1. Reductive realism and relativism

Normative judgements are mental states that can be expressed with a sentence that applies a normative predicate to something, such as:

Lying is wrong.¹

Cognitivists take such judgements to be beliefs that ascribe normative properties: they take the judgement that lying is wrong to be a belief that ascribes the property of being wrong to lying.² If we distinguish normative predicates (such as ‘is right’, ‘is wrong’, and ‘is a reason’) from descriptive predicates (such as ‘is a table’, ‘is white’, and ‘is made of wood’), we can say that

A property is normative if and only if it can non-accidentally be ascribed with a normative predicate

and that

A property is descriptive if and only if it can non-accidentally be ascribed with a descriptive predicate.³

Most cognitivists are realists, who think that normative properties exist. But realists disagree about the nature of these properties. Reductive realists think that all normative properties are descriptive properties.⁴ If so, a normative predicate such as ‘is wrong’ ascribes a descriptive property. By contrast, robust realists think that at least some normative properties are not descriptive properties.⁵

Early reductive realists often took normative predicates to ascribe descriptive properties that are constituted by our attitudes.⁶ A view of this kind could say, for example, that

But suppose that Fred thinks lying is wrong and Susan thinks lying is not wrong, and suppose that Fred disapproves of lying and Susan does not disapprove of lying. In that case, (1) seems to imply that Fred and Susan’s conflicting normative judgements are both true.⁸ This is so clearly relativist that it seems a stretch to call it ‘realism’.

Most reductive realists therefore now take normative predicates to ascribe descriptive properties that are not constituted by our attitudes. And instead of making semantic claims like (1), which tell us which descriptive property a normative predicate ascribes, most of them now only make metasemantic claims, which tell us what makes it the case that a normative predicate ascribes a certain descriptive property.⁹ They have put forward two main views of this kind. The first appeals to causation: it says that

The second appeals to descriptive information: it says that

But suppose that Susan’s community’s use of the predicate ‘is wrong’ is causally regulated by the property of failing to maximise happiness and Fred’s community’s use of this predicate is causally regulated by the property of failing to comply with the categorical imperative. In that case, (2) implies that their conflicting normative judgements can both be true. Or suppose that after considering all relevant descriptive information, Susan would apply the predicate ‘is wrong’ to actions that fail to maximise happiness and Fred would apply this predicate to actions that fail to comply with the categorical imperative. In that case, (3) also implies that their conflicting normative judgements can both be true. These versions of reductive realism therefore still seem relativist.¹²

Is it a problem for reductive realists if their view is relativist? That depends on who you ask. Robust realists will say that it is: they will point out that we tend to think that

that we tend to think that

and that when we think that something is wrong, we tend to think that

Robust realism is clearly compatible with (4) to (6). But relativist views are not: if (1) is true, these claims seem false whenever we are not mistaken about our own attitudes of disapproval, if (2) is true, these claims seem false whenever we apply the predicate ‘is wrong’ to something that has the descriptive property that causally regulates our use of this predicate, and if (3) is true, these claims seem false whenever we apply the predicate ‘is wrong’ to something that we would apply this predicate to after considering all relevant descriptive information.

Of course, reductive realists could point out that robust realism faces serious problems of its own.¹³ They may therefore regard rejecting (4) to (6) as a price worth paying for endorsing realism while avoiding these problems. Perhaps it is. But I think they can do better: I think reductive realists can show that, contrary to appearances, their view is compatible with (4) to (6).

2. How expressivists go quasi-realist

They can do this by following the example of expressivists, who take normative judgements to be non-cognitive attitudes, such as attitudes of approval or disapproval.¹⁴ Reductive realists who endorse (1) take the sentence ‘Lying is wrong’ to express the belief that lying has the descriptive property of being disapproved of by the user of this sentence. By contrast, expressivists take this sentence to express disapproval of lying. As Mark Schroeder puts it, they take the sentence ‘Lying is wrong’ to stand to disapproval of lying the way the sentence ‘Grass is green’ stands to the belief that grass is green.¹⁵

Expressivists are sometimes charged with being relativists as well.¹⁶ I agree with Schroeder that this is a mistake.¹⁷ But what gives rise to this mistake may be that simple versions of expressivism face the same problem as relativist views: they seem just as incompatible with (4) to (6) as a relativist claim like (1) seems to be.¹⁸ Unlike reductive realists, however, expressivists have a long-standing way to deal with this problem. They do this by making two moves. First, they interpret (4) to (6) as expressions of non-cognitive attitudes: for example, they may take (4) to express

they may take (5) to express

and they may take (6) to express

If (4) to (6) express these attitudes, expressivism is compatible with these claims.²⁰

