Kant and Aristotle on the Difficulty of Moral Knowledge:

Lessons from the Doctrine of Virtue*

Sean McAleer (University of Wisconsin – Eau Claire)

(Published June 2005)

PDF Icon

Printer-friendly PDF version
(43 pages, 192Kb)

The most characteristically Kantian of Kant’s objections to moral philosophies other than his own is that they are heteronomous, locating the moral law outside the will. In the Groundwork (1785) and the second Critique (1788) we find him making this objection against Wolff, Hutcheson, Montaigne, Epicurus, the Stoics, and others—but not Aristotle (e.g., KpV 33ff, G 441). In the Doctrine of Virtue (1797), however, Kant singles out Aristotle, but for a different kind of criticism altogether. Here, Kant complains that Aristotle’s doctrine of the mean cannot guide action in a sufficiently determinate way, and, since any adequate theory should do this, Aristotle’s theory should be rejected. In short, Aristotle’s ethics is prey to IAGO: the insufficiently action-guiding objection. My plan in this paper is to probe the theoretical differences underlying the objection, primarily by focusing on their different conceptions of how difficult it is—and should be—to know what morality requires of us, and secondarily on what responses an Aristotelian could make to Kant.

1. Kant v. Aristotle: The insufficiently action-guiding objection (IAGO)

In the Doctrine of Virtue Kant twice considers Aristotle’s doctrine that virtues are means between extremes of excess and deficiency, and twice he rejects it—first because it is false (DV 404), and second because it is useless (DV 432). The two claims are different but not independent: the doctrine is false because it is useless.

In both places, Kant argues that the virtue generosity[1] is properly distinguished from the vices greed and profligacy not by a difference of degree (Graden), as Aristotle asserts in NE IV.1, but by each trait’s having “a distinctive maxim, which necessarily contradicts the maxim of the other” (DV 404). If virtues and vices are not distinguished in terms of their maxims, “one vice would pass over into the (contrarie) opposite vice only through the virtue; and so virtue would be simply a diminished, or rather a vanishing, vice” (DV 432), a consequence Kant finds untenable. I suspect that metaphysical worries are looming here: how can goodness be the mere absence of badness? How could a bad thing change its mode of badness by passing through goodness? But the worries are not merely metaphysical, for Kant finds the virtue-as-vanishing-vice doctrine morally problematic, as well: “the mode of thought which sets down the absence of such vice … as being virtue … deserves to be called a radical perversity in the human heart” (Rel 37 [33]). And there are epistemic issues as well: the worry reflects Kant’s view that the logic of morality is bivalent: if acting on my maxim m1 is morally impermissible, there is one and only one maxim m2 that is the contrary[2] of m1 on which it is permissible for me to act. So every permissible action has but one contrary. On Aristotle’s view, by contrast, every permissible action has two contraries, since every virtue is opposed to two vices. This makes moral reasoning more difficult and complex, for an Aristotelian agent who realizes the impermissibility of the action she’s contemplating cannot determine what action is permissible as easily as a Kantian agent can.[3]

The underlying problem, for Kant, is that Aristotle makes virtue, and thus acting rightly, a matter of hitting the mean, without giving a formula for locating the mean:

the well-known principle (Aristotle’s) which locates virtue in the mean between two vices … does not specify (bestimmt) the degree, although it makes the conformity or nonconformity of conduct with duty depend entirely on it, [so] it cannot serve as a definition (Erklärung) (DV 404).

Since his formula contains “no determinate principles (bestimmte Principien)” (DV 404n), Aristotle cannot specify just where the mean is in a given case. There is no logos or rule for determining it; finding the mean essentially involves a kind of perception that one can acquire only by experience. Kant finds this untenable, because “the most common and unpracticed understanding” could not know “what is to be done” (KpV 36) by appealing to Aristotle’s doctrine of the mean. That Aristotle’s ethics is action-guiding for those who possess the special insight embodied in phronêsis is not enough, for Kant, since ethical theories must be able to guide conduct for any rational agent. Thus, on Kant’s view, Aristotle’s ethics is prey to IAGO, the insufficiently action-guiding objection.

Wielding IAGO against Aristotle is not peculiar to Kant. Sidgwick makes more or less the same complaint:

Nor … does Aristotle bring us much nearer to such knowledge [as will guide action] by telling us that the Good in conduct is to be found somewhere between different kinds of Bad. This at best indicates the whereabouts of virtue: it does not give us a method for finding it. (1907/1981: 376-7)

Gerasimos Santas tells us that IAGO is “a stock modern objection to Aristotelian ethics, as old as Grotius” (1996: 61), and Rosalind Hursthouse reminds us that IAGO is often made against virtue ethics generally: “[a] standard criticism made of virtue ethics is that it does not tell us what we ought and ought not to do” (1995: 57).

We should note that Kant, anachronistically but not (I assume) unfairly, is supposing that Aristotle seeks to give a criterion of right action. In the Lectures on Ethics, he characterizes the doctrine of the mean as a “principle of duty” (LE 654). It seems clear that Aristotle himself did not share the modern assumption, so well put by Mill, that “[i]t is the business of ethics to tell us what are our duties, or by what test we may know them” (1861/1979: 17); he does not see his task as one of providing a “canon of moral appraisal of action” (G 424), which “in the shortest and yet infallible way” (G 403) allows the agent “to distinguish in every case that comes up … what is in conformity with duty or contrary to duty” (G 404). Though some contemporary virtue ethicists who are eliminitivists about deonotological notions such as duty (e.g., Anscombe (1958) and Slote (1992)) would resist the supposition in question, Aristotle’s thought can, without excessive distortion, accommodate the modern concern about criteria for right action. I shall proceed on this assumption for now—but later we will see some reason to question it.

2. Two ways in which it’s hard to be good

To begin assessing Kant’s use of IAGO against Aristotle, let us distinguish two ways in which morality can be difficult. Morality can be executively difficult: it is often hard to do what morality requires of us. Morality can also be epistemically difficult: it is often hard to know what morality requires of us.

Kant’s remarks in the Groundwork indicate that he regards morality as executively difficult: it is hard to be good because our wills are affected by inclinations that draw us away from duty. We are not by nature inclined to do what we ought to do: the human condition is a condition in which “the will is not in itself completely in conformity with reason” (G 412; Kant’s emphasis). We are possessed of “a will that does not straightaway do an action just because it is good” (G 414) because “the will is exposed … to subjective conditions (certain incentives) that are not always in accord with the objective ones” (G 412). A divine will, by contrast, is a will that “by its subjective constitution … can be determined only through the representation of the [objective] good” (G 414). No imperatives hold for such a will, since it does not need to be necessitated to act rightly (G 414; KpV 72).

Whatever its executive difficulties, morality is not—and must not—be epistemically difficult, on Kant’s view, as these passages from the Groundwork and the second Critique indicate:

I do not, therefore, need any penetrating acuteness to see what I have to do in order that my volition be morally good. Inexperienced in the course of the world, incapable of being prepared for whatever might come to pass in it, I ask myself only: can you also will that your maxim become a universal law? (G 403)

[T]he moral law commands compliance from everyone, and indeed the most exact compliance. Appraising what is to be done … must … not be so difficult that the most common and unpracticed understanding should not know how to go about it, even without worldly prudence (KpV 36).

Moreover, when Kant describes the method he plans to deploy in the Groundwork, he tells us that he plans to “proceed analytically from common cognition to the determination of its supreme principle” (G 392); once he has found the principle, he will proceed “synthetically back to the common cognition in which we find it used” (G 392). Thus he is adopting epistemic ease as a methodological constraint guiding his “search for and establishment of the supreme principle of morality” (G 392).

