RESPONDING TO AESTHETIC REASONS

away from every possible source of error except a potential mismeasurement by the scales. Then if the scales weigh the object in a wholly accurate way, the values of the two parameters will be identical, but otherwise they will differ. If E is the range of values {1, 2, 3}, then there will be nine different ‘worlds’ of this kind that might obtain. As noted, we’re abstracting away from sources of epistemic error other than a failure of the scales, and so it’s harmless to assume that the agent always knows how things appear to her (if <e, f>R<e*, f*> then f=f*). We don’t want to assert the converse claim, however, that if two worlds are identical with respect to appearance then what an agent knows at one cannot rule out the other as an open epistemic possibility. That would be to allow, for example, that taking the scales at face value when they are giving a wholly accurate reading of 10 kilos in favourable circumstances could never allow an agent thereby to come to know that the object didn’t weigh a million kilos. Non-sceptics should agree that in cases of this kind, the agent can take appearance as an epistemic guide to reality. They should agree that taking accurate appearances at face value in an epistemically permissible way allows us to eliminate some worlds that were previously epistemic possibilities for us. Plausible such accounts should allow that the agent can’t take appearance as a perfect guide, however. We know that the basic epistemic methods that she employs to find out how the world is need not be, and typically won’t be, perfectly discriminating. (Even a scale which is known to be very highly accurate might not discriminate perfectly between tiny differences in weight.) We noted above that the factivity of knowledge is modelled by stipulating that the accessibility relation R is to be reflexive. An agent at a world where appearance matches reality – call it <f, f> – will thus have that world itself as an open epistemic possibility. Williamson models the fact that our epistemic methods are typically not perfectly discriminating an aspect by the stipulation that every such world accesses another distinct world. For example, if our agent is at a world where the object weighs exactly ten kilos, and the reading on the scale tells her that this is so, then it will remain an open epistemic possibility that the object weighs 10+k kilos, where k is some vanishingly small quantity to which the scale is at best unreliably sensitive to. Formally, we have the constraint that the proposition {<f, f>} is a proper subset of the open epistemic possibilities at the world <f, f>, that is: (1) {<f, f>} R(<f, f>) Responding to Aesthetic Reasons 56 Estetika: The Central European Journal of Aesthetics, LIV/X, 2017, No. 1, 00–00 Zlom1_2017_Sestava 1 24.3.17 10:46 Stránka 56

agents in selecting actions and adjusting attitudes, and help us evaluate their so doing. 1 I won't assume that any reductive account of normativity can be provided, and will generally proceed as if some form of non-reductive realism were true, although plausible forms of reductionism should be able to echo most of the central points. The reader won't be asked to accept that values are reducible to reasons or vice versa; the same holds of requirements of rationality. 2 (My inclination is to embrace a no-priority view of the former issue and a reasons-first view of the latter.) Our focus will be on aesthetic reasons in a broad, inclusive sense. We won't presuppose that any form of aesthetic empiricism is true, understood as the view that aesthetic reasons all hold in virtue of facts about experiences. Nor will we restrict our attention to some favoured class of aesthetic properties, attitudes, judgements, commitments, or concepts (such as 'free beauty' , 'aesthetic pleasure' , or 'the standard of taste'). More generally, I won't take for granted that any given historically influential treatment of the nature of aesthetic judgement or standards -Kant's say, or Hume's -is correct. I'll take that as a substantive issue, to be settled by argument. If defenders of a given position are distinctively vulnerable to (or immune from) an objection, I'll try to signal that as I go along.

II. AESTHETIC REASONS AS REASONS-FOR-AESTHETIC-ATTITUDES
When we distinguish between aesthetic reasons and other types of reasons, what kind of contrast are we drawing? And how fundamental or structural is that contrast?
Let me say a little more about the kind of questions I have in mind. Learning that somebody acted for an aesthetic reason seems to help us explain why they acted, and to know why they acted as they did. But it is very plausible to think that knowledge and explanation of this kind is closely connected with the elimination of members of a contrast set: out of an initial contextually determined set of possible kinds of reason -{aesthetic considerations; considerations of public safety} -we come to know that one set rather than the other was what guided the agent's action. When we think about things in this way, aesthetic reasons as a class will often seem to contrast with fairly specific, contextually embedded and determinate groupings of normative reasons. If Dominic wasn't wearing the hat for aesthetic reasons, he was surely wearing it because he was being paid to do so (he needed the money), or for reasons of friendship (he actually hated the hat, but the friend he was meeting thought she had chosen well), or out of vanity (to hide his bald spot), or as a sign of protest (head coverings were being unjustly banned), and so forth.
If we thought about the contrasts in this way, aesthetic reasons might seem comparatively thick and practice-sensitive in comparison with the reasons we have to value, say, free agency or self-development. But we might wonder whether aesthetic reasons constitute a more fundamental normative kind than this. After all, as Raz says in a related discussion, 'We do not think of people's behaviour towards issues involving beauty as a practice, for there is no specific action-type, performance or approval of which [which] can constitute the practice of beauty, so to speak. Our appreciation of beauty can be manifested by almost any conceivable action under some circumstances or other.' 3 There is a long tradition of thinking that the distinction between aesthetic and non-aesthetic reasons cuts at a normative joint. Can we provide adequate motivation for such a view? Let me begin by describing a common approach to this question, which I'll ultimately want to argue is at best incomplete. This strategy attempts (i) to individuate types of reasons in terms of a designated attitude or activity that they bear on and (ii) to find some such attitude or activity that is suitably related to the realm of the aesthetic. For example, a simple but attractive model might distinguish (i) epistemic reasons: reasons for belief; (ii) practical reasons: reasons for action; (iii) aesthetic reasons: reasons for appreciation.