You may think that (4) to (6) express true beliefs rather than non-cognitive attitudes. But expressivists’ second move is to endorse minimalism: they take the sentence ‘“p” is true’ to express the same mental state as the sentence ‘p’, they take a sentence to express a belief if and only if it can be true, and they take an object to have a property if and only if this property can be ascribed to it with a true sentence.²¹ If minimalism is true, the sentence ‘Lying is wrong’ expresses a belief, expressivists can take this belief to be true, they can take lying to have the property of being wrong, and they can take (4) to (6) to express true beliefs.²²

Moreover, expressivists can then also explain why we do not think that Fred and Susan’s conflicting normative judgements are both true. For according to minimalism, the sentence ‘“Lying is wrong” is true’ expresses the same attitude as the sentence ‘Lying is wrong’, which according to expressivism expresses disapproval of lying. Similarly, according to minimalism, the sentence ‘“Lying is not wrong” is true’ expresses the same attitude as the sentence ‘Lying is not wrong’, which according to expressivism expresses an attitude that is incompatible with disapproval of lying: call it toleration of lying.²³ They can therefore say that we do not think that Fred and Susan’s conflicting judgements are both true because this would involve having two incompatible attitudes.²⁴

These moves result in a more sophisticated version of expressivism that is known as quasi-realism. There are different versions of quasi-realism: instead of taking (4) to (6) to express higher-order attitudes of approval or disapproval like the ones described by (4*) to (6*), quasi-realists can also take (4) to (6) to express other non-cognitive attitudes, and instead of endorsing minimalism, they can also endorse a different expressivism-friendly account of truth, beliefs and properties.²⁵ And it is controversial whether these moves work. But suppose they do. Can reductive realists then answer the charge that their view is relativist by making quasi-realist moves as well?

3. How reductive realists can go quasi-realist

I think they can. Suppose they start by saying that

This seems to make reductive realism as incompatible with (4) to (6) as it can be. If reductive realists can answer the charge that their view is relativist while endorsing (1), they can surely also do so while endorsing (2) or (3).

Suppose next that reductive realists agree with quasi-realists that we have the following attitudes:

As we have seen, quasi-realists take (4) to (6) to express these attitudes. Since reductive realists are not expressivists, they cannot say this. But they can do something else: they can take the same relation to obtain between (4) to (6) and these attitudes that (1) takes to obtain between the sentence ‘Lying is wrong’ and disapproval of lying. They then first need to reformulate (4) to (6) as follows:

And they can then make the following claims:

What (4†) to (6†) say about these predicates is similar to what (1) says about the predicate ‘is wrong’. And if (4†) to (6†) are true, reductive realism is compatible with (4) to (6). This is analogous to what quasi-realists do: just as quasi-realists take (4) to (6) to stand to the attitudes described by (4*) to (6*) the way they take normative sentences to stand to attitudes of approval or disapproval, reductive realists who endorse (4†) to (6†) also take (4) to (6) to stand to the attitudes described by (4*) to (6*) the way they take normative sentences to stand to attitudes of approval or disapproval.

The analogy between these views can be brought out in a different way as well. Mark van Roojen has argued that if quasi-realism is true, the attitudes described by (4*) to (6*) can also be expressed with the following claims:

Quasi-realists can therefore be regarded as taking (4) to (6) to be equivalent to (4’’) to (6’’) and then applying what expressivism says about a sentence like ‘Lying is wrong’ to (4’’) to (6’’). Reductive realists who endorse (4†) to (6†) can be regarded as applying reductive realism in the same way.

Should reductive realists who endorse (4†) to (6†) also endorse minimalism? They do not need to. For if (4†) to (6†) are true, (4) to (6) already express beliefs that ascribe properties, and if we have the attitudes described by (4*) to (6*), the things to which (4) to (6) ascribe these properties already have these properties, which means that the beliefs that (4) to (6) express are already true. But as we have seen, minimalism also allows quasi-realists to explain why we do not think that Fred and Susan’s conflicting normative judgements are both true. Can reductive realists explain this as well?

