At Kant Blog, Andrew asks how many of us support "two-world" rather than "one-world" readings of Kant's transcendental idealism. One problem with "two-worlds" readings has always been that they make Kant look like a skeptic, when it seems clear that skepticism (or, as Kant puts it, "skeptical idealism," i.e. Descartes) is one of his targets. Or they just make him look flatly inconsistent – as Andrew says, the charge is that "transcendental idealism is “inconsistent Berkeleyanism,” since it seems like Berkeleyan idealism + un-Berkeleyan things in themselves". These mysterious noumena supposedly somehow
The "one-world, two-aspect" reading (e.g. Allison, Bird) is supposed to help us avoid this problem. If there is just the one world, which can be considered either as it appears to us or "as it is in itself," then this lessens our sense that when it turns out we can know only the former (that is, know the one world only under the former aspect), there is something else still out there behind the scenes that remains unknown. Instead, we know things in the only way in which it makes sense to think of us as knowing them – that's all.
I am sympathetic with this line of thought, but I think it's somewhat hasty to think that it solves our problems just like that. And I agree with Bader, the proprietor of our second Kant blog, Transcendental Idealism, that the two-worlds/one-world distinction shouldn't be thought of as one between "ontological" and "epistemological or methodological" readings of Kant. Whatever the noumena/phenomena distinction amounts to, Kant clearly has ontological fish to fry here (he'd better, or we're still stuck with the Cartesian "transcendental-realist" conception of the objective world). Bader agrees with James Van Cleve that "the distinction between appearances and things in themselves is a distinction between two separate universes of discourse - not between two ways of discoursing about the same class of objects." (Problems from Kant, p. 150). I would agree with this as well if all it means is that (as we just said) it's not true that Kant is concerned simply with knowledge and language and not with ontology. However, I take his (Kant's) point to be that the three cannot be so cleanly separated (as they are on the Cartesian picture – in fact that's what gets the skeptical problem going: the idea that we can have perfectly contentful beliefs about what might not actually be out there at all). Still, I think we have to say that there's only one world, or we run straight into the problem we started with. So in some sense we are indeed concerned with "two ways of discoursing" about the same thing (if not "the same class of objects" exactly). Again, a key Kantian point is that part of what it is to be an object at all is for it to be the object of our thought, as we conceptualize it.
Here's how I like to explain why simply pointing to the one-world reading as an answer to skepticism (i.e. so that there's no transcendent world to be worried about not being able to know) cannot work. The original question was: "why can't I know the real world instead of (just) the world of appearance?". On the one-world reading, the question does not just go away, but instead becomes: "why can't I know the world the way it really is, instead of (just) how it appears?". The virtue of the two-aspect reading is that it allows a better answer to that question (but you still have to give it, not just say "but there's only the one world, so there's no problem"!).
What we should say is this. When we know the world as it appears (say by observing the cup on the table, and inferring that the cup is on the table), we do know the world the way it really is. If you don't think that that's how things really are, then you have no business saying that you believe it; and if you don't believe it, then you need to take another look. If you still don't believe it, then you are letting Cartesian theoretical requirements dictate your beliefs (or lack thereof, qua skeptic). In particular, it is the Cartesian identification of "how things are in themselves" with "how things really are" that is causing the problem – or in other words, the Cartesian conception of objectivity as radically independent of subjectivity. That's the target of Kant's argument.
Again, objectivity is an ontological matter, not one (simply) of semantics or epistemology. But it is indeed (as the original crude identification of one-world/two-aspect readings with "epistemological or methodological" ones was meant to bring out) in explaining the role of the concept of objectivity in semantics and epistemology (that is, in speaking and knowing) that we see what objectivity is (and thus what sort of thing we should think of as "objective").
I will make these points in my own pragmatist-cum-Davidsonian language, rather than Kant's own tangled idiom, but the anti-Cartesian point is essentially the same. See for example the "Refutation of Idealism" section of the first Critique (B 274-79), where Kant speaks in terms not of language but of experience, i.e. self-consciousness. But (in this context at least) being a self-conscious agent and being a language user are the same thing, for the very Kantian reasons we are discussing: each, contra Descartes, commits you to dependence on an independent world for the content of your thoughts.
