Companions in Guilt Arguments for ethical objectivity

***

HALLVARD LILLEHAMMER
***

Hallvard Lillehammer 2006

COMPANIONS IN GUILT

ABSTRACT
Some arguments for the objectivity of ethical claims have a so-called companions in guilt form. Such arguments defend the objective credentials of ethical claims by comparing them with other claims the objective credentials of which are allegedly not in doubt, but which share the very features of ethical claims that have traditionally led to skepticism about ethical objectivity. In this book two distinct companions in guilt strategies are identified as playing an influential role in contemporary debates about the nature and status of ethical claims. For each of these strategies, three of its most prominent manifestations in contemporary ethics are subjected to detailed critical analysis. Prominent contemporary philosophers whose work is discussed in detail include Donald Davidson, Jean Hampton, Christine Korsgaard, John McDowell, Alasdair MacIntyre, Hilary Putnam, Richard Rorty, David Wiggins, and Bernard Williams. Both companions in guilt strategies are shown in the course of the book to defuse a number of wellknown metaphysical motivations for skepticism about ethical objectivity. It is also shown that neither strategy promises to defuse a number of equally wellknown motivations for skepticism about ethical objectivity based on the nature and scope of ethical disagreement. The main problem of ethical objectivity raised by the companions in guilt strategy is shown to be a problem of ethical indeterminacy.

--------------

1. Companionship in guilt Philosophers hunt in packs. The obvious pay-off is that the spread of individual and collective attention promotes the illumination of an increasing number of different areas of inquiry. The equally obvious downside is a tendency to one-sidedness. Different packs fail to communicate, sometimes at the cost of missing out on knowledge, whether of general interest or of particular relevance to their own area of interest. The latter tendency is clearly diagnosed by Hilary Putnam, when he writes: I believe that the unfortunate division of contemporary philosophy into separate fields often conceals the way in which the very same arguments arise in field after field. For example, arguments for antirealism in ethics are virtually identical with arguments for antirealism in mathematics; yet philosophers who resist those arguments in the latter case often capitulate to them in the former. We can only regain the integrated vision which philosophy has always aspired to if at least some of the time we allow ourselves to ignore the idea that a philosophical position or argument must deal with one and only one of these specific fields. (Putnam 2004, 1) Putnam thinks the tendency towards professional specialization within philosophy carries with it a threat of blindness towards shared features of different areas of philosophical inquiry. He also believes that if clearly appreciated, the relevant commonalities should lead us to improve our understanding of one field in light of our understanding of others. I shall return to the details of Putnams claim about the connection between ethics and mathematics in due course. At a highly general level, Putnam is right to draw attention to the existence of assumptions and argument forms that are shared between such different areas of inquiry as the philosophy of mathematics and moral philosophy. It follows that highly specialized professional philosophers are faced with a serious challenge. There are few contemporary professional philosophers who specialize equally in the philosophy of mathematics and moral philosophy. Moreover, given the different skills involved in ethics and mathematics, there is no guarantee that philosophers trained in one field will easily be able to acquire a reliable understanding of

the other so as to competently transfer the relevant insights without distortion (the problem here is arguably not entirely symmetric). In consequence, most attempts to make such transfers will have to be met with a high degree of initial suspicion. Even if Putnam is right about the need to aim for an integrated philosophical vision, the existence of a philosophical division of labour creates genuine obstacles to the making of illuminating comparisons between different fields, including a potential tendency to pay selective attention to superficial similarities and to ignore the existence of non-obvious differences between separate areas of philosophical concern. In this context, one would expect philosophers to proceed with particular caution when drawing comparisons between one philosophical field and another. In fact, this is not uniformly the case. Contemporary philosophy is full of appeals to similarities and differences between one area of philosophy and another, where the appeal in question is meant to support some particular view of the nature and status of the field about which the writer in question is primarily concerned. Putnams example of mathematics and ethics is one case in point, but the philosophies of mathematics and ethics are neither unique nor special in this respect. Frequently when such appeals are made, little further attention is paid to the details of either their presuppositions or their wider significance for the different areas they purport to connect. There is consequently a nontrivial case to be made for subjecting this kind of appeal to some kind of systematic study and analysis. It is to this task that the following pages are mainly directed. In what follows, I shall focus my discussion almost exclusively on the field of value or normativity in general, and ethical value or normativity in particular (I shall return briefly to the question of what we are to understand by ethics, value, and normativity below). I do not hope to produce in this book a comprehensive or integrated philosophical vision in anything like Putnams sense (Putnam is arguably one of the few contemporary philosophers to have a legitimate claim to having attempted such a thing). Yet I shall discuss a number of central comparisons that have been made between the field of ethics broadly understood and such different areas of inquiry as the mental, science, secondary qualities, logic, and elementary mathematics. A degree of integration will inevitably result, if only by a process of accidental accumulation. I shall proceed by means of a series of case studies of prominent attempts in contemporary philosophy to enhance our understanding of the nature of ethics and value by means of a

comparison between evaluative claims on the one hand, and some superficially different set of claims on the other. In particular, I shall focus my discussion on a specific form of argument that, following John Mackie and others, has come to be known as a companions in guilt strategy (c.f. Mackie 1977, 39). Companions in guilt arguments are arguments designed to defend the credentials of one set of claims by comparing them to a different set of claims with which they have some apparently problematic feature in common. Companions in guilt arguments in ethics proceed by comparing ethical thought with some set of apparently non-ethical claims, and showing that the two forms of thought are similar in some important respect that has been alleged to affect the nature and status of ethical claims. Normally, a companions in guilt argument will fix on some feature of ethical claims that has been thought to be philosophically problematic. The argument will then proceed by comparison, showing that there are other claims that are agreed by all parties to not be philosophically problematic in the same way, but that share the very same features that are thought to present a problem in the ethical case. At least with respect to the allegedly problematic features, the two sets of claims are then argued to be companions in guilt. They stand and fall together. If ethical thought is philosophically problematic on account of its possession of the relevant features, so is its companion. On the assumption that it is agreed by all relevant parties that the alleged companion of ethical thought is not philosophically problematic (at least not for that reason), we should therefore conclude that ethical thought is not problematic either (at least for that reason). From apparent companionship in guilt, we should actually infer companionship in innocence. The arguments I consider below all have a companions in guilt form. They are each aimed at defending some conception of objectivity, either as applied to some range of ethical claims in particular, or as applied to some range of evaluative claims more generally. Without offering a general philosophical analysis of the concept of objectivity in ethics or elsewhere, the kind of objectivity at issue can be at least initially characterized in terms of one or more of the following three marks. The first mark might usefully be labeled error. In order for a set of claims to count as objective in this sense, it must be possible for agents who make them to do so correctly or incorrectly. It follows from error as applied to ethics and value in general that ethical and other evaluative concepts are subject to non-vacuous conditions of correct and incorrect application. Error is arguably a necessary condition for any form of

objectivism about any set of claims whatsoever. It is arguably not, however, a sufficient condition. This is because the existence of non-vacuous conditions of correct or incorrect application is arguably a feature of any concept whatsoever, and the various debates between objectivists and subjectivists in ethics, for example, do not reduce without remainder to the question of whether there are meaningful ethical concepts in this sense (a possible exception here is traditional non-cognitivism, at least in its early and crudest emotivist phase (c.f. Ayer 1936)). Philosophers who defend some form of subjectivism about ethics or value will normally accept some form of Error as characterizing the evaluative concepts they are subjectivists about. Thus, no one who thinks that talk about fashion is a subjective matter because it is merely a matter of historically and culturally arbitrary taste will deny that the predicate is fashionable needs to be learned and mastered, and that in this sense it has non-vacuous criteria of correct and incorrect application at any given time and place. Thus, understood, a commitment to error is arguably too insubstantial a basis on which to define the controversial and philosophically problematic conceptions of objectivity that have generally been at issue in debates about the objectivity of ethics or value. In what follows I will show that some companions in guilt arguments about ethics and value have unsurprisingly made an overwhelming case for Error in their respective fields. What we should make of this fact with regard to the question of objectivity is, however, an open question. A second mark of objectivity might be usefully labeled realism. In order for a set of claims to count as objective by this mark their correct application must be answerable to substantial facts and properties in the world that exist independently of the contingent practice of making those claims and the relevant attitudes of those who make them (subject to a number of qualifications we need not go into at this point). I shall take it to follow from Realism about ethics and value so understood that, at least in a range of favourable cases, ethical and other value claims have substantial ontological correlates. Realism is arguably a sufficient condition for objectivity for any set of claims whatsoever. Indeed, a large number of objectivists about ethics or value have argued for a form of objectivism in exactly this sense (c.f. Brink 1989; Schafer-Landau 2003). Others, however, have not. There is a long tradition of ethical objectivism in the Kantian tradition, for example, that aims to vindicate ethical objectivity while avoiding any commitment to Realism so understood (c.f. ONeill 1989;

Korsgaard 1996). According to such views, the objectivity of ethics is a structural feature of rational human thought quite independent of the nature or existence of any independent reality at which our rational thought may be directed or to which it can in some sense be said to be cognitively responsive. Furthermore, many skeptics about ethical or evaluative objectivity formulate their skepticism precisely in terms of an explicit rejection of Realism as here understood (c.f. Mackie 1977). It is consequently at least controversial whether Realism should be thought of as a necessary condition for ethical objectivity. A number of companions in guilt arguments in ethics and the philosophy of value have exploited this fact to argue that the case for ethical or evaluative objectivity is of comparable strength to the case for the objectivity of other sets of claims, including those constitutive of logic and mathematics, that have historically been thought to have an analogously problematic relationship to Realism considered as a necessary condition of objectivity. I shall argue below that this is one of the areas of philosophy where companions in guilt arguments can be most fruitful. I return to the issue in detail in chapters 2, 4, 5, and 7 below. A third mark of objectivity I shall refer to as inescapability. In order for a set of claims to count as objective by this mark, it must be that the claims of reasonable agents competent with the relevant concepts would converge in favourable circumstances of rational inquiry (where the relevant notion of a favourable circumstance obviously has to be further specified). It follows from Inescapability as applied to ethics that the ethical judgements of reasonable people favourably placed would tend to converge. Two kinds of inescapability will be particularly relevant to the present inquiry. The first concerns the inescapability of the application of a given set of shared concepts among agents competent with those concepts. According to some philosophers, there are important differences between the concepts of ugliness and contradiction in this respect, in that competent speakers could reasonably differ on the question of what counts as ugly in a way they cannot differ over what counts as a contradiction. If so, claims about contradiction are inescapable in a way that claims about ugliness are not. I shall occasionally refer to this form of inescapability as inescapability in the weaker sense. The second kind of inescapability concerns the inescapability of concepts themselves. According to some philosophers, there are important differences between the concepts of sluttiness and contradiction in this respect, in the sense that suitably competent speakers could

reasonably differ in their commitment (either implicit or explicit) to apply the concept of sluttiness in a way they arguably cannot differ in the commitment (either implicit or explicit) to apply the concept of contradiction. If so, claims about contradiction are inescapable in a way that claims about sluttiness are not. I shall occasionally refer to this form of inescapability as inescapability in the stronger sense. Claims about Inescapability are obviously vague and indeterminate without further specification, in particular of the character and circumstances of the thinkers for whom a given claim is said to be, or not to be, inescapable. Many philosophers (although not all) would agree that an ethical claim that is inescapable for individuals (with a given character) in one set of historical circumstances need not be inescapable for individuals (with a similar or a different character) in different historical circumstances. In what follows, I shall repeatedly return to the question of how the notion of inescapability could be interpreted in its application to either ethical or other evaluative thought. Both the weaker and the stronger sense of inescapability have historically been associated with objectivism about ethics and value, without it always being clear exactly which sense has been in play. This fact is of some philosophical importance. Inescapability in the first sense does not obviously introduce a notion of objectivity more substantial than that introduced by the mark of Error. I shall return to the significance of this fact in due course. Inescapability in the second sense is arguably a different matter, and is plausibly a sufficient condition for a substantial form of objectivity on any plausible account thereof. Indeed, some defenders of objectivity in ethics have formulated objectivism precisely in terms of the idea that ethical claims are inescapable in this second sense, at least among fully rational agents. Indeed, this is arguably one of the more easily intelligible ideas behind the Kantian dictum that moral requirements are requirements of reason (c.f. Smith 1994; Korsgaard 1995). Yet as with realism, inescapability in either of the two senses just distinguished is not obviously a necessary condition for objectivity, whether in ethics or elsewhere. On a realist account as here understood, objective claims could be made true by facts and properties that are so esoteric and hard to discover that there is no guarantee that even the most rational or well-positioned person would be able to relaibly track them with any kind of confidence. In such a scenario, the idea that rational agents would converge on a set of correct judgements (or concepts) in favourable circumstances of rational inquiry remains, even if minimally consistent, profoundly vacuous (c.f. Pettit 2002).

I shall assume in what follows that Realism and Inescapability are logically independent features of claims, both in ethics and elsewhere. I shall also assume, however, that there is no obstacle to their logical compatibility. Without prejudging the discussion that follows, it will be fruitful to work on the pluralistic assumption that ethical and other evaluative claims have historically been classified as substantially objective where they have been thought to display the marks of Error, plus one or more of Realism and Inescapability. The arguments that follow will support this assumption. So understood, the term subjective picks out whatever is excluded by the term objective. During the course of the arguments that follow it will turn out that some companions in guilt arguments provide grounds to reject some skeptical arguments against objectivism about ethics and value insofar as objectivity is understood in terms of Realism. It will also turn out that these arguments are less promising when it comes to undermine skeptical arguments against objectivism about ethics and value insofar as objectivity is understood in terms of Inescapability, at least in the stronger sense. The picture that will emerge in this book is one according to which some of the most fundamental questions about the objectivity of ethics and value boil down to a set of issues about the inescapability of a central core of ethical and other evaluative claims. These questions about inescapability stand in clear contrast to some of the many questions about semantic or ontological status on which much of twentieth century moral philosophy has tended to concentrate. This is not to say that questions of semantics and ontology are irrelevant to the objectivity of ethics and value. It is rather to say that insofar as companions in guilt arguments in ethics or the philosophy of value have any plausibility, they are most forcefully directed at forms of subjectivism about ethics and value motivated by semantic and ontological concerns. According to my understanding of the marks of objectivity outlined above, this leaves the question of the objectivity of ethics and value still unresolved. Thus, while some of the arguments that follow will respect the spirit of Putnams claims in the book from which I just quoted, they should not be thought to support a conception of ethics or the philosophy of value in precisely the terms that the title of Putnams book suggests. My argument is not an argument in favour of ethics without ontology. It is an argument in favour of a certain understanding of the scope and limits of a particular argument form as applied to the question of the objectivity of ethical and other evaluative claims.