On this view, the distinctions between aesthetic reasons and others are as structural as the distinctions between the attitudes and activities that they are directed towards. 4 A familiar set of concerns arise here. For example, since believing falsely that P may give rise to certain practical benefits on occasion, we might worry that the above picture does not adequately distinguish epistemic from practical reasons for belief. We might then try to enrich the picture by requiring that the reasons bear on the attitude or action in a specific way. For example, we might say that epistemic reasons are those reasons for belief in P that form part of a case that P, Responding  or help us settle the question of whether P. 5 We might think of (positive) practical reasons as those that help make a case for performance of an action A, or help us settle the question of whether to do A. If we developed the view in this way, aesthetic reasons would presumably consist of those that form part of a case for appreciating an aesthetic object O, or for settling the question of whether O is worth appreciating, or whether to take up an aesthetic attitude towards O, or something like that.
Developing the view in this way raises a prima facie puzzle, however. Suppose we say that aesthetic reasons are those that help us settle the question of whether O is worth appreciating. Why aren't these reasons thereby epistemic reasons?
After all, they directly bear on the truth or falsity of a proposition. Similarly, suppose we say that aesthetic reasons are those that help us settle the question of whether to appreciate O. Why isn't this enough to make them practical reasons?
After all, they seem to help us settle the question of whether to undertake a given pattern of activity -attending to the aesthetically relevant properties of O, dwelling on their interrelations, interpreting them in the light of a range of background art-historical knowledge, and so on. Aesthetic reasons so understood might not seem to be suitably 'categorically different' from reasons for action and belief. And that might be thought to imperil the project of identifying a suitably fundamental normative joint.
Suppose, for example, we embraced Iseminger's well-known account of the aesthetic state of mind. On this view, somebody is aesthetically appreciating a state of affairs just in case (i) they believe that their cognitively apprehending something about that state of affairs is good, and (ii) that belief is not based on reasons that present that apprehension as a means to other ends. 6 On this account, presumably, one's reasons for appreciation would be reasons that (i) helped one resolve what to believe by settling the question of whether one's cognitively apprehending something about a given state of affairs is good (ii) in a way that didn't depend on their presenting the apprehension as an effective means to some other end. But, for all that has been said, those reasons might just be a subset of epistemic reasons. If so, aesthetic reasons wouldn't seem to form a natural contrast class with practical and epistemic reasons. (Compare: there are many good reasons to build bridges over rivers, but that shouldn't encourage us to talk in terms of a special normative class of 'bridge-building' reasons.) 7 There is therefore pressure on such a position, given its motivations, to find a special aesthetic attitude or activity to shore up the normative joints. We could think of them as a particular species of 'judgement-sensitive' attitude in Scanlon's sense. 8 If we thought that these states were quite different from beliefs or actions, but still comparatively practice-independent and suitably universal, we might view aesthetic reasons as belonging to a fairly basic, general, unified normative category. That would allow the distinctiveness of aesthetic reasons to 'piggy-back' on the distinctiveness of folk-psychological categories. They would be like reasons for indignation or reasons to be cheerful. They would be more like epistemic reasons, and less like reasons to build bridges.
I think that most extant developed views of this kind are ultimately unsatisfying in one of three ways. They either (a) surrender too much of the rational authority of aesthetic normativity or (b) present us as bound by overblown and implausibly demanding normative constraints, or (c) overgeneralize in a way that robs aesthetic reasons of too much of their content. I can't offer a full defence of these claims here, but in order to motivate exploration of a rival approach, let me say something about the types of ways that I think such views fall short.
A defender of such a view should want the states in question (i) to be not just beliefs or actions, (ii) to be reason-sensitive, (iii) to be suitably distinctive and unified, and (iv) to be appropriately related to everyday aesthetic talk and practice. The first condition seems to rule out cognitivist treatments of aesthetic attitudes, which identify them with beliefs whose content ascribes, say, a type of secondary quality. The second condition apparently rules out brute sensory states like aches, itches, or non-presentational sensations of pleasure. The third condition sits uneasily with treatments that approach the aesthetic attitude as one contingent, permissible, appetitive preference among others, such as a desire for ice cream. The fourth requires that the attitude in question relate in the right way to the kind of explanation that we give when we say that the placement of the bridge didn't give enough weight to aesthetic considerations. Since this is a contrastive explanation,  Judgement-sensitive attitudes are those 'that an ideally rational person would come to have whenever that person judged there to be sufficient reasons for them, and that would, in an ideally rational person, "extinguish" when that person judged them not to be supported by reasons of the appropriate kind' . T M. Scanlon, What We Owe to Each Other (Cambridge, MA: Harvard University Press, 1998), 20. the attitude in question seems as if it shouldn't be something that, say, is merited by every object or situation whatsoever.

