CHAPTER 4
NONNATURALISM
Jonathan Dancy
1. Some Distinctions
Ethical nonnaturalism is the claim that ethical properties, distinctions, and facts are different from any properties, distinctions, and facts that are worth calling natural. Ethical naturalism, as it is understood here, is the claim that all ethical properties (etc.) are also natural. The debate between these two camps is vitiated by the fact that there is no agreed account of what it is to be natural. This fact counts in favor of neither side, but it definitely counts against the sharpness of the debate; some would say that until we decide what it is to be natural, there is nothing to debate about.
Metaphysicians debate the rights and wrongs of a doctrine worth calling metaphysical naturalism: that the natural world is all there is. But one can be an ethical naturalist without being a metaphysical naturalist, and vice versa. For the debate between ethical naturalists and their nonnaturalist opponents is conducted among those who agree that there are ethical properties and facts, that is to say, among moral realists. A moral realist is someone who thinks that there are matters of ethical fact. A metaphysical naturalist who thinks that morality is not a matter of fact will neither assert or deny ethical naturalism; many noncognitivists are of this sort, and so they think that there is nothing for ethical naturalism to be about (Blackburn, 1998; Gibbard, 1990). However, most ethical naturalists are metaphysical naturalists. Indeed, as we will see, in most cases people adopt ethical naturalism as a result of some prior commitment to metaphysical naturalism.
end p.122
There is another metaethical doctrine which calls itself naturalism; this is Aristotelian naturalism (Foot, 2001; Hursthouse, 1999), so called because it holds that moral distinctions are tightly grounded in considerations of human nature. But it takes no official stand on the debate between ethical naturalism and nonnaturalism, as I have characterized it; Aristotelian naturalists could be ethical nonnaturalists. This chapter is therefore not concerned with Aristotelian naturalism.
In what follows, I first distinguish different varieties of ethical naturalism (which I will simply call naturalism from now on) and the arguments in favor of them. I then turn to the arguments on the other side. The most famous argument against naturalism appeals to the notion of normativity. I ask what normativity is and why it cannot be a natural feature or a feature of some natural thing. At the end I try to say why there is no form of naturalism that we should adopt.
2. Varieties of Naturalism
Naturalism comes in many forms. The most extreme naturalist I will consider here is Frank Jackson. Jackson (1998) offers two direct arguments for naturalism. The first of these is the simple claim that since we are natural creatures, any distinction we can grasp must be a natural distinction, one expressible (as he puts it) in descriptive terms. So if two actions are right, there must be a descriptively specifiable property that they both have; there must be a recurring descrip-tive pattern in the ways they are, despite any differences between those ways. This is not an argument that appeals to any special aspects of the moral, the ethical, or the evaluative. It is more like an argument in favor of a general metaphysical naturalism; not quite, because Jackson is arguing that for us, the natural world is all there is; so every moral property we can grasp must be natural like the others.
Jackson supports this very direct argument with another, which is much less simple and much less direct, and which appeals to special aspects of the evaluative. He argues that every evaluative predicate is necessarily equivalent to, and in fact has the same meaning as, some descriptive predicate, since for each evaluative predicate there is a descriptive predicate that it entails and that is entailed by it. He reaches this result by thinking about the relation between what he calls ethical nature (which I will call here evaluative nature, since the contrast between ethical and descriptive is peculiar)1 and descriptive nature.
Start from a representative evaluative predicate E, perhaps “is a right action,” together with an action that satisfies that predicate. There will be a complete
end p.123
description of that action and of the world in which it occurs, which is given by an enormous descriptive predicate D1. D1 will read “is an action of such and such a descriptive sort, done in a world of such and such a descriptive sort.” Now D1 entails E, by supervenience, for any action descriptively indistinguishable from the first one, and done in an indistinguishable world, must also be evaluatively similar, and so it is impossible for any other action to satisfy D1 without satisfying E. But E does not entail D1, for there may be other different actions that satisfy E, some done in the same world, others in different worlds. For each different action that satisfies E, there will, however, be a new action-in-world description D2, D3, D4, D5 ... There may even be an infinite number of such descriptions, but any action that satisfies one of them will also satisfy E. Let us form, then, the disjunction of the descriptions of all such actions-in-worlds, . What we find now is that E entails , since if E is satisfied by an action in a world, that action must satisfy one of the action-in-world descriptions D1–Dn and so satisfy the enormous disjunctive description . But also entails E, for if the enormous disjunctive description is satisfied, one of its disjuncts must be satisfied, and as I showed, each such disjunct entails E. So E both entails and is entailed by . And since E was merely a representative evaluative predicate, we can conclude that for each such predicate, there is at least one descriptive predicate that both entails and is entailed by it, that is, is necessarily equivalent to it.2
We might think that this is all fairly innocuous, but Jackson then goes on to claim that necessarily equivalent predicates pick out the same property, hence that “ethical properties are possibly infinite disjunctive descriptive properties” (1998, p. 124). And he goes further still, claiming that when enough descriptive information is in, it is impossible for two people who share the same concept of right action to disagree about whether an action is right. He allows that we humans may still need the evaluative predicate, since our access to a possibly infinite disjunctive description is probably limited. Nonetheless, he claims, the property to which mastery of that predicate gives us access is a descriptive property, and there are two analytically equivalent ways of ascribing that property: the evaluative way and the descriptive way. In descriptive terms, this property is infinite, or at least potentially infinite, in two directions, outward and inward. It is infinite outward because there may be an infinite number of different descriptive ways that right actions-in-worlds may be. It is infinite inward because there is no limit to the number of things there are to be said in describing any one of those ways. But this double enormity gives Jackson no pause.