That morality must be epistemically easy is not a thought peculiar to Kant. J. B. Schneewind suggests that it is characteristic of Kant’s philosophic milieu. By the end of the eighteenth century, he argues, a conception of morality as self-governance had replaced morality as obedience (to God’s will). If agents are to be self-governing, they must be able to know what morality demands of them, so this new conception includes an “assumption of prima facie equal moral competence” (1998: 4)—an assumption that still seems to be with us. “All of us, on this view, have an equal ability to see for ourselves what morality calls for” (1998: 4).[4]

Aristotle also regards morality as executively difficult, as the space he devotes to akrasia indicates. But characteristic of his thought is the view that morality is epistemically difficult, as well. Thus, the problem is not merely in our wills but in our (practical) intellects, as the following passage suggests:

[Hitting the mean] is no doubt difficult, and especially in individual cases, since it is not easy to determine (diorisai) both how and with whom and on what provocation and how long one should be angry … But up to what point and to what extent a man must deviate before he becomes blameworthy (psektos) is not easy to determine by reasoning (tôi logôi aphorisai), any more than anything else that is perceived by the senses; such things depend on particular facts, and the decision rests with perception (en têi aisthêsei hê krisis). (NE II.9 1109b14-23; cf. IV.5 1126a31-b4)

For Aristotle, I must have experience and insight if I am to determine what morality requires of me; thus it may be difficult for me to know what my duty is (to put the matter in Kantian terms), since I may lack the requisite experience. I may well know that my action should aim at and express the mean, but finding the mean is difficult and requires knowledge, just as solving a geometric problem does (NE II.9 1109a25). But I cannot acquire moral knowledge as I can acquire geometric knowledge: I can become “one who knows” how to find the midpoint in a circle by attending to the first proposition of Book III of Euclid’s Elements, but moral knowledge cannot “reached through abstraction (di’ aphaireseôs)” (NE VI.8 1142a18) but arises only “from experience (ex empeirias)” (NE VI.8 1142a15). Thus Aristotle would most likely regard Kant’s project in the Groundwork, “the search for and establishment of the supreme principle of morality” (G 392), as hopeless, since there is no logos or rule for determining the mean, and finding the mean involves a kind of perception which can be acquired only by experience.

So Aristotle takes morality to be both epistemically and executively difficult, while Kant regards it as only executively difficult. By contrast, Socrates’ denial of the possibility of akrasia in the Protagoras suggests that he regards morality as epistemically but not executively difficult. Thus Kant’s requirement that morality not be epistemically difficult—grounds IAGO against Aristotle in the Doctrine of Virtue.

3. Kant on the limits of IAGO: the topography of morality

Kant does not demand determinate action-guidance across the board; he requires it only for that part of morality that comprises duties whose nonperformance is blameworthy: perfect duties. The non-performance of such duties is blameworthy (verschuldet) because they are duties whose performance is owed (schuldige). On the other hand, one does not act wrongly in failing to fulfill an imperfect duty (DV 390), and thus is not culpable or blameworthy, since in fulfilling an imperfect duty one goes beyond what morality minimally requires. Though failure to perform an imperfect duty is a transgression (Übertretung), it is blameworthy only when it is based on a settled principle of nonperformance (DV 390). Failing to fulfill a perfect duty is, by contrast, always wrong and blameworthy (DR 223-4), for in failing to fulfill a perfect duty one fails to do what is owed (Schuldigkeit). These and similar passages suggest that Kant thinks of morality as organized into a core comprising duties the non-performance of which is blameworthy (perfect duties), and a periphery comprising duties the performance of which is meritorious but the non-performance of which is not blameworthy (imperfect duties). The core then specifies a moral minimum—the minimal level of decency one must live up to in order to avoid blame.[5]

Morality requires me to adopt maxims containing the morally necessary ends my own perfection and the well-being of others, but does not prescribe what actions I should take to realize them. These ends can be realized by a wide variety of actions and can be promoted to a greater or lesser degree. For example, no principle can specify in aDVance just what actions I should undertake to realize my own perfection: should I improve my grasp of Ancient Greek or become a connoisseur of Beethoven’s Late Quartets (or the Velvet Underground, for that matter) or learn how to juggle? No principle can tell me. Here I have some latitude or playroom (DV 390), and my inclinations can contribute to the determination of action (G 421n). And once I have determined what action to take, no principle can specify the degree to which I should promote my perfection: “[n]o rational principle prescribes specifically (bestimmt ) how far one should go in cultivating one’s capacities” (DV 392). The appropriate degree depends upon how meritorious one wishes to be, and how one wishes to integrate the various ends one has adopted. The same goes for the other morally necessary end, the well-being of others:

I ought to sacrifice a part of my welfare to others without hope of return … [but] it is impossible to assign determinate limits (bestimmte Grenzen) to the extent of this sacrifice. How far it should extend depends, in large part, on what each person’s true needs are in view of his sensibilities, and it must be left to each to decide this for himself… [T]his duty is only a wide one; the duty has in it a latitude for doing more or less, and no specific limits (Gränzen bestimmt) can be assigned to what should be done. The law holds only for maxims, not for determinate actions (bestimmte Handlungen). (DV 393)

Kant requires determinate action-guidance—i.e., that a theory tells me what action to perform or omit—for morality’s core because if “the most common and unpracticed understanding” is to avoid acting wrongly and thus incurring blame, it must know exactly what is required of it. He countenances indeterminate action-guidance for the periphery, by contrast, since failure to heed its demands on any particular occasion is not blameworthy.

An important difference between Kant’s and Aristotle’s tolerating indeterminate action-guidance—beyond the difference already mentioned, that Kant tolerates it only for morality’s periphery—is that for Kant the difficulty of knowing just what action to perform in the periphery (if indeed it is difficult) is not epistemic in nature. The problem is not that I cannot determine what action to perform because no principle can tell me; I am not faced with distinguishing the one right action from the many wrong actions. Rather, the “difficulty” is choosing from a plethora of right actions, any of which will promote the morally necessary end I have adopted. Kant thinks Aristotle should be embarrassed at his theory’s inability to tell “the most common and unpracticed understanding” exactly what morality requires. It would be a mistake to think that Aristotle could turn the tables on Kant, tu quoque fashion, and argue that part of Kant’s ethics—the periphery, admittedly, not the core—is prey to IAGO as well, for the only embarrassment Kant faces in this regard is an embarrassment of riches.[6]

4. A moral (and legal) argument for epistemic ease

I suspect that many of those who accept Aristotle’s view that morality is epistemically difficult do so reluctantly: it’s a shame that it’s so hard to know what morality requires of us, and it would be better if it weren’t so hard, but that’s just the way it is. Bernard Williams expresses such an attitude toward a related matter, that morality is not immune from luck, accepting “the bitter truth (I take it to be both) that morality is subject, after all, to constitutive luck” (1981: 21). But the truth—if it is a truth—that morality is epistemically difficult, would be more than bitter, it would be unfair, since it would be unfair to blame an agent for acting as she did if she were unable to know that morality required her to act differently—just as it would be unjust to blame an agent for acting as she did if she were unable to act differently (even if she knew what morality required of her). Thus morality itself demands that morality be epistemically easy.

A brief excursus into the philosophy of law might prove salutary here. The requirement of epistemic ease is expressed in the jurisprudential void-for-vagueness doctrine (see, e.g., Tribe 1988: 1033ff). In U.S. v. Harriss, for example, the Court held that

The constitutional requirement of definiteness is violated by a criminal statute that fails to give a person of ordinary intelligence fair notice that his contemplated conduct is forbidden by the statute. The underlying principle is that no man shall be held criminally responsible for conduct which he could not reasonably understand to be proscribed. (347 U. S. 612, 617 (1954))

The legal maxim nulla poena sine legem (“no punishment without a law”) begins to justify the demand for epistemic ease but by itself it is not enough, for it requires only the existence of a law, not that the law be known to or even knowable by the agent. Surely it would be unfair to punish an agent for failure to comply with an unpromulgated law—hence classical and contemporary natural law theorists such as Aquinas and Lon Fuller are especially keen to argue for the importance of promulgation (Aquinas 2000: 6 (ST I-II, Q. 90, a. 4); Fuller 1969: 106). But even promulgation is not sufficient for epistemic ease, for a law might be promulgated but be so unclear as to not be understandable and thus not followable—as Fuller’s hapless Rex discovers. Thus Fuller also champions clarity as one of the eight desiderata a legal system must realize to a sufficient degree if it is to fulfill its purpose of “subjecting human conduct to the governance of rules” (1969: 63): an average person of ordinary intelligence must be able to understand what the law requires of her if the law is to guide her conduct.

