Ethics and the vision of value Chapter 4: Normative and non-normative properties It hardly need be pointed out that scientism is the primary ideology of our age. It hardly need be pointed out that the illusions scientism engenders are so pervasive and so insidious that it is practically impossible to get anyone who is subject to them to consider the possibility that they might be illusions. Peter van Inwagen, An Essay on Free Will, 1983: 215 4.1. On “the point” of normative discourse So far in this book I have described and defended a form of moral realism which, if correct, leaves moral irrealism as an unmotivated alternative (chapter 1). Then I have criticised moral irrealism on two further counts: that it leads to an untenable relativism (chapter 2), and that moral realism gives us a more credible picture of emotion and desire (chapter 3). If normative properties are, as I have argued, patterns in reality like other properties, and if perception is pattern-recognition, then recognising normative properties can be just as much a matter of perception as, say, reading is, or understanding spoken language. But to say that is not to say that there is no difference between normative and non-normative properties. The key question for this chapter is how we should characterise this difference. First, I shall argue (4.2-4.4) that a number of familiar answers to the question “What is the difference between normative properties and non-normative properties?” are wrong. The correct answer to the question is an interesting one, but more complicated than we might have expected. Briefly, the right answer is that there is no such thing as the difference between normative properties and non-normative properties. Philosophers have looked for a single, all-or-nothing distinction between the normative and the non-normative, and have been prepared to put up with considerable theoretical contortions to sustain the result that they have found this distinction. But actually there is no one distinction between the normative and the non-normative that philosophers should have been looking for. There are two: between the prescriptive and the non-prescriptive, and between the evaluative and the non-evaluative. To put it at its simplest, the prescriptive is about deciding what to do about things, the evaluative about deciding what to think about things. Prescriptive normative thinking issues in directives, or at least in recommendations or suggestions, for action. By contrast, evaluative normative thinking issues in—well, in evaluations: in views on how things stand, in verdicts or opinions or other forms of appraisal or assessment. The prescriptive is the side of normativity that is better understood, usually more emphasised, sometimes even taken to be all there is to the normative. But this is a serious mistake; the evaluative is not reducible to the prescriptive, any more than the prescriptive is reducible to the evaluative. Of course it is true that the evaluative is an input to the prescriptive: for instance, deciding that an action is good is a step on the way to deciding that I should do it. But then, the prescriptive is equally 1 an input to the evaluative: for instance, judging that someone has done the right thing in a particular case is a step on the way to judging that he is a good person. Moral irrealism has prompted distortions of this debate. Other things being equal, irrealists would like there to be a single big clear difference between normative and non-normative properties. For if there is, then they can offer a simple and tidy diagnosis of what is going on in all normative discourse. Thus A.J.Ayer tells us not only that there is no more to the property of wrongness than the speaker’s projected disapproval of the things she calls wrong, but also that his sort of emotivist analysis covers every case of normative talk. More subtly, R.M.Hare,1 in the course of offering his well-known prescriptivist account of the meaning of “good”, is ready to agree that a phrase like “a good batting wicket” can be used descriptively, to impart information about the state of the wicket to others who share a knowledge of cricketing standards; but still insists that “a good batting wicket” is, at bottom, no more than the sort of wicket that you, speaking as a batsman, would commend to other batsmen (whom you want to bat successfully, when you are speaking sincerely, assuming they find slow straight predictable deliveries easiest to deal with, and so on).2 And in recent years, Simon Blackburn and many others continue to argue that all normative talk basically serves a single purpose: that of expressing attitudes so as to guide others’ actions. All such unifying accounts of “the point of normative discourse” seem to me wrong-headed in principle. I do not see why there has to be a the point in the first place. Like any other sort of discourse, normative discourse can have all sorts of different purposes. But even if we try and find a usual purpose, or an overall purpose, or a characteristic purpose, to such discourse, it will be my contention in 4.5 that we find two purposes rather than just one. Certainly the particularities of the irrealists’ various accounts of “the point of normative discourse” get little or no support when we examine the actual evidence about the very various ways we use normative discourse. A moment’s thought about Hare’s example of the batting wicket, for instance, shows that “This is a good batting wicket” is a quite different remark, with a quite different meaning, from “I commend this wicket to you” or “May you have such a wicket as this if you are batting!”. We make quite different inferences from the remarks “This is a good batting wicket” and “I commend this wicket to you”; and we see no incoherence in making the former and withholding the latter, or making the latter and withholding the former. Hare himself, I imagine, would have conceded these points; but that unfortunately does not seemed to have deterred him from continuing his unifying project. Again, there is not the least reason to agree with Ayer that I have to approve of generosity, or be part of a social practice of approving of generosity, or anything like that, to call generosity good. As I pointed out in 3.1, the attitudes thinking X good and 1 Hare denies that he is a moral irrealist; he claims to be unable to see the difference between realism and irrealism, and hence agnostic on the issue. However, he also denies the thesis that there are moral utterances which are assertions of facts. And this thesis is moral realism, so its denial is moral irrealism. (To deny it is not merely a rejection of moral cognitivism, the thesis that moral utterances are typically assertions of facts. One could be a moral non-cognitivist and yet a moral realist, by saying that moral utterances typically express attitudes—but also that there are facts about which attitudes one should express. But this is not Hare’s position, though it was one I tried out in Chappell 1998.) 2 I am not, of course, suggesting that Hare thinks that the use of “good” in “good batting wicket” is a moral use. But he does think that basically the same account of the meaning of “good” explains this use and the moral use: both uses are prescriptive, and the moral use is also universalisable. 2 approving of X are obviously separable—and both differ from being affectively or emotionally disposed to favour X. Further, Ayer’s idea that “Theft!” [said in a scandalised tone] counts as an analysis of the meaning of “Theft is wrong” just seems to be a non-starter. (I might say “Theft is wrong!” in a scandalised tone. If I do, is that some sort of tautology?) The problem here, as with Hare, is over-simplification in the interests of generality. As for Simon Blackburn’s thesis that all ethical talk is essentially actionguiding, let me quote Alan Thomas (2006: 122-123): Here are some ethical uses of language: first, contemplating the wrong you did a friend, who is now dead, by misinterpreting his actions as selfish when in fact he was acting for your own good in a way you could not, at the time, appreciate... Secondly, thinking to yourself about the destruction and loss of something of value. Thirdly, the contemplation, in a disinterested way, of the goodness of a person far from you in time and space… Are these instances of the ethical use of language? I would say that they were. But whereas [Blackburn] only gives examples where a person tries to change another person’s attitudes, in none of these cases do you try to change another person’s attitudes. Indeed, these cases do not seem to have much to do with action or the practical at all. The irrealists’ search for a single purpose for all normative talk is misguided in principle; their proposals about what this purpose might be are unconvincing in detail. No doubt these mistakes have helped them to make a further, and logically independent, mistake—which plenty of realists have made too. This is the mistake of mischaracterising the difference between normative and non-normative properties. I look at some popular ways of making that mistake, and in particular at the “no ought from an is” doctrine, in 4.2. 4.2. How not to tell normative from non-normative properties I: “is” and “ought” One occasionally sees normative and non-normative properties contrasted as the “moral” and the “non-moral”. Taken literally—I suspect it usually isn’t—this suggestion is obviously hopeless. If there are other sorts of value besides moral value such as aesthetic or epistemic—and there are, of course—then the moral/ non-moral contrast can hardly mark the fundamental divide between the normative and the nonnormative that philosophers often look for. A second and much commoner way of trying to capture the difference at least avoids the elementary mistake made by the first way. This is the “fact/ value” contrast. It has a simple flaw: in a discussion of moral realism and irrealism, its use is question-begging. The whole point of moral realism is to claim that there are not only facts and values, but also facts about values, normative facts; and the whole point of irrealism is to deny this. Simply as a matter of dialectical fairness, we should not even consider accepting a characterisation of the normative/ non-normative contrast according to which facts and values are exclusive categories. 