Appeared in Philosophical Studies 154 (2011) pp. 443-467.

The Unity of the Normative T. M. Scanlon Two questions that puzzle me about normativity, the subject of Judy Thomson’s excellent book, concern the scope of the normative domain and its degree of unity. I am tempted to characterize the normative domain as consisting simply of truths about reasons, including reasons for action and reasons for belief and other attitudes. But there seem to be truths about attributive goodness that are not, at least on the surface, claims about reasons. If such claims are normative, then the characterization of the normative I have just mentioned is too narrow. An alternative, to which I have also sometimes been drawn, is a (not very satisfying) dualism, or even pluralism, about the normative, according to which there at least two kinds of normative truths—truths about reasons and truths about attributive goodness, and perhaps even other kinds if a wider pluralism is correct. This view might also involve a dualism, or pluralism, about ‘ought’ judgments: some ‘ought’ judgments are clams about what people have all things considered reason to do while others are of a different kind, such as the judgment that a root ought to deliver water and nutrients to the plant of which it is a part. Thomson argues with characteristic clarity and trenchancy for a thesis of unity of the opposite kind to the one I find appealing: the thesis that the basic truths of the normative domain are all truths about attributive goodness. Truths stated by ‘ought’ judgments follow from these, she says, and judgments about reasons are ancillary to ought judgments. I will concentrate in these remarks on Thomson’s unity thesis.

2 One pull toward identifying the normative with truths about reasons comes from a way of understanding Moore’s open question argument. Moore observed that if C is a concept denoting what he called a “natural” property, then the question “X is C, but is it good?” will have an “open feel.” As Thomson observes, it is not entirely clear how the realm of the “natural” properties is to be understood, but for purposes of this discussion I will understand ‘natural’ as meaning “physical or psychological.”1 My interpretation of the phenomenon to which Moore is calling attention is that no pure claim about reasons is entailed by any claim employing only natural concepts in this sense. It is plausible to suppose that the claims about ‘good’ that Moore was considering involve claims about reasons. As Thomson observes, he seems to think that to be good is to be prima facie “to be promoted.” If so, then the “open” feel of questions of the kind Moore was talking about is explained by this fact about non-entailment. This makes it inviting to identify the domain of normative claims with the domain of claims separated from natural (physical and psychological) claims in this way: that is, to identify it with the domain of claims about reasons. But there are problems with this way of looking at the matter. As Thomson observes, many (perhaps most common) claims about what is good do not seem, at least on the surface, to involve claims about reasons. These are claims employing ‘good’ as an attributive adjective. Thomson clearly and helpfully identifies two kinds of such claims. The first are claims that something is good of a kind, K, where K is what she calls a goodness-fixing kind, that is, a kind, K, such that “what being a K is itself sets the

1

This way of understanding the “natural” may also be unsatisfactory. For example, some psychological concepts may be held to be normative (although perhaps not in the sense I am here discussing.) But I will adopt this characterization for present purposes.

3 standards that a K has to meet to be good qua K.” Goodness-fixing kinds include kinds of artifacts, such as toaster, and other functional kinds, such as seeing-eye dog. To these she adds natural kinds such as tiger and human being. “Being a good tiger,” she writes, “is being a physically fit tiger; being a good human being is being a morally good human being” (20) This difference is due, she says to differences in the capacities of humans and tigers: humans have the capacity to act well or badly, tigers don’t. This strikes me as somewhat implausible, and the phrase ‘good tiger’ (as opposed to, say, ‘healthy tiger’) seems to me linguistically odd. But I leave this aside for the moment. In addition to these properties of goodness of “goodness-fixing kinds,” Thomson calls our attention to properties of being what she calls “good-modified,” such as good at something, or good for some use, or good for some person or country. Her point might be summed up by saying that the property of being good is always a property of measuring up to certain standards. ‘Good’ as Moore uses it identifies no relevant standards and so refers to no property. But in all of the cases she mentions, there is a property of goodness because relevant standards are clearly identified. These standards are specified by goodfixing kinds, such as toaster, and in cases of “good-modified” by the modifiers, which identify the use or person for whom the thing in questions said to be good. This is bad news for the thesis that normative claims are just claims about reasons because, to put it in a nutshell, the claim that something is good in one of the ways Thomson identifies need not involve the claim that we or anyone has reason to be guided by the standards associated with these properties. We all know what it means for something to be a good dandelion root, and perhaps the fact that something is a good dandelion root gives a