Our concern is moral rather than legal, and there are some obvious differences between the domains: moral laws are not promulgated in anything like the way positive laws are; and morally (on Kant’s view), we are both legislator and subject, for example. But the points about promulgation and clarity hold, mutatis mutandis, in the moral domain as well, and thus so does the fairness argument: it would be unjust to subject an agent to blame for violating moral requirements that she could not understand. Her norm-violating action would not be justified, since she did not, all things considered, act rightly. But she should be excused from blame, for she could not reasonably be expected to have avoided acting wrongly if she could not have know her action was prohibited. Since the moral law binds every rational being, every rational being, and in particular, every average person of ordinary intelligence, must be capable of understanding what morality requires of her.

I do not know of a place where Kant explicitly makes the fairness argument being contemplated here, but an inchoate version can be found if we join (a) his holding that “justice …constitutes what is essential in this concept [i.e., punishment]” (KpV 37) with (b) his taking a punishment to be just only if “the transgressor knows within himself what his action is worth” (LE 553) and (c) his distinguishing, while discussing the notion of imputation in the Doctrine of Right, between faults and crimes as species of transgression (DR 223-4): a transgression (Übertretung) is any deed that is contrary to duty; an unintentional transgression is merely a fault (Verschuldung), while an intentional transgression (“i.e., one accompanied by consciousness of its being a transgression”) is a crime (Verbrechen).

At the very least, the burden of proof would seem to be on the proponent of morality’s epistemic difficulty. But, as noted above, those who accept morality’s epistemic difficulty do not do so with relish, so they are unlikely to demur. Williams, no friend of Kant or Kantianism, admits that the Kantian conception of morality, in which “[t]he capacity for moral agency is supposedly present to any rational agent whatsoever, to anyone for whom the question can even present itself … has an ultimate form of justice at its heart, and that is its allure” (1981: 21). The picture of epistemic ease is not unattractive, on his view—it’s just false.

5. Internal and external standards

We can better understand why Aristotle and Kant differ over morality’s epistemic difficulty by exploring a telling difference between their normative standards. Let us suppose, with Kant, that Aristotle seeks to give a criterion of right action. Where Kant would offer something like the following:

A’s φing in circumstances c is right iff A’s maxim of φing in c can be consistently universalized,[7]

Mill would offer something like:

A’s φing in circumstances c is right iff A’s φing in c maximizes total happiness,

and Aristotle would offer something like:

A’s φing in circumstances c is right iff a fully virtuous person would characteristically φ in c.

It might be thought that both (K) and (M) could be recast as versions of (A) and thus that Kant and Mill could also be thought of as virtue ethicists. Their accounts of the virtuous person would, of course, differ from Aristotle’s: for Kant, the virtuous person is a person with settled strength in doing her duty (e.g., DV 380, 383-4, 390, 395, 390, 405; KpV 128); for Mill the virtuous person is a person who is reliably disposed to maximize total happiness. This line of thought fails, however, because the reference to the virtuous person in (A) is ineliminable, but this is not the case be in the recast formulations of (K) and (M). One can state Kant’s and Mill’s theories of rightness without recourse to the notion of the virtuous person, but one cannot do this for Aristotle. This ineliminability helps explain why Aristotle and Kant differ over whether morality is epistemically difficult.

It might be objected that Aristotle’s reference to the virtuous person is also eliminable, for if we analyze the notion of the virtuous person, we find that her actions and emotions express and hit a mean between excess and deficiency, yielding something like the following:

A’s φing in circumstances c is right iff
A’s φing expresses the mean appropriate to c

While it is true that (AMEAN) makes no explicit reference to the virtuous agent, it does so implicitly, since Aristotle believes that there is no agent-independent rule or principle that can determine where the mean is (NE II.9 1109b14-23, IV.5 1126a31-b4). The mean is where the virtuous person, who possesses the intellectual virtue of phronêsis, determines it to be. Since we cannot determine where the mean lies independently of the virtuous person, Aristotle’s ethics—or, rather, the version I have anachronistically foisted upon him—does make ineliminable reference to the virtuous person.

I call standards such as Aristotle’s internal standards, to be contrasted with the external standards favored by Kant and Mill. Perhaps the best way to explain the distinction is by analogy. The usual standard for deductive validity is

An argument is deductively valid iff its conclusion must be true if its premises are true.

This is an external standard because applying it requires no special expertise, insight or judgment; one need only understand the terms in which an external standard is couched to fruitfully apply it. Such standards make no reference to persons in privileged epistemic positions—or if there are any such references, they are in principle eliminable. Of course, one may “internalize” an external standard and not consciously appeal to it while applying it, but in principle one need not have internalized an external standard to apply fruitfully it—here, to decide about the validity of the argument in question. The main point about externality is that no special insight is required to fruitfully apply the standard. Nor need one care about what the standard measures: possessing the standard, I can determine whether an argument is valid, regardless of whether I care about validity.

Contrast this external standard of validity with this internal, quasi-Humean aesthetic standard:

x is beautiful iff a true judge judges x to be beautiful.[8]

This standard can be fruitfully applied only if one is—or has access to—a true judge; merely understanding the terms in which (H) is couched is not sufficient. I can come to understand what Hume means by ‘a true judge’ by grasping his five criteria for being a true judge: delicacy of taste (perceptual and aesthetic acuity); practice (experience) in the genre in question; comparison with other genres; a mind free from prejudice; good sense (especially with regard to appreciate and judge teleologically). But understanding what a true judge is won’t allow me to fruitfully apply (H) unless I am a true judge or have ready access to one. And while the criteria indicate what steps I might take to become a true judge, there is no guarantee that I will become one, despite my best efforts and my ardent wishes. Suppose I am resolutely disinterested and free from prejudice; I read, listen, and look night and day, becoming practiced and engaging in comparison. My aesthetic and perceptual delicacy improve, but they never rise to the needed level: I can’t distinguish Miles Davis from Chet Baker; I can’t see what’s going on in the “Wandering Rocks” episode of Ulysses; I don’t really notice the camera-work in the labyrinthine entrance to the Copacabana in Goodfellas; I’m completely in the dark when it comes to Caravaggio. And as for good sense, I’m a bit of a dunce, and that just isn’t going to change. So—like many—I’m not competent to make aesthetic judgments. We are not all equally competent with respect to aesthetic knowledge, and none of us is competent to make aesthetic judgments merely in virtue of our being creatures with the capacity to be moved by beauty, let alone in virtue of our being rational creatures. Since fruitfully applying (H) requires a special kind of expertise that I lack and cannot acquire merely by my own efforts, (H) is an internal standard.

I should say at once that the distinction between internal and external standards is not a distinction between standards that are synthetically true and standards that are analytically true. If this were the case—if all standards expressing synthetic truths were internal and all standards expressing analytic truths were external—then my claim that Kant rejects internal moral standards (because of their epistemic difficulty) could be turned back on Kant, who would then be forced to reject his own preferred standard—the categorical imperative—as internal because it is synthetic (since it is synthetic a priori).[9]

A moral standard is external if any agent who understands the language in which it is written can fruitfully apply it, regardless of her experience and perspective. One merely applies a standard accessible to any agent; no special cognitive abilities are required. Nor do external standards require any special affective capacities to be fruitfully applied: I can learn how to find the center of a circle without caring about circles, and—if moral standards are external, I need not care about the difference between right and wrong to know what the difference is. A moral standard is internal, on the other hand, if applying it requires that its practitioner possess special insight or be in an epistemically privileged position that is not open to all agents in virtue of their agency. And while internal standards need not require affective structures such as care for their fruitful application, but they can readily accommodate them. For Aristotle, knowing where the mean is—and hence what “what one ought to do (ho dei poiêsai)” (NE VI.9 1142b23)—requires phronêsis. Not only can one not acquire phronêsis as one can acquire other kinds of expertise, by mastering abstract universal propositions or codifiable techniques, one cannot acquire and possess phronêsis unless, Gilbert Ryle argues, one care about the good.