That depends on who the term ‘user’ in (1) refers to. If it refers to the person who makes the relevant normative judgement, (1) says that the predicate ‘is wrong’ ascribes the descriptive property of being disapproved of by the person who makes the judgement that this thing is wrong.²⁷ This implies that Fred and Susan’s conflicting normative judgements are both true. But reductive realists can avoid this implication by taking the term ‘user’ in (1) to refer the person who assesses the truth-value of the relevant normative judgement: in other words, by taking (1) to say that

This gives rise to the same pattern of truth-value ascriptions as the combination of expressivism with minimalism. Reductive realists who interpret (1) as (1’) can therefore agree with quasi-realists that we do not think that Fred and Susan’s conflicting normative judgements are both true because this would involve having two incompatible attitudes.²⁹ More generally, they can give the same account of normative disagreement as quasi-realists. For they can say that Fred and Susan’s disagreement consists in their incompatible attitudes towards lying, which give rise to a pattern of truth-value ascriptions that is characteristic of disagreement: Fred takes his own judgement to be true and Susan’s judgement to be false, and Susan takes Fred’s judgement to be false and her own judgement to be true.³⁰

Suppose that reductive realists make all of these moves: suppose they take (4) to (6) to stand to the attitudes described by (4*) to (6*) the way they take normative sentences to stand to attitudes of approval or disapproval, and suppose they interpret (1) as (1’) in order to obtain the same pattern of truth-value ascriptions as the combination of expressivism with minimalism. They then become what we can call quasi-quasi-realists: they become reductive realists who incorporate quasi-realist moves into their view. If these moves enable expressivists to show that their view is compatible with (4) to (6), they also enable reductive realists to show that their view is compatible with (4) to (6). And I think these moves then also enable reductive realists to answer the charge that their view is relativist, since what gave this charge its bite was that it seemed to make reductive realism incompatible with (4) to (6).³¹

Robust realists may doubt that reductive realists can answer this charge in this way. I will return to this in §5. And reductive realists may not want to become quasi-quasi-realists. But they will then either have to accept that their view is relativist or find a different way to make their view compatible with (4) to (6).

4. The assumptions of quasi-quasi-realism

To be able to make these moves, quasi-quasi-realists must make certain assumptions. The first is that we have the attitudes described by (4*) to (6*). This assumption is shared with standard quasi-realism.³² If standard quasi-realism is right that (4) to (6) express these attitudes, the fact that we tend to endorse (4) to (6) indicates that we have these attitudes. In the same way, if quasi-quasi-realism is right that (4) to (6) ascribe the properties that (4†) to (6†) say they ascribe, the fact that we tend to endorse (4) to (6) also indicates that we have these attitudes.

The second assumption is that (4) to (6) ascribe the properties that (4†) to (6†) say they ascribe. This assumption does not seem more far-fetched than standard quasi-realists’ assumption that (4) to (6) express the attitudes described by (4*) to (6*). Standard quasi-realists often motivate this assumption by taking (4) to (6) to be covert normative claims.³³ Quasi-quasi-realists can do this as well. The only difference is that whereas standard quasi-realists relate (4) to (6) to the attitudes described by (4*) to (6*) with expressivist machinery, quasi-quasi-realists relate (4) to (6) to these attitudes with reductive realist machinery. Of course, whether (4†) to (6†) are true is partly an empirical question, but whether (4) to (6) express the attitudes described by (4*) to (6*) is also partly an empirical question.

Moreover, quasi-quasi-realists can drop the first two assumptions by rejecting (4†) to (6†) and instead making claims that are analogous to (2) or (3): in other words, by saying that

or that

They then no longer need to endorse (1) and no longer need to appeal to the attitudes described by (4*) to (6*). But quasi-quasi-realism’s core move remains the same: just as (4†) to (6†) say which properties these predicates ascribe by treating them the way (1) treats normative predicates, (2†) and (3†) say which properties these predicates ascribe by treating them the way (2) and (3) treat normative predicates.

The third assumption is that the structure of a predicate need not reflect the structure of the property it ascribes. For example, if (4†) is true, the predicate

‘is such that it is possible to be mistaken about whether this thing is wrong’

ascribes the property of

being such the user of this predicate approves of being willing to revise an attitude of disapproval of this thing in response to new information,

which does not seem to have the same structure as this predicate.³⁴ But standard reductive realism already assumes this: for example, if (1) is true, the predicate ‘is wrong’ ascribes the property of being disapproved of by the user of this predicate, which does not seem to have the same structure as this predicate either.³⁵ More generally, reductive views assume that a single entity in the world can have different modes of presentation and that the structure of a mode of presentation need not reflect the structure of this entity.