If everything so far is sound, Jackson has established that evaluative properties are some descriptive properties or other—possibly these doubly enormous ones. But it remains also possible that they are less enormous than this. If we want actually to isolate the descriptive property that a given evaluative property is, the method Jackson proposes is that of “Ramseyfication.” Take all the things that mature folk morality would claim about (e.g.) rightness: remove the no-
end p.124
tion of “right” from them, and see which descriptive replacement for it would lead to the least disruption (i.e., keep the most such claims true). Then identify rightness with that descriptive property, whatever it is. It might, for instance, just be the simple property of being welfare-maximizing—though it probably won't be.
I said that Jackson's position is the most extreme form of naturalism. This is because of three aspects of his view. The first is the outlandish nature of the descriptive predicates3 that Jackson claims to be analytically equivalent to evaluative predicates; most naturalists who offer necessary equivalents for “is right,” say, offer much simpler ones such as that of “maximizes welfare.” The second is his claim that the two ways of ascribing one and the same property are analytically equivalent. Most varieties of naturalism do not take this last step. Such “nonanalytic naturalisms” agree that evaluative properties are natural properties, so that there are two different ways of ascribing such a property; but they deny that the evaluative way of ascribing it has the same meaning, or invokes the same concepts, as any descriptive way of ascribing it. The third is Jackson's view that there must be a descriptive equivalent for every evaluative predicate; one can imagine a weaker view that holds merely that there might be a descriptive equivalent for some evaluative predicates, though there need not be for all; that is to say, this possibility cannot be ruled out, but there is nothing to show that things must be so in general.4
Though these weaker views are different from Jackson's view, they are similar enough to his to form a recognizable family, the members of which fall into two main groups: analytic naturalism and nonanalytic naturalism. But there is a type of naturalism that is utterly different from either of these. This is the ingenious position held by Richard Boyd (1988) and Nicholas Sturgeon (1988, 2003). I will focus here on Sturgeon. He starts by suggesting that goodness and other ethical properties appear to play a causal role in the world. All of us have at some time or other benefited from the goodness of others, and part of what led to the outcry against slavery in the United States was that the form of “chattel” slavery prevalent in the United States was much worse than the sorts of slavery found elsewhere. But Sturgeon does not move directly from this to the claim that goodness and wrongness are natural properties. To do that, he would have to claim (at least implicitly) that only the natural is capable of playing a causal role. Such a claim is dubious; it is often suggested that supernatural beings, such as God or the angels, are capable of playing a causal role, since they have the habit of occasionally interfering in the ordinary course of nature. This practice, it is suggested, does not make them any the less supernatural, or any the more natural. Sturgeon manages to avoid such difficulties. He claims merely that since moral properties are capable of playing a causal role, there is no reason to invent a special new metaphysical category, that of the “nonnatural,” for them to come in. There is no reason to think of them as other than natural.
end p.125
One can hardly say that there is a spectrum here, with Jackson's view at one end and Sturgeon's at the other. Sturgeon's view differs in style from nonanalytic naturalism and from Jackson's analytic naturalism; we could call Sturgeon's view “one-term naturalism” and the opposing views “two-term naturalism.” For the latter views accept that each evaluative distinction must be capable of being captured in nonevaluative, “descriptive” terms. They deny that the evaluative way of capturing it might be the only way of doing so. For them, there are two vocabularies, evaluative and descriptive, and anything we can say or report in the evaluative vocabulary we must also be able to say or report in the descriptive one. Now this claim leads Jackson, at any rate, into difficulties. He has to characterize the difference between the two vocabularies, so that we know which terms come in which. This is not at all easy to do without begging some question. The way that Jackson hits on is to appeal to the distinction between “is” and “ought.” He writes: “By the descriptive picture, I mean the picture tellable in the terms that belong to the `is' side of the famous `is-ought' debate” (1998, p. 113). Of course, in many cases it is controversial which side of the is-ought distinction a term will fall. Many concepts seem to have a bit of both about them, such as the concept of generosity or that of a turn (as in “it is not your turn”). Jackson deals with this by saying: “If it is unclear whether a term is or is not purely descriptive, then we can take it off the list of the purely descriptive” (p. 120). In my view, this leads to a problem. Take the whole spectrum of terms, from the most blatantly descriptive (or is-ish) to the most blatantly evaluative (or ought-ish). Jackson's initial argument, which appeals to supervenience, must allow that wherever on the spectrum we draw the line between the descriptive on the left and the evaluative on the right, what falls on the right will supervene on what falls on the left, and the distinctions drawable on the right must also be drawable on the left. But the more we reduce the scope of what is to count as descriptive, the fewer the distinctions that our descriptive vocabulary will be able to draw, but the greater the number of evaluative distinctions that need to be expressible descriptively. Eventually, one would suppose, the powers of the diminishing descriptive vocabulary will prove inadequate to its increasing task. This seems to me to cast doubt on Jackson's conciliatory definition of the descriptive-evaluative distinction. (See also Raz, 1999.)