While sharing the general aim of establishing companionship in guilt, the arguments I shall consider in this book are importantly different in structure. Some of the arguments take the form of what I shall refer to as arguments by entailment Arguments by entailment work by showing that some set of philosophically problematic claims is necessarily implied by a different set of claims that is agreed to not be philosophically problematic in the same way. Entailment is not a matter of degree. Nor is it a matter of distinguishing between different respects in which some things are similar or dissimilar to others. If one set of claims entails another, this is an all-or-nothing matter. Arguments by entailment in ethics and the philosophy of value relate evaluative claims to some different set of claims the objectivity of which is not in serious doubt, and then purport to show that the relevant evaluative claims actually follow by necessity from the truth of these underlying claims. With respect to their truth and objectivity, the two sets of claims are companions in guilt as a matter of necessity. On the assumption that we do not wish to deny the objectivity of values alleged companion, we should not deny the objectivity of ethics or value either. Companions in guilt arguments by entailment are potentially powerful and illuminating. Yet they are also subject to the following two limitations. First, entailment is harder to establish than contingent forms of dependence, association, or similarity. Thus, arguments by entailment fail if the less problematic companion can in principle be completely understood without any commitment to the claims associated with the more problematic companion. Second, the fact of entailment does not itself resolve every problem associated with the claims entailed. Rather than removing the problems associated with the more problematic companion, the allegedly less problematic companion might instead be thought to inherit them, this creating more pressure to account for it without the apparent entailment in question. Companionship in guilt may cut both ways. In what follows, I shall examine three arguments by entailment in defense of the objectivity of ethics or value. The first claims that the objectivity of value follows from the objectivity of commonsense psychological claims about mental content. This argument is probably most strongly associated with the work of Donald Davidson and Susan Hurley. The second argument I discuss claims that the objective validity of the kinds of categorical imperatives associated with moral obligation follows from the objective validity of hypothetical imperatives associated with

10

prudence or desire satisfaction. This argument is associated with the work of a number of contemporary neo-Kantian moral philosophers, including most prominently Christine Korsgaard and Jean Hampton. The third argument by entailment I shall discuss claims that the objectivity of value follows directly from the objectivity of the philosophically basic notions of truth or fact. This argument has been most prominently developed by Hilary Putnam in his influential book Reason, Truth, and History, and has been subject to various amendments by Putnam and others since then. In the case of each argument by entailment I discuss, my approach will be partly reconstructive, partly forensic. By focusing on a small number of prominent formulations of the relevant arguments in the contemporary literature, I hope to avoid the danger of arguing against straw men (or women) while remaining able to draw a number of limited yet general conclusions about the scope and limits of such arguments on the basis of a reasonably representative historical sample. A different kind of companions in guilt arguments take the form of what I shall refer to as arguments by analogy. Arguments by analogy compare two sets of claims with the aim of showing that they are similar in some important respect. Similarity is a matter of degree, as is the centrality of different features of given claims to their basic nature and function. Arguments by analogy in ethics and the philosophy of value compare evaluative claims with some set of claims generally assumed to be less problematic with respect to their objectivity, and then show that the two sets of claims both exhibit the very same features that, in the evaluative case, have generated skepticism about their objectivity. With respect to the features on question, the two sets of claims are companions in guilt. On the assumption that we do not wish to deny the objectivity of values alleged companion, we should therefore not deny the objectivity of ethics or value either. Arguments by analogy are also potentially powerful and illuminating. Yet their potential is limited by the following three facts. First, everything is similar to everything else in some respect or other. Analogies are too easy to establish to be philosophically interesting merely as such. Second, different things are not perfectly similar in every respect. The importance of the fact that two sets of claims share a certain feature therefore depends on two further factors. The first is the centrality of the relevant feature to the claims in question. The second is the presence of other relevant features that may otherwise distinguish the two sets of claims with respect to the question at issue. Thus, an argument by analogy can

11

be undermined either by the fact that the analogy is not relevant, or by the fact that the analogy, although relevant, is outweighed by the presence of further disanalogies. Third, the fact of similarity does not in itself resolve the problem the analogy is meant to illuminate. Rather than removing the problems associated with the more problematic companion, the allegedly less problematic companion could instead be thought to share them, this creating more pressure to explain away the significance of the alleged analogy. Again, companionship in guilt may cut both ways. In what follows, I shall examine three different arguments by analogy. The first draws an analogy between value and so-called secondary qualities like colour. This argument is associated with the work of John McDowell and David Wiggins, among others. The second argument draws an analogy between ethics and science. This argument is associated with the work of Hilary Putnam and John McDowell, among others. The third argument of this kind that I shall discuss draws an analogy between ethics on the one hand, and mathematics and logic on the other. Again, this argument is associated with the work of David Wiggins and John McDowell. In the case of each argument, my approach will be partly reconstructive, partly forensic. In this way, I hope to avoid the danger of arguing against straw men (or women) while drawing limited, but general conclusions about the scope and limits of these arguments on the basis of a reasonably representative historical sample. Appeals to companionship in guilt are sometimes made as part of a rhetorical strategy of philosophical intimidation. What about quantum physics?; That wont work for second order logic!; But the law is not like that at all!. Just like everyone else, philosophers of ethics and value are vulnerable to be thrown off guard by provocative appeals to apparently profound insights about their subject matter deriving from the nature of some comparatively esoteric field of philosophical inquiry about which they can usually be expected to know very little or absolutely nothing (Putnam himself is, incidentally, by no means innocent in this respect). As often as not, the application of a companions in guilt strategy will obfuscate more than it illuminates, relying as it often does on controversial interpretations of both areas of thought, with little or no attempt to defend either interpretation in any depth. There is no easy solution to this problem short of becoming an expert in all the areas of philosophy to which an appeal to companionship in guilt could be made. Most contemporary philosophers (at least those of us working in the areas of ethics and value) will be highly unlikely

12

or simply unable to choose this undeniably praiseworthy option. Even so, we have an interest in developing a reflectively robust attitude towards the nature and scope of companions in guilt arguments in our respective fields. One way of doing so, and the one I shall adopt here, is to make explicit, without going on to question, the philosophical assumptions about values alleged companion that underlie a given companions in guilt argument, and then go on to examine the claim to companionship in guilt on its own terms; questioning the alleged entailment or analogy, and examining in detail its implications and significance. On this approach, little or no more understanding of values alleged companion is required than is either explicitly or implicitly contained in the premises of the relevant companions in guilt argument. The scope and limits of the argument can therefore be assessed using only such philosophical competence as is necessary to understand these premises and the logical structure and conclusions about ethical or evaluative thought that the argument contains. As far as the ultimate nature of values alleged companion is concerned, this can be left for the division of philosophical labour to decide. Having said that, the task of making explicit the conception of values alleged companion that is assumed in various companions in guilt arguments is by no means a trivial task. Indeed, I shall argue below that a number of such arguments in contemporary philosophy are distinguished by their appeal to particularly controversial accounts of values alleged companion. While I shall not question these accounts here, the task of making them explicit and seeing them in the context of ongoing philosophical controversy outside ethics and the philosophy of value is itself a way of drawing attention to the scope and limits of companions in guilt arguments in ethics and elsewhere. As will already be abundantly clear, the approach I take in this book is highly selective. I shall examine six companions in guilt arguments in ethics and the philosophy of value in some detail, and shall place these arguments in their contemporary philosophical context only to the extent that this promotes the end of evaluating their structure and philosophical power. I do not pretend to address all companions in guilt arguments that have been given in recent philosophy, neither in ethics and the philosophy of value, nor elsewhere. For example, some companions in guilt arguments about ethical claims draw on an analogy between ethical and philosophical thought in order to ground a conception of ethical objectivity in the face of continuing disagreement about ethical claims (c.f. Schafer-Landau 2003, Putnam 2004). I shall have

13

nothing to say about this kind of argument by analogy in what follows. I have made my selection of case studies not only on grounds of contemporary relevance or interest, but also on the basis of the extent to which the objectivity of values alleged companion has tended to be taken as comparatively secure. The objective status of philosophical thought is arguably subject to as deep a controversy as that of ethical thought itself, and a discussion of the nature of philosophy in the present context would in any case exhibit an further layer of potentially clumsy reflexivity. Other possible arguments that I shall not discuss in what follows draw parallels between ethical or evaluative claims and claims from other areas of human thought without necessarily taking a companions in guilt form. This may the case with some comparisons between practical and theoretical reasoning, or comparisons between role of explanation in ethics and science, for example. Insofar as these arguments do not have a companions in guilt form, their scope and limits fall outside the domain of the present inquiry. While the arguments I do discuss in this book will allow me to draw some conclusions about the objectivity of ethics and value, I shall not be defending any specific thesis about the objectivity or subjectivity of either ethical or other evaluative claims. The argumentative tools made use of here have no serious prospect of producing general theses of this kind. What the tools in question do have the prospect of producing is further clarity about the scope and limits of one way of arguing about objectivity in ethics and elsewhere. Given the history of the subject, this may be as much as we can reasonably hope for. I am also aware that philosophical arguments, companions in guilt arguments included, have a long history. In this book I shall confine my attention almost exclusively to philosophical work published in my own lifetime (and therefore after 1970). A complementary study would investigate the use of companions in guilt arguments throughout the history of philosophical thought. At the root of the present inquiry is an interest in the perennial philosophical question about nature and status of ethical claims. In a narrow sense, I take ethical claims to essentially concern the selection and management of constraints on character and conduct that aim to serve and protect the various interests of agents who live together in more or less loosely defined social groups. The widely accepted norm of promise keeping is part of ethical thought in this narrow sense. So is being a trustworthy person. In a wider sense, ethical thought also concerns the selection and management of constraints of character and conduct that aim to

14

serve and protect agents considered in potential abstraction from the social groups in which they happen to find themselves. The widely accepted norm of being well informed is an ethical norm in this wider sense. So is a norm prescribing basic personal hygiene. Ethical thought in this wider sense forms a subset of evaluative or normative thought more generally, where this is generally agreed to include at least aesthetics and politics, and potentially also logic and epistemology. Yet as the nature and extent of the domain of value and normativity is itself part of what is at issue in some of the arguments that follow, nothing much would be gained by subjecting these notions to more precise definition at this point.

15

2. Values out of Mind 2.1. Value, objectivity, and companionship in guilt According to one version of objectivism about value, ethical and other evaluative claims have a fixed truth-value independently of who makes them or the society in which they happen to live (c.f. Davidson 2004, 42). Subjectivists about value deny this claim. According to subjectivism so understood, ethical and other evaluative claims have no fixed truth-value, either because their truth-value is dependent on who makes them, or because they have no truth-value at all, their function being instead to express affective attitudes like desire or emotion and the like. Much of the debate about the objectivity of ethics and value is focused on whether values can be located within a naturalistic worldview. Some subjectivists argue that values are unfit for a place in a naturalistically respectable ontology (Mackie 1977, 1-49). Objectivists respond that the apparently unacceptable commitments embodied in evaluative claims do either not exist or are not, after all, unacceptable. A number of objectivists who defend the latter claim employ a companions in guilt arguments in its defense.1 These companions in guilt arguments work by comparing evaluative thought with some area of thought that is assumed to be philosophically less problematic, only to show that the apparently unacceptable commitments embodied in evaluative thought are shared by this other area of thought, each area of thought therefore being neither more nor less problematic than the other in the relevant respect. The two areas of thought are companions in guilt. They stand and fall together. The strength of companions in guilt arguments in ethics and the philosophy of value depends not only on the strength of the connection between ethical or evaluative thought on the one hand
1

C.f. Mackie 1977, 39, who writes that the best move for the moral objectivist is not to evade this issue, but to look for companions in guilt. See also Hampton 1998, 125ff.

16

and their alleged companion on the other, but also on the extent to which the alleged companion is indeed philosophically less problematic than either ethical or evaluative thought. The mixed fortune of companions in guilt arguments in ethics and the philosophy of value is partly down to the perceived failure of at least some arguments of this kind to meet the latter condition. If the objective credentials of ethics or values alleged companion are as controversial as those of ethical or evaluative thought themselves, the strength of the link between ethics or value and their alleged companion can equally be used to argue against the objectivity of each. Objectivists about ethics and value therefore have a clear incentive to find a companion in guilt for ethics or value that is noncontroversially less problematic than either ethical or evaluative thought. Folk psychological claims about mental content are often regarded as less problematic with respect to their objectivity than claims about ethics or value. It is true that genuine controversy exists about the status of folk psychology (see e.g. Stich 1993, Heal 2003). Yet subjectivists about ethics or value have not historically rested their position on denying the objective truth of propositional attitude attributions. Indeed, it would be impossible to make sense of what most subjectivists say about ethics and value without assuming the objective truth of claims about minds that are somehow capable of thinking about their environment (see e.g. Ayer 1946; Mackie 1977). A sound companions in guilt argument by entailment, connecting ethical or evaluative thought on the one hand with claims about mental states on the other would therefore weaken the case for subjectivism about ethics and value by licensing the inference from a denial of the objectivity of ethics and value to a denial of the objectivity of folk psychological claims. If we are not entitled to think there are objective values, we shall not be entitled to think there are any minds. This is surely an absurd conclusion, even by philosophical standards. To many philosophers, any such argument is too to good to be true.2 Yet in a series of articles published in the 1980s and
2

In one place, David Lewis rhetorically asks: Why care about objective value or ethical reality? The sanction is that if you do not, your inner states will fail to deserve folk theoretical names (Lewis 1996, 307). Elsewhere, Lewis explicitly rejects this argument. He writes: The suggestion is intelligible and interesting, but too good to be true. For one thing, it only spreads the trouble. Instead of losing the risk that nothing deserves the name of value, we gain the added risk that nothing deserves commonplace folk-psychological names (Lewis 1989, 135). The viability of Lewiss alternative depends on the existence of an evaluatively (more)

17

1990s and recently reprinted posthumously in Volume 4 of his philosophical papers, Donald Davidson has argued that the objectivity of value follows from the objective truth of propositional attitude ascriptions, the latter being understood in terms of his influential account of mental content as determined by constraints embodied in a process of radical interpretation (Davidson 2004, 1974). According to this argument, necessary constraints on the attribution of mental content entail the attribution to agents of minimal standards of rationality, generally true beliefs, and a core set of correct (or what Davidson alternatively calls real) values (Davidson 2004, 50). A number of other philosophers have been influenced by Davidsons work to make similar claims, including Susan Hurley, Paul Hurley, and R. H. Myers.3 Davidsons argument is powerful and illuminating. Yet neither his own published work on the issue, nor that of his followers, does enough to clearly establish the exact scope and limits of his objectivist conclusion. How objective is value? How much of value is objective? In what follows, will try to make some progress on this issue. In doing so, I will not question the interpretationist account of the mental on which Davidsons argument depends.4 Instead, I will investigate what exactly we are entitled to infer from it. In particular, I will put some pressure on Davidsons claim that there would be convergence among the values of enlightened agents who fully understood one another. I will argue that for all Davidsons argument shows, there is an indefinite range of substantial evaluative claims, including paradigmatic ethical claims, that have no determinate truth-value, and are therefore not objective in the sense suggested by Davidsons use of that tterm. 2.2 Davidsons argument

neutral account of folk psychology. As my primary interest is what follows from the alternative he rejects, I shall not pursue the details of Lewiss argument in what follows (I erroneously attribute a companions in guilt argument to Lewis in Lillehammer 1999.) 3 See S. Hurley 1989, P. Hurley 2002, and Myers 2004. Susan Hurleys account is the most comprehensive of these and goes beyond the scope of Davidsons argument in a number of ways, including a thorough discussion of ethical, political, and legal deliberation. See e.g. Hurley 1989, 189ff. I shall return to the differences between Hurley and Davidson below. 4 One canonical articulation of Davidsons account is found in Davidson 1984, 123179. The view is articulated further in Davidson 2001 and Davidson 2004, 75-166.

