THE REASONS THAT MATTER Stephen Finlay [This is a preprint of an article whose final and definitive form has been published in The Australasian Journal of Philosophy © 2006 copyright Taylor & Francis; The Australasian Journal of Philosophy is available online at: http://www.journalsonline.tandf.co.uk/openurl.asp?genre=article&issn=00048402&volume=84&issue=1&spage=1 Some facts are normative reasons for action, or reasons that support a claim that we ought to act. I am disposed to claim that the fact that George W. Bush’s foreign policy undermines international law is a normative reason for voting against him in the 2004 presidential election. Some who vote against him will do so for this very reason: normative reasons sometimes motivate us.1 When they do, our normative reasons become our motivating reasons, or reasons for which we do act. But normative reasons don’t always motivate, and it is commonly maintained that not all motivating reasons are normative. Some will vote for Bush because he supports a constitutional ban on gay marriage, but I am ordinarily disposed to claim that this isn’t a reason to vote for him. What are the necessary conditions for a fact to be a normative reason? The orthodox Humean account articulated by Bernard Williams is, minimally: Reasons Internalism: a person has a normative reason for φ-ing only if φ-ing might contribute to the satisfaction of one or more or her existing motivations. In opposition reasons externalism simply denies that normative reasons require a connection to existing motivation. This disagreement has particular application to morality. Ordinary moral discourse seems committed to the truth of a form of moral 1 Michael Smith, following Donald Davidson, objects that our motivators can only be psychological states [1987: 36-41], but this has the implausible consequence that normative reasons (being facts) cannot motivate. To say that a normative reason motivates us is just to say that we are motivated by acceptance of it as a normative reason. See Dancy [2000: ch. 5.] 1 objectivism:2 our subjection to moral obligations is not contingent upon our motivations. I morally ought to refrain from spitting in my colleague’s coffee, even if this spiteful action would – and refraining from it would not – serve my desires. But do moral obligations entail normative reasons? Reasons internalists, unless they implausibly reject moral objectivism, must deny that moral oughts imply reasons:3 if moral obligations don’t presuppose existing motivation, and normative reasons do, then moral obligations don’t entail normative reasons – we may lack any reason to act as we morally ought. Derek Parfit and other externalists4 justifiably find in this a decisive objection to Williams’ internalism. Moral obligations clearly do entail normative reasons, which (they conclude) are thus neither reducible to nor dependent upon motivation. While the principal objection against Williams’ internalism is that it must reject many of our first-order judgements about reasons as erroneous,5 the main objection to Parfit’s externalism is that it makes reasons and their normative authority troublingly sui generis and inexplicable. In this paper I advance an overlooked position that is between internalism and externalism, Williams and Parfit, and is not subject to either objection. This view agrees with the letter of Parfit’s externalism (on his taxonomy I am like him an ‘externalist moral rationalist’: I believe that motivationally external moral obligations entail normative reasons) but with the spirit of Williams’ internalism. I reformulate Williams’ attack on external reasons, overcoming first order problems by claiming not that external reasons don’t exist, but rather that they don’t matter: Parfit and Williams Parfit calls this thesis ‘moral externalism,’ but for clarity I will avoid multiplying internalisms and externalisms. This thesis is commonly called ‘moral rationalism,’ but this terminology conflicts with the view advanced in this paper. I reject the common assumption that the concepts of reasons and rationality are linked by definition. 4 Parfit has illustrious company, including Thomas Nagel, T.M. Scanlon, Ronald Dworkin, Jonathan Dancy, and Joseph Raz. 5 For other error-theoretic views of moral discourse, see [Mackie 1977; Joyce 2001]. 2 3 2 both overlook an important ambiguity in the concept of normative reasons.6 I give a broad semantic account of reasons-talk that allows reasons to connect with any ends whatsoever, and an account of the pragmatics of reasons-talk that narrows appropriate ascription of reasons to a small subset of these. The internalist critique of external reasons is then reformulated with an internalist, relativistic account of what it is for a reason to matter. I. Williams vs. Parfit The case for Williams’ internalism is based chiefly on a burden of explanation argument, while Parfit’s externalism is based primarily on what I shall call the first-order argument. Williams observes that externalism does not ‘offer any content for external reasons statements’. [1995b: 191; see also 1995a: 39] While the internalist provides necessary and sufficient conditions for something’s being a reason, the externalist typically does not. This by itself is no argument for internalism, but Williams takes it to signify something more: that no plausible externalist analysis of reasons is possible, and hence that it can only be utterly mysterious what it might mean to ascribe an external reason to someone. Indeed Parfit and most other externalists agree that the concept of a reason is unanalyzable, except perhaps into other normative terms. Scanlon analyzes the concept of a reason for action as a consideration that ‘counts in favor’ of an action, and claims that the relation counting in favor of is unanalyzable [1998: 17]. Such normative analyses in internalists’ eyes merely shift the problem: what is sought is a ‘naturalistic’ analysis of normative concepts into nonnormative ones. But externalists do not agree 6 A similar view is advanced in [Millar 2002]. My view differs from Millar’s primarily in my rejection of the notion of ‘allthings-considered’ reasons. See also Smith’s discussion [1987: 38-9] of ‘minimal’ justification. 3 that this unanalyzability is problematic. Many respectable terms seem to be unanalyzable into other kinds of terms: temporal into nontemporal, modal into nonmodal, mental into nonmental, etc. We understand temporal, modal, and mental concepts well enough nonetheless, and we likewise seem to understand normative concepts without being able to reduce them. Williams is not merely seeking a definition, however. Externalists also face an explanatory burden with regard to motivation. It is essential to the concept of a reason for action that cognition of one is capable of motivating agents. Motivational influence, he claims, is something that an internalist account of reasons alone can explain: ‘I cannot see what leverage it would secure: what would these external reasons do to these people, or for our relations to them?’ [1995b: 216]. This motivational argument however has many flaws. If we are speaking merely of logical possibility, then it is compatible with the concept of reasons that an agent might be psychologically incapable of acting on his reasons. But externalists reject even the alleged psychological impossibility of motivation by external reasons. Williams’ claim that external reasons cannot motivate relies on the premise that a desire is always necessary in order to produce motivation – and hence that mere belief in an external reason alone must lack that power. But motivation is a matter of psychological causation, hence the premise has no a priori basis, as well as no adequate empirical evidence. A case can be made that being motivated entails the presence of a desire, but even this is compatible with external reasons having motivational force: the belief in an external reason may itself cause the desire, without the existence of that reason and therefore the agent’s capacity to be motivated by it being contingent on any of his desires [Nagel 1970: 29; Darwall 1983: ch. 3; Smith 1994: 179, Parfit 1997: 115, Dancy 2000: ch. 4]. 