It is commonly held that moral properties supervene on natural properties. But what is the status of this claim? And what does it tell us about the nature of morality? I want to pursue these questions by examining an argument against moral realism that Simon Blackburn has developed.) In parts 1 and 2, I consider his argument in some detail. And in parts 3 and 4,1 give a Kantian diagnosis of the argument and present my own views about moral supervenience and the role that it plays in our moral thought.

 1. BLACKBURN'S MODAL ARGUMENT

1.1. The (S)/(P) Combination

Blackburn takes off from the supervenience of the moral on the natural, and he hopes to land us with the conclusion that when we make moral judgments, we do not cognize moral facts or states of affairs, as the moral realist has it; instead, we project attitudes onto a purely natural world. He calls this Humean view "projectivism." He has also deployed the argument against Donald Davidson's position in the philosophy of mind. Blackburn argues that, unlike the projectivist, the moral realist has no "explanation" of why supervenience holds when it is combined with a certain lawlessness. The projectivist is said to cope with this combination better than the realist. In this part, I will sketch the abstract form of Blackburn's argument, and then go on to show how it applies to morality.

The abstract form of the argument is as follows. We begin with the claim that a supervenience relation holds. I shall understand supervenience as a relation between two classes of properties. I shall use italicized `F' and `G' as variables which range over pairs of families of properties, such as the moral and the natural, the mental and the physical, the dispositional and the categorical, or the macro­physical and the micro-physical. These are the supervening and subvening classes of properties. And I shall use non-italicized `F' and `G' as variables which range over particular properties within families of property. A `G*' property is a G property which is the complete subvening basis of an `F' property. Such properties are sometimes said to be `maximal' G properties and sometimes `total relevant' G properties.² These G* properties are complex conjunctive G properties. The idea of supervenience is that some such G* property of a thing fixes or determines its F properties. Blackburn formulates supervenience like this:³

(S)     Necessarily [(3x) (G*x & Fx) -+ (Vy)(G*y -+ Fy)].

I do not want to get too bogged down in different notions of supervenience; other­wise nothing else will ever get discussed! In my view, Blackburn's formulation has various shortcomings. For one thing, it fails to rule out the possibility of something with F properties but no G properties. But psychophysical supervenience, say, is usually thought to rule out Cartesian disembodied souls. Jaegwon Kim has given a formulation which is far more satisfactory on this and other scores .4 But the shortcomings of Blackburn's formulation as compared with Kim's do not matter for the moment. I shall argue in due course that Blackburn's argument goes ahead with Kim's preferable formulation.

Now, Blackburn contrasts supervenience with the following necessity:

(N)    Necessarily (Vx)(G*x -->  Fx).

If we lack (N), we have:

(P)     Possibly (3x)(G*x & -Fx)

Actually (P) cannot be quite what Blackburn has in mind, and we will need more to get the argument going. The idea of (P) is that even if it is possible that something is F and G*, it is also possible that something is G* but not R (We probably need to quantify over F and G properties in order to get the true sense of (P), but let us leave that aside.) I shall assume that this is what we have in mind when we appeal to (P). I do not want to stray too far from Blackburn's own formulations.

Blackburn asks us to suppose that there is some area in which we can argue that both (S) and (P) hold-that is an "(S)/(P) combination" obtains. Blackburn then urges us to find it problematic. In his early essay, "Moral Realism," Blackburn wrote:

... if A has some naturalistic properties, and is also good, but its goodness is a distinct further fact not following from its naturalistic features, and if B has those features as well, then it follows that B also is good. And this is a puzzle for the realist, because there is no reason at all, on his theory, why this should follow. If the goodness is, as it were, an ex gratia payment to A, one to which A is not as a matter of logic entitled in virtue of being as it is in all naturalistic respects, then it should be consistent to suppose that although goodness was given to A, it was not given to B, which merely shares the naturalistic features that do not entail the goodness.⁶

 One way of putting Blackburn's interesting worry is this: how can it be that although there is no necessity that a G* thing is F, it is necessary that a G* thing is F if some other thing is both G* and F? If F is a real property, how could it be that a G* thing is only forced to be F if something else is both G* and F? There is a puzzle here. This pattern of modalities looks mysterious. Something modally weird seems to be happening. Blackburn argues that the moral realist cannot "explain" the combination of (S) together with (P).

In his more recent publications, Spreading the Word and “Supervenience Revisited," Blackburn prefers to formulate the argument in terms of the mys­teriousness of a ban on "mixed worlds." To arrive at Blackburn's mixed world argument, we must translate the modal operator into the possible world idiom. Here is Blackburn:

In any possible world, once there is a thing that is F, and whose F-ness is underlain by G*, then anything else that is G* is F as well. However, there are possible worlds in which things are G* but not F. . . . The one thing we do not have is any mixed world, where some things are G* and F, and some are G* but not F . . . These are ruled out by the supervenience claim (S): they are precisely the kind of world that would falsify that claim. My form of problem, or mystery, now begins to appear. Why should the possible worlds partition into only the two kinds, and not into the three kinds?

 The mixed world argument is this: (P) tells us that there are possible worlds in which things are F and G* and that there are also possible worlds in which things are G* but not F. But given that there are worlds in which things are G* and F, and others in which things are G* and ~F, then it seems that there should be mixed worlds-worlds in which some G* things are F, but other G* things are ~F But this is disallowed by (S). I shall argue in part 2 that this reformulation is misguided and that Blackburn's original version of the argument in "Moral Realism" was superior.

1.2. The Combination in Morality

How does Blackburn's problem arise specifically in moral philosophy? In order to set up Blackburn's problem for moral realism, I am going to have to take the liberty of speaking of conceptual modalities. Nowadays, many philosophers distrust the idea of conceptual modality. I shall use the notion with some explanation but without a full defense against Quinean skepticism. And anyway, I shall recast the argument in Quine-speak for those who feel happier with that.

