Confronted with the contradiction between the innate intuition of free moral agency and the scientific presumption of the causal flow of nature, Kant attempted to explain free will in dualistic terms1:
Now I may say without contradiction: that all the actions of rational beings, so far as they are appearances (occurring in any experience), are subject to the necessity of nature; but the same actions, as regards merely the rational subject and its faculty of acting according to mere reason, are free.
We see this dualistic understanding replicated, albeit in nuanced form, in his Critique of Pure Reason:
At the same time, it must be carefully borne in mind that, while we surrender the power of cognizing, we still reserve the power of thinking objects, as things in themselves.* For, otherwise, we should require to affirm the existence of an appearance, without something that appears--which would be absurd. Now let us suppose, for a moment, that we had not undertaken this criticism and, accordingly, had not drawn the necessary distinction between things as objects of experience and things as they are in themselves. The principle of causality, and, by consequence, the mechanism of nature as determined by causality, would then have absolute validity in relation to all things as efficient causes. I should then be unable to assert, with regard to one and the same being, e.g., the human soul, that its will is free, and yet, at the same time, subject to natural necessity, that is, not free, without falling into a palpable contradiction, for in both propositions I should take the soul in the same signification, as a thing in general, as a thing in itself--as, without previous criticism, I could not but take it. Suppose now, on the other hand, that we have undertaken this criticism, and have learnt that an object may be taken in two senses, first, as a phenomenon, secondly, as a thing in itself; and that, according to the deduction of the conceptions of the understanding, the principle of causality has reference only to things in the first sense. We then see how it does not involve any contradiction to assert, on the one hand, that the will, in the phenomenal sphere--in visible action--is necessarily obedient to the law of nature, and, in so far, not free; and, on the other hand, that, as belonging to a thing in itself, it is not subject to that law, and, accordingly, is free2
Were it not for the fact that Kant continued to believe in the free moral agency of human beings, one would have to read the above as an academic way of papering over self-delusion. For example, his statement in Prolegomena sounds disturbingly similar to "Science tells us our actions are determined, but we sense that they are free." Indeed, he even fails to recognize that more than mere rationality is involved in free will decisions. Computers can be programmed to make decisions according to rational criteria. Only human beings are actually in dialog with the natural world (via sense experience, reflection and action) and with God.
Kant's invention of "synthetic judgment" as a distinct category of human reason was seriously flawed. His application to arithmetic, for example, was later shown by the Italian mathematician, Giuseppe Peano, to be mistaken. Arithmetic is, in Kant's sense, ultimately analytical. That is, one can prove 5 + 7 = 12 using only the relevant definitions. His notion of geometric intuition as something innate, and not conditioned by reality, is also mistaken, as the history of non-Euclidean geometry (and the non-linearity of space-time) subsequently demonstrated.
It seems to me that a major motivation for Kant's thesis is worry that the more sanitized empirical approach of David Hume closes off consideration of realities that are fundamentally beyond the reach of our senses. As twentieth century mathematics and science has come to realize, however, one need not constrain our intuitions of reality to that which is immediately observable. On the contrary, all that is fundamentally necessary is that our intuitions be consistent with that reality. And, since that reality evidently includes miraculous interventions in the causal flow of history, interventions that help us to appreciate even the nature of human free will, we need not fear the constraints imposed by a rigorously relational perspective.
Thus, Kant's distinction, rather than freeing the epistemologist to recognize noumenal realities, ultimately imposes erroneous and unnecessary restrictions on our ability to grasp relationships beyond the immediate reach of our senses.