18

Davidsons case for the objectivity of value can be formulated as a transcendental argument. The argument begins from the assumption that propositional attitudes with determinate contents can be successfully attributed to some agents. The argument then proceeds by making explicit what the necessary presuppositions of such attributions are. According to Davidson, the successful attribution of propositional attitudes presupposes the existence of at least two agents, each responding in similar ways to similar features of a shared environment, including the responses of other agents. In this way, Davidson takes the fact of interpretability to defuse philosophical skepticism both about the existence of other minds and about the external world. Propositional attitude attributions proceed by assigning contents to mental states partly in terms of what is perceived to be their relevant causes. Davidson claims that one agent, the interpreter, can only assign determinate propositional attitudes to another agent, the interpreted if at least two further conditions are met. These conditions are captured in terms of two principles that in one place Davidson calls the principle of coherence and the principle of correspondence respectively (Davidson 2001, 211). Both principles are subsumed under the heading of what Davidson calls the principle of charity. The principle of coherence concerns the internal relations between propositional attitudes. It constrains the interpreter to attribute a degree of internal consistency to the interpreted. The relevant consistency comes in two forms. First, there must be basic logical consistency between most propositional attitudes in order for these attitudes to have determinate content. Davidson claims that mental states get their content partly from their relations with each other. Thus, in order for Andy to think that Harry is a bachelor, he must also be thinking that Harry is a man, that Harry is not married, and so on. According to Davidson, the basic norms of consistency are those formulated in classical predicate logic with identity (Davidson 1984, 199-214). Second, rational agents must exhibit a basic level of instrumental consistency in order for propositional attitudes to be intelligible as reasons explanatory of action. In particular, rational agents must be interpretable as pursuing their desires in accordance with their beliefs. Thus, in order for Sheila to propose to Jimmy, she must have a desire to get married, or some alternative desire that instrumentally connects with a belief about the likely consequences of a proposal (and the content of which also depends on the content of her other intentional states). According to Davidson, the basic norms of instrumental consistency are those formulated in subjective decision

19

theory as developed by Ramsey, Jeffrey and others (Davidson 1980, 229-238; Davidson 2004, 151-166). Coherence is a matter of degree. Davidson thinks that determinate propositional attitudes can be attributed to agents who are less than perfectly coherent, but only within limits. The lower the degree of coherence among an agents attitudes, the less determinate it is what they think. According to Davidson, an agents failure to comply with the principle of coherence is a normative failure of (theoretical or practical) rationality.5 While the principle of coherence ensures that the propositional attitudes of agents form a relatively coherent set, the content of this set is underdetermined by knowledge of behaviour and its perceived causes. For any set of behavioural evidence there is an indefinite number of incompatible sets of contents that fit the behavioural evidence equally well. Some of these are quite bizarre, unnatural, or simply unbelievable (c.f. Lewis 1984, 107-8). The principle of correspondence rules out the bizarre, unnatural, and simply unbelievable contents by constraining the assignment of contents to an agents propositional attitudes by their external causes as perceived by the interpreter. It is only because the interpreter consciously correlates the responses of another creature with objects and events in the observers world that there is any basis for saying that the interpreted is responding to one feature of the world rather than another (Davidson 2001, 212-3). Davidson calls this process of external correlation triangulating. Triangulation allows the interpreter to think of the interpreted as responding to the same features of the world as the interpreter would respond to in similar circumstances. Davidson writes: The key to the solution for simultaneously identifying the meanings, beliefs, and values of an agent is a policy of rational accommodation This policy calls on us to fit our own propositions to the other persons words and attitudes in such a way as to render their speech and behaviour intelligible. This necessarily requires us to see others as much like ourselves in point of overall coherence and correctness that we see them as more or less rational creatures mentally inhabiting a world much like our own (Davidson 2004, 35).6

For a contrasting view, see e.g. Mellor 2005. It might be considered a potential problem for this account that agents do not generally interpret their own mental states on the basis of observing their own

20

Davidson claims that the principle of correspondence constrains the attribution of two kinds of mental state. First, and most obviously, the interpreted must be regarded as generally forming beliefs with the same contents as those of the interpreter when confronted with the same features of the world.7 According to Davidson, the resulting commonality of belief entails that the interpreter must regard the interpreted as having generally true beliefs. For purposes of interpretation, beliefs are intrinsically veridical (Davidson 2001, 155). Second, the interpreted must be regarded as generally forming desires with the same contents as those of the interpreter when confronted with similar features of the world. According to Davidson, the resulting commonality of desire entails that the interpreter must regard the interpreted as generally sharing her desires or values.8 For purposes of interpretation, desires and values are also intrinsically veridical. According to Davidson, the joint applicability of the principles of coherence and correspondence is a necessary condition for the existence of genuine disagreement between rational agents. It is only against the background of a shared set of beliefs that it makes sense to say that two rational agents employ a common stock of concepts in terms of which they can disagree about some determinate feature about the world. Furthermore, it is only against the background of a shared set of criteria for the application of the relevant concepts that such disagreements between rational agents can be resolved. It follows that there is a substantial limit on the degree to which the beliefs of agents can coherently conflict.9

behaviour. For Davidsons attempt to reconcile first person authority with his externalist account of mental content, see Davidson 2001, 3-14. 7 By feature of the world I shall interpret Davidson to mean event, object, or state of affairs, actual or possible. 8 In a number of places, Davidson appears to use the terms value and desire interchangeably (c.f. Davidson 2004, 36, 73). In other places, he does not. Thus, according to Davidson, evaluations come in many kinds, including wanting, desiring, cherishing, holding to be right or to be obligatory, and the negative and comparative versions of these (Davidson 2004, 20; c.f. Davidson 1980, 4). This latter use most strongly suggests a reading on which the evaluative attitudes form a wider category of which desires are an instance. In what follows, I employ the disjunctive terminology of desires or values in order to remain as neutral as possible on this interpretive issue. 9 One standard statement of Davidsons position on disagreements of belief is Davidson 1984, 183-198.

21

The main interest of Davidsons account for ethics and the philosophy of value is his application of the claim about the necessary limits of disagreement to the domain of desire. Applying the principle of correspondence to the domain of desire, Davidson draws an exactly parallel conclusion as he does for conflicts of belief. It is only against the background of a shared set of desires or values that it makes any sense to say that two agents employ a common stock of concepts in terms of which they disagree about the desirability of some feature of the world. Furthermore, it is only against the background of a shared set of criteria for the application of the relevant concepts that such disagreements can be resolved. It follows that there is a strict limit on the degree to which the desires and values of agents can coherently conflict. The more basic a desire or value is to the interpretation of an agent, the less content we can give to the idea that we could possibly disagree about that desire or value. Davidson goes on to claim that we should expect what he calls the enlightened values of rational agents to converge (Davidson 2004, 49). To the extent that they must, value judgements are true or false in much the same way our factual judgements are (Davidson 2004, 49). What makes a judgement true or false is the same in either case, namely that the same features of the world cause similar responses in different agents at different times. Where convergence fails to occur, we are entitled to assume that there is an explanation for this failure consistent with the claim that some of these responses are correct and others not. Davidson concedes that the analogy between belief and desire is imperfect in two respects. First, while paradigm beliefs are non-controversial truth-bearers, paradigm desires are not. Davidsons response is to claim that the positive attitude taken by an agent towards the object of a desire can be made linguistically explicit by formulating the desire as an evaluative judgement carrying its truth-evaluability on the surface (Davidson 2004, 37). According to Davidson, the fact that an agent who desires something desires it on the basis of some of its perceived features entails that all pro-attitudes may be expressed by value judgements that are implicit (Davidson 1980, 86).10
10

Paul Hurley interprets Davidson as holding that the content of a desire that P is given by the evaluative judgement that it is desirable that P (Hurley 2002, 6-15). What Davidson actually says is less unequivocal: The simplest view would be to identify desiring a sentence to be true with judging that it would be desirable if it were true And it is hard to see how these two attitudes can be allowed to take entirely different directions. On the other hand there is no simple

22

Commenting on the descriptive, or so-called assertoric, surface of evaluative sentences, Davidson writes: The semantic nature of such judgements is clear: we are classifying one or more things as having a certain property. The thing or things must either have that property or not There is no coherent way to avoid this conclusion (Davidson 2004, 48). On this view, desire is what Scanlon and others have called a judgement sensitive attitude (see e.g. Scanlon 1998, 18ff). It follows from this account of desire that there can be no conceptual analysis of value in terms of desire as reductively understood, contrary to what has been suggested by some contemporary naturalist (e.g. expressivist or dispositionalist) accounts of ethics and value (c.f. Blackburn 1998; Lewis 1989).11 Second, while paradigm beliefs are supposedly caused by objectively existing features of the world that give them their content, paradigm desires do not on Davidsons view have distinctively evaluative ontological correlates to provide evaluative contents for them.12 Davidsons response is to separate what he calls the epistemological issue of objectivity from what he calls the ontological issue of existence (Davidson 2004, 44). On this view, the correct application of concepts as given by their shared criteria of application is sufficient for truth, and thereby also for objectivity (Davidson 2004, 56, Note 12). To pick out the (real)
connection between our basic preferences for the truth of various sentences and our judgements about the moral or other values that would be realized if they were true. The full story here is very complicated, and touches on many disputed matters (Davidson 2004, 37). Some of these complications are noted in Myers 2004, 107, Note 2. 11 Davidsons account would arguably be problematic if it could be shown to entail that all desires have objective truth conditions regardless of their content or their conditions of possession. Thus, it is natural to think that at least some intelligible tastes (such as a preference for ice cream over pickled eels) are rationally arbitrary, and to that extent subjective. S. Hurley may be on to this when she writes: Evaluative constraints on eligibility operate at the level of pro tanto intelligibility, not correctness The evaluative assumptions about intelligibility are not of controversial solutions to ethical problems in which fundamental values conflict, but rather assumptions that it is correct to describe these problems in those value-laden terms in the first place (Hurley 1989, 105). 12 Analogous problem cases here are abstract objects like numbers, properties, propositions, logical relations and the like. For Davidson on abstract objects, see e.g. Davidson 2001, 53-4.

23

values, all that is needed is to pick out those (non-evaluative) features of the world that are correctly valued. Once we have solved the epistemological problem involved in doing so, Davidson claims, we ought to lose our interest in the residual ontological issues that have traditionally been associated with the objectivity of ethics and value. We ought therefore to reject such philosophical notions as reaction dependence or the fabric of the world that have historically been associated with debates about the nature and status of ethical and other evaluative claims. According to Davidson, such notions are infected by the inherently unintelligible question where values are (Davidson 2004, 43).13 Davidsons argument has at least three noteworthy implications for ethical and other evaluative thought in particular. First, there is agreement in value among enlightened agents as a matter of a priori necessity. In this way, Davidsons account of how values converge differs markedly from the account of many contemporary naturalists who might be otherwise attracted to various aspects of his philosophical worldview. Second, the question of objectivity is not primarily an issue in ontology, but rather as a question of convergence in ethical judgement among enlightened rational agents. Arguments for the subjectivity of ethics or value based on a commitment to a naturalist ontology are therefore in principle misguided. Third, the truth of Davidsons account would go some way to explain the centrality of appeals to pre-theoretical intuition in philosophy, both as applied to ethical or other evaluative thought and elsewhere. For if rational agents believe, desire, and value the same things as a matter of necessity, and if what they believe, desire, and value is by and large bound to be true, then it is hardly a surprise that we should consider the
13

Davidson writes: When we speak of values, we dont even seem to be referring to entities of some odd sort, and so objectivity or realism with respect to values cant sensibly be construed as an ontological issue. The same is true of weights, colours, and shapes. These arent out there or anywhere else. (Davidson 2004, 44). While Davidson is quite clear that some beliefs do have substantial ontological correlates, he is less than perfectly clear about what distinguishes those beliefs that do from those beliefs (and desires) that do not, and the extent to which this distinction is itself clear, sharp, or objective. Susan Hurley is less nervous than Davidson about ontologizing talk of objective evaluative truth, being happy to describe the value-ladenness of desire as a species of the world-ladenness of the mind in general (Hurley 1989, 55-57). Hurleys ontological nervousness appears to set in where evaluative ontology is taken to imply a theoretical role for values inside the causal order as described by natural science. It is this nervousness that is the rationale behind her claim that objectivism about values does neither depend on nor presuppose scientific realism about values (Hurley 1989, 99-100).

24

widely shared beliefs, desires, and values of presumptively rational agents as evidential in the course of ethical and other philosophical inquiry. 2.3. Objectivity and convergence On this reading of Davidsons argument, the scope and limits of ethical objectivity depends on the extent to which the values of rational agents would converge in favourable circumstances. In this sense, it would be their rational inescapability that would constitute the objectivity of ethical values. Thus, Davidson writes: I want to suggest that we should expect enlightened values the reasons we would have for valuing and acting if we had all the (non-evaluative) facts straight to converge; we should expect people who are enlightened and fully understand one another to agree on their basic values. An appreciation of what makes for such convergence also shows that value judgements are true or false in much the way our factual judgements are. (Davidson 2004, 49) In order to evaluate this claim, we need to be clear about what is meant by a) basic values, b) enlightened values, and c) converging values. a) basic values. There is more than one way to understand the notion of a basic value. First, it can be understood as a foundational value, as opposed to a derived value. Alternatively, it can be understood as a general, as opposed to a particular value. Third, it can be understood as a large subset of values, as opposed to a small subset of values. Fourth, it can be understood as a core, as opposed to a peripheral value. Whilst perhaps not directly contradicting anything Davidson says, the first and second understanding are in potential tension with Davidsons holism and his coherentist remarks about knowledge and justification (see Davidson 2001, 137-57). The third understanding is purely quantitative, and says little about the relative importance of different values. It is the fourth understanding that appears to make the best sense of Davidsons talk of holistic networks of values and norms, none of which may be analytically immune from revision, but some of which are more central to interpretation than others.14 It

14

Davidson writes: The common framework is the idea of overlap, of norms one person correctly interprets another as sharing the more basic a norm is to our

25

is therefore in accordance with this fourth understanding that I shall interpret Davidsons talk of basic values in what follows. b) enlightened values. Davidson thinks of enlightened values as values given by desires informed by the facts, where the notion of a fact is understood non-evaluatively. It is not obvious what this category includes on Davidsons view, given that on his view all mental concepts are normatively laden. On one natural understanding of non-evaluative facts, mental facts are therefore excluded. The problem with this interpretation is that it is impossible to even ask whether the enlightened values of agents converge without being able to talk about mental states like beliefs, desires, or intentions. On a second understanding, non-evaluative facts would include mental facts, perhaps because mental descriptions are true of mental events that are token-identical with physical events, and as such enter into genuine causal relations in a way that values possibly do not (c.f. Davidson 1980, 207-227; Davidson 2001, 72). This is the reading of Davidsons view I shall adopt in what follows. c) converging values. Finally, there are at least two distinct ways in which desires can be said to converge. First, Harry and Andy could both want pickled eels for their first personal selves.15 Second, Harry and Andy could both want pickled eels for Harry or Andy (or both). Yet as Hume famously pointed out, if the internal frame and external situation of agents are different, it is hardly plausible that they will all want all the same things for their first personal selves (Hume 1993, 149-152).16 Harrys sweetness of tooth could make him prefer ice-cream to pickled eels while Andy would rather have pickled eels in all their sliminess. Even if Any and Harry differ in what they want for their first personal selves, it does not follow that they are in any conflict about what either should want, given their differing internal frame and external situation. In what follows, I shall interpret Davidson as holding that while there must be some convergence in desire across agents in both senses, there need not be complete convergence across agents with respect to desires for ones first personal self. Nor shall I interpret Davidson as denying that the appropriateness of desire can
making sense of an agent, the less content we can give to the idea that we disagree with respect to that norm (Davidson 2004, 50). 15 For a discussion of different modes of self-identifying desire, see e.g. Lewis 1984, 27-50. 16 As Hume also points out, the same point applies to the same agent at different times (c.f. Hume 1993, 149-152).

26

in that sense be relative to either time, place, person, culture, tribe, or legal system (Davidson 2004, 40).17 Davidson mentions at least two conditions where disagreement in value may occur. First, the scope of disagreement in value increases in proportion to the distance of the values in question from basic evaluative commitments the attribution of which is necessary to make coherent sense of rational agents as acting on reasons at all. Davidson writes that questions of fairness, justice, and social welfare are in this sense remote from the basic values that constitute agency, and that they are therefore predictably subject to at least de facto lack of convergence (Davidson 2004, 74). These comparatively substantial moral values contrast sharply with the basic laws of logic, as well as what Davidson calls the principle of total evidence for inductive reasoning, or the analogous principle of continence (Davidson 2004, 195). According to Davidson, all rationally minded creatures are necessarily committed to the latter values, it being impossible to coherently interpret them as rational agents otherwise. Second, Davidson suggests that the scope for disagreement in value is sensitive to the differential specificity and theoretical role that different values can play in our understanding of the world and each other. In one place, he suggests that disagreement on general evaluative principles is less intelligible than disagreement about particular acts or events (Davidson 2004, 72). Elsewhere, he suggests that theoretical disagreement is in some cases more intelligible than disagreement about what he describes as more evident, including (most probably) the objects of basic (or revealed) preference and choice (Davidson 1984, 169).18 Given that genuine interpretability is consistent with a de facto lack of convergence in values, why should we think that the enlightened values of agents would converge? There are at least two options here. The first is that Davidsons convergence claim is merely a restatement of his thesis that intelligible disagreement about values presupposes the existence of agreement on the fundamental values required for basic intelligibility. There is some textual evidence for this claim. Thus, Davidson writes:

For a discussion of how Davidson can accommodate the existence of so-called agent relative values, see Myers 2004, 123-24. 18 Whether, and if so how, these claims are mutually compatible is a potentially interesting question I shall not pursue any further here (c.f. Davidson 2004, 87ff).