4 Williams’ demand for explanation acquires traction when it is addressed rather to a normative problem. We may grant that in principle any psychological state might be able to cause any other psychological state. But this applies equally to various kinds of neurological disorders: we can imagine an agent, Annie, for whom the thought of horses always causes her to desire to jump off a cliff. Reasons aren’t merely things that happen to motivate us: they are also things that ought to do so. The principal virtue of internalist theories of reasons is that they offer an explanation for why we ought to act for such reasons: if Annie’s reasons are considerations that answer to her existing concerns, then (unlike her thoughts about horses) their significance to her practical deliberations is available to her and nonmysterious. In contrast, the normativity of external reasons seems quite inexplicable. Why ought we guide our deliberations and actions by certain facts just because they possess some unexplainable property that does not connect with anything we already care about? It is part and parcel of Parfit’s nonnaturalism7 that there cannot be a noncircular explanation of the normative significance of reasons: it is something we cannot explain, yet adequately comprehend. The normative challenge is sometimes countered by definitional fiat – to ask why we ought to do anything just is to ask for reasons for doing it, so the objection is absurd. But to those like myself for whom the question seems reasonable when addressed to the nonnaturalist’s account of reasons, this suggests instead that the account must be wrong. As externalists generally deny that we can analyze the concept or reduce the property of being a reason, they have difficulty explaining how we can know that a particular fact is a reason for a particular action. As Williams puts it, 7 Parfit’s own term for his position is ‘nonreductive normative realism,’ but he opposes it to ‘naturalism’ [1997: 108] and there are forms of naturalistic nonreductive normative realism. Although they mostly avoid the term (Shafer-Landau [2003] is an exception) Moorean nonnaturalism is favoured by many contemporary philosophers [Scanlon 1998; Nagel 1997; Dancy 2000]. 5 One of the mysterious things about the denial of internalism lies precisely in the fact that it leaves it quite obscure when this form of words is thought to be appropriate. [1995a: 39] Internalism would appear to have an advantage here, as it gives us necessary and sufficient conditions for a fact’s being a reason. However here internalism encounters the main externalist objection. It follows from Williams’ internalism that an agent has no reason to act as he morally ought if he lacks moral concerns (altruistic concerns, I shall naïvely assume) – and that he does have reason for wickedness if he is motivated by a relevant evil desire. The externalist reasonably objects that this flies in the face of common sense and ordinary intuitions. There are reasons for Bush not to prevent gays from enjoying the benefits of marriage, I am disposed to say, even as I assume he lacks the concerns that such reasons might engage. And my knowledge that Jill has a strong urge to ram her car into the back of the vehicle that cut her off does not dispose me to say that there is a reason for her to do so. Indeed Jill herself may well both admit her desire and deny that she has any reason for the violent act. The internalists’ account of which facts are reasons for which actions conflicts with the first-order judgements that people – and even internalists themselves – are disposed to make. It seems we have to choose between an account of reasons that conflicts with our first-order judgements, and an account that denies the possibility of explaining what reasons are or why they matter. Or do we? II. Pseudo-reasons Williams and other internalists do not deny that speakers commonly ascribe motivationally external reasons to themselves or others, but they maintain that this talk 6 is false: it is either confused or ‘bluff’. The same game is played with different normative terms: John Mackie accepts that there are motivationally external value properties, but denies that such goodness could be prescriptive (yield ‘objectively valid’ oughts) to those for whom they are motivationally external [1977: 26-9]. We do indeed utter such ‘objective prescriptions,’ but they must all be false. Philippa Foot, on the other hand, accepts that there can be motivationally external ‘oughts’; her (now disowned) internalism is manifested in the claim that motivationally external ‘oughts’ – other than those of prudence8 – do not present us with reasons. Subsequently Williams acknowledges that we ordinarily ascribe motivationally external reasons, but denies that such reasons can exist.9 Mackie, however, contends that all ‘oughts’ entail reasons [1977: 76; see also Harman 2000: 42; Joyce 2001: 38-40]: institutions such as etiquette, law, and games like chess provide institutional standards, which generate institutional ‘oughts,’ which entail institutional reasons. Morality is an institution incorporating standards, and its ‘oughts’ likewise generate institutional moral reasons [1977: 79]. Mackie thus accepts the existence of motivationally external moral reasons, and Parfit’s taxonomy would suggest that he is an externalist moral rationalist! But Mackie denies that these reasons present us with ‘objective, intrinsic requirements’ – that we ought to be motivated by our (motivationally external) moral reasons, or, if you prefer, he denies that we have any reason to comply with our external moral reasons. ‘Rather inconsistently,’ as she describes it, she also then acknowledged reasons generated by our interests. Foot’s paper was published in 1972, Mackie’s book in 1977, and Williams’ paper in 1979. These look like successive attempts to express in recalcitrant normative terms the same ‘fugitive thought.’ Millar [2002] tries to draw the distinction by contrasting ‘justificatory’ with ‘nonjustificatory’ reasons – but see Smith [1987: 38-9] for the recurrence of the same ambiguity. These efforts are doomed to failure because they attempt to explain in normative terms an ambiguity common to all normative terms. 8 9 7 An obvious way to avoid the absurd classification of Mackie as an externalist moral rationalist is to deny that institutional reasons are normative reasons. This is problematic: institutional reasons have at least two claims to being normative. First, given the usual distinction between normative and motivating reasons, it is clear that institutional reasons are not merely motivating reasons. We can fail to be motivated by – or even aware of – our reasons of etiquette and chess. Second, institutional reasons are prescriptive: they direct us as to what we ‘ought’ to do (or minimally what ought to be the case) – and this might reasonably be taken as the hallmark of normativity. The contemporary concept of normativity combines two separate elements that we need to distinguish carefully: a content and a ‘force’ (‘weight,’ ‘authority’). There is, first, this element of prescriptive content: reasons support the claim that things ought to be a certain way; they favour certain actions, states of affairs, etc. I propose to reserve the term ‘normativity’ for this element, the favouring relation. But philosophers usually treat ‘normative reason’ as synonymous with ‘good reason’ [esp. Parfit 1984: 118]. The second element here is importance: ‘good reasons’ are reasons that matter. There are reasons that favour actions in ways that do not matter: normativity does not entail importance. (Note that my talk of importance here does not address differences in degree – the contrast is not between reasons that matter more and those that matter less, but between those that matter at all, and those that don’t. It is therefore potentially misleading to equate the ‘goodness’ of reasons with their mattering: often by ‘good’ reasons we mean not those that merely matter (minimally), but rather those that are overridingly important, and hence decisive in a situation.)10 Our problem is to determine the criteria of importance – what is it for a reason to matter? 