Aristotle says that phronêsis involves more than reason, because it cannot be forgotten (NE VI.5 1140b28-30), and Ryle argues that this is so because phronêsis necessarily involves caring about the difference, and “ceasing to care is not forgetting” (1958: 156). Standards that can yield knowledge only for those practitioners who are in certain affective states (e.g., only for those who care) are internal standards. But, it might be objected, Kant himself specifies four “subjective conditions of receptiveness to the concept of duty” (DV 399; Kant’s italics): moral feeling, conscience, love of one’s neighbor, and self-respect. These are “predispositions of the mind for being affected by concepts of duty, antecedent predispositions on the side of feeling (ästhetisch)” (DV 399; Kant’s italics). If standards requiring affectivity are internal, then it appears that Kant’s standard is internal after all, and we should then wonder not so much at the possibility of the (in)famous grief-stricken former philanthropist’s being motivated by the mere thought of his duty, but rather at the epistemic question of how the philanthropist can come to know that it is his duty to “benefit others in distress” since “their troubles did not move him” (G 398). The idea that moral knowledge requires any affective structures seems quite unKantian, but it seems that Kant himself holds this view. Kant would no doubt respond that the four specified modes of affectivity are moral or practical rather than pathological feelings (DV 399; cf. G 400, KpV 75f), the inevitable effects of awareness of the moral law. But the doctrine of feelings of non-empirical origin is one that even many Kantians find elusive and troublesome. Kant would do better, perhaps, to reply that the specified modes of affectivity are only “subjective conditions of receptiveness to the concept of duty, not … objective conditions of morality” (DV 399; Kant’s italics): they “yiel[d] no cognition” (DV 400). Rather, “it is by virtue of them that [a person] can be put under obligation” (DV 399). In other words, they are required not for moral knowledge, but for moral knowledge to have a practical effect on an agent—much as experience-sharpened judgment yields no moral cognition but provides moral principles with access (Eingang) to the will (G 389; cf. KpV 86, 151). Moreover, “every human being has them” (DV 399), and a person who lacked them would be “morally dead” (DV 400), ceasing to be an agent at all: her “humanity would dissolve (by chemical laws, as it were) into mere animality” (DV 400). So Kant need not worry that the four conditions make his standard internal.

So, not all internal standards require affectivity, and not all standards requiring affectivity are internal. It is the nature of the affectivity in question that determines whether the standard is internal or external. If our affective states were under our direct voluntary control and the required state were one that any competent agent could enter into at will (possible, but implausible)—or if the required state were one that any competent agent were in by virtue of her agency (Kant’s view)—then a standard with an affective requirement would not thereby be internal.

Another way in which Aristotle’s standard is internal is in its requiring its competent practitioner—the phronimos—be in a certain moral state, as well. Aristotle holds that “it is impossible to be practically wise without being good” (NE VI.12 1144a36-b1), since the proper end to promote by action “is not evident except to the good man; for wickedness perverts us and causes us to be deceived” (NE VI.12 1144a34-6). This implies that a bad person cannot know what she has to do to become good; she may possess cleverness (deinotêta), a capacity to promote any end she might choose (a discretionary end, in Kantian terms) (NE VI.12 1144a23-6), but lacking virtue she cannot (reliably) choose the proper end—not because her badness impairs her volitionally (though surely it does), but because it impairs her cognitively. A naturally virtuous person lacks phronêsis and thus the virtues proper, so she occasionally gets things wrong; she seems akin to the Kantian agent motivated by honor (a simulacrum of duty—see Korsgaard 1996: 56), who, if she “lights upon what is in fact … in conformity with duty” does so only “fortunately” (glücklicherweise) (G 398), as a matter of luck. Kantian honor, like Aristotelian natural virtue “can lead only contingently (zufällig) to what is good and can very often also lead to what is bad” (G 411). The person who possesses virtue proper (kuriôs arête) does not act merely according to right reason (kata ton orthon logon); her actions express the presence of right reason (meta tou orthou logou) (NE VI.13 1144b26-7)—just as the good-willed person acts not just in conformity with duty (pflichtmäßig) but from duty (aus Pflicht).

Aristotelian moral knowledge requires not only (a) a kind of expertise not possessed by all agents in virtue of their mere agency (i.e., phronêsis) and (b) a certain affective state (i.e., caring about the good) that qua affective state is not under one’s direct voluntary control, it also requires (c) moral goodness. Thus Aristotle’s standard is internal because it cannot be fruitfully applied by every agent who understands its terms. Kant’s standard, by contrast, is external, as we have seen: it requires no special insight or experience for its application. So, to put the pieces together: Kant’s objection that Aristotle’s ethics is prey to IAGO is grounded in his requirement that morality not be epistemically difficult. This requirement fuels the demand for external standards, since internal standards presuppose insight and expertise and thus by their nature render morality epistemically difficult for those who lack the experience, expertise and insight—i.e., for those who lack phronêsis.

6. Explaining the disagreement

While the distinction between internal and external standards should help to clarify Kant’s objection to Aristotle, that Kant embraces an external standard and Aristotle embraces an internal standard is not a conclusion that is likely to surprise careful readers of Kant and Aristotle. What deeper theoretical differences explain their disagreement over whether morality is epistemically difficult? Obviously, full discussions of such disagreements would merit a monograph, so here I limit myself to raising two such issues and bringing out their contours as they are relevant to Kant’s wielding IAGO against Aristotle: the difference between their theories in terms of conceptual priority (with some remarks about a priori moral knowledge), and their competing conceptions of moral value (value as essentially attractive v. value essentially imperative).

1. The first concerns the structures of their theories. Preferring the aDVantages of theft over honest toil, I shall avail myself of the helpful taxonomy of moral theories that Gary Watson offers in “On the Primacy of Character” (1990). Watson groups theories by the nature of the moral facts they take to be basic. Ethics of requirement take the basic moral facts to be facts about what actions are required—they take rightness to be prior to (i.e., explanatory of) goodness and virtue; Kant’s ethics would be the paradigmatic ethics of requirement.[10] Ethics of outcome take the basic moral facts to be facts about the intrinsic values of possible outcomes of actions or traits—they take goodness to be prior to (i.e., explanatory of) rightness and virtue; Mill’s utilitarianism is paradigmatically an ethics of outcome. Ethics of virtue take the basic moral facts to be facts about the virtues—they take virtue to be prior to (i.e., explanatory of) rightness and goodness; Aristotle’s ethics is paradigmatically an ethics of virtue.

It is very tempting to think that Kant will find any theory that fails to prioritize rightness over goodness or virtue to be epistemically difficult and thus prey to IAGO. This will be true of an ethics of outcome such as Mill’s; if we have no sure prudential knowledge, as Kant thinks we do not, we will have no sure moral knowledge, either. Happiness, for Kant, is “a rational being’s consciousness of the agreeableness of life uninterruptedly accompanying his whole existence” (KpV 22), which is reasonably close to Mill’s: “the ultimate end … is an existence exempt as far as possible from pain, and as rich as possible in enjoyments, both in point of quantity and quality” (1861/1979: 11). The problem is that “one can form no determinate and sure concept of the sum of satisfaction of all inclinations under the name of happiness” (G 399); “the concept of happiness is such an indeterminate concept that … it is impossible for the most insightful and at the same time most powerful but still finite being to frame for himself a determinate concept of what he really wills” (G 418) when he wills to be happy. Now if I cannot frame a universal rule for my own happiness, since “…the exceptions one is warranted in making upon occasion [if we take the desire for happiness to be a universal practical law] are endless and cannot be determinately embraced in a universal rule (nicht bestimmt in eine allgemeine Regel befaßt)” (KpV 28), what hope would any attempt to frame an external standard for an ethics of outcome have? I may know quite well whether this will make me happy now, but “if this [inquiry] is extended to the whole of one’s existence, [it] is always veiled in impenetrable obscurity” (KpV 36). Presumably, even graver difficulties affect me when I try to frame an epistemically easy moral standard aiming at universal happiness than affect me when I try to frame an epistemically easy prudential standard aiming at my own happiness spread over my entire life.