17

27

 [N]o matter what the subtleties involved, interpreting evaluative judgements rests on the same foundation as interpreting the evaluative attitudes: understanding depends on finding common ground. Given enough common ground, we can understand and explain differences, we can criticize, compare, and persuade. The main thing is that finding the common ground is not subsequent to understanding, but a condition of it If we understand their words, a common ground exists, we already share their way of life (Davidson 2004, 51). If this is all there is to it, Davidsons objectivism about value is less powerful and interesting than some of its formulations may initially suggest. Basic intelligibility leaves open the scope for intractable disagreement about the application of a vast range of substantial evaluative concepts, including rightness, obligation, fairness, justice, social welfare, ice-cream, pickled eels, and most of the other values that concern people the most. The fact that communists, socialists, liberals, libertarians, and the Taliban can all be said to believe in justice is likely to be cold comfort for everyone concerned unless there is some prospect that the conflict between their mutually incompatible interpretations of this allegedly common value could be reasonably resolved, at least in principle. The second option is that Davidson thinks enlightened values would converge beyond the range of basic intelligibility. There is also some textual evidence for this claim. Thus, Davidson writes: [I]f I am right, disputes over values (as in the case of other disputes) can be genuine only when there are shared criteria in the light of which there is an answer to the question who is right When we find a difference inexplicable, that is, not due to ignorance or confusion, the difference is not genuine The importance of a background of shared beliefs and values is that such a background allows us to make sense of the idea of a common standard of right and wrong, true and false. (Davidson 2004, 50-51). I shall refer to this argument as the argument from criteria. I propose to reconstruct the argument as follows. Attributing value judgements entails attributing value concepts. Value concepts are identified by the associated commitments that define their place in a wider network of concepts. These commitments fix the application criteria of value concepts. These application criteria are subscribed

28

to by anyone who competently applies the value concepts. Thus, if two agents appear to disagree in value judgement, then either at least one of them is mistaken about whether the relevant criteria obtain or they are not applying the same concept. If two conflicting agents are applying the same concept, then at least one of them is mistaken about whether the criteria of application for the concept obtain. To be mistaken about whether the criteria of application for a concept obtain is to be ignorant of some (non-evaluative) fact. It follows that the value judgements of (non-evaluatively) enlightened agents will converge.19 Thus interpreted, the argument from criteria fails to take account of at least three different forms of indeterminacy that could arise for the application of evaluative concepts. Each form of indeterminacy is arguably compatible with competent mastery of the relevant concepts. One form of indeterminacy concerns the application of agreed criteria to cases. The two others concern the status of criteria themselves. Each form of indeterminacy arises from facts about evaluative thought that are well known and not particularly controversial. First, the application of shared criteria for evaluative concepts is not as straightforward as some of Davidsons remarks may suggest. This is not only because it can be difficult to establish whether the relevant (non-evaluative) facts obtain. Such difficulties clearly apply to all concepts, evaluative or not. It is also because (as Davidson must agree) the criteria associated with evaluative concepts can themselves be evaluative. Thus, the agreed criteria for whether it is right to give someone a compliment may include whether or not it will do him any good. To establish whether this criterion obtains it is not obviously sufficient to be fully informed about the non-evaluative facts.20 Furthermore, and consistently with
19

As formulated here, Davidsons argument from criteria makes no reference to the principle of charity. It is therefore logically independent of his interpretationist account of the mind. (c.f. Davidson 2004, 55). Pettit 2002 includes a sophisticated variant of the argument from criteria in connection with a discussion of the possibility of rule following in general. Pettits argument is vulnerable to the same problems of indeterminacy that I raise for Davidsons argument below. 20 Perhaps partly for this reason, value concepts are sometimes called essentially contested. See e.g. Hurley 1989, 46-50. Gallie first introduced the notion of essential contestability, as follows: there are disputeswhich are perfectly genuine: which, although not resolvable by argument of any kind, are nevertheless sustained by perfectly respectable arguments and evidence. This is what I mean by saying that there are concepts that are essentially contested, concepts the use of which inevitably involves endless disputes about their proper uses on the part of their users (Gallie, 1955-6,169). Gallie goes on to put forward a set of seven

29

Davidsons argument, some evaluative conflicts could be subject to trade-offs with more than one coherent resolution. In such cases, the inference from commonality of concept and genuine disagreement to the non-evaluative ignorance of at least one party is unsound. Second (and as Davidson explicitly agrees) the scope of convergence is limited by indeterminacy arising from the fact that concepts are sometimes extended to cases not previously accommodated by their existing criteria of application. Thus, Davidson makes the following remark about Michael Dummett, who in one place suggests that attributions of courage to someone who has died without ever having the opportunity to display it would be lacking in truth-value: I agree that such an attribution may have no truth value, but the reason is not that verification is lacking, but that we simply have not had to make up our minds about these cases. I think the same applies to very difficult or unusual moral problems. It is consistent with objectivity that there should be no clear answers about what is right or obligatory in such cases (Davidson 2004, 51). Given the diversity in internal frame or external situation that is found across different agents and times, there could be more than one coherent and reasonable extension of a concept in such difficult or unusual scenarios. If so, the question of how a concept should be extended may have no determinate answer. To the extent that the distinction between invention and discovery blurs at this point, so does the distinction between subjective and objective. Not all difficult scenarios are unusual. Questions of justice and fairness are also quite difficult and contested, yet people are constantly forced to make up their minds about them by the need to make choices, for example when discussing the problem of abortion, issues of resource allocation, or income tax. If these problems are placed at some distance from the core of basic values about which there must be convergence among enlightened agents, it cannot be for the reason that they present competent speakers with novel cases for which criteria have yet to be proposed. They

necessary conditions for essential contestability. His examples of essentially contested concepts include art, democracy, social justice, and some moral uses of good (Gallie, 1955-6, 180-195). My argument is compatible with the essential contestability of evaluative concepts in Gallies sense, although it does not entail it.

30

most certainly have. Yet the proposed criteria remain deeply and intractably contested, even among reasonably intelligent adults. The potential for indeterminacy arises also for a third reason. The issue here is that the existing criteria of application for evaluative concepts are themselves often in dispute between competent speakers. On Davidsons view, concepts are partly defined by a network of commitments that allow speakers to identify them and distinguish them from each other. This network is neither absolutely fixed nor an all-or-nothing matter. While some commitments are more central to the identity of a concept than others (like the norm of instrumental rationality is central to the concept of rational agency), the notion of centrality is one of degree. Two agents can associate non-identical sets of commitments with a disputed evaluative concept and remain coherently describable as thinking about the same thing. If they do, they will associate different criteria of application with that concept. In some cases (although not all) they will apply the concept in incompatible ways. Where there is diversity in internal frame or external situation across different agents and times, such conflicts are possible even where the agents in question are neither obviously unreasonable nor misinformed about the (non-evaluative) facts. Where such scenarios obtain, there will be no determinate answer to the question of what to do in the relevant case.21 It would be a mistake to think that the preceding point relies on a controversial conception of concepts as delimited by essentially vague and indeterminate boundaries. A parallel conclusion would follow even if the boundaries of evaluative concepts could be assumed to be sharp. If two agents associate subtly different application criteria for the term justice, for example, it may be granted that they operate with different concepts. If so, they can agree on all non-evaluative facts, apply the term justice differently, and yet not disagree about what is just. This does not dissolve the original problem. Wherever the two concepts have incompatible action-guiding consequences the question could arise of which of the two concepts, if any, they are to be guided by. Even if the criteria of application themselves are not formulated in evaluative terms, the criteria for their selection could be. To settle on criteria of application for evaluative concepts is
21

There is no easy solution here in appealing to a set of shared higher order concepts for the resolution of disputes about the application criteria of lower order concepts. Any higher order concept would itself be subject to the same problems of indeterminacy as the lower order concept.

31

partly an evaluative matter. Once more, where diversity in internal frame and external situation across agents and times applies, there could be more than one coherent resolution of the conflict. There would then be no determinate answer to the question what to do in the relevant case.22 On Davidsons account, the scope for indeterminacy left open by the argument from criteria is obviously limited by the principle of charity. Thus, it would be impossible for agents to disagree in all their evaluative claims at the same time while remaining able to understand each other, and thus to communicate. Furthermore, there are arguably some criteria a commitment to which is necessary for being able to agree or disagree with anyone about anything. In particular, the convergence claim is most plausible when applied to the comparatively formal norms and values picked out by Davidsons principle of coherence. If enlightened convergence is a mark of objectivity, then any comparably formal values picked out by the principle of coherence are arguably objective. In this sense, if no other, the enlightened values of rational agents would converge. Of these values, Davidson writes: These are principles shared by all creatures that have propositional attitudes or act intentionally; and since I am one of those creatures all thinking creatures subscribe to my basic standards or norms of rationality. This sounds sweeping, even authoritarian, but it comes to no more than this, that it is a condition of having thoughts, judgements, and intentions that the basic standards of rationality have application. (Davidson 2004, 195) It does not follow that more substantial individual evaluative claims are equally immune from indeterminacy so long as there is compensating agreement in substantial value elsewhere. It therefore does not follow that there is a uniquely fixed and determinate set of particular features of the world the positive or negative evaluation of which all rational agents must all share if they are able to
22

Contrary to what some forms of ethical naturalism appear to suggest, this problem of indeterminacy is not easily avoided on an a posteriori naturalist account of moral concepts as natural kind concepts (c.f. Boyd 1988). Even if it were granted that some evaluative concept in fact functions to track a natural kind, the question remains what to make of this fact when deciding what to do. For criticism of a posteriori naturalism along these lines, see e.g. Horgan and Timmons 1992 and Blackburn 1998, 119-121.

32

understand each other, and are otherwise well informed about the (non-evaluative) facts.23 True, some of candidates have been proposed for the title of a priori individually non-intelligible objects of desire, including those of Elizabeth Anscombes infamous Nazi, and saucer-of-mud collector (Anscombe 1957, 73ff). One difficulty in appealing to individual examples like these in the context of Davidsons argument is that his account is arguably consistent with the existence of individual desires of this kind, provided they occur against a background of a basically coherent system of attitudes that can be interpreted as generally responding to the world as we take it to be with respect to its most basic (evaluative and non-evaluative) features. Without further specification of the (external and internal) background, it is not obvious that either a rational Nazi or a saucerof-mud collector would necessarily fail this interpretative test.24 Davidsons account is therefore consistent with an indefinitely wide range of questions of substantial value having no determinate answer. There is ample textual support for this claim among Davidsons remarks on the issue. First, on Davidsons account, the normative status of individual evaluative claims is partly a function of their place in a network of other norms and commitments. For example, Davidson claims the property of rationality is a purely relational feature of beliefs and desires. He writes: Strictly speaking irrationality consists not in any particular belief but in inconsistency within a set of beliefs I think we must say much the same about intentions, intentional actions, and other propositional attitudes They are never irrational in themselves, but only as part of a larger pattern. (Davidson 2004, 192) A parallel claim would arguably apply to any feature of the world that is judged to be valuable on Davidsons account. If no particular features of the world are valuable in themselves, but only as valued by a larger pattern of attitudes, then no particular feature of the world are determinately valuable unless valued by some suitably
23

Davidson does not regard the claim that formal values of consistency are objective as licensing the claim that they are features of the world. Davidsons list of substantial values is inclusive, and includes: good, bad, moral, blameworthy, right, courageous, courteous, trustworthy, loyal, kind, obedient, cheerful, etc. (Davidson 2004, 54). 24 For further discussion of this issue, see e.g. Velleman 1992; Blackburn 1998, 59-68; Raz 1999

33

unique or otherwise privileged pattern of attitudes. 25 Even when supported by the principles of coherence and correspondence, the argument from criteria fails to establish that any uniquely fixed and determinate set of particular features of the world is so valued. To this extent, the argument fails to establish that any unique and determinate set of particular features of the world is objectively valuable.26 Second, and as already noted, Davidsons principle of charity must be compatible with such genuine disagreement about ethics and value as there actually is. There is evidence to suggest that some such disagreements, including disagreements on matters of taste or on abstract moral issues involving justice and fairness, do obtain between rational agents who are in broad agreement about the (non-evaluative) facts. It follows that there is no straightforward route from Davidsons interpretationist account of the mental to the claim that the enlightened values of rational agents would converge. This modest conclusion receives additional support from Davidsons various remarks about moral disagreement in particular. Davidson explicitly notes that there can be real differences in norms among those who understand each other, as long as the differences are placed within a common framework (Davidson 2004, 50). He also concedes that we can make a fairly clear
The idea that the real values are those picked out by what might be described as the most coherent pattern of desire is one that, while it may be a natural extension of Davidsons argument, is not entailed by anything Davidson says. One possible move in this direction would be a so-called coherentist account of practical reasoning with a convergence claim made about enlightened values held compatibly with sound practical reasoning (c.f. Myers 2004, 134). While there is little textual evidence that Davidson is concerned to defend any such account, invoking one would considerably raise the stakes in the debate over the scope and limits of the convergence claim. While the notion of a most coherent network of attitudes is one that has been a subject of much philosophical controversy, nothing I have argued here would amounts to a proof that such a strategy is destined to failure. 26 Susan Hurley may be exhibiting awareness of this issue when she writes that the question of there being alternatives that ought to be done all things considered comes down to the question of there being a theory that does the best job of displaying coherence (Hurley 1989, 193). The identification of such a theory is left as a substantive issue, to be determined by theoretical deliberation (Hurley, 1989, 193). Hurley does not claim to have found a unique theory with this property. The reductive naturalist David Lewis is predictably less keen to claim universal convergence on the part of values in general. Lewis writes: Since our dispositions to value are contingent, they certainly vary when we take all of mankind into account The spectre of relativity within our own world is just a vivid reminder of the contingency of value (Lewis 1989, 133).
25

34

distinction between interpersonal comparisons and the normative judgements based on them, and that conceptions of fairness, justice, and social welfare have no favoured role in our attributions of propositional attitudes (Davidson 2004, 74). When making his convergence claim, Davidson often talks of basic (Davidson 2004, 49, 51), leading (Davidson 2004, 36), or strategically important values (Davidson 2004, 73). 27 When making room for de facto disagreement by distinguishing between central and peripheral values, he illustrates the former by mentioning the paradigmatically formal values of consistency and coherence. Taken together, this textual evidence goes some way to suggest that a great number of ethical and other evaluative concepts are too far distant from the basic values necessarily invoked as shared in interpretation to imply the existence of a determinate resolution to the substantial evaluative disagreements they involve. This is a significant concession to the subjectivist about ethics and value. If objectivity is understood partly in terms of convergence, and if only comparably basic values are ones on which enlightened rational agents would converge, then Davidsons argument fails to support the objectivity of an indefinite range of non-basic and substantial values. Of course, the fact that some substantial evaluative claims are indeterminate in truth-value does not entail that all substantial evaluative claims are indeterminate in truthvalue. Nevertheless, a restriction of the convergence claim to a set of comparatively basic values required to make sense of agents as minimally intelligible would amount to a significant qualification of Davidsons claim that value judgements are objective in much the same way that paradigmatically factual judgements are. This is not because no paradigmatically factual judgements have ever been thought to have indeterminate truth-values. Some clearly have. It is rather because the range of determinate and ontologically substantial fact is naturally thought to extend beyond the range of basic mutual intelligibility. To assert the contrary would be as much a way of debunking paradigmatically factual judgements as a vindication of value judgements. If one consequence of Davidsons argument is that this natural thought is profoundly mistaken, it is a
27

Yet Davidson also appears to say that the convergence claim applies to moral values (Davidson 2004, 47). And the list of values given in the context where he presents the argument for criteria includes good, bad, moral, blameworthy, right, courageous, courteous, trustworthy, loyal, kind, obedient, cheerful, etc. (Davidson 2004, 54). It is a notable feature of Davidsons list that it makes no discrimination between so-called thick and thin ethical concepts.

