The Descartes Method (Collins/Akritas, 1976) isolates the real roots of a univariate polynomial by recursive interval subdivision, using Descartes' Rule of Signs as termination criterion. The thesis to be presented provides an improved estimate for the resulting subdivision tree based on the theorem of Davenport and Mahler, from which the best known bitcomplexity bounds for integral input polynomials follow at once. Furthermore, it describes and analyzes a variant of the method that works correctly for so-called "bitstream" coefficients, which can be approximated arbitrarily well but cannot be determined exactly.