I

Let’s begin by restating some key theses. Reasons Fundamentalism, as Scanlon calls it, is the claim that the facts about reasons ‘are not reducible to or identifiable with non-normative truths, such as truths about the natural world of physical objects …’ (Scanlon Reference Scanlon2014, 2). As Scanlon makes plain, this claim of irreducibility is not the bland semantic claim that statements of the form ‘p is a reason for X to do A’ are not synonymous with statements in a non-normative idiom. It is not a claim about words concepts. It is a claim about the reason relation, the worldly feature we talk about when we talk about reasons. The Fundamentalist’s key claim is that there is no way to say in more fundamental terms what it is for p to be a reason for X to do A, in the sense in which there is a way to say in more fundamental terms what it is for a rock to have a certain density, or for a function to be continuous, or for a thing to be made of gold. As I would put the point, the claim is that unlike density, continuity and gold, the reason relation does not admit of real definition (Rosen Reference Rosen2015).

Some writers give general arguments for the irreducibility of normative properties and relations, most of which are ultimately versions of Moore’s Open Question Argument.Footnote ¹ Scanlon does not give a general argument, though he objects to particular reductive proposals on the merits, and I think that’s wise. For all we know some philosopher toiling in obscurity has just produced a counterexample-proof equation of the form:

where ϕ is entirely non-normative and facts of the form ϕ(p, x, a, c) have precisely the same ‘significance’ we normally attach to facts about reasons. Of course it is somewhat unclear what it takes for a principle of this sort to count as a reduction. According to me, the account is reductive provided it lies in the nature of the relation R that whenever R(p, x, a, c) holds in some particular case, this fact is fully grounded in the fact that ϕ(p, x, a, c) (Rosen Reference Rosen2015). But however we understand this notion, we cannot rule out the possibility of a reductive proposal that will lead us all to say, ‘So that’s what it is for p to be reason for x to do a in c!’ And if that’s right then Reasons Fundamentalism can’t be established in advance. It’s rather a bet, based on long experience, that no reduction of this sort is possible.

II

Turning now from the reason relation to truths involving it, we note that some facts about reasons are contingent on the non-normative facts. If I’ve got a headache and the pill on the table is an aspirin, then I’ve got a reason to take the pill. But if I didn’t have a headache, or if the pill were laced with cyanide, I would not have such a reason. When a normative fact is contingent on the non-normative facts in this way, Scanlon calls it mixed. Scanlon defends a view about how these mixed normative facts are to be explained that will be consequential in what follows. According to this view, whenever a mixed normative claim is true, it is backed up by – I would say, grounded in – the non-normative facts upon which it is contingent, together with a pure normative truth: a general principle about reasons that is not contingent on the non-normative facts in any sense. In this case, the relevant pure principle might be:

The claim is that whenever a mixed fact about reasons obtains (e.g. the fact that I have a reason to take the pill), it obtains in part because the non-normative facts are thus and so, and in part because some pure normative principle like this obtains.

This aspect of Scanlon’s view is not trivial. There is room in logical space for a particularist view according to which the fundamental normative truths concern the reasons had by particular agents in particular circumstances, many of which will be contingent. On this view, such general principles as there are will be grounded in these particular normative facts, rather than the other way around. This dispute between Scanlon and the (imagined) particularist is analogous to a familiar dispute about laws of nature, in which so-called ‘humeans’ say that laws are mere patterns in the facts, while ‘non-humeans’ hold that laws are explanatorily prior to their instances. In this analogy, Scanlon is the non-humean. He thinks that contingent particular facts about reasons are always grounded in contingent non-normative facts together with pure general principles that are not themselves grounded in facts about particulars.

Given a mixed fact we have a recipe for generating the pure principles that figure in its explanation. Start with the contingent normative fact R(p, x, c, a) and list the non-normative facts upon which it depends for its explanation. These will be facts about the agent, her circumstances, the actions open to her, and so on, all of which we may package as a single non-normative claim ψ(p, x, a, c). The corresponding pure normative principle is then the universally quantified conditional:

This general truth is not contingent on the non-normative facts, since every relevant condition has been bundled into the antecedent. Scanlon’s view, as I understand it, is that mixed normative facts are always grounded in pure normative principles of this sort together with contingent non-normative facts.

Scanlon doesn’t stress the point, but it’s worth noting that pure normative principles need not be fundamental or inexplicable. Suppose it’s a pure normative principle that:

1. (1) Whenever A will relieve X’s pain, this is a reason for X to do A.

And suppose it’s also true that

1. (2) Whenever A will relieve X’s nausea, this is a reason for X to do A.