4 gardener reason to take particular care to rip it out. But the claim that something is a good dandelion root does not in itself involve any such claim about reasons. This is a very plausible thesis, and Thomson argues for it forcefully. But before simply accepting it, we might look into the matter a little further. As a way of doing this I want to consider two alternative analyses of good that Thomson contrasts with her own. Paul Ziff wrote that the best analysis of good was that “‘good’ in English means is answering to certain interests,” the interests in question being supplied by the context in which something is said to be good.2 In a similar vein (but different in significant detail) John Rawls wrote that “A is a good X if and only if A has the properties (to a higher degree than the average or standard F) which it is rational to want in an X, given what X’s are used for, or expected to do and the like (whichever is appropriate.)”3 Both Ziff’s account and Rawls’s seem to work in the case of ‘good toaster.’ It is plausible to say that a good toaster is one that answers to our interests in toasters, or that it is one that has to a high degree the properties it is rational to want in a toaster given the role that toasters play in our lives. Rawls’s analysis goes wrong, however, in cases like that of a good dandelion root. This example shows that intelligible instances of goodness need not be tied to what we in fact have reason to want in things of a certain kind. Ziff’s analysis is broader, and handles such cases more plausibly, at least if we allow for sufficient elasticity in his notion of an “interest.” This term is ambiguous between a more subjective reading (what someone takes an interest in) and a more objective version (what is in someone’s interest)

2 3

Semantic Analysis (Ithaca: Cornell University Press, 1963), p. 247. A Theory of Justice, Revised Edition (Cambridge, MA: Harvard University Press, 1999), pp. 350-351.

5 But the more objective reading can also include (and I think was intended by Ziff to include) such things as the interests of a dandelion, that is to say, what is needed in order for the dandelion to “flourish.” Thomson argues against Ziff’s analysis that “answering to some interest” is not sufficient to make something good of a kind. Her example is ‘good pebble.’ Even if certain interests of ours were served by pebbles of a given kind—if, for example such pebbles, ground into dust, provided a cure for the common cold—this would not make them good pebbles, that is to say, good qua pebble. She also points out that our interests in pebbles could change, leading to changes in what made them good according to Ziff’s account, but this would not mean that what counts as a good pebble would change. Despite these objections, Ziff’s account still seems to me largely correct, and to provide the best description of most of the phenomena that Thomson is talking about. In the bulk of cases, Thomson and Ziff are, I think, describing the same thing, but from slightly different angles. Their two accounts converge because what makes kinds such as toaster, seeing eye dog and dandelion root goodness- fixing kinds is that they identify what Ziff calls “certain interests” relative to which things can be classified as (more or less) good according as they “answer to” these interests (more or less) adequately. The modifiers in what Thomson calls cases of “good-modified,” such as good tennis player, or good for some person, do the same thing. Perhaps ‘interests’ is not the best word here. ‘Purposes’ may be more accurate. But it is important that these purposes need not be one’s that we share, and calling something good in one of these ways does not involve

6 endorsing the purposes in question as ones that we or anyone else has reason to be guided by.4 Understood in this way, Ziff’s proposal seems to me to provide a deeper explanation of what is going on in almost all of the cases Thomson mentions. It handles, as I have said, the cases of goodness-fixing kinds and of good-modified. The fact that being useful in curing the common cold does not make something good qua pebble, just indicates that in that case it is not the kind, pebble that serves to identify the relevant purpose or interest. It is, rather an instance of what Thomson calls “good modified,” the modification “for treating the common cold” being, as Ziff says, supplied by the context. In the same way, we can say, on the analysis I am proposing, that although one “interest” we have in toasters is that they not be too expensive, something could be a good toaster, perhaps the best toaster in the world, even though it cost a mint. The kind, toaster, identifies certain purposes relative to which the goodness of toasters is assessed, and so long as a toaster functions in a way that serves these purposes it is a good toaster. As Thomson’s pebble example indicates, these interests can change. There might come a time at which we no longer had any use for toasters, but ‘toaster’ would continue to identify “certain interests” relative to which the phrase ‘good toaster” could be understood.5 This modified version of Ziff’s view also gives a better explanation of the idea that a good human being means a morally good human being. This is explained not, I

4

This improves on the idea of standards, which I mentioned initially, since purposes or interests provide normative explanation, without which standards seem arbitrary. 5 This makes Ziff’s account preferable to the one offered by Rawls, which seems to tie the criteria of goodness of a kind too closely to the way in which things of that kind currently figure in our lives.