10 The distinction between reasons that matter and those that do not must not be confused with the distinction between 8 It is common to distinguish between ‘pseudo’ reasons that don’t matter, and ‘real’ reasons that do [e.g. Joyce 2001], but difficult to articulate what it is about the real reasons that distinguishes them. We might say that they alone are reasons that we always have reason to heed – but this is clearly an attempt to lift ourselves out of the swamp by our hair. Or we might say that it is only these reasons that we ought to or must heed, whatever our ends – but we’ve observed that this language is infected with the same ambiguity. They are sometimes described as practical reasons – but institutional reasons also prescribe practices. The favoured discrimination is in terms of rationality: ‘real’ reasons are those which, were we aware of them, would motivate us insofar as we are rational. Parfit and Williams both put their opposed views in terms of rationality. Williams’ internalism holds that in order to have a reason that would make some action rational, I must already have some motivation towards an end that would be served by that action – hence the class of real reasons is the class of motivationally internal reasons, while the class of pseudo-reasons is the class of motivationally external reasons. Parfit’s externalism rejects this claim: we have rationally demanding reasons (including moral reasons) that are independent of our existing motivations. I wish to join Williams in denying that there are reasons lacking connections to my existing concerns that matter in a particular significant way, although I shall substitute a different account for the invocation of rationality. But the first order argument demands our attention, and has particular significance for institutional reasons. It is characteristic of institutional reasons that even where they pertain it is often appropriate to deny that there are any reasons at all for some action. I am disposed to prima facie or pro tanto reasons and ‘all-things-considered’ reasons: prima facie and pro tanto reasons still matter, even if they are overridden in particular situations. A reason that does not matter still does not matter even in the absence of other, conflicting reasons. 9 deny that there is any reason at all for me to mind my manners in the company of people I am (rightly) seeking to offend – etiquette notwithstanding. Only when we ‘speak within the institution’ is it appropriate to ascribe institutional reasons. We can thus distinguish between conversationally licensed and conversationally unlicensed reasons ascriptions. The distinctive mark of a pseudo reason might therefore reasonably be supposed to be that its ascription has only narrow conversational license (i.e. it is conversationally licensed in only a limited range of contexts). First order judgements then pose a significant problem for internalism. I am disposed ordinarily to deny not merely that there are good reasons for a bigot to support a constitutional ban on gay marriage, but further that there is any reason at all for him to do so. Reasons issued by bad desires, like those issued by institutional standards, would therefore seem to be mere pseudo reasons, while moral reasons – which have much broader conversational license – would rather be reasons that matter. This looks like evidence supporting the externalist’s claim that connection to existing motivation is neither necessary nor sufficient for a consideration’s counting as a real reason. III. A Unified Theory of Reasons All the kinds of reasons-talk can be unified by a single semantic theory. By a ‘reason’ we mean an explanatory entity: a fact that provides an appropriate answer to a ‘why?’ question. We can see this by comparing normative with nonnormative reasons. We talk not only about the reasons for which we act and ought to act, but also about the (causal) reasons why the dinosaurs became extinct, or why New York’s World Trade Towers collapsed. These reasons are facts that are appropriately cited in answer to 10 questions of the form, ‘Why is it the case that p?’11 while normative reasons are cited by appropriate answers to questions (primarily) of the form, ‘Why would it be good that p?’12 and (secondarily) ‘Why ought it be the case that p?’ ‘Good’ must take precedence over ‘ought’ here in order to make sense of overridden normative reasons. If it is false that it ought to be the case that p, then there can be no true answers as to why it ought to be the case that p, and hence no reasons why it ought to be the case that p. But there can be normative reasons favouring actions that we ought not to perform. In some instances we can resolve this problem by observing that there are different senses of ‘ought’ which can conflict, so overridden reasons may correspond to overridden ‘oughts’. But this doesn’t help with conflicting reasons arising with respect to the same normative criteria.13 Suppose (S1) that I can get to Scotland by the High Road, but (S2) I can also get there by the Low Road, which is better in all relevant respects (speed, reliability, safety, etc.) I ought therefore to take the Low Road and not the High Road, yet S1 remains a reason to take the High Road. This is because S1 explains why, with respect to getting to Scotland, the High Road is good (albeit to a comparatively low degree).14 Normative reasons answer secondarily to questions concerning why something ought to be the case, if I am right about the following relation between ‘good’ and ‘ought’: it ought to be the case that p iff p is better (more good) than all relevant alternatives.15 Explanations of the value of alternatives are relevant to enquiry into what ought to be the case, even if not individually decisive as Note that ‘a reason why it is the case that p’ doesn’t mean the same as ‘a reason for x to believe that p’ (they are answers to different questions – ‘why is it the case that p?’ and ‘why ought x believe that p?’ – although plausibly the former kind of reason always entails the latter.) 12 Contrast Scanlon’s widely discussed ‘buck-passing’ account of value, which explains goodness in terms of reasons [1998]. 13 I thank the anonymous referee who helped me appreciate this problem and revise my view accordingly. 14 By ‘good’ here I merely mean of some positive value, which is compatible with it being a poor option. 15 This does not beg the question against deontology: if an action morally ought to be performed then it is morally better than all relevant alternatives. Note that some normative reasons for p may primarily address ‘ought’ questions: those that directly explain that p is better than any alternative. 11 11 explanations of why p ought to be the case. For convenience I shall therefore connect normative reasons indiscriminately with ‘good’ and ‘ought’. It is worth noting here the confusion in the literature on motivating reasons stemming from the distinction between normative and nonnormative reasons. Talk of ‘motivating reasons’ is ambiguous: it is sometimes treated as designating (a) answers to questions of the form ‘Why did A φ?’ – i.e. as concerned with the causal explanation of A’s φ-ing. This leads to the common identification of beliefs and desires as motivating reasons. However our psychological states (motivating beliefs and desires) do not motivate us qua reasons – they generally are not (and do not figure in) considerations that lead us to act. What is normally meant by ‘motivating reasons’ are (b) normative reasons (answers to questions of the form ‘Why ought I to φ?’) that motivate us. While motivating reasons are ‘reasons’ in relation to (broadly) normative questions, because they motivate us they can also be cited in providing a (causally explanatory) reason why some action occurred. They are therefore easily confused with causally explanatory reasons. This has allowed philosophers to overlook or conceal the category of normative reasons that do not matter, by classifying them as merely motivating reasons and hence (erroneously) not normative reasons.16 The meaning of ‘a normative reason for X to φ’ is a fact that would answer a query as to why X ought to φ/it would be good that X φ. This much is nonprejudicial to Parfit’s externalism, as it makes no mention of motivating states, and to his nonnaturalism, as it retains an unanalyzed normative element. But it is then a substantive issue what makes 16 For example [Parfit 1984: 118; Raz 1978: 2-4; Schueler 1995: 45-7]. This error is encouraged by our saying that we have ‘motivating reasons’ even where our motivating belief is false, i.e. where there is an absence of normative reasons. I have to say that we can be motivated by normative reasons that don’t exist. This is not as absurd as it sounds, because it is merely a way to describe being caused to act by belief in that reason. (Consider being frightened by a bogeyman.) Being motivated by something, like believing in something, does not entail the existence of that thing. 