This paper reconstructs the evolution of Novalis' thought concerning being, nature, and knowledge. In his earlier writings (above all the Fichte-Studies) he argues that unitary being underlies finite phenomena and that we can never know, but only strive towards knowledge of, being. In contrast, in his later writings, principally the Allgemeine Brouillon, he maintains that the unitary reality underlying finite things can be known, because it is an organic whole which develops and organises itself according to an intelligible pattern. Novalis equates this whole with nature. However, because this organic whole exercises spontaneity in assuming particular forms of organisation, we can never know why it assumes just these particular forms; nature therefore remains partly unintelligible to us. I argue that Novalis' intellectual shift towards the idea that the whole can be known is motivated by his concern to explain how the modern, 'disenchanted', view of nature could be overcome. I also argue that by recognising this shift, we can resolve the dispute between Frank and Beiser as to whether Novalis thinks that the absolute can be known.