We might then entertain the hypothesis that these pure principles are explained by a more fundamental principle, for instance:

1. (3) Whenever A will relieve a disagreeable sensation, this is a reason for X to do A.

Since pain and nausea are (let’s suppose) disagreeable by their very nature, (1) and (2) follow from (3), not in virtue of any contingent truth, but as a matter of absolute necessity. So it may well be that even though (1) and (2) are pure, in the sense that they don’t depend on contingent non-normative truths for their explanation, they are nonetheless non-basic in the explanatory order.

This grounding of pure normative principles in more fundamental principles can’t go on forever, or so we normally suppose. We may explain P in terms of Q, and Q in terms of R. But as we press the demand for explanation we will eventually reach pure normative truths that cannot be explained. If (3) is basic in this sense, there will be no answer to the question, ‘Why is the fact that a sensation is disagreeable a reason to relieve it?’ Of course there may be a great deal to be said in support of (3). We may have abundant and articulable grounds for believing (3), as we will if it is supported by a reflective equilibrium argument that begins from considered judgments like (1) and (2). But these epistemic grounds for believing (3) do not explain why (3) is true, just as the evidence for Newton’s law of gravitation does not explain why Newton’s law obtains. So let’s focus on these explanatorily fundamental normative principles. The main question I want to press concerns their modal status.

III

The received view is that pure normative principles are necessary truths. If it’s true that we have reason to relieve disagreeable sensations, and if this fact is not contingent on the non-normative facts, then according to the received view, it’s not contingent at all.

Given Scanlonian assumptions, this received view is equivalent to:

To see the equivalence, suppose that Jones and Smith are alike in every non-normative respect and that some normative proposition N(Jones) is true of Jones. On Scanlon’s account, this fact is grounded in some non-normative fact about Jones, D(Jones), together with a pure normative principle to the effect that ∀x (D(x) → N(x)). Since Jones and Smith are alike in every non-normative respect, we know that D(Smith). So if the pure normative principle is a necessary truth, it follows that N(Smith), which is to say: it follows that Jones and Smith are alike in every normative respect. On the other hand, if the pure principle is a mere contingent truth, there will be cases in which Jones and Smith are alike in every non-normative respect but Jones inhabits a world where the principle is true while Smith inhabits a world in which it is false (and where Smith himself is a counterexample to it). So if the pure normative principle is contingent there will be cases in which Jones and Smith are alike in non-normative respects but different in some normative respect, so supervenience will fail.

It is important to stress that the received view is a claim about absolute or metaphysical necessity: the sort of necessity that attaches to non-indexical logical truths (If p then p), conceptual truths (Vixens are foxes), the truths of pure mathematics as standardly understood, and also to truths grounded in the essences of individuals (Socrates is human) and kinds (Lions are animals). The received view holds that pure normative principles are necessary in that sense – that just as there are no possible worlds in which water is an element, so there are no possible worlds in which the pure normative principles are different from what they are.

The problem for the non-naturalist who accepts this received view has always been to explain this necessary connection between the non-normative facts and the normative facts they fix (MacPherson Reference MacPherson2012). On a view like Scanlon’s according to which the mixed normative facts are grounded in contingent non-normative facts together with pure general principles, the challenge reduces to that of explaining why the principles are necessary. But to reduce the challenge in this way is not to answer it. The principles, recall, are generalized conditionals that say: if the non-normative facts are thus and so, then some normative fact obtains. If the normative facts contain an irreducible normative relation – the reason 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.

Suppose McQ says that colors are sui generis, irreducible properties of bodies whose natures are exhausted by the way they look, adding that when a thing is (say) canary yellow, this is always thanks to its non-chromatic features together with a law that says: if a thing has non-chromatic features C, it’s canary yellow. We then ask for the modal status of this law. Are there possible worlds in which these same non-chromatic features are paired with another color, or with no color at all? Suppose McQ says, ‘No. The color facts supervene on the non-chromatic facts, so the chromatic laws are absolutely necessary. You can describe coherent alternatives to the actual laws, and your alternative may be consistent with the natures of the colors. But the actual laws could not have been otherwise, so these coherent alternatives are impossible. Non-chromatic feature C is yoked to canary yellow with an iron yoke, not just in this world, but in every world.’ This view prompts a good question. How could the chromatic laws possibly have this status? If we can coherently describe a world in which the chromatic laws are different, and if the natures of the colors don’t encode these laws, why couldn’t God have made such a world? It is widely believed that the basic laws of physics could have been otherwise. Why should the chromatic laws be different?