3 The same goes for any attempted characterisation of the contrast which plays off “descriptive” against “normative” (or against “prescriptive” or “evaluative” or something of that sort). Unless we are already irrealists, there is no reason to accept that evaluating, prescribing, or engaging in normative talk about things cannot also be ascribing properties to them, i.e. describing them. This brings us to the “is/ ought” contrast. Considered one way, this contrast just denies the possibility of is-statements, i.e. factual statements, that are also oughtstatements, i.e. have normative content, and so begs the question in the same unhelpful and uninteresting way as the fact/ value and descriptive/ normative contrasts that I have just rejected. Considered another way, however, the “is/ ought” contrast is more interesting, and deserves closer attention. As everybody knows, the contrast is first registered as a point for philosophical attention by Hume in THN 3.1.1: In every system of morality, which I have hitherto met with, I have always remark'd, that the author proceeds for some time in the ordinary ways of reasoning, and establishes the being of a God, or makes observations concerning human affairs; when all of a sudden I am surpriz'd to find, that instead of the usual copulations of propositions, is, and is not, I meet with no proposition that is not connected with an ought, or an ought not. This change is imperceptible; but is however, of the last consequence. For as this ought, or ought not, expresses some new relation or affirmation, 'tis necessary that it shou'd be observ'd and explain'd; and at the same time that a reason should be given; for what seems altogether inconceivable, how this new relation can be a deduction from others, which are entirely different from it. As Hume’s presentation of it clearly shows, the “no ought from an is” doctrine has two parts. (A) There is the thesis that no (set of) normative proposition(s) is entailed only by any (set of) non-normative proposition(s): it “seems altogether inconceivable, how this new relation can be a deduction from others, which are entirely different from it”. This is supposed to show, as before, (B) that there is a metaphysically fundamental gulf between the normative and the non-normative: “This change is… of the last [i.e. the greatest] consequence”. The “no ought from an is” doctrine does not directly give us any account of what the normative and the nonnormative are. But if the doctrine is correct, it will surely be an important part of any such account. So let us examine the doctrine in its two parts as indicated by my letters. There has been a cottage industry of devising counter-examples to (A). Some of these counter-examples are quite implausible. For instance Searle 1964’s famous argument about promising, which proceeds from the non-normative premiss that “Smith uttered the words ‘I promise to pay Jones £5’” to the normative conclusion that “Smith ought to pay Jones £5”, cannot be made valid without the addition of a whole variety of normative premisses, such as “Promises generate obligations” and “Nothing in the situation should override Smith’s obligation to keep his promise to Jones”. 4 Likewise with “functional” arguments that purport to cross the is-ought gap: the non-normative premiss “This is a knife” does not imply the normative conclusion “So this ought to be sharp” unless we add the normative premiss “If this is a knife, then this ought to be sharp”. Also likewise with the case (originally Geach 1976’s) of the sea-captain Joe, about whom, it is sometimes said, we can validly infer from “Joe is a sea-captain” to “Joe ought to do whatever a sea-captain ought to do”. To state it briefly, and without going into the detail of the arguments, this conclusion does not follow from this premiss alone. The further premiss “If Joe is a sea-captain, then Joe ought to do whatever a sea-captain ought to do” is also needed, and this further premiss is plainly normative in content. (The further premiss is not trivial or vacuous. Contrast these two quantified conditionals: (1) For any x, (x is a sea-captain) -> (x should do whatever x should do); (2) For any x, (x is a sea-captain) -> (x should do whatever a sea-captain should do). The consequent of (1) is trivially true; so (1) is true. The consequent of (2) is neither trivial nor true, and so (2) overall is false: it says that anyone who is a sea-captain ought always to act in accordance with this role, in the teeth of the familiar fact that life’s demands are too complex to fit under any role less generic than “person”. The claim about Joe is a substitution-instance of (2), not of (1). A possible rejoinder is that “If the role of sea-captain is a role that even sea-captains ought to abandon at times, then part of what a sea-captain like Joe ought to do is, on appropriate occasions, not act as a sea-captain.” But this (rather questionable) rejoinder reduces (2) to (1).) Other counter-examples to (A) are more plausible. For instance, there is the obvious fact, first pointed out in this connection by Arthur Prior (1960), that by orintroduction any p entails p v q; so any non-normative proposition entails the disjunction of itself and any normative proposition. Hence trivially, any nonnormative proposition (Prior’s example: “Tea-drinking is common in England”) entails a proposition with normative content (Prior’s example: “Tea-drinking is common in England or all New Zealanders ought to be shot”). Of course, it can be disputed whether this latter proposition is or is not essentially normative. Certainly it is indisputable that no straightforward normative proposition is asserted here (since asserting “p or q” does not involve asserting “p”). Hence it’s also obvious that this conclusion is not much use to anyone who wants actually to assert either of the disjuncts. Perhaps the most plausible counter-example of all to (A) starts from the nonnormative premisses (1) “Everything Alfie says is true” and (2) “Alfie says ‘Murder is wrong’”. These premisses pretty clearly entail the normative conclusion (3) “Murder is wrong”. To avoid that consequence, we might try various moves, some of them fairly drastic in character. We might, for instance, take the dramatic step of rethinking our attitude to “true”: we might decide to say that “p is true” is a normative claim, perhaps equivalent to the injunction “Believe p!”, and so that premiss (1) of this argument is a normative premiss. That, I think, is too crude an analysis of “p is true”, though it does capture what 2.3 called the normativity of truth. But in any case the analysis is bound to be hard to accept for any adherent of the doctrine of the is-ought gap, since it gives us a whole galaxy of examples of is-ought inferences. On this analysis, wherever p is the case, we will be able to infer from the non-normative 5 premiss “p” to the normative conclusion “p is true”. It seems hard to square the acceptance of this possibility with the doctrine “no ought from an is”. Alternatively, we might take the very hard nominalist line that Quine 1953 takes, and refuse to quantify over things people say. Or we might allow quantification over things people say, but refuse to include any normative utterances within the scope of such quantifications, on the hardline grounds that normative utterances aren’t truth-apt. Whether a defender of the metalinguistic thesis about the is-ought gap will or should be so determined to maintain his thesis that he will pay the price implicit in such hard lines is, no doubt, another thing that we might dispute over. normative proposition(s) is entailed only by any (set of) non-normative proposition(s). Would that give us a proof of (B), the thesis that there is a metaphysically fundamental gulf between the normative and the non-normative? No, it would not. To see why not, notice a point that Snare 1992 makes in his excellent introduction to the issue: if there is an is-ought gap, then there is also a nonhedgehog-hedgehog gap. That is, our reasons for saying that there is no valid inference from a non-normative premiss to a normative conclusion are pretty well exactly as good (or as bad) as our reasons for saying that there is no valid inference from a proposition that does not refer to hedgehogs to a proposition that does refer to hedgehogs. Consider this obviously invalid argument: Suppose that these debates turn out in favour of (A), the thesis that no (set of) 1. There is something in the middle of the road. 2. The middle of the road is a dangerous place to be. 3. Therefore, there is a hedgehog in a dangerous place to be. There is almost no way of making (3) follow from (1), (2) or any number of extra premisses, unless those extra premisses refer to hedgehogs (e.g. an added premiss might say (4) “If there is something in the middle of the road, then it is a hedgehog”). Since most cases are like this, the general rule “No hedgehogmentioning conclusions from non-hedgehog-mentioning premisses” seems fairly reliable. About as reliable, in fact, as the is-ought rule. After all, hedgehog references can apparently be introduced into some few conclusions of valid arguments that have no hedgehog references in their premisses (that is why I only say “almost no way”). And the list of ways of doing this is closely parallel to the list of ways of introducing oughts into the conclusions of arguments with non-normative premisses. Most plausibly, we can argue in these two fashions: 1. Tigers are fierce. 2. Therefore, tigers are fierce or hedgehogs are prickly. 1. Everything Alfie says is true. 2. Alfie says “Hedgehogs are prickly”. 3. Therefore, hedgehogs are prickly. There is, then, a nonhedgehog-hedgehog gap just as much as there is an isought gap. And of course I make this point, not because I think it reveals a metaphysically fundamental gulf between hedgehogs and non-hedgehogs, but because on the contrary I think it shows that the is-ought doctrine fails to reveal a 6 metaphysically fundamental gulf between the normative and the non-normative. Usually it is impossible to infer an “ought” from an “is”. But that fact does not bring out anything exciting about the nature of the normative and non-normative. It merely points us toward the elementary logical truth that we usually cannot introduce objects—of any sort, not just normative sorts—into the conclusions of valid arguments unless those objects were referred to in the arguments’ premisses. To say that the is-ought doctrine fails to establish a metaphysically fundamental gulf between the normative and the non-normative is not, of course, to say that there is no such gulf. Perhaps we can bring out the difference better by some other strategy: perhaps, by appealing to a distinction between the “natural” and the “non-natural”? To make sense of this distinction, however, we need to examine it in the context of the philosophical view that makes the distinction, namely naturalism— a view which deserves some attention anyway. I consider it in 4.3. 