35

moot point whether the parity between value and fact so established is better captured by denying the subjectivity of value, the objectivity of fact, or the very mode of classification on which the traditional debate between subjectivists and objectivists about fact and value is based. In more than one place, Davidson explains his conception of objectivity in terms of the possession of some truth-value, these being true, false, or possibly neither (Davidson 2004, 42, 56). So understood, ethical and other evaluative claims would be objective even if all evaluative claims were indeterminate or false. Evaluative claims could then be objective even if the views of paradigm subjectivists like John Mackie were proved to be correct and a universal error-theory were true of ethical discourse (c.f. Mackie 1977, 1-49; Lillehammer 2004). Yet this literal reading cannot accurately represent Davidsons considered view about the objectivity of value. 28 It is beyond any reasonable doubt that Davidson thinks a significant range of evaluative claims come out true. What is not beyond reasonable doubt is the extent to which the conjunction of the principle of charity and the argument from criteria support this claim. 2.4. Conclusion The argument from criteria is unsuccessful in its most general form. While the principle of charity entails that agents must share a wide range of substantial values, it fails to establish that any uniquely fixed and determinate set of particular substantial values is objective in Davidsons sense. An indefinite range of substantial evaluative claims are therefore potentially indeterminate in truthvalue. This is not the kind of thesis normally associated with the title of objectivism in ethics or the philosophy of value. It is therefore fair to conclude that even if Davidsons argument may go some way to undermine classical motivations for value subjectivism based on crude appeals to the ontological queerness objective values, it fails to undermine some of the classical motivations for value subjectivism based on the apparently intractable disagreement on substantial ethical and other evaluative claims.

28

It is noteworthy, however, that in one place, Davidson is able to write the following: Hume, Bentham, Hare, Blackburn, and a host of others have emphasized the subjective, emotive, or projective character of value judgements. They are surely right (Davidson 2004, 57). In light of the foregoing remarks, this intriguing claim is perhaps not surprising.

36

My discussion of Davidsons argument leaves at least one serious question unanswered. This question is whether the distinction made between the comparatively formal values embodied in the principle of coherence on the one hand, and the substantial values embodied in the principle of correspondence on the other, is either a clear, sharp, or sound one. Although I have assumed this distinction throughout this chapter, there are those who might question its cogency. A fully comprehensive discussion of Davidsons argument would put this distinction to the test. I have not attempted such a discussion here. Nor have I questioned the basic tenets of Davidsons interpretationist account of the mental. This latter account is at least as philosophically controversial as the distinction between formal and substantial values. Given the limitations of Davidsons companions in guilt argument for the objectivity of values even when considered on its own terms, however, objectivists about ethical and other substantial values would be well advised to look elsewhere in support of their views.

37

3. The Normativity of Practical Reason 3.1. Humean subjectivism One influential form of subjectivism about ethics and value consists in denying the coherence or validity of a certain type of rationally inescapable norm of action. In the terms of a distinction historically associated with the work of Immanuel Kant, this form of subjectivism denies the coherence or validity of so-called categorical imperatives of practical reason, where these are imperatives that apply to agents regardless of their contingently given ends.29 Thus, when John Mackie claimed to have diagnosed a constitutive error embodied in the pre-reflective morality of his time, this was because he claimed to have shown that this morality was committed to the validity of something very much like categorical imperatives of practical reason. Thus, Mackie writes: So far as ethics is concerned, my thesis that there are no objective values is specifically the denial that any categorically imperative element is objectively valid. The objective values which I am denying would be actiondirecting absolutely, nor contingently upon the agents desires and inclinations (Mackie 1977, 29). So understood, objectivism about value is the view that there are some imperatives of practical reason that are categorically valid in
29

Kant introduces the distinction between hypothetical and categorical imperatives in Groundwork to the Metaphysics of Morals as follows: Now, all imperatives command either hypothetically or categorically. The former represent the practical necessity of a possible action as a means to achieving something else that one wills (or that it is at least possible for one to will). The categorical imperative would be that which represented an action as objectively necessary of itself, without reference to another end (Kant 1785, 414). It is a significant, but not always noted feature of this definition of hypothetical imperatives that it is literally speaking applicable to possible as well as actual ends. Hypothetical imperatives in this sense do not necessarily obtain only relative to the actual ends of agents: they are norms of consistency between means and ends, actual or possible. To this extent, it is misleading to characterize subjectivism about value as claiming that all valid imperatives of practical reason are hypothetical in Kants sense. For as that claim is normally made in the contemporary literature, valid imperatives of practical reason are said to be dependent either on the actual ends of agents or on the subset of possible ends they would possess in a set of favourable circumstances, not on any possible end (c.f. Foot 1972; Mackie 1977; Smith 1994).

38

the sense of rationally requiring some course of action unconditionally on any (actual or possible) end.30 I shall refer to the denial of this view as Humean subjectivism . 31 Mackie and subjectivists like him do not normally deny the validity or coherence of the other type of rationally inescapable norm distinguished by Kant, namely hypothetical imperatives. On the contrary, the domain of hypothetical imperatives is supposed to be definitive of the scope of practical reason on Mackies account. Mackies claim is that the scope of practical reason is exhausted by the scope of hypothetical imperatives and the instrumental reasons to which they give rise. It is precisely because Mackie thinks that practical reason is exhausted by instrumental reason alone that he also thinks the demands of pre-theoretical morality are rationally escapable, and all moral judgements pretending otherwise therefore either false or incoherent (c.f. Mackie 1977, 27-30). So understood, the subjectivity of value does not entail complete skepticism about practical reason, in the sense of the refusal to accept any norms of practical reason whatsoever. At least some norms of practical reason are objectively valid on this view, including instrumental norms that prescribe the efficient promotion of contingently given ends. Humean subjectivism might seem to be favoured by the virtues of motivational, epistemological, and metaphysical modesty. The validity of hypothetical imperatives has historically been much less controversial than the validity of categorical imperatives.32
30

One ambiguity sometimes noted in the literature about categorical imperatives is that which arises from the fact that an end can apply to agents regardless of their ends in two different ways, namely a) regardless of their contingently given ends, and b) regardless of any end whatsoever (contingent or necessary). For the purposes of the present argument, I shall read the notion of categoricity in the stronger sense, and thus in accordance with b). This means that subjectivism turns out to be consistent with the claim that some imperatives apply to agents regardless of their contingently given ends provided that some ends are possessed by agents necessarily. For the suggestion that there are some ends that agents possess necessarily, see e.g. Williams 1995, 35-45; Davidson 2004. 31 I call this view Humean in spite of the fact that it is not clear that Hume (as opposed to some of his contemporary followers) ever held it. For discussion, see Millgram 1995. 32 Kant, as well as most commentators who are sympathetic to his work, tacitly agree with subjectivists that the validity of hypothetical imperatives is less controversial than the validity of categorical imperatives. This fact is shown by the way both Kant and his followers generally devote less space to the discussion of hypothetical as opposed to categorical imperatives (one exception being Hill 1973). Thus, Kant writes: How an imperative of skill is possible requires no special discussion. Whoever wills the end also wills (insofar as reason has a

39

While many philosophers have denied that agents are rationally bound by categorical imperatives (be they ethical or otherwise), hardly anyone denies that agents are rationally required to take efficient means to their ends. While the idea that some ends of action are objectively action-guiding in themselves has seemed to many either motivationally, epistemologically, or metaphysically mysterious, the idea that some ends are required on pains of irrationality in order to efficiently promote other ends has traditionally seemed comparatively straightforward. This is especially so for Humeans like Mackie and others, who have sought to find a place for reason and value within a broadly naturalistic worldview (c.f. Gauthier 1986). In recent years, Humean subjectivism has been attacked by a number of philosophers who, inspired by what they take to be Kants view of instrumental reason, argue that Humean subjectivism is faced by an insuperable dilemma (c.f. Korsgaard 1997, Hampton 1998, 125-206). These philosophers argue as follows. Either hypothetical imperatives are normative or they are not. If hypothetical imperatives are normative, then they are valid standards of consistency in action that apply to agents regardless of their ends. If so, they entail a categorically imperative element in practical reason that is objectively valid, and therefore neither more nor less mysterious than categorical imperatives are. On the other hand, if hypothetical imperatives are not normative, then they merely register possible causal or constitutive relationships between states of affairs described as possible ends of action. If so, they are not even candidates for the status of authoritative standards of consistency in action. Humean subjectivism is therefore either implicitly committed to a form of objectivism about value, or it collapses into complete normative skepticism. Either way, there is no viable form of subjectivism about value based on the notion of pure instrumental reason alone. In what follows, I shall refer to this argument as the Kantian argument. The Kantian argument is a type of companions in guilt argument (c.f. Hampton 1998, 11). It claims that the existence of subjective value (as understood by Humeans like Mackie) entails the existence of objective value (as understood by the Kantians whose view of practical reason Humeans reject). The argument is therefore a companions in guilt
decisive influence on his actions) the indispensably necessary means to it that are within his power. This proposition is, as regards the volition, analytic (Kant 1875, 417). This claim is not, of course self-interpreting. For explication, see e.g. ONeill 1989, 89-94; Korsgaard 1997.

40

argument by entailment because the commitment to objective value is supposed to logically follow from the commitment to subjective value. The Kantian argument has two notable implications. First, and most obviously, the Humean project of accounting for practical reason in purely instrumentalist terms is bound to fail. Second, a commitment to objective values is arguably embodied not only in paradigmatically normative judgements like the claim that we should keep our promises, but also in claims that attribute to agents propositional attitudes like beliefs and desires. As noted in Chapter 2, it is a widely held view that the content of the propositional attitudes is partly defined by their basic logical and other rational relations to each other (c.f. Hurley 1989; Davidson 2004). If so, agents must be interpreted as minimally coherent in their beliefs and desires in order to count as having any thoughts at all. The consistency in question comes in two sorts, namely basic logical consistency as defined by the norms of classical logic, and basic instrumental consistency as defined by the norms of rational decision theory. Of these two types of consistency, the second is arguably committed to a normative conception of instrumental rationality. According to Donald Davidson, for example, constitutive constraints of interpretation entail that we must think of agents as being instrumentally rational in the sense of generally pursuing their desires in accordance with their beliefs.33 If the process of interpretation is, as Davidson claims, a normative process of classifying agents as successfully complying with a set of standards of rational behaviour, then the Kantian argument entails that the very idea of a propositional attitude itself contains a commitment to objective value insofar as the norms of instrumental rationality embodied in rational decision theory presuppose the validity of at least one categorical norm of practical reason. If understanding hypothetical imperatives as normative is a presupposition of agency as such, it follows that Humean subjectivism is inconsistent not only with the objectivity of value, but also with the existence of rational agency. 34 Not even the most radical Humean should be attracted to this conclusion.
33

This claim is echoed by neo-Kantians like Korsgaard for the sub-class of selfconscious action. Korsgaard writes: A rejection of the instrumental principle is a rejection of self-conscious action itself (Korsgaard 1997, 248). 34 Of course, not everyone agrees that the relevant form of decision theory should be interpreted normatively. But even those who favour a purely descriptive account of subjective decision theory are generally committed to the normative interpretation of some form of decision theory, such as objective decision theories

41

In a series of books and papers published since the 1990s, Jean E. Hampton and Christine Korsgaard have defended a Kantian conception of the objectivity value by employing a companions in guilt argument against the Humean idea of pure instrumental reason (Hampton 1992, 1994, 1995, 1996, 1998; Korsgaard 1996a, 1996b, 1997).35 Between them, Hampton and Korsgaard give at least three separate arguments against Humean subjectivism, each of which constitutes a distinct claim to companionship in guilt of the entailment form. First, they claim that the idea of a hypothetical imperative is as normative as the idea of a categorical imperative and consequently no less problematic from a naturalist perspective. Second, they claim that the validity of hypothetical imperatives entails the validity of at least one categorical imperative, namely the categorical imperative to be instrumentally consistent. Third, they claim that the normativity of hypothetical imperatives presupposes the correctness of a substantial theory of valuable ends, or the good. Their conclusion is that Humean subjectivists such as Mackie are implicitly committed to objective prescriptivity in exactly the sense that his own arguments are meant to undermine. Humean subjectivism, when fully explicated, is either unmotivated or self-defeating. It does not immediately follow, of course, that there are any valid categorical imperatives. But the price of denying the existence of such imperatives is complete skepticism about practical reason. In what follows, I shall agree that the Kantian argument succeeds in bringing out a number of frequently neglected presuppositions of Humean subjectivism. Thus, as literally formulated by Mackie, for example, the Humean position is genuinely self-defeating. Yet making this concession leaves the core of Humean subjectivism intact, or so I shall argue. The Kantian argument is relatively neutral on the content of both the categorical norms of practical reason and the substantial conception of the good to which Humean subjectivism allegedly commits us. The argument
with subjective utilities and credences replaced by objective utilities and chances (c.f. Blackburn 1998, Mellor 2005). Qua normative, objective decision theories are potential targets for the Kantian argument (c.f. Hampton 1998). 35 Korsgaard and Hampton are not the only Kantians to employ some form of companions in guilt argument against Humean subjectivism. Thus, Allen W. Wood writes: those who find nothing problematic about instrumental rationality should recognize that they cannot consistently object to the idea of a categorical imperative on the ground that it is supposed to be universally valid, necessary, and a priori, for those properties already belong to hypothetical imperatives (Wood 1999, 65).

42

is therefore consistent with a wide range of indeterminacy in the truth-value of many substantial evaluative claims. In particular, conceding the validity of some categorical norms of practical reason does not entail the acceptance of the paradigmatic kind of substantial categorical imperatives of morality the rejection of which is arguably what leads Mackie and others to endorse a Humean view in the first place.36 For all the Kantian argument shows, many of the norms of prudence, morality, aesthetics, politics and the rest on which discussions of the objectivity of value has traditionally focused are rationally escapable, and therefore not objectively binding for the agents to whom they are frequently applied. In making this argument, I shall grant the Kantian understanding of objective value proposed by Hampton and Korsgaard. I shall also grant their claim that hypothetical imperatives are normative in the sense of being universally valid standards of rational criticism. What I shall question is their specifically Kantian understanding of what this normativity amounts to, as well as their Kantian view of what follows from it by way of objective values. I shall conclude by connecting the issue of the objectivity of value to the issue of the rational inescapability of moral and other substantial norms of practical reason, and shall argue that it is partly a reasonably motivated skepticism about the range of such norms that is the guiding thought behind the Humean subjectivism of Mackie and his philosophical allies. 3.2. The normativity of instrumental reason The first strand of the Kantian argument claims that the idea of a hypothetical imperative is as normative as the idea of a categorical imperative, and consequently no less problematic from the perspective of a naturalistic worldview. To avoid confusion in what follows, I shall refer to the imperative that agents should take effective means to their ends as the formal hypothetical
36

This claim is obviously consistent with the possibility that there is no sharp distinction between formal and substantial categorical imperatives. The actuality of this possibility would not affect my argument so long as there remain differences of degrees according to which paradigmatic moral imperatives such as Give 10% of your income to charity count as (relatively) substantial, imperatives such as Promote the means to your ends count as (relatively) formal, and imperatives such as Avoid inconsistent thought fall somewhere in between. What matters for present purposes is that categorical imperatives licensed by the Kantian argument are of a (relatively) formal kind, and therefore not ethically substantial in any interesting sense. While this issue clearly merits further discussion, we do not need to pursue this further discussion here.

43

imperative. I shall refer to concrete applications of this imperative as substantial hypothetical imperatives. The claim that If you want a bribe, you should name your price is a substantial hypothetical imperative in this sense.37 The claim that rational agents will take efficient available means to their ends is a critical, evaluative standard that postulates a set of non-trivial success conditions on rational agency. According to this standard, instrumentally inefficient or self-defeating action is a form of internal incoherence in agency and thereby a paradigmatic form of irrational failure. To the extent that Humean subjectivism is motivated by commitment to a naturalistic worldview, and to the extent that a naturalistic worldview is inconsistent with the existence of normative standards of rationality, the existence of hypothetical imperatives is inconsistent with a naturalistic worldview. On these grounds, Hampton claims to diagnose in Humean subjectivism an implicit or explicit dislike of the inclusion of norms in their theory (Hampton 1998, 198). Having made the point that some normativity is inescapable even on a conception of practical reason as exhausted by hypothetical imperatives, Hampton writes: [A]ny theory of instrumental reason is just as hip-deep in normativity as any moral theory, and therefore just as metaphysically problematic as any moral theory if, as many naturalists believe, normativity is metaphysically problematic. (Hampton 1998, 206) Yet the main point of Humean subjectivism is precisely that instrumental rationality is not just as hip-deep in normativity as any moral theory. On the contrary, instrumental norms are meant to share with a wide range of other norms the absence of the feature that makes paradigmatically moral norms philosophically problematic, namely their objective prescriptivity (c.f. Mackie 1997, 29). Thus, the Humean subjectivist is not forced by consistency to reject normativity as such, but rather to reject

37

The distinction between the formal and substantial hypothetical imperatives is not always duly noted in the literature, and is a potential source of philosophical muddles about the relationship between practical rationality formally understood and the substantial reasons that agents are said to have to pursue certain ends rather than others. For a recent attempt to come to grip with these issues, see. Kolodny 2005.

