Domains and ontology

Two features of my discussion of ontology that are often seen as problematic are the vagueness of the idea of a domain and my remark that ‘the truth values of statements about one domain, insofar as they do not conflict with statements of some other domain, are properly settled by the standards of the domain that they are about’ (Scanlon, 19).

First, by a domain I meant simply a subject matter: the subject that certain statements are about. Empirical judgments are about the existence and properties of physical objects and events, and the causal relations between them. Arithmetic is about numbers and the relations between them. Normative judgments are judgments about the reasons we have for or against various actions and attitudes. Since I was concerned only with judgments of these three kinds, I might have done better simply to make my claims about these subjects and avoid the more general idea of a domain altogether. I went beyond this, and introduced general idea of a domain, in part because I was concerned that my claims about mathematical and normative claims would seem ad hoc unless placed in a more general view about ontology. Another reason for introducing the idea of a domain was that I was thinking of my view in relation to Carnap’s, and I wanted to avoid his idea of a linguistic framework. It seemed to me that the domains I have just listed – the natural world, arithmetic, and the normative – were better understood in terms of the subjects they are concerned with rather than the meaning of the words being used to refer to these subjects. The general principles and methods of argument that guide our thinking about these subjects are substantive theses about them, not rules of language.

Of course we use words expressing certain concepts in order to identify and characterize these subjects, and we use more specific concepts to make claims about them. In the case of morality, for example, these include such concepts as that of an obligation or a right, and the concept of responsibility. But a domain is the subject matter that we identify and make claims about using certain concepts, not that set of concepts itself.

One difference this makes is that while it may seem appropriate to speak of ‘adopting’ or ‘choosing’ a linguistic framework, or a set of concepts, this makes less sense in the case of a subject matter. One can decide whether to take an interest in a subject matter, and one can change one’s mind about whether a subject matter is significant, or whether claims about it are even coherent. But this is different from adopting it or creating it. One can characterize an abstract subject matter simply by defining terms in a sufficiently precise and coherent way. But even in this case, one does not create the distinctions that these terms mark out. The concepts characterizing a subject matter have content or not, and are coherent or not, whether or not we ever think about the matter.

The concepts in terms of which a domain is specified, and the principles and methods of argument that we use in our thinking about a domain, are open to revision over time. We engage in this kind of revision through reflective equilibrium reasoning about how to make best sense of what seems to be true about that domain. This process of adjustment and revision can apply even to the most fundamental concepts that we use to characterize a domain itself. For example, our current idea of ‘the natural world’ is quite different from the ideas that played the role closest to that for educated people in twelfth century Europe, and my understanding of the domain of morality has certainly changed over the 50 years that I have been thinking about moral philosophy. These changes result from ‘internal’ reasoning about the domain; they are not ‘external’ choices of which framework to adopt.

Nor can one, on my view, create reasons by characterizing a domain in a certain way. For example, we could define a subject matter in terms of a set C of rules about what is to be done in certain circumstances. If these are contentful and coherent they could characterize a domain. Its subject matter would be what it takes to be C-compliant and which things have this property, and if the rules were sufficiently clear and coherent there would be a fact of the matter about answers to these questions. But the existence of this domain need not have any normative implications.

Such implications might follow in either of two ways. First, the directives might be interpreted as making straightforward normative claims about what individuals have reason to do. In that case this domain would be a sub-domain of the normative, and one would have reason to do what conclusions about this domain direct one to do if, but only if, its directives are true claims about the reasons one has. The mere coherence of the rules says nothing about whether this is the case. Alternatively, these directives might be interpreted ‘positivistically,’ simply as giving content to a technical question about when a person’s actions are C-compliant. In this case one would have reason to follow these directives only if it were a further normative fact, independent of the domain in question, that people have reason to be C-compliant. In either case, the claim that one has reason to follow these directives would be a claim within the normative domain.Footnote ³ If these normative claims are in fact false, then the C domain is a kind of normative fiction.

Even if my view does not imply that by specifying a domain one can create new entities, or new reasons, it might seem to have other implausible ontological implications. For example, the idea that a domain exists if its subject matter can be specified in a contentful and coherent way may seem to imply an implausible profusion of domains. Enoch and McPherson ask, for example whether on my account, there would be a domain concerned with claims about the color orange, or claims about tables. My response is to ask why not? These domains will mostly be uninteresting, since they are so obviously sub-domains of the domain of the natural world. But there are questions about how to understand the relation between facts about middle-sized objects such as tables and facts about elementary particles, and this question could be put as a question about the relation between these domains. On my account, however, there would be such domains whether or not we had any reason to be interested in them. As I have said, the question of whether there is an abstract domain of a certain kind (whether the concepts that we use to characterize it have content and are coherent) is in my view an entirely separate matter from whether we have any reason to be concerned with it. I am not a pragmatist about the existence of domains, or about the existence anything else.