4.3. How not to tell normative from non-normative properties II: naturalism Naturalism and the “natural”/ “non-natural” distinction are definitionally connected, because naturalism says that any entities or properties that we want to posit in our metaphysics must be natural, and must not be non-natural: “Naturalism imposes a constraint on what there can be, stipulating that there are no nonnatural or unnatural, praeternatural or supernatural, entities” (Pettit 1992: 245). Such a definition of naturalism as Pettit’s only has content if we know what to make of his oppositions; if there is a worthwhile distinction between the natural and the non-natural. What then is that distinction?3 Etymology alone would guide us to suggest that entities or properties count as natural only if they are part of natura, nature. But that, of course, is no help at all unless we know what counts as nature. And “nature” is a deeply ambiguous word; in one good sense of the word, which Pettit as just quoted might seem to have in mind, it can mean everything that exists except God.4 There is little difficulty in identifying what counts as nature for the purposes of “naturalists” in the everyday sense of people who watch birds and count beetles. But exactly what sense of “nature” philosophical naturalists are after is much less clear. There seems to be considerable danger that naturalism will fall prey to one form or another of a false-or-trivial dilemma (Williams 2002: 23): “There is a well-known difficulty in stabilising the idea of ‘nature’ so that naturalism is not either trivially true or implausible to such an extent as to be uninteresting.” (The locus classicus for this difficulty is Hempel 1970.) Perhaps the best way forward is to suggest that entities or properties count as “natural” in the sense intended by philosophical naturalists only if they are recognised by science. This is certainly a common suggestion in the literature: For more on the distinction, or rather on its absence, in Aristotle’s political thought, see Chappell (forthcoming). 4 Another good sense of the word “nature”, if now a rather archaic one, is seen in Macbeth’s famous words: “Now o’er the one half-world nature seems dead; Wicked dreams abuse the curtain’d sleep; And witchcraft celebrates pale Hecat’s offerings” (Shakespeare, Macbeth 2.1.49). Here “nature” means everything that exists apart from witches, bad dreams, and the like. These are excluded not because they are supernatural (though they may be that too), but because they are unnatural. 3 7 [Naturalism is] the view that there is only one way of knowing: the empirical way that is the basis of science (whatever that way may be). (Devitt 1998: 45) To be a naturalist [is] to take the natural sciences as authoritative with respect to what there is. (Rea 2002: 56)5 Such formulations raise as many questions as they answer. Devitt’s, to take it first, is ambiguous. It obviously won’t do if we read it in what seems the most natural way, as meaning that a priori knowledge does not count as an “empirical way” of knowing. So read, (A) denies the existence of a priori knowledge, and that seems to me (though not to all philosophers) like a reductio ad absurdum of the view—since I take it that we know a priori that there is a priori knowledge. We not only know without appeal to any particular experience that 2 + 2 = 4; we also know without appeal to any particular experience that we know that 2 + 2 = 4 without appeal to any particular experience. To another common fear, that a priori knowledge is bound to be mysterious, I respond that as such examples show, it is a matter of a priori knowledge that a priori knowledge is perfectly familiar and unmysterious.6 Of course, we can also read Devitt’s definition another way: not as implying that there is no a priori knowledge, but as implying that a priori knowledge is a form of empirical knowledge. (To admit this possibility, just invoke Devitt’s “whatever that way may be” clause.) The trouble with this move is only that the category of “empirical knowledge” quickly comes to look like it could include almost anything. Hence, so does Devitt’s naturalism, since that is defined in terms of “empirical knowledge”. On this second understanding Devitt’s naturalism is vacuous. Another recent attempt to say what “natural” means which also brings us back to the a priori/ a posteriori distinction is Russ Shafer-Landau’s (in Horgan and Timmons 2006b: 212). He defines “the natural” by way of natural science, and then defines “natural science” thus: “a science is a natural science just in case its fundamental principles are discoverable a posteriori, through reliance primarily on empirical evidence.” One problem about this definition is the three vague notions at its heart: “fundamental”, “discoverable a posteriori”, and “empirical”. Another is that Shafer-Landau will presumably want his definition to be acceptable to as many naturalists as possible. But many naturalists do not accept the traditional distinction between the a priori and the a posteriori on which Shafer-Landau’s definition depends. A further account of naturalism that is very much alive to the danger of a false-or-trivial dilemma is offered by Michael Rea. Rea stresses that he is not offering an account of a thesis. Naturalism, he thinks, is not a thesis but a research programme (Rea 2002: 72): “there is no position—no substantive thesis about metaphysics, 5 For more in the same spirit see Rosenberg 1996. Some recent writings against the a priori are less impressive at close quarters than from a more distal inspection. Thus Horowitz 2006 presents a careful and detailed case against “classical apriorism”, but rests everything on this definition of the a priori: “a can know p independently of experience iff, in every metaphysically possible world in which a exists, p is true, and a has the concepts in p, a has enough experience to know that p” (Horowitz 2006: 129). As an account of what it is to know something independently of experience, this just seems bizarre. Horowitz’s linking of the question of a priori knowledge with the quite separate questions of certainty and incorrigibility doesn’t help, either. 6 8 epistemology, or methodology—that constitutes the heart and soul of naturalism. What unifies naturalists is just a shared set of methodological dispositions.” Rea thinks this because he shares a worry that I develop below, about whether the naturalist appeals to present or future science: “no substantive ontological thesis”, he tells us, “will do as a characterisation of any version of naturalism”, because any substantive ontological thesis that might take this characterising role will be a hostage to the future fortunes of science. There are naturalists who will just accept Rea’s characterisation of them, for instance Penelope Maddy: “naturalism, as I understand it, is not a doctrine, but an approach; not a set of answers, but a way of addressing questions… the only specifically naturalistic commitment in all this is to follow scientific enquiry wherever it might lead” (Maddy 2001: 37, 59). On the whole, however, I suspect Rea underestimates the willingness of contemporary naturalists to advance particular and supposedly definitive theses, rather than leaving naturalism as a research project. Most naturalists would think it a bit woolly to say no more about what ontology they are committed to than “Who knows? Let’s see how things turn out”. Given the choice between dogmatism and vagueness, they would rather be dogmatic. In any case Rea’s definition of naturalism is no less ambiguous than Devitt’s. One way to follow Rea’s suggestion, and “take the natural sciences as authoritative with respect to what there is,” is to deny that anything exists unless the natural sciences use theoretical terms which refer to it. People do take this position, of course (or at any rate philosophers do); but the view seems far too restrictive to be at all plausible. The other way of “taking the natural sciences as authoritative with respect to what there is” is to say that nothing exists unless its existence is consistent with the natural sciences. The trouble now is that this is nowhere near restrictive enough to make naturalism an interesting thesis. It is far from obvious that we have to assume that the scope of the natural sciences is absolutely everything. But unless we make that assumption, there is, as Michael Levine (2005) points out, no reason why even magic or miracles should not be consistent with the natural sciences: Laws of nature do not, and are not meant to, account for or describe events with supernatural causes—but only those with natural causes. Once some event is assumed to have a supernatural cause it is, by that very fact, outside the scope of laws of nature altogether and so cannot violate them. Of course, you could add to science a clause about its scope, and insist that the only explanations that are allowed anywhere are scientific explanations. This, in effect, is what Michael Wheeler’s “Muggle Constraint” does (Wheeler 2005, Ch.1): It seems clear that the most reliable check we have [on compliance with naturalistic constraints] is to ask of some proposed explanation (philosophical or otherwise), ‘Is it consistent with natural science?’. If the answer is ‘no’, then that explanation fails to pass the test, and must be rejected… In J. K. Rowling's Harry Potter books, there are two co-existing and intersecting worlds. The first is the magical realm, populated by wizards, witches, dragons, dementors, and the like… The second world is the non-magical realm, 9 populated by Muggles—people like us7... Now, if you want an understanding of how Muggles work, you had better not appeal to anything magical. So one's explanation of some phenomenon meets the Muggle Constraint just when it appeals only to entities, states and processes which are wholly non-magical in character. In other words, no spooky stuff allowed. This clause about the allegedly universal scope of science is not itself a part of science. Science is essentially a system of explanatory laws. Such a system can, of course, specify when one law in the system is overridden by another law in the system. But it is a logical truth that the system cannot specify what will be done by factors outside that system, as magic, if there is any, must be outside science. (If the system could specify this, those factors would not be outside it.) To add to naturalism a universal no-magic clause is not to give us any argument for thinking of magic, miracles, or other “spooky stuff” as inconsistent with science. It is merely to assert that they are. One sort of “universal no-magic clause” is supplied by the thesis called the completeness of physics, or the causal closure of the physical. It now looks likely that, if we want to find an argument for the present account of naturalism, we will need to find an argument for this claim. Can there be one? It is hard to see how. At any rate, there certainly can’t be an empirical argument for the completeness of physics. The assumption that physics is complete (that every explanation either is a physical explanation, or supervenes on a physical explanation) is an assumption that you bring (if you do) to experience, not one that you learn from experience. (If the completeness of physics were an item of knowledge, it would be an item of a priori knowledge.) The completeness of physics is advanced by naturalists almost as if it were a corollary of or parallel to the physicists’ principles of the conservation of matter and energy. It is no such thing. The principles of the conservation of matter and of energy say that, in a closed system, any physical or chemical transaction will always leave us with the same amount of matter and energy as we had before the transaction. Note well the italicised words. The real import of the naturalist’s principle of the completeness of physics is to assert that the world is a closed system. But this is a claim that the scientists’ two principles themselves carefully avoid making. What the scientific principles say is that we can usefully treat certain phenomena as if they were closed systems. For the practical scientist, the closedness of the systems in question is a useful idealisation: not a matter of doctrine, as it is for the naturalist. The scientist’s principles of the conservation of matter and of energy thus stand in no kind of parallel to the naturalist’s principle of the completeness of physics. Compared with the scientific principles’ modesty and caution, the naturalist’s principle shows up as a rather unscientific piece of dogmatism.8 Note here Wheeler’s assumption that “people like us” are people of whom no “spooky stuff” is true. In context, the point of this assumption seems to be its role as premiss in the following argument: we are people of whom no “spooky stuff” is true; therefore, a full explanation of how our minds work need invoke no “spooky stuff”. We are likely to find this argument disappointing if we are looking for information as to what counts as “spooky stuff”, or for an argument as to why we should disallow it. 8 This gap between what science really is and what naturalists would like science to be is seen elsewhere. Ironically enough, there have been distinguished philosophical naturalists who combine a professed devotion to science with an altogether unscientific refusal to accept what real science actually tells them about, for instance, the reality of global warming. 