The same question arises for Scanlonian pure principles, which amount to laws that assign normative features to situations on the basis of their non-normative features. The traditional supervenience problem for moral realism is to explain the necessity of these bridge laws (Blackburn Reference Blackburn1993; MacPherson Reference MacPherson2012). The problem arises most acutely for the non-naturalist because his position blocks the usual strategies for explaining why a truth is necessary. In all of the clear cases, a truth is necessary when it is logically entailed by one or more essential truths: truths about the essences or natures of the various objects, properties and relations with which it is concerned (Fine Reference Fine1994). Why is it a necessary truth that water is a compound? Because it lies in the nature of water to be H₂O, from which it follows that water is a compound. Why is it a necessary truth that whenever S knows that p, p is true? Because it lies in the nature of knowledge that whatever is known is true. The color primitivist cannot explain the necessity of the chromatic laws in this way. His position precisely that it does not lie in the nature of canary yellow to be associated with any particular underlying feature. That’s why his view is so deeply problematic. As we will see, the Reasons Fundamentalist faces a similar predicament.

IV

Scanlon appears to endorse the received view of the modal status of normative principles.

I have elided some material in this passage that points to another reading – I’ll restore it shortly. But the most natural way to understand this passage is as defense of the received view that the normative supervenes on the non-normative as a matter of absolute necessity. The explanation proceeds by pointing out that this supervenience claim is equivalent to the thesis that the pure normative truths are necessary, and then explaining their necessity by noting that since they are not contingent on facts about the natural world, they are therefore, intelligibly, non-contingent in every sense.

The crucial move is the last one, and it strikes me as highly dubious. Scanlon is right to insist that the pure normative truths are not contingent on non-normative truths. This follows almost immediately from the definition of a pure normative truth as a truth that is not grounded in, or explained by, contingent non-normative truths. But it does not follow from the fact that p is not contingent on more fundamental truths that p is absolutely necessary – that there is no possible world in which p is false. The best examples to illustrate the point are basic laws of nature, understood not as humean patterns in the non-nomic facts, but as physical principles that are prior to and explanatory of such patterns. The basic laws of physics are inexplicable. That’s why they’re basic. This gives a sense in which they are not contingent on more basic truths. But it certainly doesn’t follow from this that they are necessary. The picture is fully consistent with there being genuinely possible worlds with different laws.Footnote ³ Other examples to make the same point include the initial state of a universe with a beginning in time, or the inventory of particles in a Democritean world in which atoms are neither created nor destroyed. Suppose we have a world of this sort in which particle A exists. The fact that A exists is not contingent on more fundamental facts, but it’s still contingent.

Scanlon’s dialectical opponent at this stage is a philosopher who thinks that the pure normative principles posited by the non-naturalist look very much like laws of nature. Some may be explained by others; but the basic normative principles are brute synthetic facts with no explanation. Since alternatives to these facts are readily conceivable and thoroughly consistent with the nature of the reason relation (as we will see), this opponent wants to know why these alternatives should not be reckoned possible. In this context it’s not enough to point out that the basic normative principles are not contingent on other facts. That’s just to note that these facts are brute; and brute facts can obviously vary from world to world.Footnote ⁴

V

Scanlon says in a footnote that on his view,

The analogy is telling. To focus ideas, let the ‘mixed’ mathematical facts be the facts about impure sets, e.g. the fact that there exists a set containing just Jones and Smith. This is a contingent fact, since the set would not have existed if Jones or Smith had not existed. The mixed fact is explained by the existence of Jones and Smith together with a pure mathematical principle, the Pair Set Axiom, which says that whenever x and y exist there is a set containing just x and y. The received view is that the inventory of sets supervenes on the inventory of individuals, so that if Jones and Smith exist, their pair set must exist. In this context, that is equivalent to saying that the pure principles in the vicinity, including the Pair Set Axiom, are necessary truths. The challenge, analogous to the supervenience problem for ethical non-naturalism, is to explain the necessity of this principle. And that turns out to be non-trivial. I have argued elsewhere that given non-crazy views about the nature of necessity, it cannot be done (Rosen Reference Rosen and MacBride2006). On this view, some pure mathematical truths, including the existence axioms of set theory, do vary from world to world. If this is right, then the conventional wisdom about the modal status of mathematics is mistaken. In particular, worlds with no sets whatsoever are possible, holding the facts about non-sets fixed, so the supervenience of the sets on the non-sets fails with gusto.

The most straightforward way to resist this odd view is to hold that the elementary existence principles of set theory are built in to the nature of set – that just as it lies in the nature of water to be a compound, so it lies in the nature of set that whenever x and y exist, so does {x, y}. If this is right we have an explanation for the necessity of the Pair Set Axiom and for the supervenience of pair sets on individuals. The axiom is necessary because it flows directly from an account of what sets are, according to which sets by their very nature conform to the Pair Set Axiom.Footnote ⁵

The analogous move in the normative case would be to say that the pure normative principles are somehow built in to the nature of the reason relation; – that just as it lies in the nature of water to be a compound, so it lies in the nature of reason that we have reason to relieve disagreeable sensations, and likewise for every other basic normative principle.