12 a fact a normative reason or appropriate answer to a normative question, and thereby what ‘good’ and ‘ought’ mean. My unifying thesis is that all normative reasons – those that matter and those that don’t – are end-relational: a fact is a reason for φ-ing, relative to a system of ends E, iff it explains why17 φ-ing is conducive to E.18 (By an ‘end,’ I mean a possible aim for action or object of desire.) Goodness relative to an end E, in other words, is conduciveness19 to E, and I ought to φ, relative to E, iff φ-ing is more conducive to E than any relevant alternative.20 In defense of this claim, which offers a naturalistic reduction of the relation ‘counting in favour of’ to a relation specifiable in only nonnormative terms, observe that providing a reason for φ-ing is often explained as demonstrating the ‘point’ of φ-ing. To furnish an action with a point just is to demonstrate that it serves some purpose or end. Hence for a fact to count as a normative reason for φ-ing, it must constitute an explanation relative to some end.21 The end of prudence is personal wellbeing, so a fact is a prudential reason for me to φ iff it explains why φ-ing is conducive to my wellbeing. Epistemic ends include having beliefs that are true, so a fact is an epistemic reason for believing that p if it explains why p is likely to be true. The ends of playing chess are checkmating and avoiding being checkmated (within the rules), so a fact is a chess reason for move M iff it explains why M advances these 17 An earlier version of the paper formulated this rather in terms of indicating that φ-ing is conducive to E. A referee correctly observed that this formulation blurs the distinction between reasons for φ-ing and reasons for believing that one ought to φ. 18 End-relational analyses of normative language are not uncommon, and face the principal problem of explaining how to determine the relevant ends: linking the use of normative terms to any ends (see Ralph Barton Perry [1926]) yields unacceptable first-order results. Two traditions of ethical thought offer different semantic strategies: the Aristotelian determines the ends by appeal to an object’s natural function or telos [e.g. Bloomfield 2001], while the Humean determines the ends by appeal to an agent’s motivating desires. Both of these strategies limit normative language excessively: it is semantically unrestricted as to ends, but is always pragmatically or contextually restricted [Finlay 2004]. 19 Defining ‘conducive’ is difficult. I suggest that A is conducive to E, approximately, iff (and to the degree that) E is more likely in the presence of A than it would be in the absence of A. 20 On this view judgements of deontological or side constraints on how ends are pursued express the end-relational demands of other ends. Note that on my proposal such judgements need not assume that those further ends are themselves pursued by or matter to the agent being judged. 21 This may seem to beg the question against nonnaturalistic externalism, which denies that the normative force of reasons can be explained. But my opponent (e.g. Parfit) will agree that a reason for A to φ is a consideration that would meet A’s demand for normative explanation. He denies rather that it can be naturalistically explained why that reason qualifies as a normative explanation. 13 ends. There are as many kinds of normative reasons, therefore, as there are different systems of ends: there are reasons of faith, revenge, love, lust, and football, there are legal, medical, military, artistic, political, and scientific reasons. I am rejecting the popular doctrine that there is just one kind of real reasons for action: ‘practical reasons,’ which trade in the same currency, and form one unified domain [see also Copp 1997]. On my view there is no basis for maintaining the distinction between institutional or pseudo reasons and real reasons. Whether we are considering moral reasons or epistemic reasons or reasons of etiquette, they are all ‘reasons’ in the same broad sense, generated by various ends. Philosophers commonly assume that prudential, moral, and epistemic reasons matter, but reasons of etiquette, faith, and chess, and those generated by the ethos of the mafioso do not. (This is not to say that etiquette, faith, etc. are never important – but many will say that they matter only insofar as we have prudential, moral, or instrumental reasons to attend to them). But other people in other times or cultural contexts have taken a different view: reasons of etiquette, religion, and the ethos of organized crime can all be taken, and have all been taken, as mattering intrinsically, without regard for morality or prudence. If the institution of organized crime dictates that a godfather’s command to kill provides a reason to kill, then the mafioso under such an order is disposed to judge simply that he has a reason to kill. It is too glib to say that these are just motivating reasons: the mafioso takes his reasons to exist independently of his knowledge of them, to be normative (he is obligated) and to matter, and he is motivated because of this. Of course the mafioso might be mistaken: moral reasons genuinely matter and mafioso reasons do not. But there is no sharp difference in ontological kind between all these reasons: they have the same 14 independence of our cognition or acceptance and similar ability to motivate. The only question concerns their difference in importance. The end-relational account of normative reasons stands or falls on its ability to capture the empirical data, and its hardest test arises from the burden of accommodating first order judgements about reasons. Whereas Williams is accused of unduly restricting the normative reasons that bear on a person’s actions, the end-relational account seems to multiply them absurdly: there are indefinitely many different sets of possible ends, and therefore indefinitely many kinds of normative reasons supporting and opposing any action we might imagine. Ascription of the vast majority of the kinds of reasons here postulated are (in most circumstances) conversationally unlicensed. What about reasons simpliciter? The crucial point here is that all reasons-talk, unless explicitly qualified (as ‘epistemic,’ ‘moral,’ etc.) is context-dependent. To be adequate or conversationally appropriate, it is not sufficient that an answer to a ‘why’ question is true. Real-world talk of reasons is always talk about reasons of particular kinds. Beginning again with nonnormative reasons, observe that requests for explanation why it is the case that p seek information within certain parameters, and so always presuppose a restricted range of acceptable answers. There are many different kinds of explanations for it being that case that p, and for any genuine enquiry it is determinate which kind of answer will count as an explanation.22 Suppose I drop and smash a dinner plate on the floor, prompting my daughter to ask, ‘Why did you drop it?’ It is presumably facetious of me to reply, ‘Because it fell from my hand’ (formal cause), or ‘Because I chose to’ (efficient Whether an explanation meets an enquirer’s criteria may be something that in certain cases the enquirer herself is incapable of recognizing. For example, a complex mathematical proof might satisfy the explanatory criteria implicit in my mathematical enquiry despite my inability to work through the proof. Williams’ vulnerability on this point is exposed in Millgram [1996]. 