, as Blackburn emphasises, before that they have to be trained in the laboratories, to do experiments and interpret results. The agreements among scientists only have the normative status they do because scientists have mastered the techniques and norms of the relevant practice, which go far beyond conversations and inferences. So in ordinary descriptive cases and the ‘coastal waters’ of science, we always start from agreeing with the environment, and our success in using these discourses could only be explained by the fact that they correctly represent our environment.

So their disagreement over content boils down to their disagreement over the right explanatory order for meaning. But we can still dig deeper into their philosophical roots to help us comprehend the debate even better. The contrast is sharpest between Blackburn and Brandom. Blackburn is a naturalist whose concern is primarily to solve the placement problem and keep naturalism intact. So he does not want to abandon questions like ‘Does x really exist?’ or ‘Is this domain of discourse genuinely representational?’; his aim is precisely to answer them. Brandom, as a social pragmatist, puts much more weight on social, discursive practices than Blackburn does.Footnote 13 Moreover, he does not see himself as a naturalist, because he thinks that we cannot account for normativity with the vocabulary allowed by naturalism (2020). Blackburn starts with minds and mental states in explaining the meaningfulness of language, while Brandom starts with our pragmatic needs and social behaviours. This is also reflected in their different emphasis on what it is to be a human being: for Blackburn we are primarily natural creatures, largely continuous with other animals; for Brandom we are primarily social, ‘sapient’ creatures, and despite evolutionary continuity, the use of language fundamentally differentiates us from other animals. Price falls somewhere between Blackburn and Brandom:he is more naturalist than Brandom in that he is a naturalist and thus does not take normativity to be fundamental, and more social pragmatist than Blackburn in that he takes meaning to be internal to social practices.

4 Arguments for Going Global

I discern and discuss three main lines of arguments in favour of global expressivism. All of them are present in Price’s works, and some are also discussed by other authors. My conclusion will be that none of them could compel a local expressivist to go global.

4.1 The Semantic External Challenge: Deflationism

The first challenge to local expressivism comes from semantic deflationism.Footnote 14 Price sees deflationism not merely as a neutral semantic thesis, but a powerful weapon that deflates all attempts to say what really exists or what discourses are really representational. And since local expressivists are generally also deflationists, it is argued that they are pushed by their semantic device towards global expressivism.

Before we start discussing the arguments it is helpful to first introduce the problem of creeping minimalism. It has been noted that minimalism about truth tends to extend to other related semantic concepts such as fact and proposition, because these concepts are generally seen as closely connected or inter-definable. For example, minimalism about propositions says that the following schema captures all we need to know about proposition: ‘“S” expresses the proposition that < S > ’. Exactly how far minimalism creeps is a problem under debate. Some have argued that minimalism naturally has the tendency to creep far and wide. The argument might go like this (for example, see Dreier, 2004 and Chrisman, 2008): the primary reason for expressivists to accept minimalism about truth and proposition is that we want to recover ordinary ethical talk while adhering to the expressivist insight.Footnote 15 Now, the ordinary ethical discourse also involves expressions of ethical beliefs (‘I believe that murder is wrong’), and there is an ordinary sense of ‘representation’ that is inherent to beliefs (to be a belief just is to represent reality to be a certain way).Footnote 16 Therefore, we should be minimalists about beliefs and representations too. This is known as the ‘creeping’ tendency of minimalism. There is also a related problem: if we deflate all semantic notions, then all the traditional ways of spelling out the difference between representationalism and expressivism are gone. This is known as ‘the problem of creeping minimalism’ (Dreier, 2004).

Now consider Price’s argument, which is directed specifically towards Blackburn. He seeks to convince Blackburn to turn global with the weapon of deflationism, since Blackburn is, in Price’s term, a ‘card-carrying deflationist’ (2019, p. 10) himself. Price (1996, 2007, 2019) starts by assuming that deflationism takes the form of blanket deflationism, and that deflationism has already ‘creeped’ from the concept of truth to all other semantic concepts. It follows that no semantic notion is ‘substantial’ or ‘robust’, in the sense that no semantic notion does any explanatory work. Though it sometimes appears like he assumes that deflationism must creep to all semantic concepts, a weaker claim could be attributed to him, i.e. that local expressivists like Blackburn in fact embrace blanket deflationism. From this ‘common’ premise he urges them to recognise the full potential of this blanket deflationism: if, following truth, notions like fact, property, reference and representation all fall into the bucket and get deflated, this is simply global expressivism, which says just that there is no robust semantic notion, and thus no contrast to be drawn between terms that genuinely represent and those that do not.