Responding to Aesthetic Reasons
A 'fitting-attitude' account of appreciation might be thought to offer a means of balancing up these various tensions. 9 Such an account might aim to identify an attitude that offers a distinctive blend of cognitive and affective elements, so as to avoid the problems that, as we have just seen, would arise if we focused solely on beliefs, preferences, or 'mere' affective states. Evaluative elements might be integrated too. For example, some theorists might identify the attitude with an 'aesthetic emotion' and then characterize it as a perception of the fact that one's circumstances merit such a response, under a special affective mode of presentation. Suppose that we endorsed such a view. We could then ask whether rational agents quite generally are supposed to have aesthetic reasons to adopt or undergo such an attitude.
Suppose that we allowed that the attitude in question could vary faultlessly across a given population, in the way that emotions such as ellipsism, liberosis, or enouement might be thought to do. 10 It has seemed to many theorists that this would be to give up too much of the objectivity and generality of aesthetic response. 11 Moreover, allowing this kind of faultless variation might be thought to undermine the parallels between aesthetic, epistemic, and practical reasons, since the latter two categories arguably apply to rational agents universally.
Stressing that the attitude in question is directed towards a normative or evaluative fact doesn't seem to help here. After all, we can make sense of certain feelings of liberosis being merited or appropriate and other ones being pathological. Nevertheless, the state seems unusual or permissibly variable enough to undermine any natural contrast with belief or action.
Even regarding less unusual, more widely experienced feelings -sexual attraction, say -we might think that a plausible fitting-attitude account would not claim universality. Such states might well be thought to unify cognitive, evaluative, and affective components. We might well treat them as presenting an external, evaluative fact under an affective guise (the qualities of the person the feeling is directed at seem to merit such a response, rather than it presenting like a pathological urge). But it seems phenomenologically wrong, and 9 For a useful overview of this style of position, see Daniel Jacobsen, 'Fitting Attitude Accounts of Value' , in Stanford Encyclopedia of Philosophy, ed. Edward N. Zalta, spring 2011 ed. (Stanford University, 1997-), https://plato.stanford.edu/archives/spr2011 /entries/fitting-attitude-theories. 10 Ellipsism is a kind of regret that one will not live to see the future. Liberosis is a longing to care less about things, as one did as a child. Enouement is the feeling of wishing that you could go back in time and tell your earlier self about how things will turn out. 11 It's notable that philosophers as different as Plato, Aquinas, Hume, and Kant seem to have construed correct aesthetic response in more universal and authoritative terms. normatively incorrect to think that the meriting state in question requires everyone to respond in the way the agent does, insofar as they are rational. 12 If aesthetic reasons are like this, then I think that we should conclude that either (i) they do not require everyone to engage in appreciation, on pain of rational fault, and are in that sense rationally optional, or (ii) the fact that they do require everyone to do so remains a mystery. The first option has commonly been felt to be inconsistent with the objectivity and generality of aesthetic reasons. Even if that is incorrect, we might prefer an account that was able to demonstrate that the authority of such reasons was as robust as that of the epistemic and practical reasons it is alleged to contrast with. On the second option, the burden of proof is surely on the defenders of this kind of view to establish that they are not merely positing overblown, over-demanding normative constraints.

III. CAN A KANTIAN APPROACH HELP?
Kant's approach to aesthetic judgement is rightly regarded as the most serious, developed attempt at discharging this burden of proof. It aims at establishing the normative authority and generality of aesthetic reasons -'how the feeling in the judgement of taste comes to be exacted from everyone as a kind of duty' .
Let's consider a broadly anti-realist interpretation of the Kantian position, which offers, I believe, the most compelling explanation of the source of this rational authority. 13 On this account, Kant agrees with Hume when he tells us that the beauty of a circle is 'not a quality of the circle': 'To say, This flower is beautiful, is tantamount to a mere repetition of the flower's own claim to everyone's liking.
[…] beauty is not a property of the flower itself. ' 14 Hume thought of beauty as 'founded entirely on the particular fabric and constitution of the human species' . We might interpret this as meaning that as an a posteriori, contingent, law-like regularity of nature, certain properties of objects causally necessitate pleasure in human animals like us. Similarly, Responding to Aesthetic Reasons 12 A fallback approach might attempt to argue that everyone is so bound, as long as they meet certain enabling conditions. This might equally be true, however, of somebody who has formed a permissible longing for chocolate ice cream. Being bound by this kind of reason seems connected to human rational agency in a way that is quite different from epistemic and practical reasons. In one case, but not the other, there seems to be, for example, rational pressure to ensure the continuance of the enabling conditions. I should take steps to ensure that I am suitably responsive to evidence, reasons for action, and so on. I needn't take steps to ensure that I long for ice cream. 13 My intention here is only to offer an interpretation that respects core Kantian themes, and seems to offer a reasonable prospect of explaining the universality of the demand to share a pleasure. Elements of the position I describe may be in tension with other apparent Kantian doctrines. 14 Immanuel Kant, Critique of Judgment, trans. Werner S. Pluhar (Indianapolis: Hackett, 1986), § 32, 145 (AA 5:281-82). the prospect of the absence of defects and imperfections that prevent us feeling such pleasures also naturally pleases universally (again, as a contingent, a posteriori fact about animals like us). But Kant wants a form of normative necessitation rather than natural necessitation -a rational requirement, not a causal law. He agrees with Hume that beauty constitutively involves pleasure, but holds that this pleasure rests on 'subjective conditions for the possibility of cognition as such ' . 15 How should we understand the connection between rationally mandated pleasure and the fulfilment of the subjective conditions for cognition?