7 10 In sum: we can take naturalism, defined as Rea defines it, to be the doctrine that no property or entity exists unless it is referred to, as such, in scientific explanations. And then naturalism seems much too strong a claim to be plausible. Or we can take naturalism as the doctrine that no property or entity exists unless it is consistent with scientific explanations. But then naturalism seems much too weak a claim to be interesting, because it is not clear that even magic and miracles are “inconsistent with science” in this sense. If, despite these setbacks, we continue to look for a way of defining naturalism, and “natural” and “non-natural” in the naturalist’s sense of those terms, this dilemmatic choice between obviously false or uninterestingly trivial versions of naturalism will become something of a pattern. The pattern is repeated when we ask whether the science to which we appeal is science broadly or narrowly understood. There is no plausibility in the idea that obsesses some contemporary philosophers: than nothing can exist unless microphysics recognises it.9 There seem to be plenty of explanations and plenty of entities in such sciences as biology and chemistry—let alone psychology or anthropology—that are not strictly speaking physical explanations or entities, and which cannot easily be reduced to physical explanations. For example, two concepts that are involved in a great deal of biological explanation are the concepts of a bodily organ, and of an ecological niche. It is obvious that neither is a concept of physics; and it is not obvious that either concept can easily be reduced to purely physical concepts. So should we say that biology involves unscientific entities and patterns of explanation: that even in the sciences, there are disciplines that fail to meet the standards of naturalism? That seems to make those standards over-demanding. On the other hand, there is no bite in the idea that nothing can exist unless it is recognised by some discipline or other that we are prepared to call science: Political science and social science are by now commonplace. Many Marxists are keen to insist that historical materialism is a science… Library Science, Administrative Science, Speech Science, Forest Science, Dairy Science, Meat and Animal Science and Mortuary Science have all made their appearance on university syllabuses. (Chalmers 1997: xx.) One way of dealing with this sort of difficulty—but not a reputable one—is to maintain a studied ambiguity about the meaning of the word “science”; let it mean sometimes “physics” at one extreme, sometimes little more than “common sense” at the other, with a useful slide between these two extremes, and a powerful insinuation that the truest common sense is physics. Such, as Peter Hacker has recently remarked, seems to have been Quine’s approach (Hacker 2006: 238-239): [Quine] used the word “science” with the promiscuity characteristic of members of the Vienna Circle. Sometimes “science” means the totality of a 9 It is not only, perhaps not mainly, contemporary metaphysicians who show the greatest enthusiasm for reductive scientism. That enthusiasm crops in contemporary ethicists too, in Korsgaard for example, who is content to advocate her own ethical theory by claiming that it is the only ethics “consistent with the metaphysics of the modern world” (Korsgaard 1996: 36). Unless all the leading metaphysicians have reached a consensus without telling the rest of us, there is no such thing as the metaphysics of the modern world. 11 person’s knowledge of the external world; sometimes it means the totality of “our” knowledge of the external world; sometimes it means natural science, with especial emphasis on physics, and at others it means all natural sciences; and occasionally it means all academic disciplines concerned with truth about the world, including social sciences and history. It is very important, from context to context, to be clear what sense of “science” he had in mind… extending the term “science” to match the German “Wissenschaft” will not make history and the social sciences any more like physics and chemistry than they are, i.e. not very. The same false-or-trivial pattern is repeated again when we ask whether present or future science is the naturalist’s touchstone. The idea that there cannot be anything that is not known to present-day science is obviously stupid, while the idea that, if anything exists, then it will become known to science eventually is roomy enough to accommodate everybody, including the magician and the thaumaturge (cp. van Fraassen 1996). If we say that the world “contains just what a true complete physics would say it contains” (Mellor and Crane 1990: 186),10 then, as Mellor and Crane point out, one problem is the deeply counter-intuitive nature of this claim, given our almost unbudgeable assumption that operas, buses, gazelles and tax returns all exist, despite the fact that none of them is mentioned by any reputable physics known to us. And the other problem is that we are in no position to say what a true complete physics would contain, because we haven’t got a true complete physics. Some naturalists, David Papineau for example, face this problem about future science with equanimity: I accept that this science [the ideal future physics] will be different from current physical theory, and thus that we don’t yet know what it is. But, even so, there is no difficulty about how we know that it is complete, for we have simply defined it to be complete… [In this physics, it] seems to me highly unlikely that the psychological will turn out to be part of the physical. Current physics… aims to develop a complete theory of paradigm physical effects in terms of the categories of energy, field and spacetime structure. I am quite prepared to believe that… the categories of current physics will need supplementation before we can get a genuinely complete theory. What I do not believe is that they will need supplementation by psychological categories. (Papineau 1993: 29-31) To take such a view as this, one either has to assume that nothing is real except the physical, or that nothing is real except the physical and what supervenes on it. But again, the former view seems quite implausibly narrow, the latter too roomy to exclude anything interesting. One also has to assume a certain pessimism about the prospects for psychological categories of explanation, and a corresponding optimism about the future explanatory value of physical categories. These forecasting attitudes may or may not be intellectually respectable; but they can hardly be intellectually compulsory. Nothing obliges us, as a matter of philosophical good faith, to base our 10 Cp. Strawson 2003: 62, writing about the philosophy of mind: “Experiential phenomena… are as real as rocks, hence wholly physical”. “Hence” is good. 12 philosophical outlook on Papineau’s kind of predictions about the future. The naturalist’s position remains in part speculative, and in part dialectically unstable. The false-or-trivial pattern is repeated, too, if we try to equate naturalism with physicalism—where physicalism is the view that nothing exists that is not physical, and that any putative entity that is “non-physical” is unreal.11 Here the question, of course, is what “physical” means. If “physical property” means “the sort of property that physics is concerned with”—as it should do: this is a good clear sense of the words—then it seems perfectly obvious, for reasons I have already laboured quite enough, that physicalism is false: there are plenty of other properties and entities besides physical ones, and no sane physicist will say that it is any part of her intellectual project to deny it. If, on the other hand, “physical property” is assigned the wider and more questionable sense of “the sort of property that physics is concerned with, or any property supervenient on or emergent from that sort of property”, then physicalism, given my metaphysics of properties, will be pretty well trivially true. For as I pointed out at the end of 1.9, almost anything could supervene on the sort of properties that interest physics, including mental properties. Pettit’s response to Crane and Mellor (Pettit 1993) does not escape their version of the recurring false-or-trivial dilemma. Pettit argues for a definition on which “there may be more to the world than what is explicitly recognised in physics but all that there is… is physically constituted and physically governed”. Here everything depends on what it is for anything to be “physically constituted and physically governed”. It is not easy to see how these words might be spelled out without trivialising or a falsifying the physicalist claim. Another popular way of stating physicalism, by contrast, makes physicalism obviously false. This states it as the thesis that necessarily, if two worlds W1 and W2 coincide in all their basic physical properties, then they coincide in all their other properties. This is obviously false with the “necessarily”, since we can see no clear reason why there could not be non-physical properties in either W1 or W2 or both (especially if we don’t know what “basic” or “physical” means, and this definition does not tell us). Without the “necessarily”, this statement of physicalism is mere speculation; and not even a speculation that looks likely to be true. Take a specific application, to minds: if two minds M1 and M2 coincide in all their basic physical properties, then they coincide in all their other properties. This claim is surely false; my being in the same neuronal state as you does not guarantee that we are thinking the very same thoughts. The conclusion to draw from this catalogue of false-or-trivial dilemmas is that the word “naturalism” is best used to describe what bird-watchers and beetle-counters do. In metaethics at least,12 the word has not been attached to any thesis that is both 11 This equation of naturalism and physicalism is rejected by Chomsky 2000, who argues that naturalism, broadly conceived, is a universal aspiration for inquiry, whereas physicalism is a specific and very dubious doctrine that is supposed to follow from and be justified by naturalism, but in fact does not have any very clear justification at all. (As we might also say, Chomsky thinks naturalism is trivially true, and physicalism is substantive but false.) 