But it does seem a little odd, even to me, to think of the existence of domains as I am understanding the matter as an ontological question at all, if the existence of a domain amounts to no more than being characterized by concepts that have content and are coherent. This oddness may make it inviting to say that domains exist only in a ‘non-ontological sense,’ as Derek Parfit says of abstract entities such as numbers, sets and of normative facts (Parfit Reference Parfit2011, 480–483). The idea of ‘existence in a non-ontological sense’ is a bit obscure. Clarifying what it might mean to exist in a non-ontological sense, and assessing the case for holding that abstract entities exist in this sense will, I believe, also make clearer what is at stake the disagreement between me and more ‘robust’ realists such as Enoch and McPherson.

Simply because these entities are abstract, claims about numbers, sets, and the relations between them are much less likely to be in conflict with truths in other domains. So the criteria of existence that I have laid out will be extremely permissive with respect to these things, and for abstract entities in general. All that is required is that they be held to exist by the internal standards of coherent domains. By contrast, for Quine and Goodman, writing in the 1940s, abstract entitles were the things whose existence seemed most problematic (Quine and Goodman Reference Quine and Goodman1947). This sharp difference reflects, I assume, a difference in the ideas of existence that we have in mind. What Quine and Goodman were objecting to may well have been the idea that abstract entities exist ‘in an ontological sense;’ that is to say, that they are part of a larger world or ‘reality’ that includes the natural world but is not limited to it. If being part of such a reality is what it would be for such entities to exist in an ontological sense, then I agree that they do not exist in this sense. But I also believe that putting the matter in this way is misleading because it appears to concede too much.

As I pointed out at the beginning of this paper, fictionalism about numbers is extremely implausible. There is no contentful sense of ‘exists’ in which numbers might fail to exist despite the correctness of arithmetical claims about them. Abstract entities do not fail to exist in an ontological sense because they lack some property that other things have. Nothing exists in an ontological sense if existing in this sense means being part of a larger reality of the kind I have mentioned. Not even physical objects exist in this sense. Physical objects exist as part of the natural world. They do not, because of this, also exist in a further ‘ontological’ sense. What is involved in existing is always a purely domain-specific matter. Physical objects exist insofar as they have spatial location, mass, color, and other physical properties. Numbers exist insofar as they stand in the appropriate arithmetical relations to other numbers, and so on for elements of other domains.

This does not mean that the English word, ‘exist’ is ambiguous, and has different senses. It only means that we can use that word, with its single meaning, to make different kinds of claims. The univocal meaning of ‘exist’ is given by the logic of the existential quantifier. But the content of a claim that something exists varies, depending on the kind of thing that is in question. These are different ways of existing, not different senses of existence. The universe of all those things whose existence we are committed to is just the extension of the disjunction I mentioned earlier: the things that are either a physical object, or a number, or a set, or … To belong to the extension of this disjunction is not to have a further property of existence over and above what is attributed to a thing by a true existential statement within one or another of these domains. So nothing could be truly said to exist by a statement within one of these domains yet fail to exist in this more general ‘sense.’

This was what I had in mind in my remark that the truth values of statements about one domain, insofar as they do not conflict with statements of some other domain, are properly settled by the standards of the domain that they are about. There are, however, two problems with this way of putting the point. The first, which I have mentioned, is the ambiguity of ‘settled’ between an epistemological and a substantive reading. On the substantive reading the considerations that settle a question are the ones that make an answer to that question true or correct. On the epistemological reading the considerations that settle a question are the things we can rely on as evidence for the correctness of an answer.

I should have been clearer about this in my book, but the ambiguity does not seem to me to be a problem. The substantive reading is fundamental. Facts about the relations between numbers settle the matter (i.e. make it the case that 19 is prime.) Calculating (that is to say, carrying out the calculation) is a proper way to settling the matter in the epistemological sense insofar as this calculation is a way of coming to know these facts.

A second worry is that my statement that (conflicts aside) questions about a domain are properly settled (in either of these senses of ‘settled’) by the standards of that domain may seem to be trivially true. This appearance is enhanced by the fact that the claims in question are just the ‘pure’ claims within a domain, such as pure normative claims and pure mathematical claims. These are claims that have been ‘purified’ by eliminating any dependence on non-normative, or non-mathematical claims. In the case of normative claims of the form R(p, x, c, a) it involves eliminating dependence on whether p is in fact the case and whether the agent is in fact in circumstances c, by shifting to: ‘Were it to be the case that p and that x is in circumstances c, then p would be a reason for x to do a.’ So it would seem that pure claims, by definition, have no dependence on truths about other domains (or at least not about the natural world.)