The anti-realist interpretation replies as follows. Consider an everyday singular observational judgement, like 'This is red' or 'This is spherical' . Such a thought does at least two things, corresponding to its perceptual-demonstrative and predicative components. It (i) picks out an object, and (ii) identifies an external condition that the object has to meet if the claim is to be true. To find out whether the claim is in fact true, we have to examine the specified object and see whether it meets the external condition. Claiming that the object does meet the condition commits one to a normative claim -namely, anybody who doesn't agree that it meets the condition is making a mistake.
In the case of 'This is red' or 'This is spherical' , Kantian intuition does the job of picking out the object. Kant tells us that intuition 'refers immediately to the object and is singular' , 16 that 'an intuition is a singular representation' , and that 'since individual things, or individuals, are thoroughly determinate, there can be thoroughly determinate cognitions only as intuitions, but not as concepts ' . 17 The external condition supplied by the predicate is expressed by a concept ('mediate, by means of a mark, which can be common to several things'). 18 But the thought 'This is beautiful' works differently. We can see that by examining its epistemology. In order to tell whether something is beautiful, we don't look outward, to see if it meets some further external condition, specified in the predicate. Rather, we look inward 'to the subject and his feeling of pleasure or displeasure' . Since the predicate in 'This is beautiful' doesn't supply any external 15 Ibid., § 39. condition for the object to meet, no concept is used to cognize the object. We talk as if beauty were a property of the object, but it isn't. Thus Kant says: '[the subject] will talk about the beautiful as if beauty were a characteristic of the object and the judgment were logical (namely, a cognition of the object through concepts of it), even though in fact the judgment is only aesthetic and refers the object's presentation merely to the subject' . 19 Anachronistically, we might think of 'This is beautiful' as expressing a kind of gappy proposition that picks out an object and places it within a name-predicate structure, but leaves the predicative space empty, marking it with a kind of 'dummy' predicate 'x is beautiful' . The thought determines a singular reference by means of perception, but does not predicatively classify the object in question in any way whatsoever. The perceptual identification of the object, and its location within a structure that only requires the completion of the predicative place in order to be truth-evaluable, activates the Kantian cognitive powers. Imagination is constrained by understanding (since it has made the object available for the application of concepts) but without any particular such concept yet being applied. There's an 'attunement' between the faculty that delivers singular reference and that which delivers predicative classification. It is not implausible to think that such a process must be possible if any empirical cognition is to take place whatsoever. Thus Kant says 'this pleasure accompanies our ordinary apprehension of an object by means of the imagination, our power of intuition, in relation to the understanding, our power of concepts. This apprehension occurs by means of a procedure that judgment has to carry out to give rise to even the most ordinary experience. ' 20 What connects any of this to rationally mandated feelings of pleasure? Why is such a demand or requirement legitimate? Why, just because I take disinterested pleasure in something, am I within my rights to demand that everybody else does too? In the case of 'This is red' or 'This is square' , the normative demand was connected with the application of an external, objective condition. But that is explicitly not what is at issue with 'This is beautiful' . I think that the most defensible view is to simply identify the satisfaction of a precondition for rational activity in general with pleasure. 21 Any rational agent, qua rational agent, should will that the essential preconditions for rational agency are fulfilled, and the fulfilment of this kind of willed end -or perhaps, the recognition of such fulfilment -is just what the pleasure in question is. In the pure judgement of taste, the precondition consists in the singling out of an object, and making it available for perceptual classification. 22 The pleasure in pure beauty is thus directed at a purposiveness (singular representation perfectly poised for conceptual classification) without a purpose (no end specified by the concept, since no concept yet specified). In aesthetic experience, the world is revealed to us as thinkable, as it would be if it were made to accord with our nature as rational experiencing beings. But no particular completion of the thought is forced upon us: the content of our experience includes a predicative blank for us to fill in, without compulsion to fill it in this way rather than that. The gap-where-a-concept-ought-to-be thus offers us a conceptual affordance that speaks to us as free agents who are autonomously able to categorize our world and direct our own activity.
Suppose for the sake of argument that the foregoing was the best way to precisify and defend a Kantian position. It could hardly be accused of not doing enough to connect up aesthetic pleasure with a universal, authoritative requirement upon rational agents. But the position as described seems in danger of over-generating aesthetic reasons. After all, the view as stated placed no restrictions upon the kinds of objects that occupied the perceptualdemonstrative element of the gappy propositions. The view thus seems prima facie committed to the interpretatively unpopular and counter-intuitive view that every object can be perceived as beautiful. 23 Moreover, the 'thin' interpretation of pleasure as will-fulfilment might reasonably be thought to be phenomenologically inadequate to the felt rewards of aesthetic engagement.
Perhaps there are satisfying answers to these concerns. But the difficulties they raise, taken together with the cognate problems faced by other influential interpretations of Kant's aesthetics, at the least serve to motivate exploration of a different style of account.