Sturgeon faces none of these difficulties. He thinks that naturalism is not a doctrine about terms or about the relation between two vocabularies, and that there is no need to claim that there must be a way of expressing in one (“descriptive,” or “natural”) vocabulary what can be expressed in another (“evaluative” or “moral”). Sturgeon's view is that moral distinctions are already natural. We don't need to find a natural equivalent for them.
Here is a brief characterization of the main types of naturalism I have mentioned.
end p.126
One-term naturalism (Sturgeon): Every normative fact is a natural fact, whether or not there is a descriptive way of capturing that fact in addition to the evaluative way.Nonanalytic naturalism: For each evaluatively capturable fact, there must be a descriptive way of capturing that same fact, though the two ways of capturing it will never be analytically equivalent.
Analytic naturalism (Jackson): For each evaluatively capturable fact, there must be a descriptive way of capturing that same fact, and the two ways of capturing it will be synonymous, that is, analytically equivalent, that is, have the same meaning.
3. Initial Comments on These Arguments
Sturgeon's examples of moral properties or facts playing a causal role in the world are always going to be challengeable. Some will say that the cause of the antislavery movement in the United States was not that chattel slavery was worse than other sorts but the fact that people came to believe it worse. We might reply that they came to believe it worse because it was worse, or because it was so blatantly worse. With this reply we enter murky waters. As for the example of benefiting from the goodness of others, the response might be that what we benefit from are those features of others that make them good, for instance their concern for their fellows, or their willingness to put themselves second. It is the good-making features that are affecting the causal order, not the goodness that they make. So the examples are disputable. In a way, what is more interesting about Sturgeon's position is the combination of two claims: that there is no reason to think of moral properties as so different from (other) natural ones that we need to create a new category of the nonnatural for them, and that nonethical terms are not likely to be explanatorily equivalent to ethical ones. (See esp. Sturgeon, 2003.) Analytical naturalists will deny this last claim of Sturgeon. If evaluative distinctions are analytically equivalent to descriptive ones, for explanatory purposes it cannot matter which way one puts things. Even if all explanation is intensional, hence sensitive to one's choice of terms, it cannot be sensitive to the distinction between analytic equivalents.
Turning now to Jackson's first argument: this seems to me to be ineffective. I accept that for each learnable predicate there must be a repeatable pattern in
end p.127
the ways things can be, to which we can learn to respond, and it may be that each such pattern is a pattern of natural or “descriptive” features. But there is no reason why the pattern itself should be “descriptive.” A pattern of natural features need not be a natural pattern. The point could not be made better than it is by John McDowell, who writes:
[H]owever long a list we give of the items to which a supervening term applies, described in terms of the level supervened upon, there may be no way, expressible at the level supervened upon, of grouping just such items together. ... Understanding why just those things belong together may essentially require understanding the supervening term. (1998, p. 202)
It is important here to notice that neither Jackson's point nor McDowell's reply is especially concerned with the relation between the evaluative and the descriptive. Both the point and the reply concern any predicate that applies, where it applies, in virtue of the application of other predicates. (Such predicates are called “resultant,” “emergent,” or “consequential.”) Jackson's general idea is that the upper-level, resultant predicate can only be learnt by seeing the pattern it picks out at the lower level. But this claim seems to me to be falsified by the design structure of connectionist machines, if for no other reason. What I mean by this is that there is at least one model of rationality, the connectionist one, in which Jackson's claim is false. In fact, there is more than one; Roschian prototypes constitute another. I argue all these things in detail elsewhere (Dancy, 1999).
With Jackson's second argument, the crucial question is whether his supervenience-based argument succeeds in establishing that every evaluative fact can be reported in descriptive terms. For this it is required, not merely that for each evaluative sentence there is a necessarily equivalent descriptive one (this just means that where either is true, the other must be), but also that the two sentences express or pick out the same matter of fact (whether they have the same meaning or not). There are three steps in Jackson's argument, taken now about predicates rather than sentences. First, for each evaluative predicate, there is a necessarily equivalent descriptive one; second, those two predicates ascribe the same property. Third, they have the same meaning. If the second step can be resisted, the soundness of the third step becomes irrelevant. And if the second step fails, we lose a significant argument for nonanalytic naturalism.