What, then, was supposed to be excluded by my statement about existential claims about elements of a domain being correctly settled by the standards of those domains? What I meant to exclude was the idea, which I have just discussed, that claims about the existence of elements of a domain, although correct according to the standards of that domain, might nonetheless be false because those entities failed to exist in some broader sense.

In denying that existence claims of this broader kind have content, I am not denying that there is such a subject as ontology. My claim is rather than questions of ontology, and metaphysics more generally, are always domain-specific. Metaphysical questions about a domain are questions about its most abstract structure, including questions about its ontology – that is to say, about what its fundamental elements are. Thus, there are metaphysical, and ontological, questions about the natural world, about mathematics, and about the normative realm. The idea of metaphysics in a more general sense begins from the metaphysics of the natural world and leaves open the possibility that this universe may be extended. If being part of this extended universe does not just mean being part of the natural world, however, then the idea of this extended universe implies that even for objects in the natural world existing as part of this wider universe is a further fact beyond their existence as parts of the natural world – that is to say, their having spatio-temporal location, entering into causal relations, and so on. This seems implausible and unnecessary. The existence, or reality, of physical objects is fully accounted for by properties of the kind I have just listed, which are peculiar to the natural domain. If this is true of physical objects and facts about them, why should it not be true as well for entities and facts in other domains, such as mathematics and the normative, that their existence is a domain-specific matter?

I can now explain why it is misleading to say that sets and numbers, and normative facts, ‘exist in a non-ontological sense.’ This claim may seem to be true if by existing in an ontological sense we mean being part of a broader realm of the kind I have argued against. But on this account physical objects also would ‘exist in a non-ontological sense’ which seems, and is, absurd. My conclusion is that both physical objects and sets and numbers exist in an ontological sense if they are truly said to exist by statements that are valid according to the standards relevant to their respective domains (and their existence is consistent with valid claims in other domains).

Enoch and McPherson argue, however, that it is not easy to avoid appealing to criteria of existence that are broader than those of a particular domain. They ask, for example, whether there is, on my account, a ‘super domain,’ the domain of domains. If there is such a domain, they argue, it would involve criteria of existence broader than those of the individual domains that it deals with.

The domain of domains would be the subject matter dealt with by claims about which domains there are and about relations between them. Since I have just been making claims of this kind, I can hardly deny that there is such a subject matter. And my more general remarks about domains imply that there is such a domain, as long as the concept of a domain has content and is coherent. But recognizing such a domain of domains would not, in itself, have any implication that I need deny.

Claims within the domain of domains would be about domains, in particular about the existence of domains, but not about the existence of various things that these more specific domains involve claims about. We could, I suppose, take the claim, ‘anything exists if it exists according to the standards of some particular domain and its existence is not in conflict with the valid claims of any other domain’ to be a claim within the domain of domains. But recognizing a domain of domains does not require accepting this thesis. In addition, unless more were said about what kind of existence is being claimed for the elements of particular domains, all it would come to could be just the ‘disjunctive’ sense in which something exists ‘if it is an object in the natural world, or a number, or …’, the content of which is, as I have said, given entirely by the content of the domain-relative senses of existence involved in its disjuncts. If there were, contrary to what I maintain, contentful criteria of ‘existence’ that are broader than the domain-relative criteria I recognize, then a claim of the kind I have just mentioned, within the domain of domains, could be a claim about existence of this broader kind. But recognizing a domain of domains does not in itself give content to such criteria of existence.

We do not need general domain-independent standards to address questions about the relation between different domains, such as about whether claims about one domain, as we have most reason to understand them, entail or presuppose claims about other domains, and about what to do when there is such a conflict. Such questions can be addressed by reflecting on how the claims of individual domains are best understood and whether to accept these claims.

A theist, for example, might engage in this kind of reflection about the relation between her religious beliefs and other things that she believes.Footnote ⁴ Suppose she begins from a position of treating her theological beliefs as an autonomous domain, conclusions about which are defended or revised on the basis of interpretive judgments about scripture, various authoritative teachings, and perhaps her own religious experience. This mode of thinking might come under pressure from reflection of two kinds. On the one hand, if the theist is also a scientist, or at least accepts the methods and conclusions of science, there is a question of whether her idea of God, as a conscious being that is aware of her and other human beings and has effects on the world of some kinds, is compatible with her scientific beliefs about how the natural world operates and about how claims about it should be assessed. This may lead her to revise her idea of God and God’s relation with the natural world, or she may decide to live with the conflict, as many people seem to do.