IV. A DIFFERENT APPROACH
The model we have been discussing so far individuates types of reason in terms of the kinds of attitudes and actions that they are reasons for. That's what put pressure on the view to find a distinctive kind of aesthetic attitude or activitya special feeling of appreciation, say, or a special form of 'reflective ' judgement -22 This seems to individuate of the pleasure in terms of the content or object of the will, rather than phenomenologically. In the 'Introduction' to the Critique, Kant suggests that pleasure is typically connected with the attainment of an aim, but that unexpected fulfilment can provide a noticeably different pleasure. to contrast with the categories of belief and action. If the reasons that we had to come to accept that O is beautiful were just evidence for the belief that O had a certain kind of secondary quality, or that O ought to be appreciated in a certain way, then the intended contrast between epistemic, practical, and aesthetic reasons could not be drawn.
I do not deny that there may be distinctively aesthetic attitudes or actions. I do think that it is a bad idea to restrict the scope of aesthetic reasons to reasons we have to enter such states. To do so would be to ignore an important regulative role that such aesthetic considerations play.
On the above view, reasons help us settle whether to go into a state that is in a sense normatively and evaluatively neutral. I can believe that P, whether or not P is false, completely unjustified, concerns a matter of complete irrelevance, and so on. I can press a button, whether or not there is anything to be said for so doing, or I can utter sentences for bad purposes as well as good. Acting rationally, being reason responsive, is aiming to support one's selection of such a neutral state with considerations external to it which show it to be in good standing. 24 A competing view, which Korsgaard has ascribed to both Aristotle and Kant, sets things up a little differently. 25 On this account, the primary object of practical reflection is not a reason-independent act such as pressing the button, or wearing the hat, which one then seeks reasons for and against. Rather, it is what she terms an action, thought of as an act-end pair -to build the bridge in order to facilitate traffic flow; to wear the hat so that you can hide the bald spot. Korsgaard's discussion focuses on practical reason, but we might view epistemic activity in a similar light. On this account, the basic object of epistemic appraisal would be a state that had relationships of rational support irreducibly 'baked in' , as states of knowledge seem to have. 26

24
It is perhaps not surprising, therefore, that this view of rationality is often endorsed by those attracted to reductive-naturalistic views that identify actions with bodily movements, beliefs with causal dispositions to bodily movements, and so on. Such a view fits well with views upon which rational deliberation considers potential actions and attitudes 'under the guise of the good' . The view can be thought to involve an explanatory primacy of the 'good case' in at least two senses. First, actions under consideration include their own putative justification, an aim or purpose that is supposed to explain or assure us of their choice-worthiness or goodness. Second, the procedures of rational deliberation and choice which regulate us in our selection of action constitutively aim to distinguish the genuinely choice-worthy or good from the merely apparently so. In this sense, good action, or rationally permissible action, is explanatorily more basic than either 'act' or 'action' . We can reasonably understand pointless acts and misguided actions as failed actions-with-worth.
If we thought of things in this way, we might view acts (pressing the button) and ends (hiding the bald spot) as in some sense derivative, abstractions from a structurally more basic state. Rationally permissible actions, the right kinds of combinations of acts and ends, are not identified by locating ends and purposes that are external to them and might justify their pursuit. Rather, we mark out actions as choice-worthy by establishing that they have genuine worth, are worthwhile in themselves. The job of practical reason is (i) to distinguish between those actions (putatively reason-supported acts) that really are worthwhile in themselves and those that merely seem to be and (ii) to elect freely between the remaining range of worthwhile actions. Showing that an action is worthwhile in itself might involve showing that it is compatible with one's dignity as a rational agent, or the value of a noble life, or the goodness of happiness. 27 But the action needn't be thought of only justifiable to the extent that it is instrumental to such a valuable state. Rather, a life that is made up of such intrinsically worthwhile actions, because of the agent's own activity, might be what happiness, nobility, the dignity of humanity, and so forth, consist in. The actions would be components of a good life, and not just means to achieving one.
On the above view, prospective actions are in a sense 'filtered' by practical reason: we ought only to consider those principles of action, specifying acts-forends-in-circumstances, that are compatible with our ultimate good, given our nature as rational creatures. We can reasonably think of the Categorical Imperative test as aiming to indicate when an act-for-a-reason will be compatible in this way.
We might look at a Scanlon-type view in a similar fashion: the actions that are suitable components of our ultimate good would be those that would not be disallowed by principles that no one could reasonably reject, when allowed to make appeal to 'personal' reasons for rejection, reasons that they have 'on their own account' . This view wouldn't be compelled to insist, implausibly, that an action that was not so disallowed was only instrumentally valuable, good only insofar as it promoted the ultimate end of bringing into being practices of mutual justification between rational agents. Rather, being capable of meeting such a constraint would provide a mark of an action that was a suitable object of 27 We might think of things in a slightly more disjunctivist spirit as follows. When we feel pulled towards selecting a given act for a given reason, it typically feels as if it is the right or good thing to be doing. We might gloss such an evaluative appearance as follows: it either is the good choice, or it merely appears to be. We can think of the 'normative filter' applied by the CI-procedure, Aristotelian practical reflection, Scanlonian respect for mutual justification, and so on, as helping provide a guide to which disjunct obtains. The fact that an action passes through the filter shows that it is one that we have adequately good reason to choose now. The filter needn't be thought of as substituting or adding a different reason to choose the act. It would be more like assuring yourself that there aren't undermining defeaters for a proof, which doesn't thereby become part of the proof. choice. It wouldn't specify an additional end, over and above the reason that was already taken to justify a given act.