First supervenience: in morality, the principle of supervenience is concep­tually necessary. What this means, roughly, is that grasp of supervenience is constitutive of competence with the notion of moral value. So someone who rejects moral supervenience, rejects moralizing altogether. To deny the doctrine of moral supervenience is to change the subject away from morality. If someone appeared to deny moral supervenience, we would think that they had not grasped the concept of moral value, or perhaps that we were mistranslating their utterances as moral utterances.⁹ It is pretty incontrovertible that supervenience is an essential feature of moral thought, even if it is hard to support this with a non-question begging argument. The nearest we can get to an argument is to consider the situation of those who flout supervenience in their system of moral judgments. One example is when we make exceptions in our own case. We may do this even though we do not believe that there is a relevant difference between our own case and that of another. For example, although I may think that people ought not to accept bribes, I might nevertheless think that the bribe I took yesterday was not really as bad as all that. Were the inconsistency to be brought to my attention, no doubt I would be embarrassed, and perhaps I could eventually be brought to alter one or other of my judgments. But the fact may remain that I made two differing moral judgments about things which I believed had the same relevant subvening natural basis. So we need to make a distinction between believing that supervenience is false and a failure to conform to the principle of supervenience in the pattern of someone's moral judgments. If I think x is F on account of its being G*, but I also deny that y is F although I know it to be G* as well, I may be moralizing, but I could not, as it were, hold the two thoughts up before my mind without a certain embarrassment and discomfort, people are sometimes inconsistent in this way but when they realize it, they correct themselves. This shows that they are aware that there is a certain normative constraint on their judgments which they are failing to heed. This striving for consistency is part of what it is to moralize. But this quest for consistency rests on a commitment to moral supervenience. So moral supervenience is a conceptual truth.¹⁰

The principle which Blackburn wields together with supervenience is the principle that there are no conceptual necessities tying moral and non-moral kinds. The principle is quite weak; it says that we cannot infer that something has a certain moral property from the fact that it has a natural property, even if it is a maximal or total relevant natural property. There will always be an "open question" as to whether it really had the moral property; one could deny that it did without conceptual confusion.¹¹ (Further questions about the existence of metaphysical necessities are not to the point here. I deal with this later on.) This principle is very plausible. It has been denied but the overwhelming majority of moral philosophers have accepted it. Thus, in the moral case, and where we are dealing with conceptual modality, we lack (N) but we have (P). So we have the potentially nasty (S)/(P) combination.

It is even more obvious that there are no moral-to-natural necessities. This is because moral properties are "variably realized." Just as we owe our pains to the carbon-based neural configurations of our brains, but there might be silicon-­based Martians who also feel pain, so, similarly, some people may be evil due to a desire for power, others due to the pleasure in another's suffering, others due to envy, and others due to greed. But variable realization is beside the point. Variable realization implies the absence of necessities running from the moral to the natural or from the mental to the physical. But in Blackburn's supervenience argument we are concerned with the lack of necessities running the other way­from the natural to the moral or from the physical to the mental.¹² The existence of necessities of this sort is compatible with variable realization.

 1.3. Modal Matters

Now, philosophers sometimes distinguish different brands of modality. Black­bum's argument does not depend initially on the kind of necessity in question in (S) and (P), so long as they are of the same sort. But in different areas, claims (S) and (P) maybe acceptable in different strengths. In "Moral Realism," Blackburn speaks only of "logical necessity." But in "Supervenience Revisited," he operates with a schema according to which "conceptual" necessity implies "metaphysical" necessity, which implies "physical" necessity, which implies statistical generalizations. 13 This is reversed for possibilities. Stronger senses imply weaker senses.

It might be thought that it is an odd idea that metaphysical necessity is weaker than conceptual necessity. Surely metaphysical necessity is as strong as can be. Perhaps they are two quite different kettles of fish; metaphysical necessities being about the world, while conceptual necessities merely reveal truths about us (in particular, about the concepts that we possess).¹⁴ However it is best not to see conceptual and metaphysical modality as quite different brands of modality. It is better to say that conceptual necessities are those metaphysical necessities which we can know about in a certain way: roughly, we arrive at conceptual knowledge by following out the implications of what we must know in order to be able to possess and apply a concept. But not all metaphysical necessities can be known in this way. We can separate out those metaphysical necessities which can be known by conceptual means from those which cannot be known in that way. This would preserve the idea that conceptual necessity is stronger than metaphysical necessity, in the sense that the former entails the latter, but not in the sense of two completely different brands of necessity. The distinction is an epistemological one.

How far does Blackburn's argument depend on a particular view of modality? Not much, in my view. Imagine a philosopher whose scruples about conceptual modality mean that he prefers to paraphrase away talk of conceptual possibility altogether so that he talks in terms of whether or not we have or have not made a conceptual error if we assert some proposition. There is still the following mystery concerning the supervenience of F properties on G properties: although we make no conceptual mistake if we assert either (3x)(G*x & Fx) or (3x)(G*x & -Fx) individually, we do make a conceptual mistake if we assert the conjunction [(3x)(G*x & Fx) & (Ey)(G*y & ~Fy)].

What about those philosophers who are averse to any mention of the category of modality? It is true that an appreciation of Blackburn's argument requires that we do not hold the extreme, rabid, foaming-at-the-mouth Quinean view that there is no genuine category of modality at all and that we ought to dispense with modal considerations altogether. Such an extreme view is opposed to a moderate Quinean view according to which we use the notion to express the centrality or lack of centrality of a commitment in our web of belie£15 We can put the problem in terms that the moderate Quinean will allow as follows. It is not central to our web of belief that a G* thing is R Yet it is central to our web of belief that a G* thing is F if there is some other G* thing which is R The fact that something is F given that it is G* suddenly gets catapulted into the center of our web of belief by the apparently extraneous fact that some other thing is G* and R We could understand how it could be that x's being F if it is G* would be catapulted into the centre of the web of belief by the additional fact that Q, say, if Q somehow logically entailed that a G* thing is F That would explain why the catapulting occurs. But if Q describes some fact quite distinct from and unrelated to x's being G* and F, then why would we fix so certainly on this truth? (Presumably it is not that by induction we have established that there is a causal law to this effect.) Thus even the moderate Quinean can feel Blackburn's puzzle. Only the out-and-out, rabid, foaming-at-the-mouth Quinean has no materials with which to understand the problem. But then he has no materials to understand virtually every philosophical problem that has ever been raised. Blackburn's puzzle arises so long as we accept the category of modality. Our specific views about the nature of modality do not make any difference to the argument, so long as we allow that we can think in modal terms. So it is incorrect to say that the argument assumes modal realism, as Ian McFetridge argues at one point. 16 If the (S)/(P) combination is problematic, it cannot be relieved by some kind of anti-realism about necessity. Our views about the metaphysical nature of the relevant brand of modality do not matter for the argument so long as we accept the category of modality.