In Kant’s view, any ethics that makes goodness basic will be epistemically difficult and prey to IAGO, and he argues to this effect while discussing “the paradox of method” in the second Critique (KpV 62ff). Suppose goodness is basic. Since there is no a priori standard for what is good, the standard must be empirical, e.g., that x is good if x pleases. Any moral laws derived from an empirical account of the good would be a posteriori, not a priori, and thus could not bind with absolute necessity. But since moral laws must bind absolutely, there can be no moral laws if goodness is prior to rightness. The point may be even clearer when we focus on the ‘outcome’ in ‘ethics of outcome’. Since I can know that x is a causal outcome of y only empirically, there can be no a priori moral knowledge if facts about outcomes are the most basic moral facts.

Would the same go for ethics of virtue? Kant would presumably think so, but we should remember that Aristotle’s account of happiness as “activity of the soul in conformity with excellence” (NE I.7 1098a16-7) is the conclusion of an a priori argument. The function argument does not draw on any facts of empirical psychology; instead it is based on the barest metaphysical picture of the kind of creature we are—much as Kant often appeals to our nature as finite rational beings with needs when determining what our duties are (e.g., DV 454; cf. KpV 32, 61). But, Kant will counter, it is true that “a metaphysics of morals cannot dispense with principles of application, and we shall often have to take as our object the particular nature of human beings” (DR 216-7) if we hope to know what morality requires of us—but here we are drawing on the barest metaphysical picture of the kind of creature we are, not empirical psychological facts about our desires. Moreover, Kant would say, while we must take this picture into account when applying the moral law, we cannot take such constitutions into account when deriving moral laws—any more than we can take psychological facts about what beings like us in fact desire—for such laws would not bind all rational beings, but only beings of the kind in question.

So, perhaps unsurprisingly, Kant’s insistence that morality be epistemically easy is driven at a deeper level by his view that moral knowledge is a priori. Obviously, this is not the space for a full discussion of the nature of a priori knowledge generally or a priori moral knowledge specifically or Kant’s own version of it in particular. But we might wonder, in light of some philosophical developments of the recent past, at Kant’s view that moral knowledge must be a priori. Kant argues that moral laws must hold necessarily—indeed, they must hold “with absolute necessity(G 408). Moral laws “hold as laws only insofar as they can be seen to have an a priori basis and to be necessary” (DR 215). Perhaps Kant is correct moral laws must be seen to be necessary to be—to function as—moral laws. It is plausible that I regard a law requiring that I φ as necessary if I am to regard φing as really necessary. So perhaps there is an implicit commitment to S4 as the appropriate logic for the modality of morality: a law that necessitates my φing must itself be necessary, so if it’s necessary that I φ, it follows that it’s necessarily necessary that I φ. This seems plausible. A person who embraces cultural relativism would seem to have a less robust sense of the necessity of her φing than would a Kantian, since the relativist, upon reflection, would not regard the moral law necessitating that she φ as itself necessary—were she in a different culture, it might not be necessary that she φ, so she might well think, “I don’t really have to φ” since it’s not necessary that it’s necessary that she φ.

Kant takes necessity and universality to be “secure indications” of a priori knowledge (KrV B3), but in Naming and Necessity (1972), Saul Kripke at least called into question, and at best refuted, the connection between necessity and a prioricity. Necessity is a metaphysical concept; a prioricity is an epistemic concept. They are thus conceptually distinct and perhaps really distinct as well. On his view, the truth that water = H2O is a necessary truth, but it is obviously an a posteriori truth. Kant’s demand for morality’s epistemic ease is fueled by his thinking that moral laws must be knowable a priori. But if he is mistaken to think that they must be a priori because they are necessary, then the this support for requiring epistemic ease—and the complaint against Aristotle—are unfounded.

2. A second explanatory difference is the difference between Aristotle’s and Kant’s competing conceptions of the nature of moral value. Sidgwick (1907/1981), followed here by Charles Larmore (1990), distinguishes between imperative and attractive conceptions of moral value. Imperative conceptions take moral value primarily to be “a dictate or imperative of reason, which prescribes certain actions either unconditionally, or with reference to some ulterior end” (1907/1981: 105); such conceptions primarily ask, “What is Duty and what is its ground?” (1907/1981: 106). In attractive conceptions “this notion of rule or dictate is … only latent or implicit” (1907/1981: 105); the primary question of such conceptions is “Which of the objects that men think good is truly Good or the Highest Good?” (1907/1981: 106) or “What is the relation of the kind of Good we call Virtue, the qualities of conduct and character which men commend and admire, to other good things?” (1907/1981: 106). The attractive conception is “the fundamental conception in the Greek schools of Moral Philosophy generally” (1907/1981: 105), while modern ethical theories take moral value to be essentially imperative, as “the quasi-jural notions of modern ethics” attest (1907/1981: 106).

To claim that the pre-eminent value for modern ethics is imperative is not to claim that modern ethical theories take rightness to be prior to goodness in the explanatory sense discussed above; it is rather to claim that modern ethical theories take themselves to be articulating and defending criteria of right action. Mill, for example, takes goodness to be prior to rightness in that he explains what rightness is in terms of goodness, but he thinks that “[i]t is the business of ethics to tell us what are our duties, or by what test we may know them” (1861/1979: 17).

Aristotle’s focus on eudaimonia and his attempt to articulate a conception of the sort life a good person would want to live suggests that he regards moral value as essentially attractive. It should come as no surprise that Kant has an imperative conception moral value, given his focus on duty and given his conception of duty: “[t]he very concept of duty is already the concept of a necessitation (constraint) (Nöthigung (Zwang)) of free choice through the law” (DV 379; Kant’s italics). His beginning the Groundwork by investigating the good will tells us that he is concerned with the kind of life a person should be living, but as the Groundwork is “nothing more than the search for an establishment of the supreme principle of morality(G 392), and this supreme principle is the categorical imperative, there can be no real doubt that Kant has an imperative conception of moral value.

The difference between the attractive and imperative conceptions comes out quite clearly when we compare Aristotle and Kant on what motive gives an action moral worth. For Kant, of course, the worth-bestowing motive is duty or rightness itself: an action has moral worth only if it is done from duty—that is to say, when the agent takes the fact that she ought to φ to be sufficient reason to φ (G 397ff). Duty, and the imperative conception of moral value more generally, require submission (Unterwerfung) (KpV 86). No such submission is involved with Aristotle’s preferred motive, the fine or noble (to kalon) (NE II.3 1104b30-2): the virtuous person φs because φing is fine or—what comes down to the same thing—she chooses to φ for its own sake (di’ auta) (NE II.4 1105a31-2), and not because it is pleasurable or expedient to do so, or because she wants to avoid shame or harm. Indeed, Aristotle contrasts φing because φing is fine and φing because it is required: “we must be brave because it is fine, not because we are required (dei d’ ou di’ anagkên andreion einai, all’ hoti kalon)” (NE III.8 1116b2-3). Here ‘dei’ seems to signal an attractive, rather than an imperative, ‘must’: the reason we should be brave is not because we are compelled to do so, but because if we really understand what eudaimonia is and that acting bravely is a constituent of eudaimonia, we will see the fineness of acting bravely and thus will want to act bravely. Thus there is an aesthetic dimension to moral value to which Aristotle is sensitive: “to kalon shines through, whenever someone bears many severe misfortunes with good temper” (NE I.10 1110b30-2). Kant does not suffer from aspect-blindness here, for he remarks on the good will’s sublimity (G 425, 439, 442) and its shining like a jewel (G 394). But the aesthetic aspect is not especially prominent and is certainly not the essence of moral value. For Aristotle, by contrast, morally correct action is something essentially fine, noble or beautiful; the good person is attracted by the value such action expresses.