22 15 cause). She probably seeks the purpose for which I dropped it (final cause). Even here a correct answer is not necessarily an appropriate answer (e.g. ‘I wanted it to smash on the floor’). To the question, ‘Why did the World Trade Towers fall?’ the indefinitely many true answers include, ‘Because of an act of terrorism’, ‘Because they were struck by passenger jets’, ‘Because intense heat weakened their structural columns,’ and ‘Because of the earth’s gravitational pull’. Each of these would be an adequate answer to some enquiry or other, but there is no conceivable enquiry to which each of these would provide an adequate answer. What counts as a ‘reason’ in such contexts is determined by the enquirer’s epistemic ends: the kind of information she feels she lacks, prompting her question. I claim that talk of normative reasons is likewise context-dependent, understanding conversational context as roughly the totality of information mutually recognized by the participants in a conversation as mutually understood. Every normative conversation or discourse is framed by a set of (at least) implicit ends: purposes or desires assumed for the sake of conversation, which may variously be the purposes or desires of all, some, or none of the participants. Within a context, reasonsascriptions are normally made relative to those implicit ends, and hence are contextually relativized (talk of ought simpliciter, on this view, is always elliptical for ought in order that x). There are therefore always some actions for which it is appropriate to deny that there are any reasons at all, even where there are many different kinds of reasons in favour of them. This context-relativity explains why we are often disposed to deny the existence of reasons where certain kinds of reasons are clearly to be found. If I deny that the fact that an innocent man is facing impending execution is any reason for a murderer, who has gotten away scot free, to confess and surrender to the law, I likely do 16 not mean to deny that morality requires it, but merely to deny that it is in his selfinterest. Conversational license is not the distinguishing mark of reasons that matter, but merely signifies the relation of a domain of reasons to conversational context. While this context-dependence accommodates much of the data about our first order reasons ascriptions, there remain difficult first order data for which the endrelational account appears an uncomfortable fit. I am confident that these problems can be overcome by scrupulous attention to the complex and subtle operation of context, but logistics limit me to merely sketching the resources available for dealing with the most serious problems: (1) the evidence that there is ‘all-things-considered’ ascription of reasons; (2) the broad conversational license enjoyed by some kinds of reasons (e.g. moral and prudential reasons) that seems to enable them to transcend context; (3) the appropriateness of rejecting the existence of some end-relational reasons that are sanctioned by context. So let us consider these problems in turn, bearing in mind that the goal is defense: I am not seeking here to show that the end-relational theory is uniquely the best fit for the data, but simply that it is compatible with that data. (1) I have claimed that all reasons-talk is relativized to some ends or other. But we talk about the reasons we have all-things-considered, and it can further be argued that we need to have reasons that transcend all particular ends, because we can (and sometimes should) step back from our particular prudential, instrumental, and possibly even moral deliberations, and weigh them to reach an all-things-considered decision: what we ought to do from no particular point of view at all. I concede that insofar as we ordinarily talk about all-things-considered reasons, we are broadening our context: introducing a wider set of reasons. It doesn’t follow that we must be broadening it to include all the ends that there are, or transcending all ends altogether. ‘All-things- 17 considered’ here always quantifies over particular finite domains of reasons. It may mean, ‘All my ends considered,’ or ‘All our ends considered,’ or even merely ‘All moral ends considered’. That there can be normative deliberation and decision altogether independent of the point of view of particular ends I simply deny, and I don’t believe I am overlooking any significant evidence.23 (2) Some kinds of reasons such as prudential and moral reasons seem to be appropriately interjected into conversations in which the relevant ends are not contextually presupposed. Is this not evidence that these reasons transcend particular ends and contexts? Suppose Ann is playing chess against Ben, and Clara is giving her advice. The conversation between Ann and Clara is framed by the end of Ann’s winning the game. But Clara advises a move that is clearly suicidal, prompting Ann to demand a reason, to which Clara responds by pointing out that if Ann moves otherwise and wins, then Ben, a sore loser, could become violent. This answer is not licensed by the context, yet Ann may well accept it as an appropriate answer and as providing a reason. Sometimes what is said can drive a change in context, when it fails to be intelligible within the existing context. This unintelligibility prompts an audience to identify what has to be added to the context in order for what is said to make its intended contribution to the conversation: one way to put something into the context is to speak as if it were already there. This is most starkly evident when context changes within a sentence. When someone says ‘It’s a reason for φ-ing and a reason against it’, or ‘You ought…but you also ought not’, a change in context is forced by the contradiction in what otherwise would have been said. There is a use of reasons-talk that may seem unrelativized. Suppose I tell you that there are reasons for you to φ, and you cannot work out what I am suggesting the point of so φ-ing would be. You may then ask, ‘What reasons are there for me to φ?’ Here it seems you might be asking for any kind of reasons whatsoever. But what you are doing is borrowing my (unidentified) qualifying end, and asking me to explain it. 23 18 This response may seem inadequate due to the asymmetry between broadly and narrowly licensed kinds of reasons. We accept such introduction of certain kinds of reasons, but not of others. Consider again the game of chess, but now let us suppose that the (whispered) conversation between Ann and Clara is addressed to how they might avoid angering Ben. Clara advises a certain move, and Ann demands a reason for it. Suppose that Clara’s response is ‘Because otherwise he can take your queen’. Ann then quite reasonably responds, ‘That’s no reason at all!’ It seems legitimate brusquely to substitute prudential reasons for chess reasons, but not vice versa. This too can be accommodated. We care more about some ends – and hence some kinds of reasons – than about others: our own wellbeing is generally of more concern to us than winning a game of chess. We are therefore disposed to abandon ends when they conflict with preferred ends. For this reason we accept as appropriate abrupt changes in the context towards more preferred ends and reasons, but do not accept such changes in the other direction. This phenomenon is not evidence of a sharp difference in kind between these kinds of reasons: we can easily imagine a Chess Fiend for whom chess reasons take precedence over moral and prudential reasons (witness the toll on relationships exacted by online gaming addictions). (3) This does not yet suffice, however, to explain the appropriateness of rejecting ascription of certain reasons that according to this theory are licensed by the context (in particular, reasons generated by trivial and evil ends), and insistently ascribing reasons where the requisite change in context is resisted by others (moral reasons are paradigm here). For example: Joe comes upon Jeff and Jill plotting to get revenge on Bob. Jill suggests Joe ought to spit in his coffee, on the grounds that this will be payback. But Joe takes the moral high ground: he insists that this is no reason at all, and furthermore cites 19 the fact that Jeff and Jill would not like that to be done to them as a reason why he ought not – even though he knows that Jeff and Jill are discussing how to proceed in order to take their revenge, and that they much prefer their end of revenge to any moral ends. To explain this phenomenon we can observe that social influences on the dynamics of context and appropriate speech introduce a rhetorical element into reasonstalk. In many situations accepting a set of ends as implicit for one’s speech acts suggests that one shares (or at least has no significant objection to) those ends [Finlay 2004]. For Joe to agree that the fact that spitting in Bob’s coffee would be payback is a reason for his doing so would be to say something true but misleading: it would falsely suggest to Jeff and Jill that he does not object to their pursuit of revenge. By rather denying it is a reason he says something false in the context, but thereby communicates something else that is important and true: that he does not condone the pursuit of revenge. Conversely, there is therefore rhetorical point in behaving as if certain ends are presupposed in a context even when they aren’t: this expresses the demand that one’s audience share or respect those