Now, in order for Blackburn's argument to go ahead, the necessities of (S) and (P) must coincide. Blackburn notes that in the moral case (N) is plausible where the necessity therein is metaphysical necessity. 17 So there is no problematic (S)/(P) combination with that kind of modality. In the moral case, the modalities of (S) and (P) coincide only where they are both conceptual. I am sympathetic to the thought that there are metaphysically necessary connections between the moral and the natural. However a number of philosophers have thought that this hurts Blackburn's argument. Indeed, this is the consensus among those who have discussed it. For example, Sydney Shoemaker, James Klagge, Ian McFetridge, and James Dreier have all argued that the existence of such metaphysical necessities renders Blackburn's argument beside the point.18 But the consensus is mistaken. For Blackburn's argument goes ahead so long as there is some kind of modality where both (S) and (P) hold. In moral philosophy they both hold at the level of conceptual necessity; so that is where the problem arises. There may be metaphysically necessary connections between the natural and the moral. But Blackburn is right to think that this does not help with his problem at the level of conceptual necessity. (I shall return to this point.) It is true that the original version of the argument in "Moral Realism" was unsatisfactory in that it operated with an overly simplistic notion of "logical necessity." And it is true that the revised version in Spreading the Word and "Supervenience Revisited" is confusing in that it seems to go ahead at the metaphysical level. However, once the distinction between conceptual and metaphysical modality is made and we are clear that we are going ahead with conceptual modality, Blackburn's argument against moral realism becomes more interesting, not less.

1.4. Explaining Supervenience

Before we turn to spell out, reconstruct, and then assess this argument, it is important to get things in perspective by recognizing that it is not sufficient for the success of Blackburn's argument that realism merely has a problem with the (S)/(P) combination. Blackburn's argument requires that projectivism has a better explanation than realism. The argument is only good if the projectivist can capture this combination more easily than the realist. 19

Supervenience ought to be seen as playing a twofold role in the debate over moral realism. On the one hand, if Blackburn is right, moral realism has to struggle to explain supervenience. But on the other hand, supervenience is one of the features of our moral thought which a ("non-error-theoretic") projectivist must try to respect. This aim ought to be part of what Blackburn calls "quasi­realism."2° But it will not be easy for a projectivist to respect supervenience. For why should we not be perfectly happy to project quite different values onto things which we believe to be naturally alike, as we do in our thought about the niceness of food and in our thought about the funniness of situations? The niceness of indistinguishable cream doughnuts somehow declines after we have consumed several. Indistinguishable jokes wear thin. There is nothing wrong with this; we are not obliged to be consistent from case to case. But we are obliged to be consistent in our moral judgements. The problem for projectivism is to explain why we should not project different attitudes onto things we believe to be naturally the same, as we do in our niceness and comic thought. It is utterly unhelpful to say, as Blackburn does, that "the role of moralizing [is] to guide desires and choices among the natural features of the world."21 For a non-supervenient form of thought or discourse could also do that; it is just that such a form of thought would not constrain us to do similar things in naturally similar cases if we did not feel like it. And if we build a consistency demand into what it is to "choose, commend, rank, approve, or forbid things on the basis of their natural properties,"22 the question then becomes one of explaining how we could do that. No explanation has been provided. Explaining our respect for moral supervenience is a serious problem for quasi-realism. If quasi-realism fails to account for supervenience, it also fails to account for the problematic combination.

If projectivism cannot explain the (S)/(P) combination more easily than realism, then projectivism is not preferable on this score. Without this part of his argument it might be that Blackburn has merely located a metaphysically neutral puzzle concerning the notion of supervenience, conceived of as distinct from claims of the form of (N)-and if so, he would have nothing to wield against the realist. However, even if quasi-realism fails to explain supervenience on behalf of the projectivist, it is still important for the realist to explain it.

For a moral realist there are three possible approaches to this problem: We might deny (S); we might embrace (N) as well as (S); or we might hold (S) and (P) together while removing the difficulty. I shall assume here that there are powerful independent arguments in favor of (S) and (P). My view is that the combination does not arise at all for Davidson's view in the philosophy of mind according to which the mental supervenes on the physical but there are no strict laws connecting the mental and the physical. If the combination does not arise, Blackburn's argument is no threat to Davidson. I have argued for this claim elsewhere.23 But where the (S)/(P) combination does arise, it needs to be shown that it is nothing to loose sleep over. I shall try to defend moral realism by showing that the (S)/(P) combination is not insuperably problematic at the level of conceptual modality. I do that in part 3. But we must first consider the way Blackburn now casts his argument. Before I try to undermine the argument, I want to make it as strong as possible.

 

2. THE MIXED WORLD ARGUMENT

2.1. The New Version

Blackburn now presents his argument in terms of "mixed worlds." By the end of this part, I hope to have argued that this way of framing the issue is misguided, or at least misleading, from Blackburn's own point of view. But before I do this, I shall argue that if the mixed world formulation is legitimate, the argument can then be put in a much stronger way. Thus I might appear to be making matters more difficult for the realist. But paradoxically, it turns out to be good news for the realist. For although the problem is made worse for the realist, it becomes an equally bad problem for the projectivist. The strengthened reformulation is not one which could favor projectivism over and above realism.

I shall use `G*/F' as a predicate applying to worlds in which there are things which are both G* and F; `G*/~F' applies to a world in which there are things which are G* but not F; and a `G* world' is one in which something is G*. I shall also use `G*/F' as a predicate applying to things which are both G* and F. Blackburn's mixed world argument has five steps: we have

(a)         There are G*/F worlds

and

(b)           There are G*/~F worlds.

Both come from (P), suitably reinterpreted, since conceptual constraints do not rule out either G*/F things or G*/~F things. But

(c)         There are no "mixed worlds" in which some things are G*/F but others are G*/-F.