These competing conceptions help explain the disagreement over morality’s epistemic difficulty. It is no surprise that moral philosophers who do not regard their central question as “what is duty and what is its ground?” might offer theories that do not generate crisp, precise action-guidance. Sidgwick argues that

the chief characteristics of ancient ethical controversy as distinguished from modern may be traced to the employment of a generic notion instead of a specific one in expressing the common moral judgments on actions. (1907/1981: 105)

“φ if a virtuous person would φ” guides action, but given how embedded the notion of a virtuous person is in a broad, general conception of a good life, it is more general, and thus guides conduct less precisely, than does “φ if φ-ing is universalizable.”

In addition, because the notion of beauty is more happily a degree-concept than is the notion of authoritative requirement, philosophers who regard moral value as essentially attractive are likelier to regard morality as epistemically difficult than are philosophers who regard moral value as essentially imperative. As we have seen, Kant countenances less precise action-guidance in morality’s periphery than he is willing to countenance in its core, because the periphery comprises those duties to promote the morally necessary ends my perfection and the well-being of others, both of which can be satisfied to a greater or lesser degree: “No rational principle prescribes specifically how far one should go in cultivating one’s capacities” (DV 392). What talents to cultivate, the degree to which one cultivates them, and the actions one performs to cultivate them are “very much a matter for [one] to decide as [one] chooses” (DV 392). Again, the less determinate action-guidance is not merely that morality requires me to adopt an end rather than to perform a certain kind of action. For Kant any embarrassment in morality’s periphery is an embarrassment of riches: there are so many permissible actions that will promote the required end that an agent who genuinely has the proper end will be unlikely to go wrong (and if she does, she has only failed to be good; she has not acted wrongly). For Aristotle, by contrast, “it is possible to fail (hamartanein) in many ways … while to succeed (katorthoun) is possible only in one way” (NE II.6 1106b28-31). The parameters of “getting it right” are many: being angry is easy, but being angry at the right person, to the right degree, at the right time, for the right reason, in the right way is not easy, and it is hard to hit them all because not everyone, but only one who knows (tou eidotou) (NE II.9 1109a24-30). I can try to narrow the range of failure by avoiding the excess or deficiency, but this is only “second best” and opting for “the least of the evils” is still not opting for the good.

Philosophers who regard moral value as essentially attractive are more prey to IAGO because their central task is not specifying our duties but rather as specifying a conception of how to live. When an attractive conception of value is coupled with a medial theory of virtue and right action, the epistemic difficulty of morality follows as a practical (though not perhaps a conceptual) consequence. The Stoics seem to have an attractive conception of moral value—despite how unattractive many might find the Stoic sage!—but, Kant points out, “its disciples spoke of duties and even determined them quite well” (KpV 127n), presumably because they did not have a medial theory of virtue and they did not (as their Epicurean counterparts did) subordinate morality to happiness. Determining whether the Stoics are a genuine counter-example to my suggestion that theories with attractive conceptions of moral value are especially prey to IAGO would take us too far afield, but I will point out that there is some question about whether moral value is essentially attractive or imperative for the Stoics. Sidgwick implicitly concedes the seriousness of the question when he insists that “even the Stoics” had an attractive conception of moral value (1907/1981: 105), though he did not feel the need to explicitly include other schools of Greek ethics (e.g., the Epicureans, the Cynics).

Kant’s and Aristotle’s disagreement over conceptions of moral value helps to explain their disagreement over whether morality is epistemically difficult. Their holding fundamentally different conceptions of moral value, in turn, is explained by their competing pictures of human nature—of what human beings are capable of morally. Kant thinks that “a heart which is happy in the performance of its duty (not merely complacent in the recognition thereof) is a mark of genuineness in the virtuous disposition” (Rel 23n [19n]; Kant’s italics), but he also thinks that “human morality in its highest stage can still be nothing more than virtue” (DV 383); morality for us is always a struggle and a matter of progress—indeed, an endless progress (KpV 33, 83, 122). Kantian virtue is Aristotelian enkratia or continence. Aristotelian virtue is Kantian holiness, and holiness is not attainable for human beings, on Kant’s view. A continent person has a contrary-to-reason inclination, but the inclination “obeys reason (peitharchei tôi logoi)” (NE I.13 1107b26-7), so the continent person acts rightly after a struggle (the inconintent person acts wrongly after a struggle). A virtuous person, by contrast, has no contrary-to-reason inclinations; in her, inclination “speaks, on all matters, with the same voice (homophônei) as reason” (NE I.13 1107b27-8). Kant will have none of it. We might wish to be free of “the vice-breeding inclinations” (DV 376), which are the primary obstacles to virtue (DV 394)—indeed, “it must … be the universal wish of every rational being to be altogether free from them” (G 428; cf. KpV 118)—but we have them in virtue of our finite, needy nature.

Given this guardedly pessimistic picture of what’s morally possible for human beings, Kant proceeds to give a moral argument against attractive conceptions of moral value:

The disposition incumbent upon him to have in observing [the moral law] is to do so from duty, not from voluntary liking nor even from an endeavor he undertakes unbidden, gladly and of his own accord; … his proper moral condition, in which he can always be, is virtue, that is, moral disposition in conflict, and not holiness in the supposed possession of a completely purity of dispositions of the will. By exhortation to actions as noble, sublime, and magnanimous, minds are attuned to nothing but moral enthusiasm and exaggerated self-conceit; by such exhortations they are led into the delusion that it is not duty—that is, respect for that law whose yoke (though it is a mild one because reason itself imposes it on us) they must bear, even if reluctantly—which constitutes the determining ground of their actions… [T]hey produce in this way a frivolous, high-flown, fantastic cast of mind, flattering themselves with a spontaneous goodness of heart that needs neither spur nor bridle and for which not even a command is necessary… (KpV 84-5)

Kant’s argument, then, is that my commitment to attractive conceptions of moral value (or at least to an Aristotelian conception of virtue) results in my thinking that I can bring my inclinations into harmony with reason, that I can be sufficiently attracted by the carrot of beautiful moral goodness to no longer need the stick of duty to constrain me to right action—but this involves a dangerous self-deception, a radical lack of humility, a deep failure to know myself as a finite creature with needs.

7. Aristotelian responses to Kant

There are several ways in which an Aristotelian might respond to Kant’s attack; I will consider three. First, she might counter that Kant mistakenly takes the doctrine of the mean to be a doctrine of moderation. Kant takes “one ought not to do too much or too little of anything” (DV 433n) to represent Aristotle’s position, and this statement expresses a doctrine of moderation, not a doctrine of the mean. According to the former, if hooligans torture my cat, murder my family, and burn down my house, I should be moderately angry—the same moderate anger I should feel and act on if a salesperson treats me with minor disrespect. Not only is this view implausible, it is not Aristotle’s: the doctrine of the mean is not, J. O. Urmson reminds us, a doctrine of moderation (1973: 162; 1988: 28). Aristotle does not bid us to be moderately angry but to be appropriately angry:

…the middle state is praiseworthy (epainetê)—[it is] that in virtue of which we are angry with the right people, at the right things, in the right way, and so on. (NE IV.5 1126b5-6)

Kant’s taking the doctrine of the mean to be a doctrine of moderation is in a way puzzling. His objection is that Aristotle’s ethics is not action-guiding, but the doctrine of moderation is action-guiding—it gives bad guidance, but guidance nonetheless.

A second response is to deny that Aristotle really proposes the doctrine of the mean as an account of virtue. Rosalind Hursthouse acknowledges that Aristotle uses the doctrine of the mean in Book Two of the Ethics, when he defines character virtue and sketches the various character virtues. But when he gets down to a detailed discussion of the various virtues in the later books, he largely eschews the mean in favor of the concept of getting things right (dei):

What he usually does is operate … using the concept dei—right object, right occasion, right reason, or the very general hôs dei, as one should, without any suggestion that this concept can either be captured by, or can necessarily generate, concepts of too much or too little. (1981: 109)

To think that courage, for example, is a virtue because it is a disposition lying in a mean is to get the order of explanation wrong, she claims. Courage is a virtue because it is having the right disposition with respect to fear—the courageous person fears the right objects, in the right ways—and cowardice and rashness are vices because they are wrong dispositions. It happens that these wrong dispositions can be excessive and deficient; but to focus on this is to distort Aristotle’s theory. If Hursthouse is correct, Kant’s criticisms are aimed at the wrong target.