This is a possible view. But it’s not a plausible view in my opinion; and more importantly, it is certainly not Scanlon’s view. One way to bring this out is to imagine a case of maximally informed and reflective disagreement about pure principles. Contrast the Altruist who think that one has some reason to do A whenever A would relieve pain in oneself or in another, with the Egoist who thinks that each of us has reason to attend to his own pains, but no reason, in general, to bother with the pain of others. Since the Altruist is right about this, a proponent of the view under consideration must think that not only is the Egoist substantively mistaken about what we have reason to do; he is mistaken (or ignorant) about what it is for a fact to be a reason. Of course the claim is not that the Egoist fails to understand the word ‘reason’. He may be a competent speaker of the language. The point is rather that on the present view, the Egoist’s false view about a normative principle entails that he does not know what it is for a fact to be reason in the same sense in which someone who thinks that water is an element doesn’t know what it is to be water.

Now this is not an absurd claim. Reductive naturalists believe that reason, like water, has a hidden nature about which it is easy to be ignorant. For the reductive naturalist, this hidden nature encodes a single principle:

where ϕ is entirely non-normative. This amounts to a real definition of reason in wholly naturalistic terms. And if there is such a definition, anyone who fails to know it thereby fails to know what a reason is in the intended sense. Anyone who takes this view can gladly say that the Egoist’s substantive mistake entails a mistake about what it is for a fact to be a reason.

The trouble is that Scanlon explicitly denies that reason has this sort of hidden nature. Let’s agree that the Egoist is fully competent with the concept of a reason. He understands the word, and not just in the degenerate sense in which a non-expert counts as understanding a technical term when she uses it to mean whatever the experts mean. Since the Egoist denies the Altruist’s principle – that we have reason to relieve the pain of others – it follows that this principle is not a conceptual truth. Now according to Scanlon, properties and relations are given to us by means of concepts. When we think about a property, we think about it in a certain way. But there are two relations in which a property can stand to the concepts by means of which we think our thoughts about it:

Scanlon gives natural kind concepts like water and certain ethical concepts, like moral permissibility, as examples of concepts where there is more to be said about the corresponding property than is encoded in the concept itself. But he explicitly denies that the property of being a reason has such hidden depths:

This means that whatever the merits of the view may be, it is not open to Scanlon to explain the necessity of pure normative principles by deriving them from a non-obvious account of the nature of the reason relation. But if the necessity of these principles can’t be explained in this way, it’s quite unclear how it can be explained.

VI

It’s time to restore the elided material from the long passage quoted in §IV, since as was noted, it points in a rather different direction:

The footnote suggests a view on which pure normative facts are not metaphysically necessary, but rather merely normatively necessary. What does this mean? It would be a defensible terminological choice to use ‘p is normatively necessary’ to meant that p is an absolutely necessary truth with normative content, or an absolutely necessary truth that holds for normative reasons. But that’s not what Fine has in mind. On Fine’s view, normative necessity is a sui generis variety of necessity – a distinct modality on a par with metaphysical necessity and also with what Fine calls ‘natural necessity’. Metaphysical necessity is the sort of necessity we have been discussing until now: the full strength necessity that characterizes the truths of logic and mathematics, truths about the essences of individuals and kinds, and so on: the sort of necessity with which ‘modal metaphysicians’ like Kripke, Lewis and Williamson are centrally concerned. Natural necessity is the necessity that characterizes the laws of nature, and as we normally think, it is a weaker form of necessity. It is naturally necessary that massive bodies attract one another with a force give by (say) Newton’s law; but there are perfectly good possible worlds in which the laws are different and massive bodies do something else. Normative necessity, as Fine understands it, is analogous to natural necessity in this respect. It is given as the distinctive modality that characterizes pure normative principles, like the fact that we have reason to relieve the pain of others, and also the ‘supervenience conditionals’: statements of the form ‘If D(A) then N(A)’ where D(A) is the complete non-normative truth about some action A and N(A) is some normative fact about A.

Conventional wisdom holds that normative necessity so characterized just is metaphysical necessity – that when p is normatively necessary, p simply could not have been otherwise. Fine denies this. On his view, if it is a pure normative principle that we have reason to relieve the pain of others, then unless this case is somehow special there will be genuine metaphysically possible worlds in which this principle is false (just as there are possible worlds in which the laws of nature are different). Indeed there may be possible worlds that resemble ours in every non-normative respect in which the fact that you are in pain gives me no reason to help you out. According to Fine, pure normative principles are normatively necessary but metaphysically contingent.