 

which comes from (S). However

(d)           Once we have a G*/F world and a G*/-F world, we have enough to

construct a mixed world. So it seems that

(e)            There exist mixed worlds. But this conflicts with (c).24

Let us look as Blackburn's presentation of the first three steps of the argument. I quoted Blackburn above saying:

 In any possible world, once there is a thing that is F, and whose F-ness is underlain by G*, then anything else that is G* is F as well. However, there are possible worlds in which things are G* but not F.25

We ought to raise an eyebrow at the conception of superveience that Blackburn employs here. Following Kim, let us distinguish a weak notion of supervenience, which imposes a consistency demand only on what obtains within a world, from a strong notion of supervenience which constrains how things can be in some worlds given how they are in others. Omitting the initial necessity, one version of the weak supervenience of F on G* is this:

(WS) If something is F and G*, then anything else in that world which is G* is also F

whereas a parallel notion of the strong supervenience of F on G* is this:

(SS)   If something is F and G*, then in all possible worlds anything which is G* is also F.

We are bound to be suspicious of the fact that Blackburn operates only with weak supervenience in the quoted passage. But some of Blackburn's critics have jumped to the conclusion that his argument only appears to have force because an implausibly weak notion of supervenience is employed.27 But the strong notion is to hand. And my view is that we can recast Blackburn's argument in a more radical form with the strong notion. Blackburn's critics were right to criticize his deployment of weak supervenience. But they failed to see that a version of the argument can be mounted with strong supervenience.

2.3. The Revamped Mixed World Argument

Employing the strong notion of supervenience, the revamped argument runs like this: steps (a) and (b) still say that there are G*/F worlds and that there are also G*/~ worlds. But given strong supervenience, we can now add as the third step:

(c')        Either all G* worlds are G*/F or they are all G*/~F

The argument is then, more straightforwardly, that his is brute inconsistent with both (a) and (b) taken together. The revamped problem is that if (P) is true, it looks as if there are possible worlds which are G*/F and also ones which are G*/~F But this simply conflicts with a strong notion of supervenience. The problem is no longer merely that of explaining the source of a mysterious ban on mixed worlds but that of removing the sheer contradiction between (P), which tells us that there are G*/F worlds and also G*/~ worlds, and (SS), which tells us that there are not both these sorts of worlds.

The contradiction is most apparent if the matter is put as follows. Again, omitting the initial necessity, we can cast a strong notion of supervenience thus:

(SSS)                All G* worlds are G*/F worlds (and none are G*/~F) or else all G* worlds are `G*/~F' worlds (and none are G*/F).

This contrasts with the following weak claim, which is parallel to the one which

Blackburn employs:

(WSS)  All G* worlds are such that, either that world is G*/F (and not G*/~F) or else it is `G*/~F' (and not G*/F).

That is, no world is both G*/F and G*/~F) ("mixed"). The difference is the scope of the world quantifiers. (SSS) has two world quantifiers, the scope of each being limited to each disjunct, whereas (WSS) has only one world quantifier whose scope is the whole disjunction. In my view, a decent notion of supervenience should make the former, strong claim. But Blackburn takes supervenience in the latter, weak way, which seems to leave it open that although no world can be both G*/F and G*/~F, some worlds can be G*/F while others are G*/~F The conjunction [(Ew)(G*/Fw) & (Ew)(G*/~Fw)] is not obviously incompatible with (WSS)-we merely have a mystery; but it is obviously incompatible with (SSS). If there is a G*/F world and there is a G*/~F world then (SSS) is simply not the case. These three yield a plain contradiction, whereas with Blackburn's weak notion of supervenience we merely have a mystery. While a mystery might not seem so bad, there is no getting away from the contradiction. It is as bad as: there is at least one cow on the farm; there is at least one chicken on the farm; and all the animals on the farm are cows or they are all chickens. We have three claims which all seem to be true, yet all of which cannot be true.

To be sure, this is not a happy state of affairs. But, as promised, it is relatively good news for the realist. It changes what we can take Blackburn's argument to show. On this reformulation, it becomes very difficult to see how the argument could favor projectivism over realism. Such a contradiction is not something which a projectivist could "learn to relax with" more easily than a realist 2s The contradiction does not show that we need projectivism about the F properties. How could it?

And the argument does not show that he should reject strong supervenience and operate only with the weak formulation, for there is a general problem with holding a weak formulation without the strong one. How could it be that if something is G* and F then absolutely all G* things are F, even though any one of them could easily have been G* but not F? Wouldn't it be a real mystery if all G* things are F even though any one of them easily might not have been? Moreover, in morality in particular, the proposition `If something is bad then it could not have been exactly as it is in all relevant natural respects but not bad' is no less a conceptual truth than `If something is bad then so is anything which is relevantly similar in natural respects.' Ordinary thought is no less committed to the one case than the other.

Were we to operate only with mere weak supervemence, we might evade the nasty contradiction which the strong notion yields. However that would not make Blackburn's formulation of the mixed world problem any happier. For if anything would be mysterious, it would be weak supervenience without strong supervenience. And anyway, strong supervenience is independently plausible in morality. But the strong formulation leads straight to disaster.

2.4. The Argument Restored

The very least the strong folmulation of the argument shows is that Black­burn should not be talking in terms of conceptually possible worlds?⁹ And if this is correct, he would do better to avoid the mixed world formulation of his argument against moral realism. The original formulation of the argument in "Moral Realism" was far better. Blackburn has undersold his earlier argument by recasting it in mixed-world terms.

We should not be lulled into a false sense of security by the rejection of the mixed world formulation of Blackburn's argument. This is very far from being the end of the matter. We can restore Blackburn's argument to its former glory.

We can recast the argument once more-but this time paraphrasing away the quantification over conceptually possible worlds, so that we speak only in terms of "conceptual constraints." If we phrase things in these more polite terms, we no longer have a contradiction on our hands. Let us begin with a weak formulation. As far as conceptual constraints go, something can be G*/F; and as far as conceptual constraints go, something can be G*/~F. But, according to supervenience, conceptual constraints rule out the existence of two things, one of which is G*/F while the other is G*/~F And Blackburn's argument is then that, although we make no conceptual mistake if we assert (Ex)(G*x & Fx) and we make no conceptual mistake if we assert (E x)(G*x & ~Fx) we need an explanation of why we do make a conceptual mistake if we assert the conjunction [(Ex)(G*x & Fx) & (Ey)(G*y & ~Fy)].