Suppose that this kind of view could be made defensible. How would it differ from the reasons-for-appreciation model discussed above? On that view, genuinely aesthetic reasons were considerations apt to occupy the 'end' role in appreciation-end pairs. They were facts that normatively supported certain acts or states of aesthetic appreciation. They helped settle the question whether appreciation is aesthetically merited here. I agree that the question whether a certain object merits aesthetic appreciation is important, and that a range of interesting considerations can help us answer it. But the reasons that spring from aesthetic value may reasonably be thought to play another role, one less integrally tied to aesthetic appreciation or aesthetic judgement. They might play a regulative, filtering role, of the kind described above. They might help us identify suitable objects of choice. 28 One heuristic that is sometimes useful in identifying aesthetic reasons in their regulative role is considering an expanded version of the contractualist model described above. Suppose that we had a group of people seeking to converge in an autonomous way upon sets of principles for the general regulation of conduct.
Potential such principles are inadmissible if they can be reasonably rejected by people who are allowed to appeal to reasons that they have 'on their own account' . Conduct is morally wrong if it is prohibited by any admissible set of such principles. Now, imagine that we add to the procedure a set of spokespeople whose job is to act as 'guardians' of various forms of impersonal value. Those voices might advert to new considerations, not cited on people's own account, including reasons related to the 'impersonally good' value of knowledge, animal welfare, environmental diversity, natural grandeur, and artistic magnificence. We can get a sense of whether a reason is a regulative aesthetic reason by thinking about whether it is the kind of consideration that an able and conscientious such spokesperson ought to cite if their job is to speak up for some recognizably aesthetic value. 29 Responding to Aesthetic Reasons 28 Doesn't the claim that aesthetic reasons could play such a role conflict with the claim that the 'filtering' role of practical reason needn't be thought to provide additional reasons for action? No. Consideration of the Scanlon model and the CI-procedure show that reasons can play a role in setting up the filter, without the fact that an action passed the filtering test itself having to count as an additional or conflicting reason. 29 This is only intended as a heuristic, and not as any kind of exhaustive identification procedure or reductive analysis. Scanlon accepts that there is a wider sense of justification in which the rejection of such principles would signal lack of justifiability to others, but distinguishes them from the class of moral justifications that spring from our value as rational beings. See T. M. Scanlon, 'Reply to Gauthier and Gibbard' , Philosophy and Phenomenological Research 66 (2003): 176-89. It's clear that in many cases, we will have multiple independently sufficient grounds for rejecting inadmissible principles.
For example, quite apart from the 'personal' reasons that we have on our own behalf for objecting to the destruction of the Grand Canyon, Scanlon accepts that the loss of its aesthetic properties would make its destruction an impersonally bad result. This consideration would foreseeably be cited by a suitable spokesperson for natural beauty when rejecting various principles that might otherwise permit such destruction, and is thus an aesthetic reason. Suppose that somebody was wondering whether to strip-mine there, in order to profit from recently discovered rare minerals. The decision in question would involve settling whether such an action was a good choice in these circumstances. Due regard for aesthetic value -aesthetic reasons in their regulative role -might prohibit such an action, by 'striking down' an eligible object of choice.
Similarly, a decision to build this bridge here, and in this way, might fail to give due weight to aesthetic considerations. Making that claim need not commit one to thinking that anything in particular merits appreciation, or that things always ought to be apt for appreciation, or whatever. Aesthetic considerations on this account resemble reasons of friendship and justice. They are best conceived as reasons-of rather than as reasons-for. They are distinguished primarily by the type of value that they help us integrate into our lives, and not by a distinctive aesthetic state that they support.

V. RESPONDING TO AESTHETIC REASONS
There are at least two explanatory debts that such an account must discharge.
The first is to explain the authority of aesthetic reasons, in their regulative role. This is not quite the same task as the challenge that faced the ambitious, universalist fitting-attitude theorist. Their difficulty was in making the case that we are within our rights, in responding with feeling to a given evaluative situation, to demand that everyone ought to respond in the same way. The regulative view need not embrace this form of universalism. The demand issued in the Grand Canyon case is that the miners not damage a given landscape, not that they take appreciative pleasure in it. Nevertheless, we would like an explanation of why aesthetic value could serve as a suitable ground for such constraints on otherwise rationally permissible projects. This would provide a vindication of the claim that certain aesthetic values are impersonally good in a way that can constrain The second explanatory debt is to motivate the picture of choice that Korsgaard ascribed to Aristotle and Kant. Why shouldn't we think of aesthetic appreciation by perceivers, or aesthetic accomplishment by artists, as a selfstanding type of successful uptake of and response to independently identifiable reasons? Why view processes of rational deliberation and selection as primarily aimed at unified act-end pairs? After all, that was the model that seemed to allow us to say that aesthetic reasons could moderate the selection of practical projects, without thereby becoming the reason for those projects.