A third response is that Aristotle does not distinguish virtues from vices merely in terms of degree, as Kant claims, but by the agent’s reasons. For example, when distinguishing vulgar from magnificent persons in NE IV.2, Aristotle says that that the vulgar exceed the mean by spending more than is right (para to deon) (1123a19-20). But the vulgar “do not go to excess in the amount spent on right objects, but by showing expenditure in the wrong circumstances and in the wrong manner” (1122a31-3). The magnificent person “will spend such [large and fitting] sums for the sake of the noble (to kalon), for this is common to the excellences” (1122b6-7), while the vulgar person does not spend “for the sake of the noble but to show off his wealth, and because he thinks he is admired for such things” (1123a24-6).

So what really distinguishes the magnificent and vulgar persons are their different reasons for spending—in Kantian terms, their maxims. Aristotle makes a similar point when distinguishing the various simulacra of courage from genuine courage in NE III.8. It is not because they fail to stand firm against danger that possessors of “political courage” (andreia politikê) fail to be virtuous, for they do stand firm. They lack genuine courage because they stand firm for the wrong reason: they “face dangers because of the penalties imposed by the laws and the reproaches they would otherwise incur” (1116a18-9), not because standing firm is fine. Similarly, those who stand firm against dangers because of anger or a desire for revenge fail to be genuinely courageous since “they do not act for the sake of the noble or as reason directs, but from feeling” (1117a8-9). So it seems that Aristotle does not run afoul of Kant’s objection that he ignores the reasons on which the virtuous and vicious agents act.

The heart of Kant’s objection to Aristotle is embodied in his question, “who will specify for me this mean between the two extremes?” (DV 404n). Kant thinks that no answer can given and says, with an air of triumphalism, that those who think “we should go back to the ancients (Aristotle) … have made a bad choice in turning to [this] oracle” (DV 433n). Kant seems to regard his question as rhetorical, but he should not, for Aristotle has an answer: the phronimos (the possessor of phronêsis).

The issue here is not that Kant thinks moral standards must be rational while Aristotle does not, for Aristotle no less than Kant thinks that rightness is a rational matter: “the intermediate is determined by the dictates of reason (hôs ho logos ho orthos legei)” (NE VI.1 1138b19-20); virtue is “as right reason dictates (hôs ho orthos logos)” (NE VI.1 1138b29). Aristotle stakes out a rationalist position early in the Ethics: we should, he says, “desire and act in accordance with a rational principle (kata logon)” (NE I.3 1095a10-11), and “we must act according to right reason (kata ton orthon logon)” (NE II.2 1103b31). The brave person, he tells us, “will fear them [i.e., “the things that are not beyond human strength”] as he ought and as reason directs (hos ho logos)” (NE III.7 1115b11-2); “the temperate man craves for the things he ought, as he ought, and when he ought; and this is what reason (ho logos) directs” (NE III.12 1119b16-7); “the good-tempered man tends to … be angry in the manner, at the things, and for the length of time, that reason dictates (ho logos taxêi)” (NE IV.5 1125b34-1126a1).

So Kant’s complaint must be that Aristotle’s is the wrong kind of rationality. For Aristotle, correct reasoning about moral matters is not a matter of applying universal laws, as it is for Kant. When discussing justice, Aristotle says that “all law is universal (katholou) but about some things it is not possible to make a universal statement which will be correct (orthôs)” (NE V.10 1137b13-5). Moreover,

all things are not determined by law (mê panta kata nomon)… about some things it is impossible to lay down a law, so that a decree (psêphismatos) is needed. For when the thing is indefinite the rule (kanôn) also is indefinite, like the lead rule used in making the Lesbian molding; the rule adapts itself to the shape of the stone and is not rigid, and so too is the decree (psêphisma) adapted to the facts (pros ta pragmata). (NE V.10 1137b27-32)

Kant objects that Aristotle’s ethics is prey to IAGO because it relies on a kind of rationality—namely, phronêsis—that renders morality epistemically difficult. It might be responded that this cannot be Kant’s objection, because Kant himself mentions the need for “judgment sharpened by experience” (G 389); at most, the difference between Kant and Aristotle is merely a matter of emphasis. We should tread carefully here. Kant does think that the moral laws which moral philosophy gives to rational beings a priori

no doubt still require a judgment sharpened by experience, partly to distinguish in what cases they are applicable and partly to provide them with access to the will of the human being and efficacy for his fulfillment of them. (G 389)

Experience-sharpened judgment is a perceptual capacity that enables the agent to distinguish the morally salient features of a situation or to recognize a situation itself as morally charged, perhaps along the lines of Barbara Herman’s rules of moral salience (RMS), which “constitute the structure of moral sensitivity” (1993: 78) but “do not themselves generate duties” (1993: 79). Rather, “the function of the RMS is to guide the normal moral agent to the perception and description of the morally relevant features of his circumstances of action” (1993: 78), helping the agent to properly formulate her maxim and to “structure an agent’s perception of his situation so that what he perceives is a world with moral features” (1993: 77). Thus the protagonist of Kant’s false-promising example “still has enough conscience (Gewissen) to ask himself: is it not forbidden and contrary to duty to help oneself out of need in such a way [i.e., by a false promise]?” (G 422).

While both Kantian and Aristotelian moral judgment is discriminative, helping the agent perceive what is morally salient, only the latter—phronêsis—is verdictive, determining whether a course of action is right or wrong. Moreover, while phronêsis involves deliberation, it is primarily a perceptual faculty, so moral judgment is primarily perceptual rather than deductive:

up to what point and to what extent a man must deviate [from the mean] before he becomes blameworthy it is not easy to determine by reasoning (ouk rhaidion tôi logoi aphorisai), any more than anything else that is perceived by the senses; such things depend on particular facts, and the decision rests with perception (en têi aisthêsei hê krisis). (NE II.9 1109b20-3; cf. IV.5 1126b1-4)

This not to deny a place for general rules or principles in Aristotle’s account of moral reasoning: we need not—and should not—regard him as a radical particularist who eschews any use of moral principles, for Aristotle thinks there are general, true moral principles. For example, if I can benefit a friend or repay a loan, but not both, I should repay the loan (NE IX.2 1165a2-4). But such a principle is true only for the most part: it is true hôs epi to polu. If benefiting the friend at the expense of doing what justice strictly requires “is exceedingly noble or exceedingly necessary” (NE IX.2 1165a3-4), then I should benefit my friend. Though some actions are always wrong—e.g., adultery: there is no right way, partner, time, etc., to commit adultery (NE II.6 1107a9-17)—for the most part we shall have to be content with general rather than universal rules and principles.

8. Conclusion

At the very least, Kant is mistaken to regard his putatively rhetorical question “who will specify for me this mean between the two extremes?” (DV 404n) as either rhetorical or as decisive against Aristotle’s ethics, for Aristotle has an answer ready-to-hand: the phronimos, appealing to an internal standard, will specify the mean. This isn’t an answer Kant has much use for, and indeed against which he more than one moral argument. But perhaps moral considerations in favor of a moral theory should carry less weight than factual considerations about the nature of our moral experience. If Aristotle is correct that moral reality does not easily yield moral knowledge, then Kant’s complaint that epistemically difficult moral theories are unfair loses much of its punch, since a true theory is preferable to a fair one. Of course, that might be a pretty big ‘if’, and whether Aristotle’s picture of moral reality is correct is not something I have discussed here. What I have done, I hope, is to have shed some light on the deeper issues driving Kant’s objections to Aristotle in the Doctrine of Virtue.


*Thanks to the anonymous referee for Studies in the History of Ethics and to Stuart Rachels for helpful, penetrating comments on the penultimate draft of this paper. Thanks to Tamar Gendler, John Robertson, Laurence Thomas, Ernest Wallwork, and especially Michael Stocker for discussion of an earlier version of this paper.