On the other side, there is the question of whether her understanding of the normative significance of her religious beliefs, such as her ideas of God’s normative authority and God’s goodness, can be squared with her more general beliefs about practical reasons. (Various versions of the Euthyphro problem arise here.) Again, she may deal with this by reinterpreting her religious ideas, or by deciding it is so obvious that God’s will can be a fundamental source of reasons for action that she should revise her general understanding of normativity to allow for this.

It is, one might say, up to each person to decide what to do in such a case. There are, however, facts of the matter about what such a person has most reason to do, determined by the facts about the domains in question. I need not take a stand here on the question of what these facts are, although I naturally have views about these questions. My aim is just to describe the kind of reasoning that is involved in this process. It may seem to be inter-domain reasoning, since it aims to reconcile conflicts between claims in different domains. It is a process of seeking reflective equilibrium among all of our commitments, whatever domain they belong to. But the support for the particular commitments that are being reconciled, which needs to be reassessed in deciding whether and how to revise these commitments, is always specific to one or another particular domain.

It need not be apparent at the outset whether the concepts purporting to characterize a domain are in fact coherent. It could turn out that every attempt to work out the content of these concepts led to contradictions, in which we should conclude that we were mistaken to think that there was such a subject. Alternatively, it might turn out that there is no single way of characterizing the subject matter that seemed to be picked out by certain concepts: that these concepts were ambiguous, or did not distinguish between two quite distinct ideas. In such a case what initially appeared to be a single domain would turn out instead to be two closely related subjects. I said in my book, following a suggestion of Peter Koellner (Reference Koellner, Bueno and Linnebo2009), that this could turn out to be the case in regard to set theory (Scanlon Reference Scanlon2014, 78–79). But in order for this to be the case it would not be enough for there to be persistent disagreement among set theorists about some claims about sets (such as the axiom of choice, or the existence of large cardinals) even though they agreed about much else. It would also have to be the case that this disagreement was best explained by there being two fundamentally different concepts of set, which we had been confusing because they had some properties in common. Mere persistent unresolvable disagreement might support instead the conclusion that the realm of sets is in part indeterminate, and there is no right answer to certain questions.

Enoch and McPherson maintain in effect that a purely domain-specific idea of existence is too ‘thin’ to account for things that we want to say about normative truths. In support of this objection, they introduce the idea of ‘counter reasons.’ They ask us to imagine two linguistic communities that use different terms, ‘reasons’ and ‘schmeasons,’ to refer to things central to their normative practices. I assume this means that they each treat the things that they call ‘reasons,’ or ‘schmeasons,’ as counting in favor of actions and attitudes. But, Enoch and McPherson suppose, these two groups of people ‘tend toward quite different substantive views in their practices.’ I take this to mean that what members of one group treat as schmeasons is quite different from what members of the other group regard as reasons. Each group, they invite us to suppose, is perfectly coherent in their practices. The objection, if I understand it, is that I am committed to the view that these two groups of people are making claims about different domains, and hence that neither is in a position to claim that the other is mistaken. Enoch and McPherson maintain that an ontologically more robust account of normative truth is required in order to make sense of the idea that only the views of the reasoners, and not those of the schmeasoners ‘track the normative structure of reality.’

The terminology of reasons and schmeasons can be dropped without loss to the force of this objection. Since the people in these two groups take the things that they regard as reasons, or schmeasons, as counting in favor of certain actions and attitudes, and cite these things in justifications and criticisms, what we have here are just two groups of people who have differing views about which things are reasons. That is why they seem, despite their differing terms, to be disagreeing. They are thus, on the account I have offered, both making claims about the same subject matter – the normative domain.

I take the assumption that the views of both groups are coherent to mean that each group can cite a consistent account of reasons that fits their views. As I said in my book, in discussing reflective equilibrium, this does not rule out the possibility that one group or the other is mistaken. To decide whether this is so, we need to assess the steps through which their differing consistent views are attained. If each of these paths to equilibrium is equally sound, this would support the conclusion that there is no fact of the matter about the claims on which they disagree (Scanlon Reference Scanlon2014, 80). This possibility is the same on my view and on the more ‘robust’ view that Enoch and McPherson call Modest. On the latter view we would be concluding that ‘normative reality’ is indeterminate. But adding the word ‘reality’ makes no difference.

