The debate regarding the interpretation of Kant's idealism is usually seen as turning on the best way to understand his transcendental distinction between appearances and things in themselves: that it marks either a contrast between two types of thing (the ‘two-object’ or ‘two-world’ view) or one between two sides or aspects of ordinary empirical objects (the ‘two-aspect’ view). But, even though I have long been associated with the latter camp, I have also thought for many years that this is not the most helpful way to frame the issue. The problem lies in an ambiguity inherent in the two-aspect view. It can be understood either metaphysically, as a thesis about the kinds of properties attributable to empirical objects, that is, as a form of property dualism in which these objects are assigned both phenomenal and noumenal properties, or methodologically, as a contrast between two ways in which such objects can be considered in a philosophical reflection on the conditions of their cognition. Accordingly, I take the fundamental question to be whether transcendental idealism is to be understood in the latter way or as a form of metaphysical dualism (whether as a thing or a property dualism being a matter of relative indifference). And I have further thought that the best way of addressing that question is through a consideration of the view which Kant opposes to transcendental idealism, namely, transcendental realism. If this realism is identified with a particular metaphysical doctrine then transcendental idealism must be as well; but if, as I maintain, transcendental realism cannot be so understood, then neither can Kant's idealism.

Few of Kant's doctrines are as difficult to understand as that of self-affection. Its brief career in the published literature consists principally in its unheralded introduction in the Transcendental Aesthetic and unexpected reappearance at a key moment in the Deduction chapter in the second (B) edition of the first Critique. After blazing its trail, self-affection retreats into the background, with a discussion befitting its importance occurring only in the unfinished Opus postumum. This step out of the limelight, however, belies the doctrine's continued importance for Kant; indeed, Kant seemed to think that in self-affection was to be found the key to the project that occupied him in his last years. Thus, ‘the possibility of the transition from the metaphysical foundations of natural science to physics does not consist in the fact that the subject is empirically affected but rather that it affects itself’ (Opus postumum, 22: 405). As he continued to struggle with this doctrine and with the pivot-point on which to work this vital transition, Kant himself would surely come to rue his confident statement in the B Deduction: ‘I do not see how one can find so many difficulties in the fact that inner sense is affected by ourselves’ (B156n).

The role of transcendental idealism in Kant's theory of knowledge has been both deliberately underrated1 and inadvertently exaggerated. If conceivably not necessary, its role in Kant's explanation of the possibility of a priori knowledge in the Critique of Pure Reason is at least pivotal to the success of the explanation. On the other hand, though transcendental idealism depends on Kant's epistemological criterion of an existing object, or, simply, his criterion of existence, the criterion for its part is actually independent of the idealism. In fact, it may be because this independence has hardly been recognized that commentators have been unaware of the role the criterion may actually be playing in the continuing controversy over the correct interpretation of the idealism. Altogether, this article addresses both shortcomings – the underestimation and the exaggeration of the role of the idealism in Kant's epistemology. While it places the idealism at the centre of the epistemology, it also separates the criterion of existence from the idealism. In highlighting this contrast, the article explains how the criterion may actually be contributing to the persistence of the ongoing dispute over the correct interpretation of the idealism.

Perfect ethical duties have usually puzzled commentators on Kant's ethics because they do not fit neatly within his taxonomy of duties. Ethical duties require the adoption of maxims of ends: the happiness of others and one's own perfection are Kant's two main categories. These duties, he claims, are of wide obligation because they do not specify what in particular one ought to do, when, and how much. They leave ‘a latitude for free choice’ as he puts it. Perfect duties, however, such as the duties of respect, to avoid suicide, lying, and servility, do not appear to require the adoption of ends but only the performance or omission of specific types of actions. The puzzle is how these duties can be ethical, and therefore wide. Faced with this difficulty, Mary Gregor denies that perfect ethical duties are wide. She claims that they are an ‘anomaly’ and that they do not belong to ethics proper but to moral philosophy in general. She argues that these duties are derived from the categorical imperative, instead of, as Kant himself appears to have thought, the first principle of virtue. Taking a very different approach, Onora O'Neill finds the perfect/imperfect distinction of little importance and suggests doing without it altogether. Most other interpreters also assume that ‘wide’ is opposed to ‘perfect’ so that a wide perfect duty is a conceptual impossibility.

In this article, I want to argue that scepticism for Kant must be seen in ancient and not just modern terms, and that if we take this into account we will need to take a different view of Kant's response to Hume from the one that is standardly presented in the literature. This standard view has been put forward recently by Paul Guyer, and it is therefore his view that I want to look at in some detail, and to try to correct.

In her essay (in Kantian Review 10 (2005), 82–111), Shell wants to demonstrate that 1. Kant's theory of the right of nations ‘can furnish us with some much needed practical help and guidance’, and 2. ‘Kant is less averse to the use of force, including resort to pre-emptive war… than he is often taken to be’ (p. 82). The first claim is unconvincing. The second one is in need of clarification. Shell turns Kant into a kind of realist and just-war theorist, into a liberal who is prejudiced against illiberal regimes. In the end, her Kant is closer to Locke, Vattel and other early liberal international lawyers than to himself. Almost all that is unique in Kant's theory of the right of nations gets lost. In this, Shell follows a general trend among some Kant interpreters: the interpretation is only loosely based on Kant; it claims to follow his ‘spirit’ and offers creative ‘Kantian perspectives’. Amidst interpretational creativity, Kant's texts more or less disappear in the mist.