On a view of this sort, the supervenience problem as traditionally conceived does not arise. The traditional problem is to explain why the normative facts do not ‘vary on their own,’ or in other words, to explain why, if the normative facts are distinct from the natural facts, the latter should fix the former as a matter of absolute necessity. If metaphysical supervenience is false, as Fine maintains, the explanandum evaporates and the problem disappears.

The mere existence of this view is important, if only as an antidote to the common view that metaphysical supervenience is a conceptual truth, or simply a truth that no reasonable person can deny. Fine denies it, as do I; and Fine at least is clearly reasonable. It is also an antidote to Blackburn’s claim, endorsed by Scanlon, that supervenience functions as a constraint on competent moralizing without which ‘nothing recognizably ethical could be approached at all’ (Blackburn Reference Blackburn1993, 146, quoted in Scanlon Reference Scanlon2014, 41). The Finean can agree with common sense about what we in fact have reason to do, and about what we would have reason to do if the non-normative facts were different. In particular he can agree that if we were to encounter a situation just like some given situation in non-normative respects, we would have the same reasons in both cases. Counterfactuals about the reasons we would have if the non-normative facts were thus and so always invite us to consider the nearest worlds in which the relevant non-normative facts obtain. Since the nearest such worlds involve no gratuitous departures from actuality, they will agree with the actual world about pure normative truths that mediate the connection between the non-normative and the normative. This is enough to justify holding the actual principles fixed in our reasoning about what we would have reason to do if the facts were different, and so to satisfy the demands of common sense and moral theory. In these respects, the view is orthodox. Fine’s distinctive claim is that in addition to this ‘inner sphere’ of worlds in which the actual normative laws are held fixed, there are also remote and yet still genuinely possible worlds in which the laws are different.

There is thus a tension between Scanlon’s endorsement of Fine’s view in the footnote and his claim in the text that the pure normative truths ‘do not vary on their own’, since on Fine’s view they do vary; they are contingent. Of course they are not contingent on the non-normative facts. There are no true counterfactuals of the form, ‘If the non-normative facts had been different in such and such a way, the pure normative truths would have been different’. Still they are contingent in the standard sense: they could have been different from what they are.

VII

Let me briefly sketch a broadly Finean view that Scanlon might adopt, and which in my opinion he should adopt given his non-naturalism. The first plank in the platform is Fine’s account of metaphysical necessity, according to which for p to hold of metaphysical necessity just is for p to be an essential truth: a logical consequence of the natures or essences of things (Fine, Reference Fine1994). It is metaphysically necessary that water is a compound because it lies in the nature of water to be H₂O, which entails that water is a compound. It’s metaphysically necessary that 3 is greater than 2 because like in the nature of 3 to be 2 plus 1, and in the nature of greater than that n + 1 is greater than n (when n is finite), etc. Pure normative truths are metaphysically necessary on this account only if it is built in to the nature of the reason relation, perhaps together with the natures of other items, that these particular truths should hold. But if reason has no hidden nature – the second plank – then the substantive synthetic principles are not encoded in its nature. Scanlon’s ‘no hidden depths’ thesis, together with Fine’s account of metaphysical necessity, thus entails that many pure normative truths are metaphysically contingent.

Such principles are, however, normatively necessary. Fine himself does not define this notion, though it seems to me that it can be defined (Rosen, Reference Rosen and Dumitruforthcoming). For p to be normatively necessary, on my account, just is for p to be a true proposition that would still have been true no matter how the non-normative facts had been.Footnote ⁶ Truths that have this status are modally resilient in the following sense: they would still have been true no matter how hard we had tried to falsify them, no matter what we had done or thought, no matter how the contingent history of the natural world had unfolded, and so on. This modal resilience, I claim, amounts to a perfectly good species of necessity.

That the pure normative truths are normatively necessary in this sense follows almost immediately from their purity. To say that p is pure is to say that p is a normative truth that does not depend for its explanation on any contingent non-normative fact. It can be shown given plausible assumptions connecting explanation and counterfactuals that when p is pure in this sense, it is counterfactually independent of the non-normative facts, which is just to say that p is normatively necessary in the sense just given (Rosen, Reference Rosen and Dumitruforthcoming §7).