It might be objected that the disagreement between these two groups, which may be quite extensive, means that on my view they have views about different subject matters, of the kind I allowed for in discussing the possibility of set theoretic pluralism. But as I said in that context, pluralism of this kind is a plausible interpretation only if the disagreement is best explained in terms of some systematic ambiguity in the concepts in question. Disagreement alone is not enough. I don’t see that there is room for this kind of ambiguity about the idea of a reason, holding fixed that both groups are talking about what ‘counts in favor of’ actions or attitudes. Enoch does not explain what such an ambiguity might be. But, in any event, whether the disagreement is best understood in this pluralist way would depend on substantive interpretation of the content of the judgments involved. The possibility is not ruled out by adding the term ‘reality’ to our description of what they have views about. Grounds for thinking that these two groups of people are talking about different things would remain, in the form of grounds for taking them to be referring to (what we come to see as) two different normative realities.

In the course of discussing this objection, Enoch and McPherson say that my view is indistinguishable from a form of fictionalism. To explain why this is not the case, I will return to my brief remarks about fictionalism at the beginning of this paper. A fiction involves speaking of things of a certain kind as if they had properties appropriate to things of that kind, but with the presupposition that they are not in fact things of that kind with those properties. The kind of existence and the properties in question are domain specific. For example, a fiction about human beings talks about human beings that have properties such as spatio-temporal location, height and weight, beliefs and desires, with the presupposition that there are actually no things that have these properties.

The fact that the properties that are only fictionally attributed are domain-specific is what seems to rule out fictionalism about arithmetic. As I said above, if a person engages in counting, and in arithmetical calculations, and believes that there are correct and incorrect ways of doing these things, then there is no property of numbers left for that person to be a fictionalist about. Similarly, if a person regards various considerations as counting in favor of certain actions and attitudes, there is nothing left for that person to deny that would make him or her a fictionalist about reasons. (By contrast, someone might write ‘normative fiction,’ imagining that, say, the fact that a thing is blue would count in favor of treating in certain ways (the supposition being that being blue is not actually such a reason.) Such a person would not be a fictionalist about reasons in general, since what he or she is doing presupposes that there are facts about what is or is not a reason.

In any event, I am not a fictionalist about reasons. I believe that there are truths about what counts in favor of what. All I deny is that in order for there to be such truths they have to be part of ‘the world’ in some sense that is broader than the normative domain itself.

Epistemological problems

Enoch and McPherson say that the epistemological problem for normative realism is the problem of explaining why there should be any correlation between these facts and our beliefs about them. One possibility that they mention would be that normative facts are causally efficacious and we have some kind of causal interaction with them that can produce correct beliefs. Since I do not believe that normative facts are the kind of things that can have causal powers, this alternative is not available to me. If normative facts were entities of a special kind but did not have causal powers, it would be mysterious how we could acquire information about them. This is an alternative that I discuss and reject. There seems to me no reason to think that normative facts involve entities of this kind.

But, Enoch and McPherson say, this leaves the question what kind of explanation there could be of the correlation between normative facts and our beliefs about them, which we hope holds in cases of veridical normative belief. There are two things here that might need to be explained. The first is how we came to have the capacity to make judgments about normative questions, such as judgments about which aims are worth pursuing and what to do in the light of those aims. This is the difficulty raised by Sharon Street (Reference Street2006) and others, who hold that normative realism is incompatible with evolutionary explanations of our normative attitudes. For reasons set out by Selim Berker, I do not believe that this is an insuperable problem for realist views (Berker Reference Berker, D’Arms and Jacobson2014).

The second question is how to explain the fact that we (at least some times) form true beliefs about normative questions. I will take this to be the problem of explaining in what way our normative beliefs are responsive to the normative facts (assuming these facts are not causally efficacious). To answer this question it is helpful to begin from the recognition that our normative beliefs are often false. We can then think about how it is possible for us to correct these beliefs when we do. We do this by noticing that we held our earlier beliefs only because we were failing to make some significant distinction, or were failing to notice a relevant alternative. It seems to me that when we revise our beliefs in this way we are responding to normative facts by thinking about a normative question in the right way (just as when we correct our mistakes in arithmetic, by engaging in more careful reasoning about numbers). All I am claiming is that revisions of this kind are plausibly seen as improvements in our thinking that is responsive to the normative facts. This is not to say that the revised beliefs we arrive at are necessarily correct. Significant errors in our thinking may remain.

How successful we are in this kind of self-correction, and whether we even engage in it at all, depends on the conditions in which we are placed. It can, for example, depend on what others around us are generally thinking and saying about the matters in question, and on whether we have the time and inclination to think more carefully and reflectively about them. Coming to have correct normative beliefs is thus in no small part a matter of luck. This may mean that our normative beliefs are not ‘safe’ in the sense that Clarke-Doane discusses. But that’s life.