7 would say, by the kind, human being but by our particular interest in human beings as people to live with. We might also make sense of ‘good tiger’ as “physically fit tiger” insofar as the purpose in question is taken to be achieving tiger-flourishing, or something like that. (Although I still would express this by saying “doing well as a tiger” rather than “good tiger,” which sounds as if I were, ill-advisedly, patting the tiger on the head.) This analysis also explains nicely the way in which such judgments about goodness are normative, and how they are related to but differ from claims about reasons. As I explained in pointing out the difference between Ziff and Rawls, the claim that something is good of a kind, or that it is good-modified, does not, when these are understood in the Ziffian way I am proposing, entail any claim about the reasons anyone actually has. But, on the other hand, such claims are not simply claims of physical or psychological fact. They might seem to be such claims insofar as they seem to be simply claims about causes and effects. So, for example, a good thermometer might be thought to be one that controls a furnace in such a way as to maintain a set temperature. This would simply be a claim of physical fact. But something would not be a good thermostat if it were the size of the Empire State Building, or took as much energy to operate as the furnace itself. The “purposes” or “interests” relative to which we judge something a good thermostat include a variety of more specific considerations, and such a conclusion about goodness requires a judgment about the proper balance between these considerations. In this example, the interest in question is one that we actually have, but the point I am making generalizes. Whether the purposes or interests are ones that we share or endorse, we can make a judgment about goodness only if we understand them and can engage in the necessary form of hypothetical practical reasoning, taking them as reason-providing

8 for the purposes of this exercise. So such judgments about goodness differ from judgments of physical or psychological fact because they involve judgments about reasons, at least of a hypothetical kind. In addition to evaluatives—judgments of goodness of the kind I have so far been discussing—the domain of the normative as Thomson understands it includes oughtjudgments, which she calls directives. Her account of normativity is impressively unified because she interprets directives as entirely dependent upon evaluatives. In her view, something ought to phi if it belongs to a kind such that it is defective as a thing of that kind if it does not phi. ‘Defective’ here is an evaluative (a negative one.) Something is defective qua K in having property P if having P prevents it from being a good K. So, for example, toasters ought to make bread warm and crispy, and a toaster that does not is a defective toaster. In the case of human beings who, as she says, lead more complicated lives than toasters, there are two kinds of ‘ought’s. On the one hand, If A is a human being , and V-ing is a physical or mental norm for the species, then for it to be the case that A ought to V is for it to be the case that a human being is a physically or mentally defective human being if he does not V. (215) This applies to such V-ings as reasoning, and walking on two legs. It is, she says, up to biological scientists to determine what counts as a defect in the relevant sense. But when V is acting, things are different: If A is a human being, then for it to be the case that A ought to Vact is for it to be the case that if A knows at the time what will probably happen if he

9 Vacts and what will probably happen if he does not, then he is a defective human being if he does not. Thomson mentions as relevant forms of defect in this case such things as unjust, imprudent, ruthless, cruel, and sanctimonious. (218) She suggests that it is up to moral theory to determine what things are defects in this sense, although she allows that there are cases of goodness and defect that go beyond the moral. I said earlier that Ziff’s analysis seems to me to apply to almost all of the cases Thomson discusses. Cases I had in mind in which this analysis seems to apply less well are instances of “defect” in the first of the two senses that Thomson mentions. For example, a human being ought to have ten toes, and I would be defective as a human being if I had fewer than ten, or more. These judgments seem quite correct even if all that is meant by the relevant ‘ought’ is that it is normal for human beings to have ten toes. But when such an ‘ought’ judgment is understood in this way, the Ziffian analysis fails. There need be no interest or purpose that is served by human’s having ten toes rather than eight or twelve. But the relevant application of ‘good’ also seems misplaced in these cases. The claim that “a good human being has ten toes” seems to me odd. The oddness may be removed if by ‘good human being’ we mean “good example of a human being.” In this case, however, the Ziffian analysis does, after all, apply, the relevant “interest” being that of providing an example what humans are usually like. Alternatively, we might understand the relevant ‘ought’ and the corresponding idea of defect to claim more than merely what is normal or usual. So, for example, it might be claimed that humans have ten toes because this promotes success in walking