Quasi-quasi-realism’s claims about the properties that (4) to (6) ascribe may seem counterintuitive. But if a single entity can have different modes of presentation, which property a predicate describes can be a surprising discovery. Moreover, students who encounter standard quasi-realism for the first time often find its claims about the attitudes that (4) to (6) express similarly counterintuitive. To the extent that standard quasi-realism’s claims seem less counterintuitive, this may simply be because we are more familiar with this view than with quasi-quasi-realism. And whether standard quasi-realism or quasi-quasi-realism is true should be settled by considering these views as a whole rather than by our intuitive responses to some of their claims.

The fourth and final assumption is that which property a predicate ascribes can depend on who assesses the truth-value of a sentence that applies this predicate. Though this assumption is controversial, it is independently defensible.³⁶ Which property a predicate ascribes can depend on who utters a sentence that applies it: for example, which property the predicate ‘has the property that I am currently thinking about’ ascribes depends on who utters a sentence that applies this predicate. Which property a predicate ascribes can also depend on who reads a sentence that applies it: for example, which property the predicate ‘has the property that the reader of this sentence is currently thinking about’ ascribes depends on who reads a sentence that applies this predicate. This suggests that it should similarly be possible that which property a predicate ascribes depends on who assesses the truth-value of a sentence that applies it. For it suggests that the meanings of predicates can be Kaplanian characters that make which property a predicate ascribes depend on different features of the context: who utters a sentence that applies this predicate, who reads a sentence that applies this predicate, or who assesses the truth-value of a sentence that applies this predicate.

Of course, none of this shows that quasi-quasi-realists’ assumptions are true. But I think it does show that these assumptions are either shared with standard quasi-realism, or no more far-fetched than a similar assumption made by quasi-realism, or already made by standard reductive realism, or independently defensible. I therefore think that these assumptions do not threaten my claim that if quasi-realist moves enable expressivists to show that their view is compatible with (4) to (6), then such moves also enable reductive realists to answer the charge that their view is relativist.

Moreover, quasi-quasi-realism also has two advantages over standard quasi-realism. First, it does not need to appeal to minimalism or to any other expressivism-friendly account of truth, beliefs and properties. For as I have said, if (4†) to (6†) are true, (4) to (6) already express beliefs that ascribe properties, and if we have the attitudes described by (4*) to (6*), the things to which (4) to (6) ascribe these properties already have these properties, which means that the beliefs that (4) to (6) express are already true. Second, it does not face the Frege-Geach problem to the same extent as standard quasi-realism.³⁷ Like standard quasi-realists, quasi-quasi-realists may need to explain why disapproval is incompatible with toleration, which is often regarded as part of the Frege-Geach problem.³⁸ But since quasi-quasi-realists are cognitivists, explaining this is enough for them to solve this problem. By contrast, standard quasi-realists need to do much more.

Quasi-quasi-realists can make the moves I have described whether or not standard quasi-realists take (4) to (6) to express higher-order attitudes like the ones described by (4*) to (6*). For any version of standard quasi-realism is going to say that (4) to (6) express certain attitudes, and quasi-quasi-realists can make claims about these attitudes that are analogous to (4†) to (6†) no matter what these attitudes are.³⁹ Quasi-quasi-realists can therefore make the moves I have described no matter which version of quasi-realism is most defensible. Moreover, as we have seen, they can make their key move even if they reject (1) and endorse (2) or (3) instead, since they can then endorse (2†) or (3†) rather than (4†) to (6†). Quasi-quasi-realists can therefore make their key move no matter which version of reductive realism is most defensible. But if reductive realists can answer the charge that their view is relativist while starting from (1), there may be no need for them to start from (2) or (3) instead. For in that case taking normative predicates to ascribe descriptive properties that are constituted by our attitudes, as early reductive realists often did, is much more defensible than it may have seemed.⁴⁰

Some expressivists have recently moved towards reductive realism by incorporating the claim that normative properties are descriptive properties into their view.⁴¹ Quasi-quasi-realists are reductive realists who move towards these expressivists from the opposite direction, by incorporating quasi-realists moves into their view. These expressivists and quasi-quasi-realists are therefore travelling to similar destinations. But they use different machinery and make different assumptions along the way.

5. Robust realists’ doubts about quasi-quasi-realism

Robust realists may doubt that reductive realists can answer the charge that their view is relativist by becoming quasi-quasi-realists: they may feel that there is ‘a deeply uneasy gap … between the spirit of the theory itself and the spirit it supposedly justifies’, as Bernard Williams puts it in a different context.⁴² They may therefore regard quasi-quasi-realism as merely mimicking what robust realists say.