If we are capable of thinking about normative questions in the right way and we arrive at a normative belief by doing this (consciously or only implicitly), then it is no accident that our belief correlates with the normative facts. In other cases, of course, a correct belief may be merely a coincidence, if we believe it without considering the matter properly. In a common class of cases, we may arrive at a belief because we conform our normative beliefs to those of some other person or persons, whom we trust in such matters. If we have good reason to trust them, and their beliefs are not merely matters of coincidence, then our correct belief is not simply a matter of good luck either. If, on the other hand, we have no reason to trust them, or their belief is just a coincidence, then our belief as well is just a matter of luck.

Clarke-Doane distinguishes two forms of epistemological challenge to normative realism. The first is the justificatory challenge: to explain how our normative beliefs can be justified. The second he calls the reliability challenge: to explain the correlation between our normative beliefs and the normative facts. After some careful discussion, he interprets the latter challenge as the challenge to explain why our normative beliefs are safe: why we could not easily have false normative beliefs (using the method we have to form these beliefs) (9). He accepts that I have a plausible answer to the first of these challenges, but is concerned that I do not have an adequate response to the second.

Clarke-Doane considers the possibility that what he calls my ‘metaphysical pluralism’ – my permissive attitude regarding the existence of distinct domains – might provide such a response, which would proceed along the following lines. If our normative beliefs are coherent, then they mark out a domain, which according to my permissive standards exists, and these beliefs are true of this domain. As long as we could not ‘easily’ have had normative beliefs that were incoherent in the relevant sense, our normative beliefs could not easily have been false. So these beliefs are safe, in the sense that Clarke-Doane has in mind. But this response may, depending on how the relevant idea of ‘coherence’ is understood, either fail, or entail an unacceptable lack of objectivity in our normative beliefs.

Ideas that might be called coherence (although I would not use that term for both) play two distinct roles in the view I defend in my book. The first role is that of a sufficient condition for the existence on an abstract domain. As indicated above, I hold that if a domain is specified by a set of concepts in a coherent way, and does not involve claims that conflict with those in any other domain, then there is such a domain. The relevant notion of coherence is in this case a minimal one: logical and conceptual consistency. The second role is that of a standard that we aim to meet in trying to form justified beliefs about a domain. As Clarke-Doane recognizes, this is a more demanding standard. In my view, the method to be used in forming normative beliefs is, ultimately, the method of reflective equilibrium. This method is sometimes described as a process of seeking coherence in our beliefs. But as I explain in Chapter 4 of my book, this term is misleading (Scanlon Reference Scanlon2014, 79–80). What is sought in striving for reflective equilibrium is neither mere logical consistency nor consistency with some privileged set of basic beliefs. For one thing, there is no set of privileged beliefs. All of our beliefs are open for reconsideration. For another, as Clarke-Doane observes, coherence in either of these senses would be too easily achieved.

The question one must ask in carrying out the method of reflective equilibrium is not, ‘which alternative, modifying my considered judgments or looking for different principles to account for them, would be more coherent?’ The question is, rather, a substantive question of relative plausibility: ‘How plausible are the particular judgments we have been accepting, given what we now see to be the more general principles that would be required to account for them?’ What we strive to do in seeking reflective equilibrium is to carry out a process in which all of these judgments are made in the right way. This is, of course, an extremely high standard. In order for our belief to be justified, in the sense that figures, for example, in the idea that knowledge is justified true belief, it is sufficient to meet a lesser standard. A person’s belief is justified in this sense if there is no reason for questioning it (for making a different choice in the pursuit of equilibrium) that that person should have been able to see, given his or her circumstances. Obviously, a belief can be justified in this sense even though it does not correspond to the facts about the subject in question.

Let me now bring this clarification to bear on Clarke-Doane’s discussion of pluralism and the safety of our normative beliefs. First, two sets of normative beliefs that are minimally coherent may characterize two distinct domains, each set being true of the domain it characterizes. But as I have pointed out above, in responding to Enoch and McPherson, it does not follow that these are two sets of true normative claims – claims about the domain of reasons. I thus agree with Clarke-Doane that the view he calls radical pluralism, and even his moderate pluralism, would deprive our normative views of the objectivity we take them to have. But neither form of pluralism follows from my views about ontology.

Normative claims are more like claims about geometry, understood as claims about the physical world rather than as ‘pure mathematical conjectures’ (Clarke-Doane Reference Clarke-Doane2017, 13). The same is true, I would say, of claims about sets. We should distinguish here between the possibility that coherent sets of claims can define distinct domains that are, as it were, alternatives to set theory, and the kind of ‘bifurcation’ of the domain of sets that I discussed in my book, and which Clarke-Doane mentions. This was the possibility that we might conclude that what we had taken to be a single subject matter, the realm of sets, was in fact two distinct, similar and perhaps overlapping subjects. But in order for such a conclusion to be justified, it is not sufficient that there be distinct bodies of ‘set theoretic’ statements that are internally consistent and even overlap on some basic claims. It is necessary, in addition, first, that the steps of seeking reflective equilibrium that lead to these two sets of claims are, as far as we can judge, equally sound. This might, however, merely support the conclusion that the facts about sets were not entirely determinate – that there are some questions about sets about which there is no single right answer. In order to support the different conclusion of ‘bifurcation,’ it must also be the case that the two divergent paths to equilibrium reveal some systematic ambiguity in the notion of set that we have been dealing with. As I have said above, whatever may be the case with set theory, this possibility strikes me as very unlikely in the case of the normative domain.

