Levels of "thing" talk in the first Critique:
- Things in general.
- Things in themselves (AKA transcendental subjects/objects).
- Noumena (some overlap with (2), but note that space and time of different dimensionality than {3 + 1} would also count as noumenal in a weaker sense).
- Phenomena, or things as they appear to us.
At one point in the Transcendental Dialectic, Kant says:
The question, "What is the constitution of a transcendental object?" is unanswerable—we are unable to say what it is; but we can perceive that the question itself is nothing; because it does not relate to any object that can be presented to us. For this reason, we must consider all the questions raised in transcendental psychology as answerable and as really answered; for they relate to the transcendental subject of all internal phenomena, which is not itself phenomenon and consequently not given as an object, in which, moreover, none of the categories—and it is to them that the question is properly directed—find any conditions of its application. Here, therefore, is a case where no answer is the only proper answer. For a question regarding the constitution of a something which cannot be cogitated by any determined predicate, being completely beyond the sphere of objects and experience, is perfectly null and void.
This is the English translation, from J. M. D. Meiklejohn, of the B-edition. Here is the German base:
Man kan zwar auf die Frage, was ein transscendentaler Gegenstand vor eine Beschaffenheit habe, keine Antwort geben, nemlich was er sey, aber wol daß die Frage selbst nichts sey, darum, weil kein Gegenstand derselben gegeben worden. Daher sind alle Fragen der transscendentalen Seelenlehre auch beantwortlich und wirklich beantwortet; denn sie betreffen das transsc. Subiect aller inneren Erscheinungen, welches selbst nicht Erscheinung ist und also nicht als Gegenstand gegeben ist, und worauf keine der Categorien (auf welche doch eigentlich die Frage gestellt [479] ist) Bedingungen ihrer Anwendung antreffen. Also ist hier der Fall, da der gemeine Ausdruck gilt: daß keine Antwort auch eine Antwort sey, nemlich daß eine Frage nach der Beschaffenheit desienigen Etwas, was durch kein bestimtes Prädicat gedacht werden kan, weil es gänzlich ausser der Sphäre der Gegenstände gesezt wird, die uns gegeben werden können, gänzlich nichtig und leer sey.
When I fed the first part of the first sentence into Google Translate, I got a peculiar result: "One can indeed answer the question of what a transcendental object is..." So far, though, the rest of the start of the quote and the end of the whole bit, do seem to compare pretty well with the given English version.
So why would an empty thesis be necessary? In the Groundwork for the Metaphysics of Morals, Kant makes a lot out of the fact that understanding and reason are "spontaneous" or proactive, with reason the most spontaneous (compared to sensation/perception and understanding). And things-in-themselves, he says in the first Critique, would be objects of intellectual intuition, which is how God would intuit things. But so now things-in-themselves are spontaneous enough, in relation to us, to be "recognized" by their proactivity (grounding phenomenal representations), but this "recognition" does not and cannot have any implications more specifically, here.
Yet Kant "dreads" (with good enough reason) that even this lonely sentence about things-in-themselves might get the wheels of transcendental realists turning, to try to deduce a whole ensemble of metaphysical claims from that sentence; or, worse, to attribute the potential for such deductions to Kant himself, then. He knows for reasons of general logic, the logic of things in general, that he is compelled to differentiate the concept of things-in-themselves from things-as-they-appear-to-us, noumena from phenomena, and he's understandably concerned that this compulsion, which is abstract, will not be cautiously satisfied by others. For transcendental illusions are as unavoidable as optical ones, and must be corrected for by reflection.
So when Kant says that the distinction between things-in-themselves and phenomena is necessary, but that the question of what things-in-themselves are is "empty," is he contradicting himself? Perhaps. But perhaps it's an easy enough fix: just take him to be denying that the distinction, by itself, carries substantial meaningfulness. Or c.f. Wittgenstein's distinction between senselessness and nonsense, then.