But since robust realists have the same doubt about standard quasi-realism, this does not undermine my claim that if quasi-realist moves enable expressivists to show that their view is compatible with (4) to (6), then they also enable reductive realists to answer the charge that their view is relativist. Moreover, if standard quasi-realists are right that (4) to (6) express the attitudes described by (4*) to (6*), their view does not merely mimic what robust realists say but really incorporates it. The same applies to quasi-quasi-realists: if they are right that (4) to (6) ascribe the properties that (4†) to (6†) say they ascribe, their view really incorporates what robust realists say.

This will not satisfy robust realists, since they will deny the antecedents of these conditional claims: they will say that

and that

How should quasi-quasi-realists respond to this denial? This is a large question that I cannot settle here. But I think there are at least two possible responses. One is to agree with Jamie Dreier that quasi-realist moves only aim ‘to vindicate ordinary moral thought and language’.⁴³ Quasi-quasi-realists can then reject (8) by denying that this claim is part of ordinary normative thought. But it may be difficult to draw a clear line between claims that are part of ordinary normative thought and claims that are not.⁴⁴

Alternatively, they can keep making quasi-realist moves. I have argued elsewhere that standard quasi-realists can take (7) to express

If so, quasi-quasi-realists can similarly say that

Robust realists could also express their doubts about quasi-quasi-realism in more general ways. For example, they could say that if quasi-quasi-realism is true, it is not the case that

But I have argued elsewhere that standard quasi-realists can take (9) to express

If so, quasi-quasi-realists can similarly say that

And they can keep making such moves in response to any other claim that robust realists could make to express their doubts.⁴⁷

I have argued elsewhere that if standard quasi-realists keep making such moves, they end up with an extreme version of their view that we can call super-quasi-realism. If quasi-quasi-realists do the same, they end up with a similarly extreme version of their view that we can call super-quasi-quasi-realism. Since anyone who tries to endorse super-quasi-realism will endorse every claim that robust realists endorse and reject every claim that robust realists reject, anyone who tries to endorse this view will become a robust realist. But I have argued that this does not mean that super-quasi-realism is false or that it collapses into robust realism. Instead, it merely means that anyone who tries to endorse super-quasi-realism will end up believing a different view.⁴⁸ I think the same applies to super-quasi-quasi-realism.⁴⁹

Since super-quasi-realism and super-quasi-quasi-realism both use anti-realist building blocks to enable us to endorse (4) to (9), robust realists may continue to think that these views are not realist enough. But these views do enable us to say everything robust realists say. They are therefore as close to robust realism as it is possible for anti-realist views to get.

6. Is the remaining difference between these views important?

Suppose there are two exactly similar communities, except that in one of them standard quasi-realism is true and in the other a version of quasi-quasi-realism that starts from (1) is true. There will then be many similarities between these communities: which attitudes of approval or disapproval people have will be the same, which normative sentences they endorse or reject will be the same, and which truth-values they ascribe to these sentences will be the same. Moreover, in both communities people will have the following attitudes:

And in both communities they will endorse the following claims:

The only difference between these communities will be the machinery that relates people’s attitudes of approval or disapproval to their endorsement or rejection of normative sentences, to their ascription of truth-values to these sentences, and to their endorsement of (4) to (6): in the standard quasi-realist community this will be the expressivist machinery I outlined in §2, but in the quasi-quasi-realist community this will be the reductive realist machinery I outlined in §3.

Suppose that you know you belong to one of these communities, but that you do not know which one. How could you find out which community is yours? Not by finding out which attitudes of approval or disapproval people in your community have, or which normative sentences they accept or reject, or which truth-values they ascribe to these sentences, or whether they endorse (4) to (6). For these things will be exactly the same in both communities. Moreover, if there is a solution to the Frege-Geach problem, you also cannot find out which community is yours by asking people to make normative inferences. For these inferences will then also be exactly the same in both communities.

You may think you could find out which community is yours by asking people whether lying is wrong because they disapprove of it. As quasi-realists often point out, people in the standard quasi-realist community will deny this: they will say that lying is wrong because of the descriptive facts that make it wrong, such as the fact that lying undermines trust, and they will deny that these facts include the fact that they disapprove of lying. But people in the quasi-quasi-realist community will say exactly the same thing. For if (1) is true, having the descriptive property of being disapproved of by the user of the predicate ‘is wrong’ is what it is to be wrong, which means that the fact that lying has this descriptive property cannot also be among the descriptive facts that make lying wrong. People in this community will therefore agree that lying is wrong because of other descriptive facts, such as the fact that lying undermines trust.⁵⁰

You may also think you could find out which community is yours by investigating whether people are motivated to act in accordance with their normative judgements. For whereas in the standard quasi-realist community people who think lying is wrong disapprove of lying, which seems closely connected to motivation, in the quasi-quasi-realist community people who think lying is wrong believe that they disapprove of lying, which seems less closely connected to motivation. But for standard quasi-realism to be plausible it must posit a fairly weak connection between disapproval and motivation: it must say that people are generally but not always motivated to act in accordance with their normative judgements. This difference will therefore probably be too small to enable you to find out which community is yours.