This alleged victory looks too easy not to be suspected. After all, if a blanket semantic deflationism is conceptually so close to global expressivism, there is no good reason to believe that local expressivists actually accept it. A careful interpretation of Blackburn shows exactly this, and Blackburn’s equivocation among terms reveals a deeper lesson about deflationism and the creeping problem.

In his earlier works Blackburn seems to agree with a deflationary treatment of representation (‘Minimalism seems to let us end up saying, for instance, that ‘kindness is good’ represents the facts. For ‘represents the facts’ means no more than: ‘is true’.’ [1998a, p. 79]). Later, however, he rejects deflationism about reference and representation, and draws the lesson that, in order to deflate some semantic notions, we must have some other inflated semantic notions that do the work.Footnote 17He starts comparing his deflationism to early deflationists like Frank RamseyFootnote 18 and writes: ‘early deflationists contrast deflationism about truth with whatever thick or robust stories might be given about reference. Typically, it was the elusive nature of facts that bothered them. Things are fine, and so in principle are the relations whereby words pick out things’ (2012, p. 208).

His criticism of Horwich’s (1990) and Field’s (1999) deflationism illustrates this point. In his view, Horwich’s deflationism relies on an inflated conception of reference, and Field’s deflationism relies on an inflated conception of truth conditions:

The minimalist may gain the position that there is no last chapter waiting to be written about truth. But that is because rather substantial first chapters needed to be written about the semantics of the components of the sentence… the theory of the proposition cannot be given without a theory of representation, and this in turn will provide the material for subsentential semantic relations. An across-the-board deflationism is not possible. (ibid., p. 209, emphasis added)


In parallel, though in early works he embraces a minimalist theory of ethical cognition and thus could talk about ethical beliefs and ethical knowledge (1998a, p. 79), later he subtly suggests that this might not be the ideal way of putting things:

[After an expressivist treatment of, say, ethical thoughts] If you want to call the result ‘belief’, well and good—but it won’t necessarily be much like belief in other areas. And, I would say… these beliefs can equally properly, and metaphysically much more illuminatingly, be thought of in other terms. (2010b, p. 5)


This might seem like a substantial change in Blackburn’s philosophy, but it is not. As the paragraph above suggests, what term you choose does not matter that much. The central lesson remains: you can deflate whatever word you like, but something needs to remain inflated to do the work, to secure a semantic relation between us and the world, to secure that our words are meaningful; if you want to deflate some of the terms I use, well and good, I’ll just use other terms to express the distinction I want to express. His motto is Wittgenstein’s ‘I’ll show you differences’ (recall 2.2: his functional pluralism is ‘messier’ than Price’s), and he is resolutely against a theory that ‘is the denial of differences, the celebration of the seamless web of language, the soothing away of distinctions’; he wants to clearly mark the differences ‘even in a postmodern, minimalist, deflationary world in which almost nothing could be said’ (1998b, p. 157–8). Therefore, he does not, as Price assumes, endorse a deflationary treatment of all semantic notions, and Price’s argument cannot get off the ground that way.

4.2 The Semantic Internal Challenge

What Price and Macarthur calls the semantic internal challenge (Price, 2019; Price & Macarthur, 2007) comes to this: if local expressivists could show why, in the hard cases, we can have truth without having representations, then why not in easy cases too? If ‘the explanation of our truth talk in the case of ethical language is that it encourages us to align our affective attitudes in a useful way, why not say the same about other mental states, such as the ones [the local expressivist] thinks of as genuine beliefs?’ (Price, 2019, p. 145) Alternatively put, there is a dilemma for local expressivists. Either the notion of representation becomes an idle cog: given that the theory of content is already given by the general account that applies to all assertions alike, adding the claim that they are representational does not help explain their meaning at all. Or they say that there is something in the genuinely representational discourses that other discourses cannot emulate, and in that case it shows that the other discourses cannot ‘deliver the goods, just where it really matters’ (Price & Macarthur, 2007, p. 245).

