Langton’s (1998) and Allais’ (2015) metaphysical interpretations of Kant’s idealism have given special relevance to Kant’s analysis of the inner/outer dichotomy in the Amphiboly of the Concepts of Reflection, for they agree that this dichotomy is key to correctly grasping Kant’s distinction between appearances and things in themselves. In this article I argue that Langton’s and Allais’ accounts of Kant’s analysis of the inner/outer dichotomy have major limitations, and therefore that the text should not be read in the way they propose. In order to show these limitations, I examine the overall structure of Kant’s argument in the Amphiboly. Furthermore, I aim to establish the contribution brought by the Amphiboly to the issue of noumenal knowledge and the nature of things in themselves. Langton’s and Allais’ accounts of the relation between appearances and things in themselves as the inner nature of things not only prove to be unwarranted, but indeed to some extent opposite to what I claim to be Kant’s actual stance on things in themselves in the Amphiboly.

The problem of Kant’s Neglected Alternative is that while his Aesthetic provides an argument that space and time are empirically real – in applying to all appearances – its argument seems to fall short of the conclusion that space and time are transcendentally ideal, in not applying to any things in themselves. By considering an overlooked passage in which Kant explains why his Transcendental Deduction is ‘unavoidably necessary’, I argue that it is not solely in his Aesthetic but more so in his Deduction where he intends to provide his argument for the transcendental ideality of space and time. His Deduction shows that space and time do not have a valid application to any things in themselves by arguing that the categories do have a valid application to everything in space and time, but that the categories do not have a valid application to any things in themselves.

Matter’s real essence is a ground for certain features of phenomena. Things in themselves are likewise a ground for certain features of phenomena. How do these claims relate? The former is a causal essentialism about physics, Stang argues; and the features so grounded are phenomenally nomically necessary. The latter involves a distinctive ontology of things in themselves, I argue; but the features so grounded are not noumenally nomically necessary. Stang’s version of Kant’s modal metaphysics is admirable, but does not go far enough. Kant’s causal essentialism involves the essences of fundamental properties, as well as of matter. And things in themselves are grounds, because they are substances, the ‘substrate’ of phenomena.