Your only remaining option seems to be to ask people in your community how their attitudes of approval or disapproval are related to their endorsement or rejection of normative sentences, to their ascription of truth-values to these sentences, and to their endorsement of (4) to (6): is this the expressivist machinery I outlined in §2 or the reductive realist one I outlined in §3? But the vast majority of people will have no idea what you are talking about. And those who do understand what you are talking about will be philosophers whose answers to such questions will be just as divided as they are among actual philosophers. There therefore does not seem to be any way for you to find out whether you belong to the standard quasi-realist community or the quasi-quasi-realist community.⁵¹

Moreover, robust realists will take the same thing to be missing from both communities. They can make this clear by appealing to John MacFarlane’s distinction between preclusion of joint accuracy, according to which

the accuracy of my attitudes (as assessed from any context) precludes the accuracy of your attitude or speech act (as assessed from that same context),

and preclusion of joint reflexive accuracy, according to which

the accuracy of my attitudes (as assessed from my context) precludes the accuracy of your attitude or speech act (as assessed from your context).⁵²

Robust realism precludes joint reflexive accuracy, but MacFarlane argues that relativism at most precludes joint accuracy.⁵³ Robust realists will say that the combination of expressivism with minimalism also at most precludes joint accuracy: they will say that this combination does not make it impossible that conflicting normative judgements are both true when their truth-values are assessed from different people’s normative perspectives.⁵⁴ We can mark this similarity between relativism and the combination of expressivism with minimalism by regarding them as versions of a more general view that we can call perspectivalism.

Quasi-realist moves affect these versions of perspectivalism in the same way: they make what is in fact only preclusion of joint accuracy look from within people’s normative perspectives like preclusion of joint reflexive accuracy. These moves therefore make it hard to see from within people’s normative perspectives that perspectivalism is true, and instead make it look from within people’s normative perspectives as if something approaching robust realism is true instead. And super-quasi-realism and super-quasi-quasi-realism do this to the greatest possible extent: they make people conflate preclusion of joint accuracy with preclusion of joint reflexive accuracy from within their normative perspectives. That is how these views make those who try to endorse them reject perspectivalism and endorse robust realism instead.

Suppose I am right that these views are both versions of perspectivalism that use different machinery to relate our attitudes of approval or disapproval to our endorsement or rejection of normative sentences, to our ascription of truth-values to these sentences, and to our endorsement of (4) to (6). And suppose I am right that they both use this machinery in similar ways to make it hard or even impossible to see from within our normative perspectives that this is what is going on. Is the difference between standard quasi-realism and quasi-quasi-realism then philosophically important? Perhaps it is not: perhaps what is philosophically important is only whether perspectivalism is true. Perhaps we should leave the question which version of perspectivalism is true to linguistics. And perhaps linguistics will tell us that these two versions of perspectivalism are equally admissible interpretations of the same worldly and mental facts and linguistic data.

7. Conclusion

I conclude that if the quasi-realist moves that many expressivists make work, reductive realists can copy these moves to answer the charge that they are relativists. Reductive realists should welcome this conclusion. But so may robust realists. For they may say that if quasi-realist moves enable even a view with a clearly relativist starting point like (1) to answer the charge that it is relativist, we should not conclude that this charge has really been answered. Instead, robust realists may say, we should conclude that there is something wrong with these moves. And they may take this to show that there is something wrong with these moves when expressivists make them as well.

Acknowledgements

For comments on previous versions of this paper I am grateful to Umut Baysan, Matthew Chrisman, Daan Evers, Sam Mason, Silke Mast, Berend Pot, Michael Ridge, Neil Roughley, Peter Shiu-Hwa Tsu, Justin Snedegar, Christine Tiefensee, Pekka Väyrynen, two very helpful referees for this journal, and audiences at the University of St Andrews, the University of Edinburgh, and the MetaEssen workshop at the University of Duisburg-Essen.