This argument will not move Blackburn and Gibbard. We can most clearly recognise this by referring back to the discussion in 2.3 on their disagreement over explanatory order. What Price is proposing here is essentially to start from a general, across-the-board account of assertions and the norm of truth, and say that they constitute what we need to know about content. The local expressivists, instead, think that piecemeal explanations are more appropriate, and that in some cases agreeing with the world is prior to agreeing with our peers. They would happily accept the second horn of Price’s dilemma: yes, they would say, there is indeed something in the genuinely representational discourses that others cannot emulate, but equally there is something in the moral discourses, say, that representational discourses cannot emulate. This is just functional pluralism. Their point is not that all discourses are the same; for them, a minimalist account of truth and truth-aptness merely provides a very thin foundation underlying all discourses, and different kinds of discourses add different things on top of that. And this is indeed how Gibbard directly responds to Price’s dilemma: ‘As for the second alternative… that is indeed what I am saying, and I don’t see anything wrong with it. Quasi-representation is like full, non-quasi representation in some ways and not in others. That’s what makes it quasi.’ (2015, p. 217).

Note that this doesn’t, as Price sometimes puts it, make the non-descriptive discourses ‘second-class citizens’. It is one thing to keep an eye on differences, quite another to make a hierarchy out of them. It might be that tree talk is genuinely representational while value talk isn’t. But similarly, value talk is genuinely action-guiding while tree talk isn’t. As Gibbard remarks, ‘normative thinking and language have kinds of significance that the naturalistic lacks. Naturalistic findings, after all, don’t by themselves settle what to do. In this respect, are they first class where naturalistic language is second class?’ (ibid., p. 216).Footnote 19

4.3 The Pragmatic Challenge: Rule-Following Considerations

Price calls his third challenge the pragmatic challenge (2019). There he gives nothing more than a sketch, but a more complete account could be put together given what he says in two other places (1988, 2017). Roughly, according to Price, the central insight of expressivism is that it links certain assertoric vocabularies (like moral vocabulary) to particular pragmatic grounds—to the practical features of speakers on which the use of that vocabulary depends (like affective attitudes). Now Price proposes to combine this insight with the later Wittgenstein’s famous rule-following considerations—and a range of similar considerations, such as Quine’s indeterminacy thesis—which raises the question of how meaning could be determinate given that we always acquire language on the basis of finite experience. He takes these considerations to show that the pragmatic grounds which local expressivists appeal to have to apply to all of language—which essentially means that they should go global –otherwise they cannot avoid a form of meaning scepticism.

Briefly, the rule-following considerations that Price appeal to comes to this. We learn about the meaning of terms through training and experience, and, be it our own experiences or our ancestors’, they are always finite. They do not determinately tell us how to use the term in the future. Take Kripke’s (1982) famous example regarding the rule of addition. Suppose, along Kripke’s line, that all our previous trainings regarding addition were based on numbers less than 100. Now we can make up a different function, ‘quus’, which behaves like plus when the inputs are less than 100, but gives the output ‘5’ when the inputs are not less than 100. Our previous experience, given its finitude, cannot tell us whether we are really using the plus or the quus function. There seems to be no fact of the matter to which we can appeal in order to show that plus, instead of quus, is the right way of going on. As Wittgenstein puts it himself, ‘no course of action could be determined by a rule, because every course of action can be made out to accord with the rule’ (1953, §201). Evidently, the idea goes beyond mathematical calculations and applies to meaning generally.