The central concept of epistemology is that of belief. The truth of your belief that p depends not on anything having to do with you, but instead on whether p is really (objectively) the case. My belief that the cup is on the table is true iff the cup really is on the table, independently of my believing it. The concept of belief marks the conceptual difference between actual knowledge of the objective world and possibly false subjective impressions of it. Belief, we say, doesn't make it so; yet if things are indeed as you believe them to be, then your knowledge is perfectly objective. This is why relativism is false: the truth of our beliefs depends on how the world is.
Analogously, the central concept of semantics is that of meaning. My statement "the cup is on the table" means what it does in virtue of its being true when things really are as it says they are when it is so construed (here, as meaning that the cup is on the table). Or, in a non-homonymous case (and omitting a few complications), taking "the slod is on the kovep" as being true in virtue of the cup's being on the table (and not otherwise, whether I said so or not), would be to take it as meaning "the cup is on the table" (and, presumably, that "slod" means "cup" and "kovep" means "table"). This is why idealism is false: the content of our thoughts depends on how things are when our beliefs (that they are that way) are true.
That's all there is to the concept of objectivity (i.e. things being really one way and not another, independently of what we say or believe). It is a Cartesian distortion to identify "objectivity" with the world as seen from no particular point of view (itself a distortion of the idea of the world as seen through mathematical eyes, as the new science demands). This is the point of the Kantian rejection of noumena (i.e. as "nothing to us"). For Kant, the phenomenal world is the (one and only) objective world; and when we speak truly, we see things as they really are. (There are different kinds of "noumena," and Kant sometimes backslides, but I think it undeniable that this is generally Kant's view: that noumena are to be rejected as a Cartesian fantasy.)
Seen from a Cartesian point of view, this looks like idealism – that what is objectively the case depends essentially on our practices of speaking and thinking, rather than being so "anyway" (as Bernard Williams puts it). But of course that's one of the things wrong with the Cartesian view: that seen from it, the correct view looks unacceptably idealistic. (But it isn't.)
We can indeed put the Cartesian fantasy in terms of there being an "independent world" (i.e. "beyond appearances," metaphysically speaking), as in the two-worlds version of Kant's philosophy. That's why it's hard to see that reading as helping: it seems to leave the Cartesian thought in place. Then it looks like the mystery of the noumenal world is Kant's doing, not Descartes's. (For example of this sort of reading, see Barry Stroud's uncharitable reading of Kant in The Significance of Philosophical Skepticism).
I have thus interpreted Kant's "transcendental idealism" as not being idealism at all (as I would Hegel's "absolute idealism" too, for that matter). It's as "realist" as we should want, once we give up the Cartesian fantasy of "absolute" objectivity. Some philosophers embrace Kant's "empirical realism" as allowing us to put metaphysics behind us (and become "scientific realists"). However, I tend to reserve the term "realism" for the "transcendental realist" position Kant enables us to reject (or, in contemporary terms, "metaphysical realism"). So we should embrace neither realism nor idealism, which should be rejected together as jointly committed to an untenably dualist picture. ("Scientific realism" has its own problems.)
This doesn't mean that I think we can accept Kant's position as it stands. From our present position on the far side of the "linguistic turn," we can see Kant as a key predecessor, but only once we know what to look for. To be "Kantian" nowadays is to apply Kantian insight to our contemporary situation, not to drag Kantian doctrines whole into the present (or to join them in the past). For various reasons (including his obsession with the problem of freedom (agency, rationality, morality, etc.) and his somewhat less successful solution to it, not to mention his equally obsessive systematicity), Kant seems to have left a certain amount of residual Cartesianism in place (or even, in some places, reinforced it), and it is natural for us to turn to subsequent critics (Hegel, various Romantics, Kierkegaard, Nietzsche) for help. Still, the "critical turn" remains an essential part of the glorious Nietzschean story (even if Nietzsche himself understates Kant's role in it, relegating him to an earlier phase than I would) of How the True World Became a Fable.