The upshot of this discussion is that I am grasping the second horn of the dilemma that Clarke-Doane poses: our normative beliefs are not all that safe, because they could quite easily be mistaken, and often are.

The phenomenon known as supervenience

The comments by David Enoch and Tristram McPherson, and especially those by Gideon Rosen, make me realize that in my book I may have mischaracterized my own view about the phenomenon of supervenience, by not being sufficiently clear about the relation between my view and what Rosen calls ‘the received view’ of the matter. Let me review the situation.

Common sense holds that the following two things are true. First, normative facts can vary when non-normative facts vary. This alone may seem to call for some explanation, given that normative statements are not logically or conceptually tied to non-normative ones. Second, the normative facts do not vary ‘on their own:’ they are the same as long as all the non-normative facts remain the same.

The ‘received view’ in metaethics is that these two features of normative facts are metaphysically necessary (I doubt that this is so according to common sense, since metaphysical necessity does not seem to me a common sense notion). As Rosen puts it: ‘If two metaphysically possible things are alike in every non-normative respect, they are alike in every normative respect’ (Rosen Reference Rosen2017, 4). In addition, the received view is that the connection is one between two different realms of facts, the normative and the non-normative, and that this is part of what makes it puzzling. In Rosen’s words, ‘If the normative facts contain an irreducible normative relation – the reasons relation – then to regard such principles as necessary is to affirm a necessary connection between facts of one sort and facts of an entirely different sort. And many philosophers have found it puzzling how there could be such a connection’ (Rosen Reference Rosen2017, 5).

In my book, I took issue with the latter aspect of the received view – that the connection is between two realms of facts of quite different sorts – and I claimed that this change made the connection less puzzling. Enoch and McPherson question this claim.Footnote ⁵ Rosen does not mention it, but some of the things he says seem to cast doubt on it. So I will revisit the matter.

The principles that Rosen is referring to in the passage I just quoted are what I called pure normative claims. I introduced the idea of such claims as a way of addressing a slightly different question about the relation between normative and non-normative truths. My aim was to clarify the sense in which it is true that there is a ‘gap’ between normative and non-normative claims. Many normative claims are in fact entailed by non-normative claims. This is true even of very simple claims such as that some fact p is a reason to do some action a. The fact that p cannot be a reason to a unless it is a fact that p. So the denial of the non-normative claim that p entails the denial of the normative claim that p is a reason. The only normative claims that are independent of non-normative claims are what I called pure normative claims, claims that have been, so to speak, purged of their non-normative content and presuppositions. These claims isolate the essentially normative content of mixed normative claims such as the claim that p is a reason to do a.

Pure normative claims have two features. The first, which I have just been discussing, is their ‘purity’ – their lack of dependence on any non-normative facts. The second feature, which they share with mixed normative claims, is that, as Rosen puts it, they ‘assign normative features to situations on the basis of their non-normative features’ (Rosen Reference Rosen2017, 6). It is this second feature that is most relevant to the current question about the relation between normative and non-normative facts.

The covariance between normative truths and non-normative truths is most likely to seem puzzling if we think of normative truths of the form ‘x ought to be done,’ ‘x is good’ ‘x is wrong’ or, in Enoch and McPherson’s example, ‘x is obligatory.’ It is, as I said, common sense that the truth values of such claims depend on, and therefore co-vary with, non-normative properties of x. Things are good in virtue of other properties that they have, and actions ought to be done, or are right or wrong, in virtue of other features of these actions, such as the consequences of performing them or not performing them. But this dependence is not made explicit in claims of the kind I have just listed. Insofar as it is not explicit, the covariance of their truth values with non-normative facts may appear to be a puzzling connection between facts of two ‘entirely different sorts.’

But this appearance should not be puzzling. Given that, as Rosen says, normative claims ‘assign normative features to situations on the basis of their non-normative features,’ why should it be puzzling that the truth values of such claims depend on whether the non-normative features to which they assign normative significance actually obtain? This is true even of claims such as ‘x is good’ or ‘x is obligatory’ in which this dependence on non-normative facts is left implicit. It is more obvious in the case of normative claims that explicitly assign normative features to certain non-normative facts, by specifying which properties make a certain action wrong, or make some state of affairs good.

