Recent discussions of structuralist approaches to scientific theories have stemmed primarily from Worrall's (1989), in which he defends a position (since characterized `epistemic structural realism') whose historical roots he attributes to Poincare. In the renewed debate inspired by Worrall, it is thus not uncommon to find Poincare's name associated with various structuralist positions. However, Poincare's structuralism is deeply entwined with both his conventionalism and his idealism, and in this paper we explore the nature of these dependencies. What comes out in the end is not only a clearer picture of Poincare's position regarding structuralism, but also two arguments for versions of epistemic structuralism different in kind from that given by Worrall.