Here I think that the defender of regulative aesthetic reasons should emulate the strategy of knowledge-first theorists in epistemology. 31 Williamson famously argues that the history of 'Gettierology' gives us inductive grounds for thinking that theoretical rationality is directed at unified states of knowledge. But he also offers a demonstration that quite minimal and plausible constraints on epistemic model theory have the result that there will be cases of justified true belief that aren't knowledge. 32 Given this demonstration, the hope that more plausible future theory could close the gap between warranted true belief and knowledge seems remote. Accordingly, it is reasonable to take knowledge as a primitive in Justification helps settle the question of whether P, and provides reason to think that P is true. Williamson's argument, if it works, entails that successfully according with such a reason for belief (having excellent evidence for the truth of P, and correctly believing that P as a result) won't entail knowing that P. We'll have to add an unreduced clause 'according with it in the right way' or 'in the right circumstances' , that we don't know how to elucidate without appeal to knowledge itself. This frustrates the reduction, and encourages us simply to start with knowledge as a primitive. Analogously, my argument entails that successfully according with aesthetic reasons (making an attempt at an aesthetic outcome that I recognize that I have reason to bring about, and achieving it) won't amount to the kind of skilful accomplishment that we know is possible in

VI. WILLIAMSON ON EPISTEMIC MODEL THEORY
Williamson's account proceeds as follows. Let an epistemic model be an ordered pair <W, R>, where W is a non-empty set, and R a set of ordered pairs of members of W. The members of W are thought of as worlds -maximally specific states. (For exegetical purposes I'll follow Williamson in concentrating on a more specific set of models that deal with very simple 'worlds' , but we'll get to that soon enough.) We want enough structure in our model theory to represent propositions that can be true, known, justifiedly believed, and so on. Williamson adds most of this structure in a very standard way. Propositions in a model are subsets of W.
A proposition p is true in a world w just in case w is a member of p, and false otherwise. One proposition entails another if the former is a subset of the latter, and so propositions are identical just in case they are mutually entailing. An agent counts as knowing whatever is true in all the worlds that, for all she knows, she is in. We model that by means of R, construed as a relation of epistemic accessibility: a world x is accessible from a world w(wRx) just in case whatever the agent knows in w is true in x. The proposition that the agent knows that p is identified with the proposition Kp that is true in a world w just in case every world x accessible from w is a world in which p is true. We build in the factivity of knowledge by requiring that R be reflexive. We'll model the intuitive conception of what an agent knows in w by identifying it with the strongest proposition known by the agent, which we'll call R(w) = {x:wRx}. We can think of an important goal of an enquiring agent being to advance knowledge by ruling out epistemic possibilities -that is, by eliminating worlds from R(w). For example, if what an agent knows doesn't settle whether Helen is speaking or not, and they come to know that she is, then every non-Helen-speaking world will thereby be eliminated from R(w), and what the agent now knows will now be represented by a stronger, smaller proposition R'(k).
Williamson is going to deal with issues of justified belief as well as knowledge, so he needs to add a bit more structure to the theory. It's easiest to see the role that this structure is playing if we follow him in considering a set of very simple 'worlds' that just encode a set of facts about how an agent takes her environment as being, and how it really is. Abstracting from irrelevant complications about indeterminate appearances and the like, let's think of the worlds in question as being ordered pairs of two parameters <e, f> from a value-set E, where e represents the real value of some parameter and f represents the value it appears to the agent to have. For example, suppose that we have a set of aesthetic agency. We'll have to add the modifying clauses. As with the epistemic case, this will in effect frustrate a reduction of action to 'act' and 'reason' , and encourage us simply to take the former as a primitive.
possible weights E, and the agent is reading off the weight of a given object by using a set of electronic scales that she is entitled to presume are accurate. Let's abstract away from every possible source of error except a potential mismeasurement by the scales. Then if the scales weigh the object in a wholly accurate way, the values of the two parameters will be identical, but otherwise they will differ. If E is the range of values {1, 2, 3}, then there will be nine different 'worlds' of this kind that might obtain.
As noted, we're abstracting away from sources of epistemic error other than a failure of the scales, and so it's harmless to assume that the agent always knows how things appear to her (if <e, f>R<e*, f*> then f=f*). We don't want to assert the converse claim, however, that if two worlds are identical with respect to appearance then what an agent knows at one cannot rule out the other as an open epistemic possibility. That would be to allow, for example, that taking the scales at face value when they are giving a wholly accurate reading of 10 kilos in favourable circumstances could never allow an agent thereby to come to know that the object didn't weigh a million kilos. Non-sceptics should agree that in cases of this kind, the agent can take appearance as an epistemic guide to reality.