This, I claim, is how Blackburn should put his argument. His problem does not get off the ground at the level of conceptually possible worlds; but it does get off the ground if it is framed more modestly, in terms of conceptual possibility and conceptual mistakes, as it was in "Moral Realism." We no longer have a contradiction if we paraphrase away the talk of conceptually possible worlds. But we still have a problem. The problem is that a conceptual possibility seems to have disappeared.

We can also formulate a version of the argument which employs strong supervenience in the present manner, so long as we are clear that the worlds quantified over in strong supervenience are metaphysically possible worlds. The question is: why is it that although we make no conceptual mistake if we assert (Ex)(Ew)(G*xw & Fxw) and we make no conceptual mistake if we assert (Ex)(Ew)(G*xw & ~Fxw), we do make a conceptual mistake if we assert [(Ex)(Ew)(G*xw & Fxw) & (Ey)(Ew')(G*yw' & ~Fxw')]? This is not a problem of metaphysically "mixed worlds." (I have indulged in world talk only for the embedded metaphysical necessity.)

We can now see why it is an error to say that Blackburn's argument only appears to go ahead because he formulates the argument in terms of a "weak" rather than a "strong" notion of supervenience. Blackburn himself encouraged this incorrect criticism by setting up his problem with the weak notion. But he need not have done so. The argument goes ahead just as well with the cross-world notion3° This is one reason why I did not spend too much time fussing about the notion of supervenience when I introduced Blackburn's problem in part 1.³¹

We noted earlier that (N) is plausible in morality at the level of metaphysical necessity. Blackburn himself admits this. So the supervenience argument is impotent against moral realism at the level of metaphysical modality. If there are no metaphysically possible G*/~F worlds because there are only G*/F worlds, then there is no problem about metaphysically mixed worlds, for there are none. To say this is not to set about explaining supervenience, but just to assert that it holds. We should not attempt to explain a ban on metaphysically mixed worlds, since if we hold to supervenience, we should not admit both G*/F and G*/~F worlds in the first place. So there is nothing to fret about at the level of metaphysical necessity. And as I have just argued, given a G*/F world, we should not be allowed to say that even though there are no G*/~F metaphysically possible worlds, there is, nevertheless, a G*/~F conceptually possible world. The problem should not be allowed to get off the ground in terms of conceptually possible worlds. Yet for all that, there is still a problem with the joint conceptual possibility of (S) and (P) in morality.

If we insist on representing the conceptual modalities in the world lingo, then we should say that steps (a), (b), (c) and (c') are all true but that does not create a problem. There is indeed a ban on mixed worlds just as there is a ban on one world being G*/F while another is G*/~F. There is nothing mysterious about that. Only a confusion with metaphysical worlds could make someone think that there is a problem. This comes out when Blackburn attempts to argue that given (a) and (b) (that there is both a G*/F world and a G*/~F world) then there should be a mixed world. He appeals to spatial and temporal analogies 3² If we can imagine a G*/F world and a G*/~F world, he says, then surely we could imagine a world in which one half is G*/F while the other half is G*/~F, or we could imagine a world which is G*/F until a certain time but G*/~F thereafter. But this argument reveals a slippage between conceptual and metaphysical modality because the mereological principles it invokes are only plausible for metaphysical modality, and yet the argument is supposed to be going ahead at the conceptual level. So the culprit of the mixed world argument is (d). If we are dealing with conceptual modality it has no plausibility at all. It only appears to be plausible because we illegitimately slip into interpreting the modality as metaphysical. There remains something puzzling about supervenience constraints, but the mixed world argument fails to bring out what it is.

For ease of exposition, I have been talking rather naively in terms of conceptual modality as if it were a variety of modality all of its own. However, as I suggested, it is more likely that a conceptual necessity is just a metaphysical neces­sity which we can know in a certain way. The method of acquiring knowledge is, roughly, knowledge which is arrived at by following out the implications of what we must know in order to grasp a certain concept. This means that, since we have (N) at the level of metaphysical necessity, the original possibility which caused all the trouble was, in fact, merely a certain sort of epistemic possibility. (Epistemic modality, presumably, comes in different varieties corresponding to different sorts of knowledge.) Just as some materialists replied to Saul Kripke's anti­materialist argument that the alleged Cartesian possibilities were in fact merely epistemic possibilities and not genuine metaphysical possibilities,³³ similarly, we can imagine the realist replying to Blackburn that although both G*/F and G*/~F worlds are merely possible for all we can know by conceptual reflection, they are not both genuinely metaphysically possible. But this does not defeat Blackburn's argument. Rephrased in terms of what knowledge leaves open, his problem is this: how can it be that (Ew)(G*/Fw) is left open by the knowledge built into the concept of morality; and the same goes for (Ew)(G*/~Fw); but the conjunction [(Ew)(G*/Fw) and (Ew)(G*/~Fw)] is not left open by the knowledge built into the concept of morality? How can it be that it is possible as far as the concept of moral­ity is concerned that (Ew)(G*/Fw) is true, and it is possible as far as the concept of morality is concerned that (Ew)(G*/~Fw) is true, and yet it is not possible as far as the concept of morality is concerned that both (Ew)(G*/Fw) and (Ew)(G*/~Fw) are true? What is the source of this constraint? What justifies us in accepting it? Blackburn can be construed as presenting us with a puzzle about a certain variety of epistemic possibility. It may be metaphysically necessary that there are only G*/F worlds, but so long as we admit that a G*/F world and G*/~F world are each conceptually or epistemically possible, Blackburn's argument goes ahead.

   

3. DIAGNOSIS

3.1. The Blunt Response

In this part I shall try not so much to rebut the argument but to diagnose what is attractive about it. It will turn out that this serves as some protection against it. I shall first try to unsettle the intuition that Blackburn's argument, as I have reconstructed it, is impossibly problematic for the realist. To recall: the argument is that if a G*/F thing is conceptually possible and a G*/~F thing is also conceptually possible, then why is it not also conceptually possible that there is one thing which is G*/F as well as one which is G*/~F? Or-to put the question another way-how come something is not conceptually forced to be F given that it is G*, unless something else is both G* and F?