44

specific forms of it.38 To see why, consider the normative status of social conventions like norms of etiquette, such as If someone offers you their hand, take it. This norm is not objectively prescriptive in Mackies sense. It is possible to reasonably act against it without having misunderstood either its content or social significance. Refusing to take someones hand may be rude, but it does not thereby constitute having made any kind of mistake (except trivially, from the perspective of the norm in question). We must have room in our worldview for such conventional norms, otherwise we could convict people of irrational behaviour by arbitrary stipulation. The Humean subjectivist should not deny the existence of conventional social rules that embody internal standards of correct behaviour. What he or she is really interested in denying is that any such conventional norms are objectively prescriptive in the sense of constituting categorical imperatives that are rationally inescapable regardless of any end (actual or possible) that they may serve to promote. It is plausible that some norms are rationally inescapable in this sense. Arguably, these include the basic laws of logical inference (within limits, because the extent to which an agent has reasons to be consistent may depend on what her other ends are). Logical inconsistency in reasoning is arguably a categorically prescriptive form of irrationality, all else equal. Somewhat more controversially, the set of rationally inescapable norms arguably also includes the basic epistemic norms of belief formation (again within limits, because the extent to which an agent has reasons to have true beliefs may depend on what her other ends are). Wishful thinking is a categorically prescriptive form of irrationality, all else equal. Humean subjectivists like Mackie generally accept the rational inescapability of basic logical and epistemic norms. To that extent they are objectivists about logic and epistemology.39 The
38

It is equally true that Humean subjectivists have traditionally accepted a commitment to only such forms of normativity as can in some sense be naturalised. Mackie commits himself to this view when he writes in response to the companions in guilt charge: The only adequate reply would be to show how, on empiricist foundations, we can construct an account of the ideas and beliefs and knowledge that we have of these matters I can only state my belief that satisfactory accounts of most of these can be given in empirical terms. (Mackie 1977, 39). The question of whether all forms of normativity can be naturalized is orthogonal to the present inquiry. I shall therefore not discuss the issue further here. 39 Hampton extends the Kantian argument to epistemology and science (c.f. Hampton 1998, 207-214). Korsgaard equally draws an explicit link between value on the one hand and epistemology and logic on the other (c.f. Korsgaard 1997,

45

question for Humean subjectivists is whether it is possible to be an objectivist (in this sense) about logical and epistemic norms while being a subjectivist (in the same sense) about a wide range of norms of practical reason, including the categorical norms of pretheoretical morality. The main issue between the Kantian and the Humean is therefore whether the norms of practical reason are inescapable in the same way as the basic norms of logic and belief formation apparently are. It is arguably impossible to coherently opt out of the basic norms of instrumental rationality in the way one can obviously opt out of the conventional norms of etiquette. In one uncontroversial sense, therefore, the basic norms of instrumental rationality are as inescapable as the norms of logic and belief. As we have seen, minimal instrumental consistency is arguably a precondition of rational agency. Failures of instrumental consistency are paradigm cases of practically irrationality. It is therefore impossible to rationally escape the formal hypothetical imperative that one should take effective available means to ones ends. On the other hand, almost any substantial hypothetical imperative is rationally escapable because, as is often pointed out, there are two ways in which one can comply with it: either by taking the means or by abandoning the end.40 It is partly the latter form of escapability that the Humean subjectivist appeals to when he denies that the norms of morality are objectively prescriptive. The point of Humean subjectivism so understood is that agents can comply with the inescapable norms of instrumental rationality without thereby being committed to any particular subset of substantial ends, including the norms of commonsense morality. Unless it can be independently established that the norms of morality are likewise rationally inescapable, the objectivity of instrumental rationality is compatible with the subjectivity of morality in the relevant sense.41 On of the most natural ways of understanding Humean subjectivism is as claiming that there are no rationally inescapable ends apart from the ends of formal consistency in attitude and action. The issue is not
248). The question of the ontological consequences of this commitment is orthogonal to the question of the success of the Kantian argument. I return to the connection between ethics and epistemology in Chapter 4, ethics and science in Chapters 4 and 6, and ethics and logic in Chapter 7, below. 40 The obvious exceptions are ends that involve respecting basic norms of formal consistency, and any necessary ends (if such there be). 41 Of course, Kantians also argue that the norms of morality follow from pure practical reason. But their case for this claim has not traditionally been thought to rest solely on the kind of companions in guilt argument at issue here.

46

one about normativity as such. Nor, in spite of Mackies unhelpful formulation, is it one about whether there are any norms at all that are objectively prescriptive in the sense of being rationally inescapable. Instead, the issue is one about how far the domain of rational inescapability extends.42 While Hampton is right that Humean subjectivists such as Mackie are committed to the rational inescapability of some norms on pains of inconsistency, this does not undermine Humean skepticism about specific substantial normative systems like the commonsense morality of Mackies late 2oth Century culture. At best, Hamptons argument shows that moral scepticism should not be supported by complete skepticism about normativity as such. Yet this is by no means the must charitable way to understand the core position of Humean subjectivists like Mackie and the like.43 3.3. The categoricity of instrumental reason Even if Humeans are right to reject the rational inescapability of moral and other substantial norms, they are arguably committed to the rational inescapability of the formal hypothetical imperative that agents ought to take efficient means to their ends. This fact gives rise to the second strand of the Kantian argument. This second strand attacks the Humean dictum that reason obtains relative to desire. In her book The Authority of Reason, Hampton claims that this dictum is incompatible with the rational inescapability of the formal hypothetical imperative because even though it may depend on your desires whether or not you have a reason to promote a given substantial end, it is not contingent on your desires whether or not you are bound by the formal hypothetical imperative (Hampton 1998, 125-206). According to Hampton, the normativity of the formal hypothetical imperative entails the existence of at least one
42

As Hampton writes (my italics): As Ive noted, some normativity is inescapable for example, with respect to the coherence constraints and with respect to defining such things as excused failures or best means to ends. (Hampton 1998, 194-195). An ambitious Humean would deny the claim that there can be no reductively naturalistic account of instrumental rationality. This, however, is a topic for a different occasion. 43 Here I agree with James Dreier, who writes: As a Humean myself, I think we should be upfront about this. The special status of instrumental reason is due to its being the sine qua non of having reasons at all. We shouldnt be embarrassed to take the insights of Kantian philosophizing to heart Our scepticism should consist in doubts that the content of practical reason is anything like the content of morality. We should be contesting the normative ground, not contesting its very existence (Dreier 1997, 98-99). Dreiers argument is also relevant to the discussion in Section 3.3. below.

47

norm of practical reason the authority of which is not desiredependent, namely the norm of instrumental consistency in action. Pure instrumentalism about practical reason, and any form of value subjectivism based on it, is therefore incoherent. Hampton writes: [T]he authority of this instrumental norm has to be understood noninstrumentally. Because it is the foundation of the idea that we ought to act on means appropriate to the achievement of our ends, it cannot itself be defended consequentially. Thus, understood as an imperative, it is categorical and not hypothetical (Hampton 1998, 140, Note 22)44 To claim that the normative authority of the hypothetical imperative is not itself hypothetical does not entail that the adoption of any arbitrary end makes it irrational or unreasonable to not promote it. Hypothetical imperatives can be complied with by giving up the end. This point is sometimes put by saying that the consequent of a hypothetical imperative is not automatically detachable (c.f. Greenspan 1975, Broome 1999). To mark this distinction, the idea of a reason is sometimes contrasted with the idea of a normative requirement (c.f. Broome 1999).45 A normative requirement, on this
44

Korsgaard makes a related, but distinct, point about the incompleteness of a purely instrumental account of practical reason. She writes: [T]he instrumental principle cannot stand alone. Unless there are normative principles directing us to the adoption of certain ends, there can be no requirement to take the means to our ends. The familiar view that the instrumental principle is the only requirement of practical reason is incoherent (Korsgaard 1997, 220; c.f. Kolodny 2005). Korsgaards point is that unless an agent has a reason to promote an end, the fact that some action is a means to that end fails to constitute a reason for the agent to promote the means to it. Now this fact itself (if it is a fact) does not refute Humean subjectivism. The subjectivist will say that it is the very fact that an agent has an end that makes that end reason giving for her. Korsgaard shows some recognition of this possible response when she writes [H]ypothetical imperatives cannot exist without categorical ones, or anyway without principles which direct us to the pursuit of certain ends, or anyway without something which gives normative status to our ends (Korsgaard 1997, 250). There are different ways in which this claim can be understood, some of which are compatible with Humean subjectivism and some of which are not. A number of alternatives are explored by Korsgaard herself (c.f. Korsgaard 1996a, 49-130). I shall return to this issue briefly in Section 3.4 below. 45 Another way of marking this distinction is to confine talk of rationality t o constraints on consistency and coherence in attitude, and to confine talk of reasons to substantial constraints on correctness in attitude in response to the world. This way of marking the distinction has been subject to much debate (c.f. Scanlon 1998, 25-33; Kolodny 2005). I shall not pursue this terminological issue further here.

48

view, is a standard of coherence on rational attitudes and their relationship to action. The formal hypothetical imperative on this view is a standard of coherence on rational action, specifying the relationships in which rational attitudes stand to each other regardless of their content. A reason on this view is a standard of substantial correctness of rational attitudes in response to the world, specifying the content of ends agents ought to pursue. The formal hypothetical imperative is not in this sense a standard of correctness on rational action, as it fails to set any substantial ends for the agents to whom it applies. It is therefore blind with respect to the question of the substantial correctness of an agents attitudes. The fact that an agent is normatively required to effectively promote the means to her ends does not, on this view, entail that she has any reason to promote the means to any of the ends she actually has. Thus, a completely crazy agent may still be normatively required to show consistency in action even if he has no reasons to promote any one of her silly, deranged, or evil ends.46 In the terminology of reasons and normative requirements, the second strand of the Kantian argument can be reformulated as follows. Even if all reasons obtain relative to desire, reasons do not exhaust the domain of practical rationality. The norms of practical rationality also include the normative requirements of consistency in attitude and action embodied in the formal hypothetical imperative. This normative requirement does not obtain relative to desire. It follows that the norms of practical rationality are not all
46

Broome 1999 accuses Korsgaard of confusing reasons with normative requirements when she writes: For the instrumental principle to provide you with a reason , you must think that the fact that you will an end is a reason for the end. (Korsgaard 1997, 245). This interpretation of Korsgaard is not compulsory. Korsgaard shows awareness that the instrumental principle is conditional and can be complied with either by pursuing the means or abandoning the end when she writes: The hypothetical character of the principle implies that you can actually conform to it in either of two ways: you may take the means, or you may cease to will the end There is no irrationality in this (Korsgaard 1997, 237). When Korsgaard talks of willing an end as taking there to be a reason to promote it she is arguably using the term willing not to pick out the formation of any arbitrary desire, but rather the normative act of legislating for oneself. This is explicit in the passage from her paper quoted by Broome, where she writes: The instrumental principle can only be normative if we take ourselves to be capable of giving laws to ourselves or, in Kants own phrase, if we take our own wills to be legislative (Korsgaard 1997, 246). Korsgaard is here stating her own account of the source of reasons (c.f. Korsgaard 1996a, 90-130). I shall return to this account in Section 3.4 below. The detaching point is also noted by Hill, who says: what the Hypothetical Imperative prescribes is Take the necessary means or else give up the end (Hill 1973, 436).

49

instrumental. Any purely instrumental account of practical rationality is incoherent. As defined by Broome, for example, the distinction between reasons and normative requirements is arguably no deeper than the distinction between formal ends like consistency on the one hand, and substantial ends like prudence or morality on the other. This distinction is arguably one of degree. Even if normative requirements do not set agents any interesting substantial ends, such as being nice to their neighbors or caring about the past, they do impose genuine restrictions on the ends of agents, namely to retain instrumental coherence among their attitudes. While formal coherence can be retained while remaining neutral between most substantial ends, it still requires agents to make structural choices between them. Thus, normative requirements are genuine standards telling agents to be, or to do, some things rather than others. It is consequently neither plainly incoherent or ungrammatical to say that agents who are normatively required to retain consistency in their attitudes have reasons to be consistent, reasons which are not instrumental in deriving from the satisfaction conditions of their ends. Thus formulated, the second strand of the Kantian argument can be reformulated as follows. In order for agents to have instrumental reasons to promote their ends, they must have noninstrumental reasons to be consistent in their attitudes. The rational inescapability of the formal hypothetical imperative therefore entails the rational inescapability of a categorical imperative of consistency in attitude. While this categorical imperative might be blind with respect to the content of the substantial ends of agents, it is not blind with respect to the content of their formal ends, including the end of being consistent in ones attitudes. To the extent that Humean subjectivism is motivated by skepticism about the validity or coherence of any categorically imperative element embodied in evaluative discourse, it is therefore badly motivated. One response to this strand of the Kantian argument is to claim that the force of the hypothetical imperative is itself hypothetical. Thus, agents should obviously promote the means to their ends effectively if they desire to promote their ends. Consistency in attitude is a necessary means to successful action, and (equally obviously) to consistency in attitude itself. This response does not, however, take the Humean subjectivist very far. The first problem is one of explanatory impotence. To account for the normativity of the formal hypothetical imperative by appealing to the formal hypothetical imperative itself quickly threatens with triviality, regress, or circularity. The second problem is one of

50

incoherence. It is hard to see what it could possibly mean for anyone to reject the formal hypothetical imperative outright. True, some ends may be better served by a limited degree of instrumental irrationality. But no ends can be served without being served. Thus, any minimally rational agent would comply at least partially with the formal hypothetical imperative, regardless of their ends. Instrumental rationality is rationally inescapable regardless of the substantial contents of desire. While this may not entail that instrumental rationality is categorically prescriptive in the sense of applying to agents independently of any ends whatsoever, it does entail that instrumental rationality is categorically prescriptive in the sense of being one necessary end. To that extent it is rationally inescapable and objectively valid. A second subjectivist response to the second strand of the Kantian argument would be to universally reject all normative claims, regardless of discourse and context. Independently of the credentials of its ontological cousin (which I like to refer to as normative antirealism), this position (which I like to refer to as normative scepticism) is arguably incoherent. Agents must be committed to some norms of thought and action in order to coherently think or act at all.47 These norms include basic norms of logical consistency. They arguably also include normative requirements on consistency in attitude such as the formal hypothetical imperative.48 The Humean subjectivist is therefore arguably stuck with some genuinely inescapable norms of practical reason on pain of not being able to make coherent sense of thought and action at all. The question is whether this concession refutes the Humean subjectivist position. The answer to this question is less clear than the second strand of the Kantian argument may initially suggest. One potential source of Humean scepticism about the categorical prescriptivity of moral norms is the apparent absence of a coherent explanation for their existence (c.f. Mackie 1977). Where coherent explanations exist, the subjectivist concedes that categorical presciptivity obtains. Thus, a Humean subjectivist like Mackie can coherently accept the categorical prescriptivity of logic without accepting the categorical
47

The case for normative antirealism and normative scepticism comes apart on a conception of normative objectivity as inescapability, where such inescapability carries no commitment to the existence of ontological substantial correlates for the inescapable norms in question (c.f. Davidson 2004; Chapter 2 above) 48 Korsgaard could be making a similar point when she writes: There is no position from which you can reject the government of instrumental reason: for if you reject it, there is no you. (Korsgaard 1997, 254)

51

prescriptivity of morality so long as the explanation of the categorical prescriptivity of the former does not extend to the latter. If the categorical prescriptivity of logic must be presupposed in order for coherent thought or action to be possible, this may provide all the explanation of categorical prescriptivity that the Humean subjectivist needs. If the Kantian argument is well founded, the same explanation is available to account for the objective prescriptivity of normative requirements like the formal hypothetical imperative. Not only is minimal compliance with the formal hypothetical imperative arguably a necessary condition for being interpretable as a rational agent (Davidson 2004). A commitment to the formal hypothetical imperative is also a necessary condition for any form of minimally coherent practical deliberation, whether from the first or the third person point of view. In other words, the formal hypothetical imperative is genuinely inescapable, regardless of the substantial contents of the desires of agents.49 Not so for the substantial norms of morality or prudence about which paradigm Humean subjectivists have historically been skeptical. Thus, moral norms take a categorical form if they are applicable to agents regardless of their substantial ends. Compliance with moral norms in this sense is not a necessary condition for the pursuit of any substantial end whatsoever. This is why the rational inescapabilty of moral norms has been a subject of intelligent controversy among reasonable people for so long. Normative requirements like the formal hypothetical imperative may equally take a categorical form if they apply to agents regardless of their substantial ends. But compliance with the formal hypothetical imperative is arguably a necessary condition for the pursuit of any substantial end whatsoever. This is why the rational inescapability of the formal hypothetical imperative has not been a subject of intelligent controversy among reasonable people in the same way for so long. If so, there is enough of a genuine explanatory asymmetry between basic norms of instrumental rationality and paradigmatically moral or prudential norms to make Humean skepticism about the categorical prescriptivity of moral norms a prima facie reflectively coherent position.