12 Outside metaethics, philosophers have had more success with the word, since elsewhere in philosophy, “naturalism” can at least name the denial of supernaturalism (the belief in God or other transcendent beings). This thesis is, at any rate, substantive. 13 substantive and likely to be true; nor, therefore, have “natural” and “non-natural”. It follows that we cannot characterise the difference between the normative and the nonnormative by way of the supposed distinction between the natural and the non-natural; for that distinction is illusory. So if you ask me whether the patterns in reality that I take moral properties to be are “natural” or “non-natural”, my answer is simply that I do not understand the question—and doubt that there is anything for me to understand. Some readers might have noticed by now that so far my discussion of the “natural” and the “non-natural” has left out one obvious point of reference: G.E.Moore, the inventor of this terminology. I turn to him in 4.4. 4.4. How not to tell normative from non-normative properties (III): the open question argument Moore writes: [I]t may easily be thought, at first sight, that to be good may mean to be that which we desire to desire13… But, if we carry the investigation further and ask ourselves “Is it good to desire to desire A?” it is apparent, on a little reflection, that this question is itself as intelligible, as the original question “Is A good?”—that we are, in fact, now asking for exactly the same information about the desire to desire A, for which we formerly asked with regard to A itself… It may indeed be true that what we desire to desire is also always good; perhaps, even the converse may be true: but … the mere fact that we understand very well what is meant by doubting it, shews clearly that we have two different notions before our minds… [It is similarly intelligible to doubt any definition at all that is offered of good.] Good, then, is indefinable… so far as the meaning of good goes, anything whatever may be good… [but if we] start with the conviction that good must mean so and so, [we] shall therefore be inclined either to misunderstand our opponent’s arguments or to cut them short with the reply, “This is not an open question: the very meaning of the word decides it: no one can think otherwise except through confusion.” (Moore, Principia Ethica Ch.1, §§13-15) There is an important and correct idea in this famous passage. So it is unfortunate that Moore commits a number of obvious mistakes in his presentation of it, for which he is frequently taken to task.14 In particular, Moore commits himself to the absurd view that any definition or analysis can only be correct if it is unintelligible to doubt its correctness. (This mistake Moore was prepared to defend explicitly, under the name of “the paradox of analysis”.15) The argument’s flaws should not distract us from the much more promising idea that is its real driving force. Moore’s idea is that we can imagine dialogues like these: For my critique of the “dispositional view of value”, see 3.4. See Miller 2003, Ch.2, for useful further discussion of Moore and the literature on Moore. 15 See.Schilpp 1942: 660-667, and Max Black 1944’s solution of the supposed paradox. 13 14 14 A. X is the action which will maximise the level of preference-satisfaction involved in the consequences. B. Yes, but is X the best thing to do? A. Y is an action which promotes flourishing. B. Yes, but is Y good? A. Z is the action which is most in accordance with nature. B. Yes, but is Z right? A. W is an action which does not breach the Ten Commandments. B. Yes, but is W permissible? Every one of B’s questions is a reasonable one for him to ask. B’s questions do not display ignorance of synonymies or of simple logical consequences, or anything like that. These are reasonable questions for B to ask, because (to use Moore’s terminology) they are open questions. There is nothing in A’s claims to make any particular answers to them logically or rationally inevitable. Intuitively, saying that an action maximises the level of preference-satisfaction is not enough to deliver the verdict that it is the best thing to do. Saying that it promotes flourishing is not enough to deliver the verdict that it is the right thing to do. Showing that it is most in accordance with nature will not show that it a good thing to do. And showing that it does not breach the Ten Commandments will not show that it is permissible. True, the things that A says give evidence that might help us reach verdicts about what to do. Maybe actions that (e.g.) promote flourishing are usually good actions; if so, knowing this does give us something to go on in our deliberations. But even if actions that promote flourishing are always good actions, their being good does not seem to follow from their promoting flourishing. Such claims as A’s still leave open the verdicts that B is looking for: they do not settle them. Even if we know that an action will maximise the level of preference-satisfaction involved in the consequences, it is still what Moore calls an open question whether this action is the best thing to do. In the four dialogues above, let us call the gap between A’s terms and B’s terms “the open question gap”. Here then are our questions. First, how should we characterise the terms on either side of the open question gap? And secondly, what is the relation between them? To start with the second question, I think Moore himself must have supposed that the relation holding between the terms on either side of the open question gap was simply difference. This certainly seems to be the point of his epigraph to Principia Ethica, Bishop Butler’s “Everything is what it is, and not something else”. About this, as we shall see, Moore is more or less right, and importantly right.16 16 If a difference between two concepts, or sorts of concepts, is all that an open question argument aims to identify, then it will be hard to mount a complex argument for such a simple claim; and it ought to be unnecessary. In this case, moreover, it will not be to the point for naturalists like Ridge 2003 (cp. Harman 1977: 19-20, Frankena 1973: 99-100) to call it “question-begging” to run an open question argument. If a demonstration that two concepts are different is all the open question argument aims to give, then there isn’t really space for there to be a question to beg. (Cp. Huemer 2006.) 15 On the first question, about what kind of terms we have on either side of this gap, Moore is more explicit. He is talking about a distinction between what he calls the “natural” and the “non-natural”—the failure to mark that distinction being the “naturalistic fallacy”. What did he mean by these terms? The good news is that he was not defending anything like the implausible modern distinction between “natural” and “non-natural” that I have just criticised in 4.3. The bad news is that Moore’s own account of that distinction, in chapter 2 of Principia Ethica, is even less plausible. According to Moore the natural is the temporal, the non-natural the timeless: as if a man’s or a deed’s goodness should last for ever, while the man or the deed goes the way of all flesh; and as if “Friendship is good”, “Lilac is a colour”, and “2 + 2 = 4” had something in common that “He is my friend”, “Monet is fond of lilac”, and “I gave you two two-pence pieces” lack. To the purposes for which the natural/ nonnatural distinction has usually been wanted, including the purposes for which Moore himself wanted it, his proposal about what the distinction is a distinction between just seems wildly irrelevant. Moore is wrong, then, in his answer to the first question. He is prevented from answering it right by a false presupposition that in fact underlies both questions. This is the presupposition that there is just one open question gap, and hence just one distinction between kinds of things lying on either side of that gap. Moore himself, unfortunately, does nothing to draw attention to this presupposition, and much to help us to swallow it without even noticing it. In particular, his single-minded determination to apply the method of analysis to quarry out the supposed implications of the concept “good” obscures from us the variety of other moral concepts that he might also have subjected to analysis. Moore’s strategy is to focus exclusively on “good”, and assume that once we know what we want to say about that, we will be able without too much difficulty to work out what to say about all the other moral terms. (Compare Richard Hare’s strategy of focusing pretty well exclusively on “ought”, with consequences for Hare’s broader ethical views that are closely parallel, and just as disastrous.) We need no more than the four little dialogues given above to see a crucial point that Moore apparently misses, because of his narrowness of focus: namely that there is not just one open question gap, there are lots of them. Moore was right, of course, to notice that descriptive terms such as “maximising preference-satisfaction”, “promoting flourishing”, “being in accordance with nature”, and “not breaching the Ten Commandments” all face the open question argument, because there is a logical and a semantic gap between saying that these terms are true of something, and saying that that thing is good. What he apparently didn’t notice is that, if his open question test generates one open question gap, then it generates swarms of the things. For a start, my four little dialogues between A and B introduced four open question gaps: between non-normative terms such as “promotes flourishing” and four different kinds of normative terms, “best”, “good”, “right” and “permissible”. Again, it will always be an open question in Moore’s sense to ask, “I know this is best/ good/ permissible/ just… but is it right?”