Kant devotes part of The Metaphysics of Morals to attacking traditional accounts of virtue (6:403-409).* Something like this is essential for what Kant is doing, since the traditional concept of virtue is a significant obstacle to the Kantian approach, but it's not discussed all that much because it is definitely not Kant at his best. He garbles the traditional theory of virtue quite a bit, although in ways that are perhaps not surprising given his own view. Kant takes the account to include three major claims:
(1) There is one virtue and one vice.
(2) Virtue lies in the mean between opposing vices.
(3) Virtue requires experience.
(1) seems to be Kant's oddly stated version of the unity of virtues thesis, the idea that to have one virtue, or at least one fundamental virtue, you must in some sense have them all; at the very least he seems to take it to mean that reasons for one virtue can be reasons for another as well. Kant argues in opposition to this that every duty can have one and only one ground of obligation. If you have a duty, there is one specific reason for doing it, and nothing more. Since Kant takes duty to be more fundamental than virtue, this carries over to virtue. The duty to be truthful can have only one proof from one ground; if you say that you should be truthful out of self-respect and also out of the harm that lies cause, you are actually confusing completely different things; truthfulness is simply different from benevolence.
(2), of course, is the most important idea in virtue ethics, the one that has to be considered seriously even by approaches to virtue that reject the golden mean. Unfortunately, Kant completely bungles his account of it. As he interprets it, it means that the virtue is itself a midpoint between two vices. Call them vice L and vice R. He takes it that the claim is that you can move smoothly from vice L, through the virtue, to vice R, and vice versa. Let's suppose vice R is the vice of excess, he takes it that this means that vice R is the virtue taken too far, whereas vice L is not going far enough with the virtue. Excess and defect just mean that virtue and vice differ by degree. Despite the fact that Kant purports to be giving Aristotle's doctrine of the mean, Aristotle's account is, of course, inconsistent with all of these claims. Virtue is not a mean in the sense that it is simply a midpoint between two vices, because merely lacking vices in both directions is not virtue -- virtue is not merely a lack of vice but a positive disposition or habit in its own right, and so must be developed itself. Likewise, the vice of excess is not going too far in the virtue for the simple reason that the virtue includes the mean in its definition -- it is impossible to go too far or not far enough with virtue, because virtue is just right by its very nature. This is not actually very far from Kant's insistence that vice L, vice R, and the virtue are all associated with different maxims that contradict each other, although, of course, maxims go with actions rather than dispositions.
Only (3) is reasonably close to its real historical version, and Kant's reason for rejecting it ties directly into the central features of Kant's approach to ethics: duties must not be based on empirical features of human nature but on a law that commands categorically for all rational beings.
Ultimately, of course, since he takes duty to be more fundamental than virtue, Kant has to have a very different account of virtue than someone like Aristotle. Virtue for Kant is a kind of strength for attaining moral objects; you can consider it an aptitude, but because he rejects the doctrine of the mean, he needs an alternative way to identify which aptitudes in particular are virtues. To be moral, it has to connect to the Kantian conception of freedom, so Kant takes virtue to be an aptitude to determine oneself to act in conformity with law through the thought of the law -- or, in short, an aptitude for freely willing moral maxims.
---
* It makes sense, of course, for The Metaphysics of Morals to consider the question of what virtue is, but I wonder if there is more to it than that. It is not often enough recognized that MM is a thoroughly anti-Catholic work. In the course of it, Kant argues for the absolute right of the state to seize any and all church property (6:324-325), against the validity of liturgical law and canon law (6:327), against the idea that the church is independent of the state (6:368-369), for the claim kneeling, prostration, and veneration of icons in church is idolatrous (6:437), and against "monkish ascetics" and penance (6:484-485). He insists that we have no natural duties to God (6:443-444, 486-491). He can hardly be unaware that Catholic moral theology tends to proceed on the basis of a broadly Aristotelian account of virtue and an insistence on our natural duties to God, and in a book in which he takes the trouble to insist, explicitly, that a number of Catholic claims are false, it seems plausible that he is deliberately taking aim at Catholic moral theology as one of his major targets. I forget who first said that Kant was preeminently the philosopher of Protestantism, but The Metaphysics of Morals could very well be Exhibit A in the argument for that claim.