10 upright, and is thus conducive to flourishing as a human being. But this again brings in a goal or purpose. So it seems to me that the cases in which the Ziffian analysis fails to apply are ones—like “humans (normally) have ten toes”—in which ‘ought’ and the corresponding idea of defect are not really normative. This conclusion seems to be supported by the fact that ‘good K’ sounds odd in these cases, unless it is understood to mean just “good example of a K.” I turn now to Thomson’s discussion of ‘ought to Vact’. It seems implausibly heavy to say of every true statement about what I ought to do, that I am a defective human being if I don’t do the thing in question. Perhaps I ought to buy a new computer now, before the price goes up; maybe I ought to give up philosophy and take up literary theory, or economics instead; and maybe I ought to give up the cello and devote more time to reading novels. It does not seem to me, however, that insofar as these things are true I am a defective human being if I fail to act in accord with them. Or at least, this would be so only if a non-defective human being is understood to be one that responds correctly to the all reasons that he or she has. Thomson clearly does not accept this interpretation of ‘ought,’ but this leaves us with the question of how else the relevant notion of a defect is to be understood. This brings me back to what I said earlier about the hypothetically reasoninvolving character of claims about goodness-fixing kinds. I said that kinds are goodnessfixing because they identify some purpose or interest, and that claims that things are good qua things of that kind are clams about what is required to advance that purpose or serve that interest, but that such claims do not involve endorsing the idea that anyone in

11 particular has reason to be guided by that purpose, or to advance that interest. To be a defective instance of a kind is to fail to have features that are needed to promote the relevant purpose. This leads to several questions about the idea of a defective human being. (1) Are there purposes or interests with respect to which this idea is to be understood? (2) If not, what gives this idea its content? (3) In the light of the answers to (1) and (2), do facts about what is defective in the specified sense support conclusions about what any particular human being has reason to do? I suggested earlier that morally good human being might be understood relative to the interests we have in the character of the human beings we are going to associate with. This would provide an interpretation of “moral defect.” My question (3) would then be just the familiar question of “Why be moral?” Perhaps Thomson believes that the idea of a human being can be appealed to as a way of settling, or dismissing this question. I don’t see how. Here I should note that Thomson’s view about reasons for acting is quite different from mine. She believes that all reasons are reasons for believing, and that reasons for action are dependent on ‘ought’s rather than the other way around. In her view, (146) for X to be a reason for A to Vact (objectively interpreted) is for X to lend (epistemic) weight to the proposition that A should Vact. Taking into account her analysis of ought and should statements, this means that for X to be a reason for A to Vact is for X to lend epistemic weight to the proposition that A would be a defective human being if he failed to Vact. I will mention two worries I have about this proposal.

12 The first is that the proposal seems to me to run together two quite different kinds of reasons: reasons for acting and reasons for accepting some view of human nature. Suppose that fact, p, that epistemic weight to a certain view of human nature. If failing to Vact would be defective according to this view of human nature, then on Thomsons’s view, p would lend epistemic weight to the proposition that A ought to Vact,. But it would not follow that p was a reason for p to Vact.. The second worry concerns the relation between reasons on the one hand and ‘should’s and defects on the other. Thomson considers (147) the suggestion that the fact that X would enjoy Vacting is a reason for Vacting but does not lend support to the truth of the proposition that A should Vact. She replies that this person is giving ‘should’ an unwarranted moral reading, one to which enjoyment is not relevant. This is somewhat puzzling, since Thomson’s own examples of ought and should, relying on defects such as unjust, imprudent, ruthless, cruel, and sanctimonious, seem suffused with morality. In any event, however, the same objection seems to me to arise in cases where morality is not in question. Suppose A is considering going to several different films, each of which he would enjoy, or having a relaxing evening at home. The fact that he would enjoy film F is a reason to go to that film. It would lend support to the idea that should go to that film if by that one simply meant that going to the film was what he had most reason to do. But as I have said, Thomson rejects this reading. (159ff) For her, the claim that he should go to the film is the claim that he would be defective if he did not. What would the defect be? It cannot, on Thomson’s view lie simply in the fact that he is failing to respond to a reason

13 he has. But I don’t see any way of understanding the defect involved in this case that would be independent of such claims about reasons. Let me conclude with a hypothesis. The difference between Thomson’s view and mine has to do with the relative priority we assign to truths about reasons and truths about what makes a human being defective. She thinks the latter are primary, whereas I do not. It sometimes seems to me that she is drawn to this position because it seems to her that if reasons for action are not explained by facts about defects and, ultimately, about goodness of a kind, then the only alternative is a subjective account of reasons for action that bases them in what an agent in fact wants. I of course believe that there are truths about reasons for action that are not subjective in this sense. As I have argued, it seems to me that these are what give content to claims about goodness of a kind. She believes that the reverse is true. Working through Thomson’s arguments has helped me to understand better the relation between claims about reasons and claims about goodness, and hence to understand better the unity of the normative. For that, as for her sheer goodness qua philosopher, I am very grateful.