The truth values of most normative claims – the ones that I called ‘mixed normative claims’ – depend on the truth of non-normative claims and thus vary when the truth values of the latter vary. Insofar as non-normative claims and mixed normative claims are claims of different kinds, it is true, in a way, that the covariance of these normative truths with non-normative truths is a connection between facts of two different sorts. But when we look more closely at the ‘sort’ of facts that normative facts are, namely that they are facts about the normative significance of certain non-normative facts, this connection should not seem puzzling. Moreover, it is misleading to say that the principles specifying this dependence are ‘bridge principles’ linking the normative and non-normative domains. Rather, they are themselves normative claims of the most common kind.

Thus, insofar as it is a part of the ‘received view’ of supervenience that it involves such a connection between two different domains, it may have been misleading for me to say that I was offering an explanation of supervenience. Perhaps I should have said instead that I was offering an explanation of the phenomenon formerly, but misleadingly, known as the supervenience of the normative on the non-normative. To put the matter as a relation between the two domains is misleading in two ways. First, the way in which normative facts depend on non-normative facts is a general feature of the normative domain itself, specified by the truth of central normative claims. Second, it is not true that the normative domain as a whole supervenes on the non-normative. Some normative truths, namely pure normative truths, do not depend on, vary with, or supervene on non-normative truths.

A second feature of the received view is that the connection in question is necessary, and it may be that I made a second mistake in venturing an opinion about the kind of necessity involved. I could simply have rested with the point that the ‘connection’ in question is specified by pure normative truths, and left it open whether these pure normative truths are necessary and if so what kind of necessity they have.

The nature of this necessity is the question that mainly occupies Rosen in his careful discussion. He is correct that I was not sufficiently clear about the difference between normative and metaphysical necessity. The choice he offers me is between holding that pure normative truths are metaphysically necessary – that is to say, they hold in all metaphysically possible worlds – or that they are only normatively necessary – holding in all normatively possible worlds but not in some worlds that are metaphysically possible. I confess that I find it difficult to conceive the possibility of a world in which the pure normative facts are different from those in our world. It is difficult to conceive, for example, of a world that was like our world in all non-normative facts but in which pain was something we had a reason to seek rather than a reason to avoid.

Rosen says that in order to maintain that a world of this kind is not possible, I would need to explain how it is ruled out by the nature of the reason relation itself. As I said above, it does seem to be that it is ‘in the nature of’ goodness, rightness, wrongness, and obligatoriness’ that these properties obtain in virtue of other properties of the things or actions involved. I am even inclined to say that it is in the nature of these properties that they depend on facts about the reasons that individuals have in virtue of these more particular properties. If so, then these connections are metaphysically necessary in the sense that Rosen specifies.

I might even say that it is ‘in the nature of the reason relation’ that particular facts are reasons for certain actions in virtue of non-normative properties of those facts. The question, then, is whether it is ‘in the nature of the reason relation’ exactly which properties of facts are reason-giving. When I said, in my book, that I did not believe that the property of being a reason had any further explanation beyond that given by the concept of the reason relation, what I meant to exclude was a further explanation of the normativity of the reason relation, not an elucidation of its content – of the things to which it applies – which is what is now at issue. That said however, I do not see how the nature of what it is to be a reason could be filled out in a way that would specify its content in this way. So I do not see how to account for the metaphysical necessity of pure normative truths about reasons in this way even though, as I said, I cannot conceive how worlds could differ in these basic normative facts.

My aim, however, was to account for what I took to be common sense truths about normative truth, not to show that normative truths supervene on non-normative truths in the way that the received view holds. I have offered a description of the relation between normative and non-normative claims in the light of which these common sense truths are not puzzling.Footnote ⁶ Whether supervenience in the form understood by the received view holds or not I leave to others.

I will conclude by responding to Enoch and McPherson’s question (Enoch and McPherson Reference Enoch and McPherson2017, 20) about the reasons I offer in support of my view. First, I take normative realism to be supported by common sense and to fit well with the roles that claims about reasons have in our lives. I defend this view mainly by rebutting the main reasons that philosophers have offered for rejecting it, reasons based on metaphysical concerns, on the alleged inability of cognitive states to motivate agents, and on doubts about how we could come to know normative facts on a realist interpretation. In addition, as Enoch and McPherson note, I argue against some alternative accounts, such as a reductive desire theory, an expressivist theory, and a constructivist account of the normative domain as a whole. I hope that what I have said above makes clear why I also believe that fictionalism about normative or mathematical truth should be rejected and that there is nothing to be added by a more ‘robust’ form of normative realism.