. This implies that there are also open question gaps between “right” and four other sorts of normative terms, namely “best”17, “good”, “permissible”, and terms for the sorts of “thick” ethical 17 In Principia Ethica sections 17 and 89, Moore develops the suggestion (which he later rejected) that “right” is to be defined as “productive of the greatest possible sum of good”. If this is how to define “right”, then there is no open question gap between “good” and “right”. However, Moore’s own test 16 properties that are exemplified by the virtues. Similarly, “I know that this is best, but is it good?” and “I know that this is best, but is it permissible?” and “I know that this is good, but is it permissible?” and “I know that this is just, but is it kind?” (…etc.) all seem to be what Moore calls open questions too. So it turns out that there are open question gaps all over the metaethical landscape. That, of course, is hardly surprising if Moore was more or less right, as I said he was, in his answer to the question about the relation between the two sides of open question gaps. The answer that I read into Moore was that the relation between the concepts on either side of any open question gap is simply difference. This “more or less right” answer becomes exactly right if we strengthen difference to independence, where two concepts are independent if and only if they are different concepts that do not interdefine and neither of which is entailed by the other.18 Then it is hardly surprising if there are lots of open question gaps in our metaethics, because there are lots of independent concepts in our metaethics. Indeed there are lots of independent concepts in our thought in general. Compare here what the last section pointed out, that the same arguments as establish an is/ ought gap also establish a hedgehog/ non-hedgehog gap. The right conclusion about the open question argument is the same as about the doctrine of the is-ought gap. Proponents of both aim to show that there is a logical and semantic gap between the normative and the non-normative; and their arguments succeed. However, it is not clear how much follows from this success: for such gaps are commonplace. There is not only an is-ought gap but also a hedgehog-nonhedgehog gap; not only is “It maximises well-being, but is it good?” an open question, so is “It is coloured, but is it red?”.19 The difference between the normative and the non-normative is not fully captured by anything like Moore’s open question argument. The doctrine of “the naturalistic fallacy” is not of uniquely metaethical significance. 20 The essence of that fallacy is the mistake of failing to see independent concepts as independent. But independent concepts occur throughout the conceptual landscape; so there is nothing specifically metaethical about this mistake. All the same, Moore’s attempt to characterise the normative/ non-normative relation gets us further than the other attempts that I have looked at so far in this chapter. One reason for this is the greater generality of Moore’s argument, compared tells against this definition. The question “I know this act will produce the greatest possible sum of good, but is it right?” seems just as open as the question “I know this act accords with what I desire to desire, but is it good?”. This lacuna in Moore’s theory gives evidence, it seems to me, of just the kind of hasty assimilation of what we say about “right” to what we say about “good” that we might expect to follow from Moore’s over-exclusive focus on “good”. 18 But remember here the important qualification noted in 1.6 about what entails what. There can be entailment relations between properties (and concepts of properties) that we simply do not see, and could not see without a complete theory of the interrelations between the patterns in the world that properties are. In many cases such a complete theory is inaccessible to us—in some cases, perhaps inevitably so. 19 I suspect this is why some passages of Chapter 1 of Principia Ethica (section 12, for instance) tell us that there is an open question gap between the term “yellow” and any proposed analysis of that term. Perhaps it is a merit of my interpretation of Moore that it explains this puzzling remark. 20 Though Moore himself is less than clear on this issue, he apparently did not think this: [the mistake with yellow is the very same mistake] ref. 17 for instance with Hume’s. The is-ought gap is just one instance of conceptindependence; the open question gap is the phenomenon of concept-independence. Also, Moore’s argument puts us on to the idea of the independence of concepts—an idea that, of course, we might also have been alerted to by Hume’s question “how this new relation can be a deduction from others, which are entirely different from it”. And in the notion of an open question, Moore gives us a way to test for conceptindependence. As I will show in 4.5, this notion of concept-independence is an important part of what we need to make sense of the relations between the normative and the non-normative. 4.5. The prescriptive, the evaluative, and the merely descriptive The open question argument turns on the idea of concept-independence, a phenomenon of quite general philosophical interest, in no way particular to the study of metaethics. What is of specifically metaethical interest is the grouping of independent ethical concepts around two sorts of activity: the guiding of action, and the formulation of ethical assessments. These are the two activities that I summarily call prescription and evaluation. My thesis about prescription and evaluation is that they are irreducibly different sorts of normative activity, and as such generate two irreducibly different groupings of independent normative concepts. Both prescriptive and evaluative concepts are reason-giving; furthermore, there are concepts which are both prescriptive and evaluative. The two sorts of reasons are, nonetheless, reasons of different sorts. An evaluative reason is a reason to take a particular view of something—to evaluate it a particular way, and, if you are a good person, to have corresponding attitudes of approval and corresponding emotions (which will not necessarily follow unless you are a good person: 3.1). A prescriptive reason, by contrast, is a reason to act. As I said in 4.1, the evaluative is not reducible to the prescriptive, nor the prescriptive to the evaluative: we use evaluations to help us formulate prescriptions, but we also use prescriptions to help us formulate evaluations. This claim gives us a diagnosis of what goes wrong with Hare’s view that even talk about a “good batting wicket” is ultimately commendation rather than description. Hare assumes that all ethical talk has a single purpose, practical commendation or action-guidance, and interprets every use of moral language as serving this purpose—no matter how strained this interpretation turns out to be in particular cases. If we keep the evaluative separate from the prescriptive, and simply give up on the misguided attempt to reduce all normativity to prescription, it is easy to avoid this mistake. We can recognise talk about a “good batting wicket” as what it patently is: a descriptive assessment made according to the standards of cricket, which needs no relation at all to any practical upshot to give it sense—in short, an evaluation. Likewise with the examples of non-practical moral reflection that Alan Thomas produces (4.1) as evidence against the emotivist/ expressivist thesis that all moral utterance is essentially about tugging at people’s attitudes with a view to action-guiding: thinking about past wrongs, assessing the virtues of some character from history, and so forth. It is very contrived to treat these as examples of prescription in some distal or diluted sense; it is entirely natural to treat them as examples of evaluation. If we take evaluative normativity to be something quite 18 different from prescriptive normativity, as I am proposing, we can avoid the contrived treatment. That evaluative and prescriptive terms are irreducible to each other is best shown, of course, by demonstrating that there is just as much an open question gap between the prescriptive and the evaluative, as there is between either the prescriptive or the evaluative and the descriptive. And so there is: the fact that an action satisfies such evaluative descriptions as “brave” or “just” or “wise”, for instance, does not show that there is reason to do it. An action can satisfy some virtue-description while also satisfying some vice-description: if an action is brave, but also unjust, unwise, and unkind, it is pretty likely not to be one that anyone should do. Even the evaluative fact that an action is good need not imply the prescriptive fact that it is to be done (other salient21 options might be better); nor indeed does the evaluative fact that an action is bad imply the prescriptive fact that it is not to be done (other salient options might be worse). Thus there is an open question about how to get from evaluations to prescriptions. And conversely there is an open question about how to get from prescriptions to evaluations: that someone “does the right thing”, or “acts as he ought”, or “does what is required”, does not imply that he acts well or virtuously. “But surely saying that an action is e.g. wise gives us some reason to do it, even if not a decisive one; and surely saying that an action is e.g. right gives us some reason to think that it is good, even if not a decisive one.” This is correct, but it does not show that prescription and evaluation are two quite different sorts of normativity. What it shows is that evaluation is an input to prescription, and prescription is an input to evaluation. That they are inputs to each other does not make them the same sort of judgement. After all, non-normative description too is an input to evaluation and to prescription, but the difference between non-normative description and evaluation or prescription is clear if any difference is. “But formulating evaluations is really a step on the way to prescription: evaluations are non-decisive prescriptions. Saying that some action is e.g. brave, or wise, is an evidential claim; saying that the action is right, or should be done, is a verdictive claim. What you call evaluation is therefore just a preliminary to prescription.” This is a popular picture of the relation of the evaluative and the prescriptive. It is the picture implicit, for example, in many treatments of the relation of “thick” to “thin” ethical terms.22 However, the only part of this popular picture that I accept is the distinction between verdictive and evidential terms. There is indeed a verdictive/ evidential distinction; in fact there are two such distinctions, because evaluative and prescriptive judgements alike can be either evidential or verdictive. The classic virtue terms—“brave”, “wise”, “just”, self-controlled”—are all evidential evaluative terms; “good” and “virtuous”, are verdictive evaluative terms. “Advisable”, “prudent”, “sensible”, “a good idea”, “wise”, are all evidential prescriptive terms; “right”, “the thing to do”, and “morally required” are verdictive prescriptive terms. (Notice here the interesting point that some evidential terms, such as “prudent” and “wise”, can be used both evaluatively and prescriptively. This point is no threat to my thesis that evaluation and prescription are fundamentally different sorts of 21 22 The purport of this word will become clear in 5.5. For two interesting discussions of the thick and the thin see Scheffler 1987, Blackburn (online). 19 normativity; as already pointed out, I just accept the claim that the normative and the evaluative are inputs for each other.) We can cash out these two verdictive/ evidential distinctions by talking about reasons. All normative assertion is intrinsically reason-giving; that is what distinguishes the normative from the non-normative. (Here, then, is my answer to the question posed at the end of chapter 3:“What is the difference between a valueproperty and any other sort of property?”