Here, then, is my initial blunt reply to Blackburn: it just isn't, and this is a conceptual truth with an independent source. Consider an imaginary card game, by way of an analogy. It may be possible according to the rules of the game to lay down a pair of green cards, and it may be possible according to the rules to lay down a pair of blue cards. But that does not imply that it is also possible according to the rules to lay down one green card and one blue card. It may be that an independent rule rules this out.

Of course, Blackburn will not be impressed with this initial blunt response. What will turn out to be significant is the diagnosis of why he will notbe impressed. The remainder of my diagnosis is an enquiry into this. The blunt response needs to be placed in some kind of context.

 

3.2. Analytic/Synthetic

McFetridge points out that it is not always true that possibilities can be conjoined, and he gives the obvious example of `It is possible that p' and `It is possible that ~p' (where p is contingent).³⁴ This does not itself block Blackburn's argument, but it points us in the right direction. It is not true that for all p and for all q, if p and q are individually possible then the conjunction of the two is also possible. Two familiar examples of unconjoinable possibilities not so far from McFetridge's, are these: It is possible for something to have four sides. It is also possible for something to be triangular. But it is not possible for something to be a four-sided triangle. Similarly, it is possible that I marry, and it is possible that I remain a bachelor. But it is not possible that I marry while remaining a bachelor. It is necessary that triangles have three sides and that bachelors are unmarried.³⁵

With one eye on these examples, let us turn to morality. What is the relevant disanalogy between moral supervenience and the triangle and bachelor cases? I suspect that Blackburn would find the source of their necessity unmysterious because that source is simply that of analytic necessity. What I mean by `analytic' is, roughly, that a truth is analytic if and only if its negation is or implies a self­contradiction. (Quine says that an analytic truth is one which "can be turned into a logical truth by putting synonyms for synonyms.")<sup>36</sup> It is crucial that I do not use the broader and vaguer sense of `true in virtue of meaning'. There are undoubtedly problems with giving a satisfactory account of analyticity. For example, what is a logical constant? What kind of implication is involved? There are many problems. But I shall here assume something like the traditional notion.

Blackburn will say that there is no problem about explaining why there are no married bachelors or four-sided triangles, for they are ruled out as a matter of analytic necessity. But-and this is the disanalogy-this is not true of the negation of the supervenience principle. And this is what the problem of explaining supervenience consists in. If the negation of supervenience were a self­-contradiction or else implied one, we could explain it as an instance of logical or analytic necessity. But since it isn't, we can't. So we need some other explanation.

 

3.3. Synthetic A Priori Supervenience

With this observation, we approach the heart of the matter. Casting all caution to the wind, I conjecture that what is different about moral supervenience which explains why Blackburn might be unimpressed by the triangle and bachelor analogies is that the truth of moral supervenience is synthetic a priori while the triangle and bachelor cases are analytic a priori truths. (I apologize for throwing around these hallowed and weighty terms, but my purpose here is diagnostic.) Moral supervenience is a synthetic a priori truth because its negation does not imply the contradiction which is necessary for analyticity. The conjecture is that it is. the synthetic a priori status of moral supervenience which makes it seem more mysterious than the analytic a priority of the triangle and bachelor cases.37

The moral supervenience constraint is a principle which parallels those high-level principles that Nathan Salmon has described as governing our thought about natural kinds, origin, and constitution, and which generate necessary truths in these areas.³⁸ For instance, there seems to be a modal principle to the effect that given that an actual sample of a natural kind has a certain constitution, then all actual and possible samples also have that constitution. Perhaps moral super­venience has a status rather like Salmon's essentialist principle governing natural kinds. Salmon agrees with Keith Donellan in thinking that the general modal prin­ciples which are operative in cases such as natural kinds, origin, and constitution are known empirically. But this is not plausible in the case of morality. Moral supervenience is more akin to the case of proper names, where it seems to be an a priori matter that identity statements between proper names are necessary if true.

These days, many American moral realists-who have come to be called "Cornell Moral Realists"-believe that moral beliefs are based on broadly em­pirical grounds. The non-observability of morality is supposedly explained away by the "holism" of empirical knowledge: moral beliefs, they say, are somewhat remote from the experiential periphery, but they answer to it all the same. But even if one were this kind of moral realist, one could not hold that the general principle of moral supervenience is known empirically. For that would be to open the possibility of discovering evidence for thinking that morality does not need to conform to the supervenience constraint after all. The conceptual status of supervenience would thus be lost. Blackburn is absolutely right if he is assuming that if one is a realist, one is stuck with a potentially mysterious synthetic a priori principle. In this respect, moral supervenience is unlike the cases that Salmon discusses where it is half plausible that we are dealing with high-level empirical principles.

Now, if my conjecture about the synthetic a priori status of moral supervenience is correct, we are then entitled to cough discreetly and wryly observe that there has been and there still is a general philosophical problem about synthetic a priority which might fairly be described as "big." The mysteriousness of moral supervenience would be part of that wider problem.

Blackburn's argument conceals its true epicenter. There are deep Kantian under-currents swirling beneath it. This diagnosis should seem apt as applied to a self-confessed, born-again Humean.

 

3.4. Partners in Crime?

The diagnosis, then, is that at the root of Blackburn's discomfort lies the fact that he finds the necessity of logical or analytic truths unmysterious, but other sorts of a priori necessary truths he finds mysterious. So his argument is only as good as synthetic a priority is dubious. But maybe synthetic a priority is perfectly alright.

 Where do we go from here?

 One direction would be to challenge Blackburn to explain why logical or analytic necessity is so unmysterious. This is far from obvious. Perhaps the law of non-contradiction is as mysterious as moral supervenience. I shall not pursue this.

A second response we might give is a stubborn one At the end of his paper, McFetridge imagines a moral realist (who is also a modal realist) who says that,

Since ... there must be some necessities which are brute, why should the supervenience of the moral on the naturalistic not be one such.

This amounts to a rejection of the demand for an explanation of the necessity of supervenience. We might sometimes explain a necessity by going from one or more other necessity to the one to be explained; but this cannot always happen. Perhaps moral realists are within their rights to insist that the necessity of supervenience is just a "brute" conceptual necessity.

While I am sympathetic with this stubbornness, I think the realist can do better. A third plan of action is to look for "partners in crime."