[1]Kant actually considers “good management” (gute Wirthschaft) to be the virtue between greed and prodigality.

[2]For simplicity’s sake, let maxims be structured triplets comprising a circumstantial component, an action to be performed, and an end to be realized: <(circumstance), [action], {end}>. (The agent’s motive is her reason for adopting this triplet.) Suppose I am considering acting on maxim m1 = < (when I need money), [I will falsely promise to repay a loan], {in order to get the money I need} >. I see that acting on m1 is not permissible because it leads to a conceptual contradiction; I can find the particular maxim which is permissible by negating the action-component: m2 = < (when I need money), [I will not falsely promise to repay a loan], {in order to get the money I need} >.

[3] Thanks to the anonymous referee for Studies in the History of Ethics for this insight.

[4] Thanks to the anonymous referee for SHE for suggesting this point.

[5] The Groundwork’s distinction between perfect and imperfect duties largely gives way to the distinction between juridical and ethical duties in the Metaphysics of Morals (and in the second Critique we meet the distinction between essential (wesentliche) and nonessential (außerwesentliche) duties (KpV 159)). Determining whether these are merely different ways of making the same distinction would take us beyond the ambit of this essay. Moreover, the crucial point, that Kant allows for indeterminate action-guidance for morality’s periphery but not for its core, can be appreciated independently of this question of the relationship between the perfect-imperfect and juridical-ethical distinctions.

[6] Thanks to the anonymous referee for SHE for forcing me to be clearer on this point.

[7] This is a simplification, since the consistent universalizability of the maxim would tell us only that A’s φing in c is permissible; φing would be required if the contrary of A’s maxim (see note 2, above) is not universalizable. If both the maxim and its contrary are universalizable, the actions are permissible: neither is required, neither forbidden. Kant, of course, believes that a case where neither maxim is universalizable (i.e., a case in which the same action would be both permitted and forbidden) can never arise.

[8] Hume actually considers “the joint verdict of [a panel of true judges] … [to be] the true standard of taste and beauty” (1757/1985: 241). Mill offers a similar standard for distinguishing between higher and lower pleasures in Utilitarianism, though where Hume requires a “joint verdict” Mill requires merely a majority: “On the question of which is the best worth having of two pleasures, or which of two modes of existence is the most grateful to the feelings … the judgment of those who are qualified by knowledge of both, or, if they differ, that of a majority among them, must be admitted as final” (1861/1979: 11); my italics).

[9] My thanks to the anonymous referee for SHE for raising the analytic-synthetic worry, and forcing me to be clearer on this issue.

[10] This is not to deny that many authors have argued that Kant is a virtue ethicist and at least one that Kant is a consequentialist. But I trust that my assumption that Kant’s ethics is at least presumptively an ethics of requirement, as Mill’s is presumptively an ethics of outcome and Aristotle’s an ethics of virtue, will suffice for the purposes of this article; querying the validity of these presumptions would be fascinating but is well beyond my ambit here, especially for Aristotle, for it is not clear that Aristotle has a virtue ethics on Watson’s taxonomy, given that virtue and goodness seem to be coordinate. But since Kant finds both ethics of outcome and ethics of virtue to be prey to IAGO and to make morality epistemically difficult, we need not worry about the proper categorization of Aristotle’s theory for our purposes.


Kant’s texts are cited using the following abbreviations along with the corresponding page number of the relevant volume of Prussian Academy edition of Kant’s works, which appear in the margins of most translations. The translations from which I have quoted are listed with the abbreviation.

Anthropology from a Pragmatic Point of View (1798), tr. Victor Lyle Dowdell. Carbondale: Southern Illinois University Press, 1978.
The Doctrine of Right (1797), in Mary Gregor, ed. and tr., The Metaphysics of Morals. New York: Cambridge University Press, 1996.
The Doctrine of Virtue (1797), in Mary Gregor, ed. and tr., The Metaphysics of Morals. New York: Cambridge University Press, 1996.
Groundwork of the Metaphysics of Morals (1785), tr. Mary Gregor. New York: Cambridge University Press, 1997.
Critique of Practical Reason (1788), tr. Mary Gregor. New York: Cambridge University Press, 1997.
Critique of Pure Reason (1781, 1797), trs. Paul Guyer, Allen Wood. New York: Cambridge University Press, 1998.
Lectures on Ethics (1775-80), tr. Peter Heath, New York: Cambridge University Press, 1997.
Religion within the Limits of Reason Alone (1793), tr. Theodore M. Greene and Hoyt H. Hudson. New York: Harper Torchbooks, 1960. (As this edition of the Religion does not have the marginal Academy pagination, the page number of the translation is provided in brackets in the citation, e.g., (R 27 [32]).)

Aristotle’s Nicomachean Ethics is cited by the abbreviation NE, with references to Book and chapter followed by the Bekker pagination. Translations are from J. Barnes, ed., The Complete Works of Aristotle (The Revised Oxford Translation). Princeton: Princeton University Press, 1984.

Anscombe, Elizabeth. 1958. “Modern Moral Philosophy.” Philosophy 33: 1-19.

Aquinas, Thomas. 2000. Treatise on Law (R. Regan, ed.). Indianapolis: Hackett Publishing Company.

Fuller, Lon. 1969. The Morality of Law (rev. ed.). New Haven: Yale University Press.

Herman, Barbara. 1993. The Practice of Moral Judgment. Cambridge: Harvard University Press.

Hume, David. 1757/1985. “Of the Standard of Taste” in E. Miller, ed., David Hume: Essays Moral, Political and Literary. Indianapolis: Liberty Classics.

Hursthouse, Rosalind. 1981. “A False Doctrine of the Mean.” Proceedings of the Aristotelian Society 81: 57-72. Reprinted in N. Sherman, ed., Aristotle’s Nicomachean Ethics: Critical Essays (Lanham MD: Rowman & Littlefield, 1999).

Hursthouse, Rosalind. 1995. “Applying Virtue Ethics” in Hursthouse et al., eds., Virtues and Reasons. Oxford: Clarendon Press, 1995.

Korsgaard, Christine. 1996. Creating the Kingdom of Ends. Cambridge: Cambridge University Press.

Kripke, Saul. 1972. Naming and Necessity. Cambridge: Harvard University Press.

Larmore, Charles. 1990. “The Right and the Good.” Philosophia 20: 15-32

Mill, John Stuart. 1861/1979. Utilitarianism. G. Sher, ed. Indianapolis: Hackett Publishing.

Ryle, Gilbert. 1958. “On Forgetting the Difference Between Right and Wrong” in A. I. Meldon, ed., Essays in Moral Philosophy (Seattle: University of Washington Press, 1958).

Santas, Gerasimos. 1996. “The Structure of Aristotle’s Ethical Theory.” Topoi 15: 59-80.

Schneewind, J. B. 1998. The Invention of Autonomy. Cambridge: Cambridge University Press.

Sidgwick, Henry. 1907/1981. The Methods of Ethics (7th ed.). Indianapolis: Hackett Publishing. (Reprint of 7th ed. published in 1907 by Macmillan.)

Slote, Michael. 1992. From Morality to Virtue. New York: Oxford University Press.

Tribe, Laurence. 1988. American Constitutional Law (2nd ed.). Mineola, NY: The Foundation Press.

Urmson, J. O. 1973. “Aristotle’s Doctrine of the Mean.” American Philosophical Quarterly 10: 223-30. Reprinted in A. O. Rorty, ed., Essays on Aristotle’s Ethics. Berkeley and Los Angeles: University of California Press, 1980).

Urmson, J. O. 1988. Aristotle’s Ethics. Oxford: Basil Blackwell.

Watson, Gary. 1990. “On the Primacy of Character,” in O. Flanagan and A. O. Rorty, eds., Identity, Character and Morality (Cambridge: MIT Press, 1990).

Williams, Bernard. 1981. Moral Luck. New York: Cambridge University Press.