They should agree that taking accurate appearances at face value in an epistemically permissible way allows us to eliminate some worlds that were previously epistemic possibilities for us. Plausible such accounts should allow that the agent can't take appearance as a perfect guide, however. We know that the basic epistemic methods that she employs to find out how the world is need not be, and typically won't be, perfectly discriminating. (Even a scale which is known to be very highly accurate might not discriminate perfectly between tiny differences in weight.) We noted above that the factivity of knowledge is modelled by stipulating that the accessibility relation R is to be reflexive. An agent at a world where appearance matches reality -call it <f, f> -will thus have that world itself as an open epistemic possibility. Williamson models the fact that our epistemic methods are typically not perfectly discriminating an aspect by the stipulation that every such world accesses another distinct world. For example, if our agent is at a world where the object weighs exactly ten kilos, and the reading on the scale tells her that this is so, then it will remain an open epistemic possibility that the object weighs 10+k kilos, where k is some vanishingly small quantity to which the scale is at best unreliably sensitive to. Formally, we have the constraint that the proposition {<f, f>} is a proper subset of the open epistemic possibilities at the world <f, f>, that is: (1) {<f, f>} R(<f, f>) Adopting an anti-sceptical position with respect to whether appearance can ever be a guide to reality shouldn't commit us to the implausible view that knowledge is only possible when appearance and reality match perfectly. It's much more plausible to think that, even if light-refraction causes a mountain to look to me to be a little larger than it actually is, I can still come to exclude certain epistemic possibilities on the basis of my visual access to it -I can rule out the hypothesis that the mountain is the size of a grape, for example. Moreover, my capacity for such exclusion seems to be inversely proportional to the degree to which appearance and reality come apart. In situations where the refraction is more and If the distance between appearance and reality is greater in one world than another, then the agent knows less at the former than she does at the latter.
Williamson argues that the above two constraints establish that the epistemic accessibility relation R is non-transitive (corresponding to failure of the KK-principle) and non-symmetric (in a way that suggests that an exact match between appearance and reality is an epistemically open possibility in bad cases where appearance varies widely from reality, but that the agent is in a position to eliminate such bad cases when occupying the good case of close match between appearance and reality). I refer interested readers to his discussion of the issues. More important for our purposes here is his distinctive treatment of belief and justification in the theory.
Let's continue to follow Williamson in only considering the highly simplified <reality, appearance> type worlds introduced above. We'll also follow him in simplifying the case further in restricting our attention to cases where the agent's beliefs depend only on how things appear to her. Since belief will thus be indifferent as to whether or not appearances match reality or not, we can consider the good case of the wholly accurate <f, f> world. By the belief condition on knowledge, whatever the agent knows at <f, f> she will believe at <f, f>. Given our anti-sceptical approach to knowledge, it is overwhelming natural to think that when she forms her belief that the real value is f in the good case <f, f> world, (2) d(e, f) ≤ d(e*, f) if and only if R(<e, f>) R(<e*, f>) If the gap between conception and execution is larger in one situation than another, then (holding our idealizing assumptions fixed) the agent skilfully accomplished less in the former than the latter. We can piggyback on Williamson's arguments to establish the result that the relation R will be non-transitive and non-symmetric.
What is the correlate of belief here? We can think of the content of the agent's reasonable attempt as being the set of worlds that she would have brought about had she skilfully accomplished her aim. We can argue for that in a way that mirrors our reflections on belief and knowledge above. Adopt the optimistic assumption that we do sometimes skilfully accomplish things -that apt performance is possible. And make the further plausible assumption -the correlate of the belief condition on knowledge -that every skilful accomplishment requires a reasonable attempt. Finally, let's continue with our simplifying assumption that the attempt of the agent is governed only by her selection of an aim, and that The generalized interpretation of the model theory that Williamson provided us with offers natural interpretations of the very same theses that he uses to construct Gettier cases. So exactly the same structures will arise in the practical case. Suppose that the graphic artist recognizes the aesthetic reason she has to draw an elegant line. Consider a series of three worlds where her drawing of the line manifests less and less skill. At w1, her drawn line is a clear case of skilfully produced elegance; at w2, her drawn line is just within the bounds of elegance, but is so close to inelegance that she could easily have drawn an inelegant line, given the limitations of her skill; and at w3 the line is inelegant. That is, we have: w1 = <f, f> w2 = <f' , f> w3 = <f'' , f> such that w1Rw2 and w2Rw3 but not w1Rw3.
Consider the proposition R(w) -the content of the agent's skilful accomplishment at w1. The agent made an attempt to bring about this result in w2. Moreover, her attempt successfully brought about the attempted result, since R(w) is true at w2.
(w2 is a member of R(w).) Moreover, the agent's attempt was motivated by her successful recognition of the reason(s) she had to draw an elegant line. But the agent does not count as having skilfully accomplished that result, since it is consistent with all the agent did skilfully that the line was inelegantly drawn, as at w3, which is an R-accessible possibility at w2. We have a successful responsiveness to reasons which does not amount to skilful accomplishment.

VIII. CONCLUSION
We began with two questions. What kinds of contrasts are we drawing when we distinguish aesthetic considerations from other kinds? And what is involved in responding to aesthetic reasons? My answers here have been primarily negative in two respects.
First, I've argued against a certain way of construing the contrast between aesthetic, practical, and theoretical reasons. This held that (i) aesthetic reasons resemble the other categories in being essentially reasons-for, and that (ii) the specific difference that marks them out as aesthetic consists in there being reasons for some distinctively aesthetic attitude, as opposed to acts or beliefs.
I've suggested that this view does not give due weight to the regulative role of aesthetic reasons.
Second, I've argued that skilled practical accomplishment, including aesthetic achievement, should not be analysed as a successful response to reasons.