At this point we should reach for our well-thumbed copy of Kant's Critique of Pure Reason. But consider first color incompatibilities which are-to some extent-parallel with the moral case. If Colin McGinn is right, they are conceptual constraints governing color attributions; and this is analogous to the role of supervenience in our moral thinking 40 Yet they do not look like analytic truths. Consider this familiar pattern: it is possible for something to be red all over; it is also possible for something to be green all over; but it is not possible for something to be both red and green all over ("mixed objects"). It might have been either red or green, but once it is red all over, it cannot be green-and vice versa. Is this mysterious? Perhaps it is. But then this may be an area where being "mysterious" or "queer" is nothing to be ashamed of. Or at least we need an argument that it is. Of course, a worry with this case might be that it is an inappropriate analogy for a moral realist to use because if McGinn is right, these incompatibilities spring merely from the nature of our experience and not from the nature of the world. If so, the explanation of the incompatibilities is not a realist one. But then there are other analogies on hand.

It is plausible that there are synthetic a priori principles embedded in our temporal thinking. Take the transitivity of temporal relations: if A occurs before B, and B occurs before C, then A occurs before C. Or take what Michael Dummett calls "truth-value-links:"⁴¹ if the thought "X is F now" was true ten years ago, then the present thought "X was F ten years ago" must be true now. And there are other such principles. These principles are not analytic in the sense of being reducible to a logical truth by substituting synonyms. But they do not look like empirical truths either. So it is reasonable to put them down as synthetic a priori. Now it is also constitutive of temporal thinking that we grasp these truths. We could not be thinking in temporal terms if we failed to grasp them. They are principles which we presuppose in thinking temporally. And yet they are neither analytic nor empirical. Similarly, moral supervenience is not analytic and it is not something which we come to believe on empirical grounds; but it is constitutive of moral thought that we grasp moral supervenience. In a sense, then, such propositions are "conceptually necessary" because they are "framework principles" which are presupposed by all of our thought in the area. Such principles are part and parcel of the concept. This might strike some as odd, for it turns out that not everything which is conceptually necessary is analytic. Conceptually necessary truths seem to divide into two sorts: those which are analytic a priori and those which are synthetic a priori. But this ought not to seem particularly puzzling 42 An analytic truth is one which can be reduced to a logical truth by substituting terms with the same meaning. But we obviously cannot do this with moral supervenience or temporal truth-value-links. I invite the reader to try!⁴³ Similarly, consider Kant's `every event has a cause' or `all bodies are divisible'. Perhaps these doctrines are synthetic a priori constraints regulating our causal and spatial thought. These principles are not analytic; and it seems unlikely that our grounds for believing them are empirical. They seem to be presupposed by causal and spatial thought. So it is reasonable to put them down as synthetic a priori. They seem to be truths which have some kind of non-analytic conceptual necessity. These cases seem to have the same synthetic a priori conceptual status as moral supervenience.⁴⁴

If, in these various other cases, we have a pattern of modalities which is analogous to the moral case, it is enough to show either that the pattern is unmysterious or at least that we cannot glibly proceed in moral philosophy, assuming a certain upshot from a pivotal issue at the center of philosophy.⁴⁵

 

3.5. Clash of the Titans

This, then, is how I recommend taking the sting out of Blackburn's argument. According to my diagnosis, moral supervenience is an instance of the more widespread phenomenon of synthetic a priori constraints on our thinking. The problem about moral supervenience devolves upon this larger problem. What we really need to do is to tackle the big Kantian question head-on.

One reaction to the problem of how synthetic a priori knowledge is possible is to abandon realism. Kant reacted to his synthetic a priori propositions with his "transcendental idealism;" and this is rather like the way that Blackburn reacts with his moral projectivism. But no reason has been given to persuade us that this reaction is obligatory (or even that it helps).

We must first reject Blackburn's explanatory demand by saying that the conceptual necessity of supervenience is just "brute." This should then be supple­mented with an appeal to "partners in crime." There may be other brute synthetic

a priori conceptual necessities, such as color incompatibilities, or doctrines about space, time, or causality. In lieu of a solution to the "big" Kantian question, such doctrines should make us feel more comfortable with moral supervenience.

Now it may be that in this clash of philosophical titans Hume is right and that there is, after all, no synthetic a priori knowledge; the principles in question are in fact bastards of experience. Or perhaps Kant is right, and the fact that there is synthetic a priori knowledge requires that we take a more idealistic stance. Or perhaps, as I would hope, there is a coherent "transcendental realist" position: there is some source of synthetic a priori knowledge which is compatible with robust realism. But we can put what remains of Blackburn's supervenience argument into cold storage until we have sorted out this fundamental issue.

Blackburn's argument is more interesting than many have thought. But the argument has turned out not to be a metaphysical problem about the relation between moral and natural properties, but rather an  pistemological problem   If synthetic a priority is acceptable, there is nothing epistemologically mysterious; about moral supervenience, and moral realism is not queer in this respect.

 

4. MODAL MORALS

4.1. Mixing Modalities

That is the end of the main dialectic over Blackburn's argument. But what we have been through invites further reflection on the nature of the principle of moral supervenience.

The moves I have made towards defusing Blackburn's argument have nothing as yet to do with the plausibility that (N) holds with metaphysical necessity. Call that `(Nm)', where the subscript indicates the kind of modality. Blackburn's argument can get going because he has (P) at the level of conceptual modality-that is `(Pc)'. But although (N_m) is irrelevant as an initial response to Blackburn's argument, I think that it must be bound up with moral supervenience. But how?

Let us formulate supervenience roughly in the way that Kim did it in his paper, "Concepts of Supervenience." Liberally quantifying over properties, what we need is:

(Scm)      Necessarily { (EF in F)(3x)(Fx -->

(EG in G)[Gx & necessarily (Vx)(Gx -->  Fx)]) }

The two subscripts by `S' mark the kind of necessity involved at each of the two occurrences of the necessity operator. The consequent of (Scm) just is (N_m). So (Scm) says that it is conceptually necessary that (N_m) is bound up with every instantiation of an F property. That is, it may be quite contingent that something actually has a certain G* property; but given that it has the G* property, it is metaphysically necessary that it has the F property it has-that is, it must have it in all metaphysically possible worlds in which it has the G* property. So I am inclined to agree with Blackburn in thinking that supervenience would be mysterious without (Nm).⁴⁷ (Scm builds (N_m) into every F truth.