On this view, the truth in the vicinity of supervenience may be framed as follows. Each world w is associated a class of worlds that are normatively possible relative to w, where w* is normatively possible relative to w iff every proposition that is normatively necessary at w holds at w*. Considered as an unrestricted claim about all possible situations, supervenience is false. Considered as a restricted claim about normatively possible situations, it is true:

Normative Supervenience:

This is a restricted supervenience thesis. It says, in effect, that when two situations have the same normative laws, they are normative duplicates if they are non-normative duplicates. Normative Supervenience follows immediately from a thesis that Scanlon accepts, and which I accept as well., viz., that it lies in the nature of the reason relation that contingent truths about reasons are always grounded in contingent non-normative facts together with pure normative principles.

Normative Supervenience is clearly much weaker than the metaphysical supervenience thesis that metaethicists have found self-evident. The chief metaphysical difficulty for non-naturalism has always been to explain that thesis. The present proposal, which I offer Scanlon in a spirit of non-naturalist solidarity, is to concede that this cannot be done and to embrace Normative Supervenience as a replacement thesis.

VIII

If metaphysical supervenience is false, why have so many philosophers found it so obvious? One possible diagnosis begins with the commonsensical point that when a particular thing has a normative feature, this is never brute fact:

If Jones has a reason to take the pill, this is because Jones possesses some (complex, possibly extrinsic) non-normative feature ϕ in virtue of which he has reason to take the pill. But from this it seems to follow that if Smith is like Jones in every non-normative respect. He too must have reason to take the pill. He has the feature ϕ in virtue of which Jones has this reason, after all. And this sounds a lot like Metaphysical Supervenience.

From the present point of view, however, this tempting argument trades on an equivocation. The principle that underlies the last step may be put as follows:

This is unassailable when ‘in virtue of’ means wholly in virtue of – the relation Fine (Reference Fine, Correia and Schnieder2012) calls ‘full metaphysical grounding’. But when the phrase is read in this way, (*) is false. The truth is rather that when X possesses a normative feature N, X is N partly in virtue of its non-normative features and partly in virtue of a pure principle that mediates the connection between those features and N. When ‘in virtue of’ is read to mean partly in virtue of, on the other hand, (**) is clearly false. So one diagnosis of the tendency to endorse metaphysical supervenience is that an intuitive argument for it trades on a conflation of full and partial ground.Footnote ⁷

A second and more likely diagnosis is epistemological. When we set out to test the truth of Metaphysical Supervenience, we naturally try to imagine – i.e. picture – a situation that is just like ours in non-normative respects but which differs in some normative respect. The common response is that people cannot do this. But that should come as no surprise even for those us who believe that there are such situations. When we imagine possible situations for the purposes of normative inquiry, we typically picture a situation in which the relevant non-normative features hold and then apply our background normative view to flesh out the normative features of the case. We do this in part because the normative facts are invisible and hence literally unimaginable except by picturing the non-normative facts that underlie them, but also because this is the appropriate method when our aim is to determine how things would be in normative respects if the non-normative facts were different, which is what we normally care about when we go in for counterfactual reasoning about the normative.

But this is the wrong procedure if we’re interested in Supervenience. The procedure, after all, always takes us to the nearest worlds with the stipulated non-normative features, and it should come as no surprise that among these nearby worlds, situations that are non-normatively alike are also normatively alike. When the topic is Metaphysical Supervenience, however, we want an answer to a different question, viz., whether there exist (perhaps quite remote) possible situations in which the pure normative laws are different. And for this purpose the right procedure is simply to stipulate a world where the pure normative truths differ in some concrete way and then to ask whether the stipulated situation should be reckoned possible. One might ask one’s interlocutor to consider a world in which the non-normative facts are held fixed but in which the Egoist is right about our reasons to help others, or (more plausibly) a Homeric world in which martial glory is worth pursuing for its own sake. Invitations of this sort provoke a familiar psychological resistance (Gendler Reference Gendler2000). But once the resistance has been overcome, we can point out that it’s easy to reason about how things would have been if these situations had been actual; that no contradiction or incoherence follows from the supposition; that such worlds are consistent with everything we know about the natures of the properties and relations that figure in the stipulations, and so on. And having done this we can ask: Why couldn’t things have been as these coherent world-descriptions say they are? The result of this exercise may be unclear. (For what it’s worth, having completed the exercise I find it obvious that many such counter-normative worlds are possible.) But the important point is that this way of testing the plausibility of Metaphysical Supervenience is both legitimate and quite different from the illegitimate method outlines in the last paragraph.

IX

This heterodox view of the modal status of normative principles is consistent with almost everything Scanlon says. I believe it’s the best view for him to hold given his non-naturalism. However it appears to be at odds with another distinctive aspect of Scanlon’s theory. I’ll close with a discussion of this apparent tension.