49

Some ends require instrumental incoherence for their realisation, including most obviously the end of being instrumentally coherent. Yet even the realization of this end arguably presupposes instrumental coherence in some respect insofar as it genuinely constitutes the realization of an end (c.f. Williams 1995, 35-45)

52

The categorically imperative element embodied in the formal hypothetical imperative is therefore not a source of incoherence in Humean subjectivism as such. Categorical prescriptivity is acceptable where it is explicably inescapable. This is arguably the case with the formal hypothetical imperative, but arguably not with moral and other substantial imperatives. This distinction is related to the fact that whereas the former concerns a formal issue of coherence among an agents attitudes, the latter concerns the substantial specification of an agents ends. One (potentially misleading) way of making this point is to say that the prescriptivity of normative requirements is not of the same kind as the prescriptivity of reasons. Another (and to my mind more straightforward) way of making this point is to say that such categorical prescriptivity as exists on the rational selection of ends is a relatively formal, and therefore relatively neutral, constraint on the choice of substantial ends. Put either way, Humean subjectivism is compatible with the categorical prescriptivity of the formal hypothetical imperative. While a complete normative skeptic would go further and reject all forms of categorical prescriptivity, this option is not made compulsory for the Humean by the second strand of the Kantian argument. It follows that while Mackies official formulation of Humean subjectivism demonstrably overstates the subjectivist case, that case is not self-defeating in the way that some neo-Kantians such as Hampton and others have claimed. 3.4. Instrumental reasoning and the good Both Hampton and Korsgaard argue that the normativity of instrumental reason is incompatible with complete evaluative neutrality. If this is right, then the objective prescriptivity of instrumental reason cannot be purely formal, since claims about value (or what Hampton calls the good) are claims about the substantial ends that agents ought to pursue. The basic thought behind this third strand of the Kantian argument is as follows (c.f. Hampton 1998, 167-206). In order for the formal hypothetical imperative to be action guiding, instrumental reasoning must be guided by three different kinds of constraint. First, there are structural constraints on how ends can be rationally ordered. Not every structure of ends is rational because not every structure of ends is coherent. Yet even if this generic description of the rational demand for coherence is granted on all sides, there are different ways of being coherent and different views about what rationality demands by way of coherence. Thus, Hampton writes:

53

 [A]nyone interested in defining instrumental rationality is going to have to take a normative stand on what coherence constraints are a stand that simultaneously involves taking a stand on the structure of a rational agents good. (Hampton 1998, 172) 50 If this is all, however, the point made in Section 3.3 about the substantial neutrality of instrumental reason would still hold. Norms of coherence are relatively silent on the substantial content of an agents ends. Second, there are formal constraints on the source of ends to be rationally ordered. Not every possible object of desire is consistent with an agents present ends. Present ends can conflict with (actual) past and (determinate, probable, or possible) future ends of either the agent herself or other agents (c.f. Nagel 1970). Furthermore, some non-actual ends are potential candidates for inclusion in a rational structure of ends. These may include the ends one would have if one were better informed or fully impartial, imaginatively aware, and so on (c.f. Williams 1982; Lewis 1989; Smith 1989). Deciding on which ends to include in practical reasoning cannot be purely a matter of applying the formal hypothetical imperative, since the application of this imperative presupposes an answer to the question of which ends to apply it to. Thus, the Humean default position that only actually occurring present desires count amounts to a substantial claim about what matters from the point of view of practical reason, and does arguably not follow analytically from the meaning of the formal hypothetical imperative itself. Thus, Hampton writes: [T]he instrumental reasoner, in order to specify his conception of reasoning, has to say what a good-defining preference is, and doing so means that he must take a stand on the sources of the set of good-defining preferences. (Hampton 1998, 188) These so called source-defining constraints on preferences go beyond the purely structural constraints postulated above insofar as
50

In a similar vein, Korsgaard writes: [T]he normative force of the instrumental principle does seem to depend on our having a way to say to ourselves of some ends that there are reasons for them, that they are good. (Korsgaard 1997, 251). See also Quinn 1993, 210-255.

54

they not only impose a standard on how ends are to be rationally ordered but also on the selection of ends to rationally order. Even so, they share the relative content neutrality of the purely structural constraints considered above, and therefore also share their comparatively formal status, even if the question of formality is one of degree. No further conclusions follow from these constraints about the existence other, and substantial, values embodied in the idea of instrumental reason alone. The third set of putative constraints on rational preference does have substantial evaluative implications. Hampton and Korsgaard both argue that the normativity of instrumental reason presupposes the normative discrimination between the substantial content of ends. This claim can take at least two different forms. In its first form, the argument is that rational agents must take up a normative stand of endorsement towards some substantial ends in order to genuinely will (as opposed to merely wish or desire) their realization. In this vein, Korsgaard writes: I must have something to say to myself about why I am doing that something better, moreover, than the fact that this is what I wanted It looks as if the end is one that has to be good, in some sense that goes beyond the locally desirable (Korsgaard 1997, 250)51 Korsgaards thought in this passage is that from the first person perspective the selection of ends is usually not rationally arbitrary. We select ends on the basis of features they have that make them attractive to us. At bedrock, these features cannot consist purely in the fact that some things rather than others will serve as means to our ends. It would appear to follow that the normativity of instrumental reasoning goes beyond the purely instrumental. Hampton appears to concur when she writes:
51

Elsewhere, Korsgaard goes beyond this claim to suggest that the act of willing is itself a source of the normativity of ends. Se writes: Its not exactly that there has to be a further reason; its just that you must take the act of your own will to be normative for you that your willing the end in a sense makes it good. The instrumental principle can only be normative if we take ourselves to be capable of giving laws to ourselves (Korsgaard 1997, 245-6). As noted in Section 3 above, Korsgaard is here taking a distinctive view on the issue of source of rational ends. In line with her Kantian (and broadly antirealist) sympathies, her position remains relatively formalistic in that the notion of rational willing is defined without reference to any interesting set of substantial ends. For further discussion, see Korsgaard 1996a, 90ff.

55

Our pursuit of a conception of instrumental reason requires not only structural constraints on a conception of the good, and not only source defining constraints on a theory of the good, but also content-defining constraints on a conception of the good To put the point succinctly, what counts as a good-defining preference isnt self-defining; hence to know what counts as a potentially good-defining-preference one must have a theory that picks out such a thing, and this theory must take a stand on the nature of a good associated with rational action. (Hampton 1998, 204-205).52 In its second (and stronger) form, the argument is that for at least some ends, rational agents must engage in substantial evaluation in order not only to rank them but also to make up their minds about what they are. The ends in question are the objects of what John Broome has called nonpractical preferences, namely values, desires, or other attitudes that are not defined in terms of what is directly revealed by an agents behaviour. For such ends, the first person process of discovery is a form of evaluatively informed decision-making. Broome defends a version of this claim when he writes: How is the person to know that she desires to desire, the object? I can think of no plausible answer if this state is one that simply imposes itself on the person causally Most plausibly, a person will find out whether she desires to desire something by considering whether she has a reason to desire it. This is a matter of estimating its goodness, and it is a rational process. (Broome 1993, 85-86) Broomes claim does not entail that evaluative judgement is a necessary condition for knowledge of desire as such. He can concede that there is a practical sense of desire in which desire is revealed in behaviour. Furthermore, some desires do simply impose themselves on persons causally. Nevertheless, Broome is arguably right that there is a further sense in which agents can be said to only desire what they evaluatively endorse - a state the paradigm knowledge of which is the act of endorsement itself. Furthermore, it is arguably the content of such evaluatively loaded (or what in
52

Hamptons reference to a theory of the good may overstate her case here. I shall return this issue below.

56

Chapter 2 I referred to as judgement sensitive) attitudes that are in question when agents apply the formal hypothetical imperative to their ends in order to discover what instrumental reason demands of them. Even if there are alternative modes of epistemic access to the ends that agents endorse (for example from the third person perspective), the act of substantially evaluating them is often a part of what produces the content to which the formal hypothetical imperative is applied in practical reasoning from the first person point of view. The claim that the normativity of instrumental reason depends on taking up a substantial view of the good is not equivalent to the previously discussed claim that the normativity of instrumental reason is committed to categorically prescriptive norms. While both claims deny that practical rationality is purely instrumental, they do so in different ways. Where the former postulates the need for rational agents to value some substantial ends, the latter postulates the need for agents to regard some ends as categorically prescriptive. The latter claim asserts that practical reason is not purely hypothetical. The former claim asserts that practical reason is not purely formal. It does not follow that the nonhypothetical aspect of practical reason extends beyond the purely formal. Yet while not equivalent, the two claims have similar limitations as potential premises in a conclusive argument against Humean subjectivism. First, the rational inescapability of having at least some substantial values does not entail the rational inescapability of any specific substantial values. The fact that rational agents need to have substantial values in order to coherently apply the formal hypothetical imperative is itself relatively neutral on the content of these values (whether they include values of prudence or etiquette, for example). The same applies equally to the aforementioned source-defining constraints on ends and what Hampton calls the structure of an agents good. The fact that rational agents need to select a subset of ends from an indefinitely large source set of possible ends is itself relatively neutral on the content of that subset (whether they include the objects of either past, or merely hypothetical, desires, for example). The fact that rational agents need to order their desires in accordance with some constraint of coherence is itself relatively neutral on the exact nature of that constraint (how far it extends beyond bare logical consistency, for example). Yet it is partly the idea that there is a high degree of content neutrality embodied in the normativity of instrumental reason that has historically made Humean subjectivists like Mackie

57

into skeptics about the categorical prescriptivity of paradigmatically moral norms (such as norms prohibiting lies, theft, or promise breaking, for example). Pointing out that the normativity of instrumental reason cannot be purely formal does not in itself undermine the Humean subjectivist view in this respect. What it suggests instead is that the issue between the Humean and her opponent is not an all-or-nothing matter.53 Second, both the Humean subjectivist and her opponent can agree that rational agents must evaluatively endorse some substantial ends in order to apply the formal hypothetical imperative in action. It does not follow that all rational agents must do so in the same way. If there is more than one way to coherently structure ones ends, different agents (or the same agent at different times) may differ over which to choose. This applies not only to the choice of substantial values (prudence versus morality, say), but also to source-defining preferences (present versus future ends, say) and conceptions of coherence (minimal consistency versus explanatory coherence, say). While Hampton is right that an objective theory of practical reason in general (and of instrumental reason in particular) would have to decide between these, the mere fact that such decisions must be made by agents in practical reasoning does not guarantee that there is only one reflectively coherent way to make the choice. Nor does it entail that there is only one reflectively coherent way to theoretically systematise such choices. To the extent that the choice between coherent sets of ends is rationally indeterminate, the subjectivist claim that practical reason is relatively neutral between substantial values remains unrefuted. Separate argument is required to decide whether there is an interestingly small subset of ends that is uniquely privileged from the perspective of practical reason. The main issue in the present context is whether the fact that the normativity of instrumental reason is not purely instrumental refutes Humean subjectivism outright. With respect to the third strand of the Kantian argument, as with the first and second strand, the answer is negative. In fact, the claim that some element of subjectivity in value is left open by
53

It may be tempting to think otherwise if one reads too literally such claims as Hamptons that: defining the instrumental role of reason will involve taking a variety of normative stands, including a way of defining an agents good . (Hampton 1998, 174). While this claim is supported by the third strand of the Kantian argument if good is understood along the lines of favoured by norms genuinely endorsed, it is not equally supported if good is understood along the lines of what is objectively in the agents interest (or what is objectively good for the agent).

58

the Kantian argument is strongly suggested by the following remark from the final section of Korsgaards paper The Normativity of Instrumental Reason, where she writes: I have shown nothing so far about the content of those principles [T]he reason to pursue the end which is needed to support the reason to take the means can be as thin and unsubstantial as the agents arbitrary will, his raw and unmotivated decision that he will take a certain end to be normative for himself, for no other reason than that he wills it (Korsgaard 1997, 252)54 These remarks suggest that Korsgaard at least should be prepared to concede that the third strand of the Kantian argument is unable on its own to refute Humean subjectivism. Third, some formulations of the third strand of the Kantian argument give a potentially misleading picture of what it means to regard some feature of an end as valuable or good. The remark quoted from Broome above, for example, may suggest that there are only two ways of thinking about whether an end should be adopted as an object of will or not: either we observe whether we find ourselves gripped by it as a matter of pure causal receptivity or we attribute to the end some evaluative property in virtue of one of its (intrinsic or extrinsic) features. Yet one way of deciding whether something speaks in favour of adopting a certain end is to discover whether it fits coherently with ones existing attitudes, potentially including all of ones beliefs, desires, and other values. Thus, agents rarely (if ever) suspend their commitment to all ends completely in order to only select ends instantiating some appropriate evaluative feature. On the contrary, ends are generally evaluated against a background of existing attitudes our commitment to which is held relatively (but not indefeasibly) constant. This fact leaves open the possibility that what constitutes the goodness or source-defining status of an end is nothing other than its coherence with the agents existing attitudes, even if the process of deliberating about the goodness of ends does not only consist in directly comparing for coherence potential ends with the agents actual attitudes. In fact, in many cases the most effective way to adopt ends coherent with ones actual attitudes could be to form desires for ends based on the
54

Elsewhere, Korsgaard moves beyond this claim to argue (on different grounds) that substantial ends, including those of the commonsense morality of her time, are required by practical reason. See e.g. Korsgaard 1996a, 131-166.