ends. (This depends, of course, on the audience being able to infer what those ends are likely to be – the evidence for which may be the social status of those ends, as is plausibly the case with morality, or alternatively the known values of the speaker, as in Williams’ example of Owen Wingrave, whose father insists he has reasons for military service even though Owen cares not a whit for family tradition, on which those reasons are based). It has been objected that this response is insufficient, and the end-relational theory implausible, because it sanctions far too many true reasons claims. Never mind what we can appropriately say to those ascribing reasons to persons based on evil or 20 trivial ends, we should consider what we appropriately say about their reasonsascriptions from outside their conversation. A reviewer comments, Suppose two sadists are discussing whether Hitler had reason to order the Final Solution, and they conclude that he did (it caused so much pain, after all). How shall we evaluate their claims? According to the account, their claims are true, since Hitler’s action did serve conversationally implicit ends. But endorsing as true the claim ‘The fact that it caused so much pain was a reason for Hitler to order the Holocaust’ seems almost a reductio of a theory of reasons. The end-relational theory entails that many such reasons claims, understood within their context, are indeed true. Surely that is something to which we cannot assent? In reply, I wish first to observe that saying that someone’s statement is true usually expresses in part that one endorses or confirms that statement. For me to say, ‘A’s statement that there was a reason for Hitler to order the Final Solution is true’ is therefore like saying myself, ‘There was a reason for Hitler to order the Final Solution’. In my conversational context, this will typically communicate that I endorse sadists’ ends, for the reason noted above that ascription of implicitly qualified reasons typically communicates speaker endorsement of the relevant ends. It is therefore a statement I generally cannot make nonmisleadingly. Of course, if a speaker says something true, it should be possible to report that he does. My point is that when we report others’ statements we cannot always simply repeat their words, because their conversational context is not similarly replicated. Here’s a trivial example: (1) I did not have sexual relations with that woman. 21 (2) Clinton’s statement that I did not have sexual relations with that woman is true. Clinton’s utterance of (1) occurs in the context of Clinton’s being the speaker, hence his use of ‘I’ refers to Clinton. Because I do not (and cannot) utter (2) in this same context, I cannot appropriately utter (2), but rather (falsely) (3) Clinton’s statement that he did not have sexual relations with that woman is true. The lesson is that in order to appropriately and nonmisleadingly report another’s statement as true, I must reword that statement so that it expresses the same proposition in my context that his expresses in his context. As sadistic ends are not implicit in my context, to report the sadist’s utterance as true, I must make its index explicit: (4) A’s statement that [there is a reason why in order that there be enormous suffering in the world Hitler ought to have ordered the Final Solution] is true. This contains a fully explicit articulation of the sadist’s utterance, as understood by the end-relational theory. According to the theory (4) is indeed true, but I do not find it counter-intuitive. The objection is generated by the fact that contextually relativized ascription of reasons and ‘oughts’ typically expresses endorsement of the implicit ends. This is what distinguishes the sadist’s utterance from a fully explicit utterance with the ellipsis removed and makes it the case that I cannot appropriately report as true the sadist’s exact words. A remaining objection: I have only addressed what we can say. But what about what we can think? The judgements we make in the privacy of our own minds, one might think, are surely removed from the operations of conversational context and rhetorical force. And surely we rightly think simply that sadism provides no reasons at 22 all for Hitler’s having ordered the Final Solution? Here’s my response. Thinking is either linguistic or nonlinguistic. If this thinking is linguistic (‘saying it to yourself’), then arguably it is conversational, and operates within conversational context. I do not say every component of the complete proposition when I say something to myself any more than when I speak to others. (I do not prefix my thought, ‘I ought to take a bath’, with ‘In order not to offend others’, even when that is clearly my meaning). If, on the other hand, this thinking is nonlinguistic, then the objection cannot be immediately obvious, and I maintain that the end-relational theory best expresses the content. The end-relational theory of reasons will accommodate the first order data, if I am right, by appeal to the many resources supplied by the dynamics of context. Really making good on this promissory note would require a long and scrupulous investigation that I cannot undertake here: it is enough for present purposes that I have shown the theory to be plausible and defensible. I now direct attention back to the dispute between Williams and Parfit, internalism and externalism. The end-relational theory says that Williams is right that ascription of reasons is made relative to certain ends but errs in his internalist claim that appropriate ascription of reasons to agents is determined by just those ends towards which those agents are motivated. The reasons that are appropriately ascribed to me are determined not by what my ends happen to be, but by the speaker’s conversational context. The internalist error is easily understood because often the ends presupposed in a context are indeed the agent’s ends: a common form of normative discourse is adoptively instrumental, 24 in which a speaker conversationally adopts the practical perspective of an agent’s 24 In a way I am claiming that all reasons are instrumental. But usually when we describe a reason as ‘instrumental’ we mean that it has an instrumental relation to the agent’s ends. 23 motivating ends, and ascribes reasons to the agent relative to those. But this is only one kind of practical context, and the semantics of reasons provided here permits many others. When we ascribe moral reasons we enter a context (or pretend to) where certain moral ends are presupposed, and agents’ actual motivations play no role in determining these ends. They can be reasons nonetheless: answers to questions of the form ‘Why ought he φ?’ that meet the implicit criteria of the enquiry. So Parfit is right: there are external moral reasons. There is a consideration in favour of Williams’ internalism, however, that I must address. Williams himself introduces his argument by observing the difference between the locutions ‘There is a reason for A to φ’ and ‘A has a reason to φ’, and suggests the former invites a motivationally external reading, the latter an internal reading. While he challenges the sense of external reasons ascriptions, I have offered an account that explains and defends them. But talk of A ‘having’ a reason, and perhaps also talk merely of a consideration being a reason ‘for’ A may seem to imply an answer to a normative question potentially asked by A, and hence may presuppose that only reasons motivationally internal to A would ever be conversationally licensed. We can supplement these semantic points with a practical one: it is unclear what would be the point of ascribing reasons to a person for whom awareness of them could not influence him. I admit that both locutions are suggestive of a motivationally internal connection, but the suggestion is not strong enough, and ordinary use does not conform closely enough (as Williams concedes) to base a significant philosophical distinction on it, even though Williams may be right that there is an element of rhetorical ‘bluff’ in their motivationally external application. First, our normative enquiries as to how we 24 ourselves ought to act are not necessarily motivationally internal. An amoralist can ask why he morally oughtn’t φ. This could be from simple intellectual curiosity, but we must also bear in mind that it is not always clear what matters for a person or what they care about. In other words, we may not always encounter reasons as obviously internal or obviously external. Owen Wingrave may ponder the reason for military service provided by his father, unsure whether family tradition really matters for him. Second, I suggest that ‘a reason for A to φ’ is elliptical for ‘a reason for it being the case that A φs’, which eliminates the privilege of A’s point of view. It is commonly held that all normative reasons are indexed to particular persons and particular actions, but this claim is theory-driven. It makes perfect sense to say that there ought to be a world without evil, without this being reducible to the claim that anybody in particular or in general ought to make it so. An answer to the normative enquiry, ‘Why ought there be a world without evil?’ is a normative reason that is not intrinsically a reason for any particular person’s acting. A reason for Bush not to prevent gays from enjoying the benefits of marriage, therefore, is simply a true and conversationally appropriate answer to the question, ‘Why ought not it be the case that Bush prevents gays from enjoying the benefits of marriage?’ This question can be important for people besides Bush, and for reasons other than influencing Bush’s actions. IV. How Reasons Matter Although I have nominally conceded reasons externalism and that moral ought implies reasons, externalists like Parfit will not be happy. The ‘normative reasons’ countenanced here are humble company for moral reasons. Reasons of etiquette and chess are not reasons that matter in the right way – one can ignore them with impunity, 25 ‘rationally,’ we might say. We need to distinguish between the unimportant normative reasons and the important normative (or ‘good’) reasons that matter – and we need to explain the difference. At this point rationality is usually invoked – but what is meant by ‘rationality’? I am deeply suspicious that rationality is an occult quality into which it is fashionable and convenient to stuff all the difficult puzzles about normativity. But let us see what is signified by this move. As Parfit notes, internalists and externalists favour different accounts of rationality. Internalists favour procedural accounts according to which to be rational is just to deliberate in certain ways: to reason or draw inferences in accordance with logical or other formal constraints. Externalists however advance substantive accounts on which rationality incorporates particular ends or motivations. Whether or not rationality is thought to require only motivation grounded in preceding motivation thus depends on what is built into the account of rationality. If the concept of rationality is to serve the purpose of explaining the difference between the reasons that matter and those that don’t, then both procedural and substantive accounts have to be taken as synthetic accounts of rationality, rather than as definitions or explanations of what ‘rationality’ means. If to say that an instance of deliberation is rational is merely to say that it follows the canons of logic, or additionally that it is guided by certain ends, then it can be legitimately questioned why rationality matters, and why it is precisely the reasons that motivate rational people that matter – particularly if we are not ourselves rational people. Talk of ‘rationality’ clearly suggests something more: it is intrinsic to the concept of rationality that it is normatively required (in a way that matters) of everybody capable of it. Rational requirements are such that, once we understand them as such, we cannot coherently or intelligibly challenge their 26 authority over us. This seems to be all that is uncontroversial about the concept: but in that case all that it uncontroversially means to say that real reasons are rationally demanding on us is that they are reasons that matter. The real difficulty – concerning what it is to matter – has merely been transposed into the question of what rationality is. Invoking rationality here is thus opaque and unhelpful, and I shall now sketch another way of distinguishing the reasons that matter, with affinities to Harry Frankfurt’s classic work on importance [1988]. My aim here is to reinterpret and reinvigorate Williams’ attack on external reasons by claiming not that they do not exist, but rather that they are not important/do not matter. I shall not however argue here for this view of importance, nor shall I provide a careful analysis of the concept. This sketch is broadly compatible with (or at least reformulable as) variously a relational analysis (there are facts about importance, but it is always importance-for-x), an expressivist analysis (judgements of importance are not fact-stating, but rather function to express the concerns of the speaker), and (less comfortably) a projectivist/error theory (ordinary judgements of importance are all false, because they posit a property of importance simpliciter that does not exist). The suggestion here is that reasons are important only if they are motivationally internal. I am claiming only that this is a necessary condition, not that it is sufficient for importance – we must accommodate the fact that we can be motivated by whims, impulses, compulsions etc. that do not confer importance on their objects. Rather, importance arises from our cares and concerns. This is to relativize importance: the claim is not that facts about motivational connections are themselves important simpliciter (a claim no less mysterious than those made by externalists like Parfit, Scanlon, Dancy, and Raz) but rather that there is no such thing as importance simpliciter, there is only 27 importance for particular subjects. There is nothing more to something’s being (intrinsically) important for A than its ‘seeming’ (intrinsically) important to A, and nothing more to this ‘seeming’ than A’s caring intrinsically about it. Our intrinsic concerns are the source of importance for us not by constituting intrinsically important facts, but by making it the case that their objects matter intrinsically to us. An adaption of the Euthyphro challenge may be pressed against me here: are things important for us because we care about them, or do we care about them because we judge them to be important? The account offered here seems to claim the former, contrary to common sense and the preponderance of philosophical opinion. My answer, however, is ‘Both’. Things can be instrumentally and/or derivatively important for us even if we don’t care about them – and we can care about them because of this importance – if only they stand in a normative (i.e. end-relational) relationship with things that we do care about intrinsically. But things can only be intrinsically important for us insofar as they are objects of our intrinsic concerns. (I am intrinsically concerned for my children. This makes them important for me. I intrinsically care that I be respected and liked by my fellow humans. This makes the respect and liking of my fellows important for me.) The implication of this view is that conceivably any kind of reasons may matter, and any kind may not matter, for some person. Moral reasons matter (intrinsically) only for those with moral concerns (which for all I have said here could be almost everyone). A morally unconcerned person can still have moral reasons, but they cannot be reasons that matter for him. There is however a significant ambiguity here: there are multiple dimensions of importance. If you ascribe a reason to me, it can be motivationally internal in at least two ways: it can be motivationally internal to me, the agent judged, or it can be 28 motivationally internal to you, the judge. Even if the reason is motivationally external to and hence unimportant for me, it may still be a reason that matters – for you – if you care about the relevant end. This is not to say that it is a reason for you to act, rather that it is a reason for which it matters for you that I act on it. To illustrate: (1) Suppose you ascribe feng shui reasons to me in my home renovations, without either of us caring about the ends of feng shui: these reasons have no importance for either judge or agent (they matter for neither of us); (2) You ascribe to me instrumental reasons to conceal from you that I poisoned your dog: these reasons are important for me, but not for you; (3) You ascribe to me prudential reasons not to kill myself from despair over losing the love of my life: if you are my friend, and I have ceased to care about my wellbeing, then these reasons are unimportant for me, but important for you; (4) You ascribe to me epistemic reasons to accept moral rationalism: if we are both philosophers seeking the truth and you are correct, then these reasons are important for both of us. If we are enquiring as to which reasons are appropriately judged to be ‘important’ without explicit relativization, we must consider whether it is the concerns of the agent judged or of the speaker judging that are relevant. What is jarring about Williams’ view is the suggestion that the monster who is untouched by moral concern isn’t ignoring any reasons that matter – i.e. Williams ascribes importance as if from the perspective of the monster’s concerns. I concur with Parfit that Williams errs here: the monster violates moral reasons, and because we are motivated by moral concerns, these reasons (and this violation) matter for us, and therefore it is appropriate for us to judge that he violates ‘important’ reasons, and generally inappropriate to deny it.25 An expressivist account of such judgements is a natural choice here, although I would champion instead the pragmatic account I suggest for normative judgements on [p. 20] and in [2004]. 