In response to familiar worries about our epistemic ‘access’ to non-natural normative facts, Scanlon defends a ‘Carnapian’ view according to which, as he puts it, ‘truth values of statements in one domain, insofar as they do not conflict with statements of some other domain, are properly settled by the standards of the domain they are about’ (Scanlon Reference Scanlon2014, 19). Scanlon distinguishes three main domains: natural science, mathematics and the normative. It’s a commonplace that if we privilege the natural sciences as a source of knowledge, we face a puzzle about our epistemic access to causally inert aspects of reality that can’t be poked and prodded in the lab. One standard response is a kind of holism that sees the postulation of mathematical objects, for example, as ultimately licensed by the standards of the natural sciences, which allow for the postulation of new entities, even inert ones, when this improves the unity and power of one’s total theory (Quine Reference Quine1976). But whatever its merits, Scanlon rejects this Quinean approach. The alternative is a form of pluralism that accords the best methods in each domain the same default authoritative status as the Quinean accords the scientific method. On this view, the mathematical methods that yield the near universal acceptance of standard set theory are automatically legitimate as ways of fixing belief in mathematics, even if set theory turns out to be dispensable for scientific purposes.Footnote ⁸ And similarly, the best method for arriving at normative judgments – the method of reflective equilibrium, according to Scanlon – is perfectly okay as it stands and requires no deeper justification. Just as the scientific method is not answerable to any extra-scientific tribunal, so the best methods of mathematics and practical reasoning are not answerable to standards drawn from other areas.

This ecumenical view can be understood in two quite different ways. On one reading, the relevant claim is purely epistemological, viz., that when a statement concerns a single domain D, the claim is justified simpliciter iff it is justified by the standards proper to D. On the other, the relevant claim is metaphysical, viz., that a statement about D is true when it is fully justified by the standards proper to D. Since justification is one thing and truth another, these views are obviously distinct. The Axiom of Infinity is clearly justified by the standards internal to mathematics. On the first view it follows from this that we are justified in believing the Axiom. On second, it follows from this that infinite sets exist.

Both readings of Scanlon’s ‘pluralism’ are supported by the text. Thus Scanlon writes:

And later:

These passages point to a thesis about the conditions under which domain-specific claims are justified. This reading is further supported by the general claim that the doctrine is at bottom commonsensical:

It may be commonsensical to say that scientific claims are ipso facto justified when they are justified-by-scientific-standards. But it is hardly commonsensical to say that scientific claims are true when they are justified in this sense. (In fact this is pretty clearly false. A false claim about the center of the sun may be brilliantly justified by scientific standards given the evidence we have – indeed, given all the evidence we could have. Science is fallible even in the ideal limit.)

At the same time there is ample evidence that Scanlon intends the metaphysical reading. Thus in a footnote Scanlon writes:

And later, discussing a possible objection:

The most important evidence for the metaphysical reading of Scanlon’s pluralism comes from his suggestion that this aspect of his view raises a question about the determinacy of the facts about reasons. Scanlon argues convincingly that we have no guarantee in advance that careful reasoning about a practical question will always yield an answer. Of course the same might be said about physics or Roman history. But in the case of reasons, Scanlon draws a moral:

If Scanlon’s pluralism were an epistemological thesis it would have no tendency to suggest that the facts give out where our methods for ascertaining them fall silent. Only when it is read as a metaphysical thesis does it have this implication. So, since Scanlon clearly believes that his thesis raises the specter of metaphysical indeterminacy in the normative domain, we must conclude that he intends it as a claim in metaphysics.Footnote ⁹

X

I bother with this in order to raise question about whether this conception of normative truth is consistent with the Finean package I have offered Scanlon as the best account of the modal status of normative principles. The Finean package says that there are possible worlds just like the actual world in non-normative respects but in which the pure normative truths are different. Let w be such a world where the fact that A would relieve X’s pain is a reason for X, but not for an arbitrary other, to do A. Since w is just like our world in non-normative respects, it agrees with our world about what people believe about reasons, and about what they would believe under various non-normatively specified conditions. But of course in our world we are not Egoists. Not only do we believe that people have reason to relieve the pain of others; we would continue to believe this after careful reflection. So the same is true in w. But this is just to say that our normative beliefs in w are systematically false, and would remain false no matter how carefully we thought things through.

The worry is that this possibility, to which the Finean appears to be committed, sounds inconsistent with Scanlon’s claim that normative truths are ‘settled’ by the standards that govern normative reasoning. This thesis is presumably meant to be a necessary truth. Scanlon’s idea is that once we understand what reasons are, we see that by their very nature the facts about reasons are fixed by our best reasoning about them. But this entails that there cannot be normative truths that are inaccessible to our best methods, whereas the Finean view appears to say just the opposite.