.) However, some normative assertions only give us prima facie or pro tanto reasons: these are the evidential ones. While others give us decisive reasons: these are the verdictive ones. Evidential evaluative assertions give us prima facie reasons to think something good (or bad); verdictive evaluative assertions give us decisive reasons to form such evaluations. Likewise evidential prescriptive assertions give us prima facie reasons to act (or abstain); verdictive prescriptive assertions give us decisive reasons to act or abstain. In a table, with examples: 20 Evidential Verdictive Non-normative Normative I: (merely evaluative descriptive) “The litmus has “He is a gone red” person”/ “It brave action” “There is acid in “He is a the flask” person”/ “It good action” Normative prescriptive II: brave “It is an advisable is a action” good “It is the is a action” right “But all this is too complicated; we have lost the simplicity of the analysis of moral language offered by authors like Hare.” –Well, there is no theoretical virtue in simplicity unless the theory is also correct. “And the metaphysics is bound to be baroque.” Not at all; as I argued in chapter 1, all the metaphysics that moral realism needs is the thesis that properties, including moral and other normative properties, are patterns in the world. The essential simplicity of the metaphysics generated by that thesis is not undermined by distinctions between different sorts of normative property such as I have offered here. But how, someone might ask, does my theory of properties explain the present claim that a pattern constituting a merely descriptive property can ground a pattern constituting an evaluative property, or that a pattern constituting an evaluative property can ground a prescriptive property (or vice versa)? This question is rather like the question that drives Blackburn’s puzzle about supervenience (1.6). Indeed, the answer to it is so similar to the solution that I offered to that puzzle that I can state it in almost the same words. The claim that, if A and B share one set N of properties of one sort, then they must also share another set M of properties of a second sort, comes out on my analysis as a claim about the necessary relations of patterns. The claim will take the form: “If A and B both exemplify one pattern NP, then they must both exemplify another pattern MP”. Pace Mackie (1977: 41: “But just what in the world is signified by this ‘because’?”), we can explain this explanatory connection, this “must”, without any difficulty at all. We cash it out by explaining e.g. how, if anything is a sixteen-dot matrix like the one displayed in 1.5, then it must be possible to see a cross pattern in it, and a four-square pattern, and a three-corridor pattern (vertical or horizontal) etc.; or how, if anything is a duck-rabbit, then it must be possible to see a duck in it; and also a rabbit; etc. “But this response makes the relation of different sorts of properties a matter of entailment—thus undermining the claim that there are open question gaps between the different sorts of properties.” Yes and no. Yes, because the relationship is indeed entailment, or something very like it, at the level of the analysis of the patterns. No, because the relationship is nothing like entailment at the level of pattern-recognition. Theoretical inference and representational perception go their different ways to the same destination, knowledge. Working out a complete, God’s-eye metaethics would mean so thoroughly explaining the relations that hold between the merely descriptive, evaluative, and prescriptive properties that those relations would, in the theory, 21 become (very complicated and long-winded) entailments. But theoretical inference is one thing, and representational perception is another. Since perception’s ambition is not to turn into theory, but to get more accurate as perception, there is no reason to say that someone who does not grasp these inferential relations is cognitively defective (as opposed to: unversed in the complete true metaethics). Since, on the other hand, the relation between the various sorts of property is something that can in principle be explained by the sort of realistic moral theory that I am developing here, there is no (in-principle) mystery about why the perceptions of certain patterns of different sorts go together. Thus the present theory can explain both why we get open question gaps (evidencing relations) between the properties of the different sorts, and also how, at least in principle, those properties are interrelated. Questions remain, of course, about how these sorts of property interrelate. In particular, notice a certain “narrowing down” that happens between the evaluative and the prescriptive: the disparity between the very rich, “thick”, and various stock of evaluative notions that we work with, and the rather small and “thin” range of prescriptive notions that are available.23 At the extreme practical end of the spectrum, things thin right down to uniqueness: an agent can only do one action at a time. How then do we get from the rich variety of the evaluative to the thin selectiveness of the prescriptive—and from there to the unique determinacy of action itself? This question—call it the question of the good-to-right function—is the main question that I want to address in chapters 5-6.24 Before I address it there, let me offer a preliminary study of two other sorts of normativity—the epistemic and logical kinds. This chapter’s picture of the varieties of moral (or better, practical) normativity will be reinforced if it turns out that there are parallel varieties in the epistemic and logical cases: and it does. The next chapter’s picture of how practical judgement works will also be supported if it turns out that epistemic and logical judgement work in a parallel way: and it does. 4.6. Epistemic and logical normativity25 We might expect there to be an evaluative/ prescriptive distinction, or something like it, in areas other than ethics. After all, logic and knowledge, for instance, are normative realms too. (Frege [1997]: 228: “Like ethics, logic can also be called a normative science”.) In the case of knowledge we have the term “true”, in the case of logic the term “entails”. There seems prima facie to be a good case for saying that these terms are verdictive prescriptives: what is true, you might say, we have to believe, and what is entailed by what else we believe, we also have to believe. There may be yet other normative realms as well. Perhaps the semantic, for instance:26 surely the fact that Jean-Pierre knows that Miquet is a mouse gives Jean“Theory typically uses the assumption that we probably have too many ethical ideas, some of which may well turn out to be mere prejudices. Our major problem now is actually that we have not too many but too few, and we need to cherish as many as we can” (Williams 1985: 117). 24 Obviously it is also possible to ask how we get from the prescriptive to the evaluative; but I shall not focus on that question here. 25 There is some (unavoidable) overlap between what I say here about the normativity of truth, and my remarks about the same topic in 2.3. 23 22 Pierre reason to assent to Miquet est un souri, and maybe Hans’s saying truly Dies ist grűn gives Hans reason to say of anything else that is sufficiently similarly coloured Dies auch ist grűn. Or the mental: maybe attributing a belief to someone that x is F commits you to interpreting him, not as someone who is indisposed to assenting to “x is not F”, but as someone who should be thus indisposed.27 Or again there is the modal: saying that something p is necessarily true is, perhaps, saying that we have reason to take a particular attitude to p’s truth; we should not only hold p true, but hold it true in a very particular way (with the greatest possible intransigence?). For present purposes, however, I will restrict myself to contrasting the epistemic, logical, and moral cases, since these are less contentious and less technical (though still very contentious, and quite technical). So is it right to say that “what is true, we have to believe”, or that “what is entailed by what else we believe, we also have to believe”? No, it is not. But the reason why neither of these claims is correct sheds no doubt on the thesis that “true” or “entails” are normative terms, and so essentially reason-giving terms. What it sheds doubt on is the thesis that either is a decisive-reason-giving term—a verdictive term— rather than a reason-giving but not decisive-reason-giving term—an evidential term. To see this, consider the positions I’ll label epistemic maximalism and logical maximalism: Epistemic maximalism: ( p) ( x) p → O (Bx, p) (Read as: “For any proposition p and any believer x, if p, then x is obliged to believe p.”) Logical maximalism: ( p) ( q) ( x) p & (p →q) & Bx, (p & (p → q)) → O (Bx, q) (Read as: “For any propositions p and q and any believer x: if p and if p then q, and if x believes that p and that if p then q, then x is obliged to believe that q.”) What should we make of these positions? Let us take epistemic maximalism first. As is frequently pointed out, this position is obviously insane. No one can be required to believe every truth, however trivial, useless, or irrelevant to his or her life. (There was some particular number of hairs on Augustus Caesar’s head at 4 pm on the Ides of March 24 BC. Am I obliged, do I even have reason, to have a true belief about how many?) So “true” is not a decisive-reason-giving term: the fact that a belief is true does not, all on its own, give anyone decisive reason to adopt it. Does it follow that “true” is not a reason-giving term at all? That would be a plain non sequitur: just because “It is a central ingredient in understanding an expression to grasp that there are associated with it conditions for its correct application. Put another way, it is essential to any expression’s possessing whatever meaning it does, that there are rules for its correct use. In this sense, meaning is normative” (Hale and Wright 1997: 674). 27 That there is this sort of normativity about the mental is brought out nicely by Hacker 2006: 249: “Maybe some specific neural state is a necessary condition for someone’s believing that p, but his believing that p could not be identical with that mental state. Otherwise, inter alia, one would be able to say ‘I believe that p’ (referring thus to one’s mental state), but it is not the case that p’.” Any belief is a commitment (among other things, to not contradicting the belief by coming out with Mooreparadoxical utterances); but no neural state is a commitment. 