59

attractiveness or perceived interest of their (intrinsic or extrinsic) features. On the assumption that the agents ends already are largely coherent, this procedure would be as likely as not to retain that coherence. Furthermore, exclusive attention to the coherence or otherwise of potential with existing ends could be as likely as not to draw attention away from the features of those ends on which the question of coherence depends. The application of a coherence constraint on the choice of ends is therefore consistent with maintaining a focus on apparently good-making features of ends in practical deliberation (c.f. Lillehammer 2002). If so, the claim that the normativity of instrumental reason is evaluatively loaded would have to be re-interpreted. For on the present assumption, while the point would remain that the formal hypothetical imperative is normative and has categorically prescriptive authority, it would no longer be the case that the normativity of instrumental reason extends beyond the purely formal constraints embodied in the rationally inescapable norms of logical and instrumental consistency and coherence. 3.5. Conclusion The Kantian argument fails to refute Humean subjectivism about value. There is a consistent form of Humean subjectivism about value that accepts the categorical prescriptivity of the formal hypothetical imperative. This form of subjectivism is neither committed to complete normative skepticism nor to the categorical prescriptivity of substantial norms like those of the commonsense morality of our time. To the extent that some formulations of Humean subjectivism have been based on a complete rejection of all forms of categorical prescriptivity, however, they can arguably be shown to be genuinely inconsistent. But a form of Human subjectivism that accepts categorical prescriptivity where it is explicably inescapable (and is skeptical about it where it is not) is not only minimally coherent, it is also a theoretically respectable view about the scope and limits of practical reason. The preceding discussion suggests a reading of the debate between objectivists subjectivists about value as being composed of a series of interlocking strands, some of which are epistemological, some metaphysical, some descriptive, and some normative. In particular, the discussion in this chapter suggests that one of the central issues at stake in debates about the objectivity of value is the scope of rationally inescapable norms, and the range of indeterminacy of value caused by limits to the scope of reasonable convergence among the internally coherent values of different

60

agents at different times, and in different social or historical contexts (c.f. Harold 2003). While determining the precise scope and limits of reasoned convergence in values across all agents and times would fall way outside the remit of the present enquiry, it is relatively safe to suggest that it is unlikely to be an all-or-nothing matter. One possible upshot of the Kantian argument is therefore a refocusing of the debate about the objectivity of value from a focus on concepts like truth or fact on the one hand, to a focus on concepts like rationality and necessity on the other. Quite independently of this suggestion, however, the present discussion leaves one important question unanswered. This question is whether the notions of truth or fact themselves make sense in the absence of any reference to substantial goods like epistemic, prudential, or ethical values. According to some philosophers, a negative answer to this question provides the basis for a different kind of companions in guilt argument for the objectivity of ethics and value (c.f. Putnam 1981). If these philosophers are right, there could be life in the companions in guilt strategy yet, regardless of the fate of the Kantian argument. The scope and limits of this form of companions in guilt argument is the focus of the next chapter.

61

BIBLIOGRAPHY Altham, J. 1995 Reflection and Confidence in J. E. J. Altham and Ross Harrison (eds.) World, Mind, and Ethics Cambridge: Cambridge University Press, 156-169. Anscombe, E. 1957 Intention Oxford: Basil Blackwell. Anscombe, G. E. M. 1958 Modern Moral Philosophy Philosophy 33, 1-19. Audi, R. 2004 The Good in the Right Princeton: Princeton University Press. Ayer, A. J. 1947 Language, Truth and Logic, 2 edn. London:Victor Gollancz. Blackburn, S. 1992 Through Thick and Thin Proceedings of the Aristotelian Society, Supplementary Volume 66, 285-299. Blackburn, S. 1993 Essays in Quasi-Realism Oxford: Oxford University Press. Blackburn, S, 1994 Enchanting Views in P. Clark and B. Hale (eds.) Reading Putnam Oxford: Blackwell, 12-30. Blackburn, S. 1998 Ruling Passions Oxford: Oxford University Press. Blackburn, S. 1999 Is Objective Moral Justification Possible on a Quasi-Realist Foundation? Inquiry 42, 213-228. Boghossian, P, and Velleman J. D. 1989 Color as a Secondary Quality Mind 98, 81-103. Boyd, R. 1988 How to be a Moral Realist in G. Sayre-McCord (ed.) Essays on Moral Realism Cornell: Cornell University Press, 181-228. Brandom, R. (ed.) 2000 Rorty and his Critics Oxford: Blackwell. Broome, J. 1993 Can a Humean be moderate? in R. G. Frey and C. Morris (eds.) Value,Welfare, and Morality Cambridge: Cambridge University Press, 68-87. Broome, J. 1997 Is Incommensurability Vagueness? in R. Chang (ed.) Incommensurability, Incomparability, and Practical Reason Cambridge: Harvard University Press, 67-89. Broome, J. 1999 Normative Requirements Ratio 12, 398-419. Brower, B. 1993 Dispositional Ethical Realism Ethics 103, 221 249. Budd, M. 2002 The Aesthetic Appreciation of Nature Oxford:

62

Oxford University Press. Cavell, S. 1969 Must We Mean What We Say? New York: Charles Scribner & Sons). Conant, J. 1997 Wittgensteins Philosophy of Mathematics Proceedings of the Aristotelian Society 97, 195-222. Davidson, D. 1980 Essays on Actions and Events Oxford: Oxford University Press. Davidson, D. 1984 Inquiries into Truth and Interpretation Oxford: Oxford University Press. Davidson, D. 2001 Subjective, Intersubjective, Objective Oxford: Oxford University Press. Davidson, D. 2004 Problems of Rationality Oxford: Oxford University Press. Dreier, J. 1997 Humean Doubts about the Practical Justification of Morality in G. Cullity and B. Gaut (eds.) Ethics and Practical Reason (Oxford: Oxford University Press), 81-99. Dworkin R. 1977 Taking Rights Seriously Cambridge: Harvard University Press. Dworkin, R. 1996 Objectivity and Truth: Youd Better Believe it Philosophy and Public Affairs 25, 87-139. Foot, P. 1972 Morality as a System of Hypothetical Imperatives Philosophical Review 81, 305-316. Frankfurt, H. 1971 Freedom of the Will and the Concept of a Person Journal of Philosophy 68, 5-20. Franklin, J. 2004 On the Parallel between Mathematics and Morals Philosophy, 79, 97-119. Gallie, W. B. 1955 Essentially Contestable Concepts Proceedings of the Aristotelian Society 56, 167-198. Gauthier, D. 1986 Morals by Agreement Oxford: Oxford University Press. Geuss, R. 2001 History and Illusion in Politics Cambridge: Cambridge University Press. Gibbard, A. 1990 Wise Choices, Apt Feelings Oxford: Oxford University Press. Gibbard, A. 1992 Through Thick and Thin Proceedings of the Aristotelian Society, Supplementary Volume 66, 267-283. Greenspan, P. 1975 Conditional Oughts and Hypothetical Imperatives Journal of Philosophy 72, 259-276. Hampton, J. E. 1992 Rethinking Reason American Philosophical Quarterly 29, 219-236. Hampton, J. E. 1994 The Failure of Expected Utility Theory as a Theory of ReasonEconomics and Philosophy 10, 195-242.

63

Hampton, J. E. 1995 Does Hume have an Instrumental Conception of Practical Reason? Hume Studies 21, 57-74. Hampton, J. E. 1996 On Instrumental Rationality in J. B. Schneewind (ed.) Reason, Ethics, and Society Chicago: Open Court, 84-116. Hampton, J. E. 1998 The Authority of Reason Cambridge: Cambridge University Press. Harman, G. 1977 The Nature of Morality Oxford: Oxford University Press. Harman, G. 2000 Explaining Value Oxford: Oxford University Press. Harold, J. 2003 Practical Reason and Companions in Guilt Philosophical Investigations 26, 311-331. Heal, J. 1989 Ethics and the Absolute Conception Philosophy 94, 49-65. Heal, J. 2003 What are Psychological Concepts for? in her Mind, Reason, and Imagination Cambridge: Cambridge University Press, 225-249. Hill, T. E. Jr. 1973 The Hypothetical Imperative Philosophical Review 82, 429-450. Hookway, C. 1986 Two Conceptions of Moral Realism Proceedings of the Aristotelian Society, Supplementary Volume 60, 189-205. Hookway, C. 1995 Fallibilism and Objectivity: Science and Ethics in J. E. J. Altham and R. Harrison (Eds.) World, Mind, and Ethics (Cambridge: Cambridge University Press), 46-67. Horgan, T and Timmons, M. 1992 Troubles for New Wave Moral Semantics: the Open Question Argument Revived Philosophical Papers 21, 153-175. Hume, D. 1993 Selected Essays Oxford: Oxford University Press. Hurley, P. 2002 A Davidsonian Reconciliation of Internalism, Objectivity, and the Belief-Desire Theory The Journal of Ethics 6, 1-20. Hurley, S. 1989 Natural Reasons Oxford: Oxford University Press. Jackson, F. 1998 From Metaphysics to Ethics Oxford: Oxford University Press. Jardine, N. 1995 Science, Ethics, and Objectivity in J. E. J. Altham and R. Harrison (eds.) World, Mind, and Ethics (Cambridge: Cambridge University Press), 32-43. Johnston, M. 1989 Dispositional Theories of Value Proceedings of the Aristotelian Society, Supplementary Volume 63, 139 174.

64

Kant, I. 1785 Groundwork of the Metaphysics of Morals Trans. H. J. Paton. New York:Harper & Row, 1956. Kneale, W. 1952 Objectivity in Morals in W. Sellars and J. Hospers (eds.) Readings in Ethical Theory New York: Appleton-Century-Crofts, 681-97. Kolbel, M. 2002 Truth without Objectivity London: Routledge. Kolodny, N. 2005 Why Be Rational? Mind 114, 509-563. Korsgaard, C. 1996a The Sources of Normativity Cambridge: Cambridge UniversityPress. Korsgaard, C. 1996b Creating the Kingdom of Ends Cambridge: Cambridge UniversityPress. Korsgaard, C. 1997 The Normativity of Instrumental Reason in G. Cullity and B. Gaut (eds.) Ethics and Practical Reason Oxford: Oxford University Press, 215-54. Kripke, S. 1980 Naming and Necessity Oxford: Blackwell. Kripke, S. 1982 Wittgenstein on Rules and Private Language Oxford: Blackwell. Lang, G. 2001 The Rule-Following Considerations and Metaethics: Some False Moves European Journal of Philosophy 9, 190 209. Lear, J. 1983 Ethics, Mathematics and Relativism Mind 92, 38-60. Lear, J. 1984 Moral Objectivity in S. C. Brown (ed.) Objectivity & Cultural Divergence Cambridge: Cambridge University Press, 135-170. Lewis, D. 1984 On the Plurality of Worlds Oxford: Oxford University Press. Lewis, D. 1984b Putnams Paradox Australian Journal of Philosophy 62, 221-36. Lewis, D. 1989 Dispositional Theories of Value Proceedings of the Aristotelian Society, Supplementary Volume 63, 113137. Lewis, D. 1996 Desire as Belief II Mind 105, 303-314. Lillehammer, H. 1999 Normative Antirealism Southern Journal of Philosophy 37, 221-225. Lillehammer, H. 2002 Moral Realism, Normative Reasons, and Rational Intelligibility Erkenntnis 57, 47-69. Lillehammer, H. 2003 Debunking Morality: Moral Error Theory and Evolutionary Naturalism Biology and Philosophy 18, 566-581. Lillehammer, H. 2004 Moral Error Theory Proceedings of the Aristotelian Society 104, 93-109. McDowell, J. 1986 Critical Notice of Bernard Williams Ethics and the Limits of Philosophy Mind 95, 377-386.

65

McDowell, J. 1998 Mind, Value & Reality Cambridge, MA: Harvard University Press. MacIntyre, A. 1985 After Virtue London: Duckworth. Second Edition. Mackie, J. L. 1977 Ethics: Inventing Right and Wrong Harmondsworth: Penguin. Mellor, D. H. 2005 What Does Subjective Decision Theory Tell Us? in H. Lillehammer and D. H. Mellor (eds.) Ramseys Legacy Oxford: Oxford University Press, 137-148. Millgram, E. 1995 Was Hume a Humean? Hume Studies 21, 75 93. Moore, A. W. 1997 Points of View Oxford: Oxford University Press. Moore, A. W. 1999 Conative Transcendental Arguments and the Question Whether There Can Be External Reasons in R. Stern (ed.) Transcendental Arguments Oxford: Oxford University Press, 271-292. Moore, A. W. 2003 Williams on Ethics, Knowledge, and Reflection Philosophy 78, 337-354. Myers, R. H. 2004 Finding Value in Davidson Canadian Journal of Philosophy 34, 107-136. Nagel, T. 1970 The Possibility of Altruism Oxford: Oxford University Press. Nagel. T. 1986 The View from Nowhere Oxford: Oxford University Press. ONeill, O. 1989 Constructions of Reason Cambridge: Cambridge University Press. Peacocke, C. 2004, The Realm of Reason Oxford: Oxford University Press. Pettit, P. 2002 Rules, Reasons, and Norms Oxford: Oxford University Press. Putnam, H. 1981 Reason, Truth, and History Cambridge: Cambridge University Press. Putnam, H. 1990 Realism with a Human Face Cambridge: Harvard University Press. Putnam, H. 1993 Objectivity and the Science-Ethics Distinction in M. Nussbaum and A. Sen (eds.) The Quality of Life (Oxford: Oxford University Press), 143-164. Putnam, H. 1994 Words & Life (ed. J.Conant) Cambridge: Harvard University Press. Putnam, H. 1996 On Wittgensteins Philosophy of Mathematics Proceedings of the Aristotelian Society, Supplementary Volume 70, 243-264.

66

Putnam, H. 2002 The Collapse of the Fact/Value Distinction Cambridge: Harvard University Press. Putnam, H. 2004 Ethics without Ontology Cambridge: Harvard University Press. Quinn, W. 1993 Morality and Action Cambridge: Cambridge University Press. Railton, P. 1986 Moral Realism Philosophical Review 95, 163207. Railton, P. 1997 On the Hypothetical and Non-Hypothetical in Reasoning about Belief and Action in G. Cullity and B. Gaut (eds.) Ethics and Practical Reason (Oxford: Oxford University Press), 53-79. Raz, J. 1999 Agency, Reason, and the Good in his Engaging Reason Oxford: Oxford University Press, 22-45. Rawls, J. 1999 The Law of Peoples Cambridge: Harvard University Press Rorty, R. 1989 Contingency, Irony, and Solidarity Cambridge: Cambridge UniversityPress. Ross, W. D. 1930 The Right and the Good Oxford: Oxford University Press. Scanlon, T. M. What We Owe Each Other Cambridge: Harvard University Press. Schafer-Landau, R. 2003 Moral Realism: A Defence Oxford: Oxford University Press. Sibley, F. 1967/8 Colours Proceedings of the Aristotelian Society 68, 145-66. Sibley, F. 1968 Objectivity and Aesthetics Proceedings of the Aristotelian Society, Supplementary Volume 42, 31-54. Smith, M. 1989 Dispositional Theories of Value Proceedings of the Aristotelian Society, Supplementary Volume 63, 89-111. Smith, M. 1994 The Moral Problem Oxford: Blackwell. Smith, M and Stoljar, D 1998 Global Response Dependence and Noumenal Realism The Monist 81, 85-111. Sosa, E. 1993 Putnams Pragmatic Realism The Journal of Philosophy 110, 605-26. Stich, S. 1983 From Folk Psychology to Cognitive Science Cambridge: MIT Press. Sturgeon, N. 1985 Moral Explanations in D. Copp and D. Zimmermann (eds.) Morality,Reason, and Truth Totowa: Rowan and Allanheld, 49-78. Tanner, M. 1968 Objectivity and Aesthetics Proceedings of the Aristotelian Society, Supplementary Volume 42, 55-72.

67

Taylor, C. 2003 Ethics and Ontology The Journal of Philosophy 100, 305-20. Velleman, J. D. 1992 The Guise of the Good Nous 26, 3-26. Werner, R. 1983 Ethical Realism Ethics 93, 653-679. Wiggins, D. 1987 Needs, Values, Truth Oxford: Blackwell. Wiggins, D. 1990 Moral Cognitivism, Moral Relativism, and Motivating Moral Belief Proceedings of the Aristotelian Society 91, 61-85. Wiggins, D. 1995 Categorical Requirements: Kant and Hume on the Idea of Duty in Hursthouse, R., Lawrence, G., and Quinn, W. Virtues and Reasons Oxford: Oxford University Press, 297-330. Wiggins, D. 2005 Objectivity in Ethics: Two Difficulties, Two Responses Ratio 18, 1-26. Williams, B. 1978 Descartes: the Project of Pure Enquiry Harmondsworth: Penguin. Williams, B. 1982 Moral Luck Cambridge: Cambridge University Press. Williams, B. 1985 Ethics and the Limits of Philosophy London: Fontana. Williams, B. 1995 Making Sense of Humanity Cambridge: Cambridge University Press. Wittgenstein, L. 1953 Philosophical Investigations Oxford: Basil Blackwell. Wood, A. W. 1999 Kants Ethical Thought Cambridge: Cambridge University Press. Wright, C. 1992 Truth and Objectivity Harvard: Harvard University Press. Wright, C. 2002 What Could Antirealism about Ordinary Psychology Possibly Be? Philosophical Review 111, 205233.

68