I believe that Scanlon can responds to this worry in a way that would make it possible for him to accept the Finean metaphysics I have offered while retaining his view about the relation between the normative facts and our best reasoning. The key point is this. Scanlon does not hold that the normative truths are settled by what we actually believe, or by what we would believe under non-normatively specifiable conditions. He holds that the pure normative truths are settled by the upshot of sound practical reasoning, or in other words, by what we would be justified in believing. Put schematically:

This connects one normative notion – the notion of a (practical) reason – with another: the notion of epistemic justification. The latter notion may or may not be analyzable in terms of the generic notion of a reason. But be that as it may, we can suppose that (P) is grounded in the nature of the reason relation and is therefore a necessary truth. This means that whenever the pure normative principles vary from world to world, as they do according to the Finean, the truths about what we are justified in believing must also vary. But that’s ok! Nothing in the Finean view precludes there being normative truths that are constrained to march in lockstep from world to world.

With this is in mind, return to w, the world in which the natural facts are just as they are but Egoism is true. (P) tells us that if this is to be a possible world, it must also be a world in which sound practical deliberation supports Egoism. But that’s fully consistent with our stipulation that w is a world in which people reject Egoism and would continue to reject it upon reflection. It simply forces us to describe w as a world in which our stable non-Egoistic beliefs, which we regard as justified, are in fact unjustified.

This is consistent with the letter of Scanlon’s view; but it’s palpably at odds with its Carnapian spirit. This way of squaring (P) with Fine’s modal metaphysics allows a radical disconnect between what people reflectively regard as sound reasoning and what is in fact sound reasoning. And for many philosophers, perhaps including Scanlon, this will be hard to swallow.

If we hold (P) fixed, the only way to avoid this result within a Finean framework is to suppose that the notion of epistemic justification has hidden depths. Suppose it lies in the nature of epistemic justification that a belief is justified only if it satisfies some non-normative condition K, and that a belief in Egoism fails to satisfy this condition in the actual world. (Maybe it lies in the nature of justification that a belief is justified only if clear-headed people would regard it as justified upon ideal reflection.) A belief in Egoism will then fail to satisfy K in w, since w is like the actual world in all non-normative respects. And this fact, together with (P), will entail that w is not a possible world after all.Footnote ¹⁰

This is a powerful strategy for limiting the weirdness of the Finean package, but it has its limits. For even if the concept of epistemic justification has hidden depths, it should still be highly unlikely, given Scanlon’s general perspective, that these depths suffice to yield non-normative necessary and sufficient conditions for justified belief. For if that were so, it would lie in the nature of epistemic justification that B is justified iff K(B), which would be tantamount to reductive naturalism about epistemic justification. Together with (P), this would entail the sort of naturalistic constructivism about the normative that Scanlon explicitly rejects (97 ff.), according to which the pure normative truths obtain because certain normative beliefs satisfy some non-normative condition. Scanlon must therefore think that even if the nature of epistemic justification places non-trivial non-normative constraints on when a belief can count as justified, there remains sufficient slack in these constraints to block this constructivist reduction. And so long as there is slack of this sort, the Finean view will entail that the principles of epistemic justification can vary from world to world, even holding the non-normative facts fixed. And from this it will follow, given (P), that the pure normative truths can likewise vary from world to world.

XI

I have offered Scanlon a package that jibes with almost everything he says about the metaphysical issues we’ve been discussing. The view maintains that the reason relation is irreducible; that mixed normative facts are always partly grounded in pure normative principles; that these principles are necessary in a sense; and that a pure principle obtains if and only if sound practical deliberation supports it. The view departs from Scanlon’s stated view only in insisting that these pure principles can sometimes vary from world to world. This sounds shocking, since it amounts to the rejection of (metaphysical) supervenience. But once we see what this means we should not be shocked. The view departs from orthodoxy mainly in the modal label it attaches to remote worlds in which the pure normative principles differ from those that obtain in actuality, deeming such worlds possible whenever they are consistent with the nature of the reason relation (and any other items that might be relevant). Its main advantage over Scanlon’s stated view is that it dissolves the last metaphysical objection to non-naturalism: the challenge to explain the supervenience of the normative on the natural. The objection is neutralized by conceding that the normative does not supervene as a matter of metaphysical necessity, but rather only as a matter of normative necessity. This latter supervenience thesis is then readily explained. (It’s a simple consequence of the Scanlonian premise that mixed facts are explained by non-normative facts plus pure principles, together with the observation that the pure principles are, almost by definition, counterfactually independent of the non-normative facts, hence normatively necessary.) The view thus explains every necessary connection that it posits, and that is progress.