Now Blackburn seems to assume that we are either dealing with (S_cc) or (Smm), as does James Dreier in his discussion of Blackburn's argument ⁴⁸ But in morality we must mix modalities. Moral supervenience, taken as a whole, is conceptually necessary. The first necessity of (Scm) is one we can know by conceptual means. But (Scm) contains an embedded metaphysical necessity in the consequent of the overall conditional4⁹ The crucial thing is our conceptual grasp that there are some metaphysical necessities in the offing-although we may not know which.

It is not part of our competence with the concept of moral value to know that causing-pain-for-fun is evil. It is conceptually possible that the causing pain­for-fun is good and it is conceptually possible that it is evil. Yet it may be metaphysically necessary that causing-pain-for-fun is evil. Were we to know that this natural-to-moral necessity obtains, we would have a substantial piece of moral knowledge. Which metaphysical necessity obtains is, as it were, conceptually contingent. But even though it is not built into our concepts that everything which is an instance of causing-pain-for-fun is evil, it may be necessary all the same. This could not have turned out otherwise. But it might have turned out

otherwise for all we know as a consequence of what we must know in order to make moral judgements. This merely epistemic possibility might generate an illusion of metaphysical contingency in the unwary. By contrast, it is part of our competence with the concept of moral value to realize that if one instance of causing-pain-for-fun is evil, then all actual and possible relevantly similar instances of causing-pain-for-fun are also evil. In morality, we know conceptually that if a G* thing is F then it is metaphysically necessary that it is. So once we know that one case of causing-pain-for-fun is evil, we can infer that all relevantly similar actual and possible cases of causing-pain-for-fun are evil because we then know which metaphysical necessity obtains. One cannot infer that anything is evil solely from its naturalistic constitution. But if we know that one naturalistic set­up is evil, then we can infer that there are no metaphysically possible worlds in which a relevantly similar naturalistic set-up is not evil.

Contrast the relation between the mental and the physical: while it is plausible that there are metaphysical necessities connecting the mental and the physical as in the moral-natural case, it is plausible that the first necessity of psycho-physical supervenience is not knowable conceptually, as it is in morality. Competent deployers of the notion of a psychological state may fail to realize that the mental supervenes on the physicals° They may be wrong but they are not conceptually confused. So in the philosophy of mind, we have (Shun) but not (S.). Similarly with natural kinds. Of course, in the case of natural kinds, the consequent of (S) can be converted to a biconditional, giving us "reduction." But then supervenience still holds. Salmon argues that in the case of natural kinds, the reduction principle is probably held on empirical grounds. Someone who knows what 'water' means may not accept a reduction principle, even though one might be true. (As Blackburn says: "Uneducated people still need to drink and wash.")⁵¹

 

4.2. Mixed Modals and Explanation

To satisfy Blackburn, we had to explain how it could be that although each of the conjuncts of [(Ex)(Ew)(G*xw & Fxw) & (Ey)(Ew')G*yw' & ~Fyw')] is individually conceptually possible, the conjunction is conceptually impossible. What explains this-if it can be called an explanation-is the brute conceptual fact that it is built into the relevant concepts that one of the conjuncts is metaphys­ically possible while the other is metaphysically impossible. It is conceptually necessary that one of the conjuncts is metaphysically necessary while the other is metaphysically impossible, even though each of them, considered in itself, is conceptually possible. Thus, it is conceptually necessary that either in all metaphysically possible worlds G* things are F or in all metaphysically possible worlds G* things are ~F, even though neither disjunct is conceptually necessary taken by itself.

So despite (N) at the level of metaphysical necessity, the original conceptual possibility which caused all the trouble remains intact-though we should be wary of putting it in possible world terms. Blackburn's worry is that it is mysterious that it is conceptually possible for something to be G* and ~F only so long as nothing else is G* and F. He is worried about the disappearing conceptual possibility. He urges us to find it mysterious. But given the conceptually necessary principle of supervenience, once we know that there is a possible world where there is a G*/F thing, we have a ("conceptually contingent") metaphysical necessity, which, by disjunctive syllogism, renders unmysterious the disappearance of one of the problematic conceptual possibilities.

If we are asked what explains the conceptual status of supervenience, the answer is "nothing." But then as a brute synthetic a priori conceptual necessity, with companions in guilt, that may be acceptable. There are indeed problems about exactly how such synthetic a priori conceptual knowledge is possible. We need to give an account of how we know such things. Maybe there is a mystery here. But at this stage of the debate no one is justified in asserting that such knowledge is definitely not possible.

Is (Scm) a doctrine which might be called "essentialism"? Well, it is plausible that moral properties can be variably realized in natural properties; and so moral properties have no natural essence. But that leaves open the possibility that particular natural things have essential moral properties. I am inclined to think that no person is essentially good or evil (except perhaps God). But it is necessary that a person is good or evil (to some degree) so long as he has certain natural properties. So if there is an essentialist doctrine here, it is a "weak" rather than a "strong" one. Mortal individuals do not have essential moral properties. No mortal individual is necessarily evil but it might be necessary that someone is evil if he has certain natural properties. I would only be committed to a strong form of essentialism if it turned out that people have certain naturalistic properties essentially and those naturalistic properties determined moral properties. A strong form of moral essentiallim would then follow by modus ponens.

As I suggested in part 3, supervenience is a conceptually necessary modal principle constraining our moral thought which is best seen as having a synthetic a priori status. And such a synthetic a priori principle need not be thought to be beyond the pale. This modal doctrine concerning morality is not a rejectable piece of dubious philosophical speculation; it is an assumption built into moralizing. To moralize is to make cross-world commitments. Quit modalizing and you quit moralizing. Moralizing is a modal pasttime.

 

CODA

Blackburn's argument is more interesting than many have thought, and he himself sells it short. His argument is interesting and seductive because it rests on and reveals a curious and fundamental feature of our moral thought. That is the diagnosis I offer. But Blackburn has not shown that this fundamental feature cannot be explained by a moral realist. There are some puzzles remaining over moral supervenience which I have not considered in this paper. But these are not such as to threaten moral realism. As far as supervenience is concerned, we can go so far as to conclude that it has not been established that there is anything wrong with moral realism.⁵²