25 29 I contend that this is sufficient to accommodate our condemnation of the monster, but I am well aware that many – including Parfit, no doubt – will disagree. ‘The normative force of moral judgement,’ one may object, ‘requires that an agent’s immorality violates reasons that matter for that very agent.’ But this strikes me as merely an expression of the futile hope that there be some philosophical means of persuading every individual to virtue, or minimally, that every villain has failed to recognize important truths.26 Why must the force of judgement depend upon its significance for the agent judged? It is case (3) above (the reasons matter only for the judge), rather than case (2) (the reasons matter only for the agent) that more closely approximates the weight and authority of moral judgement. (The additionally harsh tone of moral condemnation is due to the different attitude towards the person judged: when you criticize a friend’s self-destructiveness, you do so motivated out of concern for your friend, so your criticism is benevolent. When you criticize the moral monster, you typically do so motivated out of concern for others who stand to be harmed: your criticism is likely to be hostile and accusatory.) Why accept this importance internalism? I submit that it tracks our first order judgements of the reasons that matter, and for whom they matter, and I also submit that it alone provides a satisfactory explanation of importance or normative authority.27 26 Why should this matter? A philosopher’s prejudice: it rankles morally upstanding philosophers to think that the moral enemy might not be ignorant of some relevant truth, and comforting to think that he is, at least, mistaken. (To a philosopher, what could be worse?) In former times philosophers found it necessary to appeal to a villain’s self-interest in order to justify morality, and the claim that the villain is merely mistaken would have seemed absurdly insipid: I am inclined to attribute this change to the increasingly insular nature of academic philosophy. 27 There are two rival internalist accounts of reasons. Williams eventually opts for the hypothetical motivation model: R is a reason for A to φ iff A would be motivated to φ if rational and informed. Parfit rightly objects [1997: 126] that (even hypothetical) psychological facts cannot substitute for normative facts. But the better model, which Parfit ‘sets aside’ [1997: 100], is the end-relational model: R is a reason for A to φ iff R explains why φ-ing is conducive to some contextually determined end. This is a nonpsychological, naturalistic and normative fact. (Williams’ choice is due to worries about desires based on false beliefs. But the end-relational account can avoid these problems by specifying only intrinsic desires: my ‘desire’ to drink the stuff in the bottle is derivative on my intrinsically desiring to drink the gin I believe it contains, which doesn’t give me any reason to drink the petrol it actually contains.) 30 V. Conclusion I have argued that contemporary debate over normative reasons fails to observe a significant distinction between normativity and importance (or normative content and normative force), and have reformulated Williams’ skepticism about external reasons by identifying important reasons with motivationally internal reasons, unimportant reasons with motivationally external reasons – although I have also observed that being motivationally internal and hence mattering is a relative and ambiguous affair. This account avoids Williams’ problems with first order judgements because it does not deny the existence of motivationally external normative reasons. Its advantage over Parfit’s nonnaturalistic externalism is that it allows us to explain nonmysteriously why certain facts are reasons for certain acts, how and why normative beliefs motivate, and also (without recourse to mysterious invocations of ‘irrationality’) why sometimes they don’t. Understanding the landscape of reasons in this way, the orthodox and simple classifications of normative and motivating reasons appears hopelessly naïve. Any ascription of reasons has to be situated on multiple dimensions: whether the alleged reasons answer to normative or nonnormative questions, which question and end they answer to, whether they are really facts, whether they are facts that are really reasons, whether they matter (i.e. are they motivationally internal?), and for whom they matter.28 28 I am grateful for comments from Chrisoula Andreou, Marshall Cohen, John Dreher, Pamela Hieronymi, Aaron James, Sharon Lloyd, Steve Munzer, Chris Naticchia, Calvin Normore, Andy Reath, and several anonymous referees, including those of the Australasian Journal of Philosophy. 31 REFERENCES Bloomfield, Paul 2001. Moral Reality, New York: Oxford University Press. Copp, David 1997. The Ring of Gyges: Overridingness and the Unity of Reason, Social Philosophy and Policy 14: 86-106. Dancy, Jonathan 2000. Practical Reality, Oxford: Oxford University Press. Darwall, Stephen 1983. Impartial Reason, Ithaca, NY: Cornell University Press. Finlay, Stephen 2004. The Conversational Practicality of Value Judgement, Journal of Ethics 8 (3): 205-223. Foot, Philippa 1972. Morality as a System of Hypothetical Imperatives, Philosophical Review 81: 305-316. Frankfurt, Harry 1988. The Importance of What We Care About reprinted in The Importance of What We Care About: Philosophical Essays, Cambridge: Cambridge University Press. Harman, Gilbert 2000. Relativistic Ethics: Morality as Politics reprinted in Explaining Value and Other Essays in Moral Philosophy, Oxford: Oxford University Press. Joyce, Richard 2001. The Myth of Morality, Cambridge: Cambridge University Press. Mackie, J. L. 1977. Ethics: Inventing Right and Wrong, Harmondsworth & New York: Penguin Books. Millar, Allan 2002. Reasons for Action and Instrumental Normativity in Reason and Nature: Essays in the Theory of Rationality, ed. Jose Luis Bermudez & Allan Millar, Oxford: Clarendon Press. Millgram, Elijah 1996. Williams’ Argument Against External Reasons, Nous 30: 197-220. Nagel, Thomas 1970. The Possibility of Altruism, Oxford: Clarendon Press. Nagel, Thomas 1997. The Last Word, New York: Oxford University Press. Parfit, Derek 1984. Reasons and Persons, New York: Oxford University Press. Parfit, Derek 1997. Reasons and Motivation, Proceedings of the Aristotelian Society (suppl.) 71: 98-146. Perry, Ralph Barton 1926. General theory of value, its meaning and basic principles construed in terms of interest, New York: Longmans, Green & Co. Shafer-Landau, Russ 2003. Moral Realism: A defense, New York & Oxford: Clarendon Press. Scanlon, T. M. 1998. What We Owe to Each Other, Cambridge, MA: Harvard University Press. Smith, Michael 1987. The Humean Theory of Motivation, Mind 96: 36-61. Smith, Michael 1994. The Moral Problem, Malden, MA: Blackwell. Williams, Bernard 1981. Internal and External Reasons, reprinted in Moral Luck, Cambridge: Cambridge University Press. Williams, Bernard 1995a. Internal Reasons and the Obscurity of Blame, in Making Sense of Humanity, Cambridge: Cambridge University Press. Williams, Bernard 1995b. Replies in World, Mind, and Ethics, ed. J. E. J. Altham & Ross Harrison, Cambridge: Cambridge University Press. 32