26 23 you don’t seek as much as possible of a given value, it doesn’t follow that you don’t hold it to be a value at all. So in place of epistemic maximalism, we should propose something like the following claim: Epistemic non-maximalism: ( p) ( x) p & P (x, p?) v O (x, p?) & (x, p?) → (OBx, p) (Read as: for any proposition p and any believer x: if p, and if x is epistemically permitted or obliged to inquire whether p, and if x does inquire whether p, then x is epistemically obliged to believe that p.)28 What the familiar line of argument against the property of truth as a value (cp. 2.3) shows is the falsity, indeed craziness, of epistemic maximalism. However, rejecting epistemic maximalism is not tantamount to rejecting the thesis that truth is a value, because epistemic non-maximalism is also available. Epistemic nonmaximalism gives us a non-promoting or non-maximising attitude to the value of truth: it doesn’t oblige us to believe anything and everything that happens to be true. Truth being an evidentially normative property, there is always some reason to believe what is true; but we are only obliged to believe what’s true where it falls within the scope of our obligations or permissions that we should have some belief about a given subject-matter. How are those obligations and permissions determined? The global goal of inquiry is not truth, nor even knowledge, but wisdom and understanding; this stands to particular decisions about what to inquire into as eudaimonia stands to particular decisions about what to do. The idea that I should believe everything that’s true is just as unreasonable as the idea that I should do everything that’s good, and in the same way: simply, there are too many good things for me to do them all, and too many true beliefs for me to hold them all. What I must do is what’s right, which is defined partly by what any agent is required to do on any project of eudaimonia, and partly by the constraints of my personal project of eudaimonia. What I may do is what’s permitted, and focal to my personal project of eudaimonia. Likewise what I must believe is that part of the totality of what’s true which is fixed by the constraints (a) that it is what any believer would have to believe to attain understanding, and (b) that it is what I as a believer have to believe to attain the particular kinds of understanding that I particularly aim at. So “good” stands in ethics roughly where “true” stands in epistemology. “Good” is, as we saw in 4.5, is a verdictive evaluative term, but only an evidential prescriptive term: while it constitutes a decisive verdict in evaluative assessment, it is only apt to justify assertions about what we ought to do. Likewise “the true” is not “what we must believe”; but truth, as well as being of interest in its own right, is an indispensable part of the ground for believing anything. What term in epistemology stands where “right” stands in ethics—in the verdictive prescriptive role? I am not Doesn’t epistemic non-maximalism needs something more alongside the condition that “x inquires whether p” to generate an obligation for x to believe that p? After all, x’s inquiry might, through no epistemic fault of x’s, get absolutely nowhere, and then surely x would have no obligation to believe either p or not-p. The answer to this objection (which I thank Carrie Jenkins, Gary Kemp, and Philip Percival for raising) is that certainly x can incur no subjective epistemic obligation in that sort of unsuccessful inquiry. But x can incur an objective epistemic obligation. Parallel points will hold in the entailment case. 28 24 sure any term plays this role in ordinary language. Some term could be invented: “tobe-believed”, perhaps. The point about epistemic prescriptivity, to repeat, is that it is not merely a matter of truth: what we have to believe is not whatever is true, but rather whatever is true and relevant to our epistemic goals—and those goals are always more specific than “the whole truth”. To anticipate the argument of chapter 5, there are striking analogies that can be developed between the way we get from the evaluative to the prescriptive, and from the evidential to the verdictive, in epistemology, and the way this happens in ethics. We have discretion, within limits, about which truths to inquire into; we also have discretion, within limits, about which goods to pursue. Our discretion about which truths to inquire into is not best exercised arbitrarily, but with an eye on the overall goal of the epistemic life—understanding.29 Likewise, our discretion about which goods to pursue is not best exercised arbitrarily, but with an eye on the overall goal of the ethical life—eudaimonia. Again, there are bounds on what epistemic attitudes we may adopt to any truth whether or not we are inquiring it; there is a sharp distinction between not believing a truth (often but not always epistemically permissible) and believing an untruth (always epistemically impermissible). Similarly, there are limits to we may do to any good whether or not we are pursuing it; there is a sharp distinction between omitting to pursue a good (often but not always permissible), and actively violating it (always impermissible). And what, very briefly, about logical maximalism? I take it this position is just as insane as epistemic maximalism. No one can possibly be required to believe every consequence of all his true beliefs, because no one ever could believe all these consequences. Instead we should adopt a logical non-maximalist position, according to which we are obliged to believe the logical consequences only of those beliefs whose logical consequences we are permitted or obliged to inquire into: Logical non-maximalism: ( p) ( q) ( x) p & Bx, p & (p → q) & O (Ax, (p → q)?) v P (Ax, (p → q)?) & (Ax, (p → q)?) → O (Bx, (p → q)) & O (Bx, q) (Read as: “For any propositions p and q and any believer x: if p, and x believes p, and p entails q, and if it is rationally obligatory or rationally permissible for x to inquire whether p entails q, and x does inquire whether p entails q, then x is epistemically obliged to believe that p entails q30, and x is rationally obliged to believe q.”) The upshot, as in the epistemic case, is not that entailment is not a normative notion at all. Rather, it is that “entails” is an essentially reason-giving term, but not a decisive-reason-giving term. The truth of “p entails q” essentially gives us reason to believe q if we believe p. But it does not give us a decisive reason to believe q if we believe p, since the question whether q is a consequence of p may be one that we simply have no reason to consider. Truths have all sorts of consequences; but only “Wisdom is the principal thing; therefore get wisdom; and with all thy getting get understanding”: Proverbs 4.7 (Authorised Version). 30 As the formula says, this obligation to believe that p entails q, because it is true that p entails q, is an epistemic obligation, for which the rationale is given by epistemic non-maximalism. Contrast the logical or rational obligation to believe q, because p is true and q is entailed by p. 29 25 some of these are interesting. (For example, any logical truth is a consequence of any truth; but logical truths are only interesting under very particular conditions.)31 All this might seem to have an interesting incidental consequence: it might seem to mean that logical non-maximalism offers comfort to opponents of scepticism. Many anti-sceptics argue for the denial of the Closure Principle—the principle that knowledge is closed under known entailment, so that no one can know (1) “I do not know that I am not being tricked by an evil demon” and (2) “If I do not know that I am not being tricked by an evil demon then I do not know that I have hands” without also knowing (3) “I do not know that I have hands”. Now it seems clear from the last paragraph that logical non-maximalism involves a denial of the Closure Principle too. So is logical non-maximalism just what the anti-sceptic needs? No, it isn’t, because what logical non-maximalism says is that we have permission not to accept logical consequences where we are permitted or obliged not to inquire about them. But presumably, we are all permitted (and professional epistemologists are perhaps obliged) to inquire about sceptical arguments like the one I’ve just given. As soon as I so much as raise outlandish sceptical questions like “Do I really have hands?”, they pop up within my project of inquiry. And then, precisely because I have turned my attention to the consequences that worry the sceptic, I don’t have permission to ignore these consequences. Thus the permission that logical non-maximalism gives us to fail to satisfy Closure runs out just in those cases where the anti-sceptic needs it. Logical non-maximalism’s slogan is “If you inquire into the consequences of a proposition, then believe what it entails”; epistemic non-maximalism’s slogan is “If you believe anything about something, then believe the truth about it”. These slogans are a way of saying that the truth is worth having, without saying the crazy thing that epistemic and logical maximalism say—that we must, altogether indiscriminately, go for as much truth about as many different things as possible. Epistemic nonmaximalism thus treats truth as a value—yet not in the crude and implausible way that epistemic maximalism treats it as a value. As we shall see in the next chapter, there is an important parallel between the structure of epistemic rationality outlined by epistemic non-maximalism, and the structure of practical rationality that I shall argue for there. Just as epistemic reason 31 Will open question arguments be available in the logical and epistemic cases, as they are in the moral? Yes, wherever evidential logical notions such as entailment or necessity, and evidential epistemic notions such as truth or knowledge or justification, are mistakenly treated as verdictive. In the logical case, it is well known for instance that, while we can if we like just stipulate that material implication is what we mean by entailment, paradoxical consequences follow on this decision. It is also well known that the notion of necessity does not readily admit of analysis as truth in all worlds (or indeed of any other analysis). The same is true in the epistemic case, indeed the point is clearer there: the notion of truth is famously resistant to analysis as correspondence, coherence, or anything else; the notion of knowledge notoriously cannot be analysed as justified true belief, except by making the notion of justification no less mysterious than that of knowledge; and the notion of warrant or justification likewise evades all attempts at definition by a formula. Thus it will be possible for us to ask, e.g., “I know this belief is justified and true, but is it knowledge?”, or “I know this belief is part of a maximal consistent belief-set, but is it true?”, or “I know this belief is based on the evidence of my normally-operating senses, but is it justified?”; and these will be just as much open questions as the questions we asked in the ethical case. In all these areas alike, it seems quite impossible for the naturalist to eliminate essentially reason-giving properties by analytically reducing them to other sorts of properties. The open question argument spells out, or more simply is, the intuition of this impossibility. 26 knows of other ways of responding to truth besides that suggested by epistemic maximalism, so practical reason knows of other ways of responding to practical values besides maximising them. Or so I shall argue as I turn from metaethics to normative ethics in chapter 5. 27