The lesson that Price draws is that ‘[t]he indeterminacy of meaning thus seems to be the basis for a form of universal non-factualism [an earlier term of his, which comes close to his later preferred ‘global expressivism’]’, that ‘in virtue of the nature of meaning there can be no genuine facts about anything’ (1988, p. 194–5). Later he clarifies that this was not, by his lights, an argument for global scepticism. It is rather an argument for the global bankruptcy of the representational picture of language, within which the implications of the rule-following arguments appear to be sceptical (2017, p. 156). For, if the meaning of terms is primarily the result of word-world representational relations, then, according to Price, the indeterminacy of how we go on means that such relations cannot be firmly established, which leads to unacceptable meaning scepticism. This global failure of representationalism means that local expressivism, which also embraces local representationalism, is untenable. On the other hand, Price’s favoured picture is one in which meaning is grounded in socially coordinated dispositions, in agreements and the forging of agreements (as is described in 2.3), in which, as William James famously puts it, ‘the trail of the human serpent is thus over everything’ (quoted in Price, 2017). To rephrase the idea in my framework here: what Price is saying is that, you may want to say (for some discourses) meaning is about agreeing with the world, but rule-following considerations show that we cannot mean anything at all unless we agree with our peers first.

The problem with this argument, as with before, is that Price and Blackburn draw very different lessons from rule-following considerations, each in accordance with their other, general philosophical commitments. Blackburn emphatically disagrees with the idea that the rule-following considerations should be—or indeed, could be—dealt with socially. In his (1984) he denies Wittgenstein’s conclusion that public rule-following is possible while private rule-following is not, and carefully demonstrates that the community standard for meaning fares no better in front of the Kripke-Wittgenstein sceptic: the public use of terms is no less finite than that of the private, and members of a community stand to each other just as the momentary time-slices of an individual do. His own preferred answer appeals to notions like practice and technique. Note, as he emphasises, that neither practice nor technique need be social; what matters is that they are stable, reliable, and successful. A born Crusoe who finds a Rubik’s cube washed onto his island and learns to solve it is someone who follows a rule, for all by himself he has a reliable technique, a practice, which he can repeat with success on demand. And he is so, insists Blackburn, regardless of whether he is in any community and of whether any community had thought about him (1984, p. 298). His later dispute with Rorty on the importance of experience and experiment over conversation, as he illustrates in his (2013) and discussed in Sect. 3.3, echoed this early passage almost perfectly: ‘[Wittgenstein] skips the intermediate case where the classification is given a putative private use. It fits into a project—a practice or technique—of ordering the expectation of recurrence of sensation, with an aim at prediction, explanation, systematisation… To someone engaged on this project… [s]ystem soon enforces recognition of fallibility.’ (1984, p. 299–300).

This should all sound familiar now. Price’s ‘pragmatic challenge’, after all, comes to only another illustration of their differences, not a successful argument for going global. Once again, on the topic of rule-following considerations, they already disagree from the start, and their disagreement here is essentially epiphenomenal to their disagreement on expressivism. Both could be traced back to the fundamental disagreement on whether we first agree with the environment (reliable disposition, technique, experiment) or with our peers (social practice, intersubjective agreement, conversation).

5 Conclusion

As such, I propose to reframe the debate between global and local expressivists, and see them as having a more fundamental disagreement than intending to carry the same project to different extents. Essentially, I reverted Simpson’s (2020) argument: while he argued, from the presumption that global expressivism is trying to do globally what local expressivism tries to do locally, to the conclusion that they are not really distinct doctrines, I turned it into a modus tollens and argued that we could only see their difference by rejecting that presumption, and moreover by seeing their disagreement as deeper than is usually supposed. This reversal is justified, not only by a careful interpretation of the sometimes-confusing texts of the philosophers involved, but also by its fruitfulness in helping us understand the debate and evaluate its arguments in a new light.

Now to sum up its fruits: local expressivists aim to solve the Placement Problems by arguing that some kinds of talk are not representational; global expressivists begin with a wholesale rejection of representationalism and a global replacement. Local expressivists propose that, in ordinary descriptive and scientific talk, our causal connection with the environment is sufficient and primary in accounting for their meaning; global expressivists acknowledge the role of causal connections, but argue that they are neither sufficient nor primary for meaning, and that meaning is always primarily inferential. In ordinary descriptive and scientific talk, local expressivists argue that agreement with the environment explains our mutual agreement, while global expressivists argue for the reverse. Thus understood, none of the existing arguments for going global is successful: they all underestimate the depth of this discrepancy, and assume seemingly common grounds that turn out to be mirages.