Hypertext has grown rapidly in popularity, from being a relatively obscure concept in the late 1980’s to near ubiquity with the advent and spread of the World Wide Web. Many forms of information are naturally and easily expressed as hypertexts, especially those that draw connections among various works. Academic work is relatively highly structured work, both internally, and in how it is related to existing work. Nonetheless, aside from a few systems that provide simple footnote and reference linking, no serious attempts have been made to publish academic articles in hypertextual form, with robust, first-class structures comprising a large part of the product. Previous investigations into the requirements for supporting the publishing of academic articles as fully functional hypertexts have noted numerous reasons for slow progress toward this goal. In this paper, we look closely at models for publishing hypertexts. We do so by pointing out several axes along which publishing models may differ, providing a systematic way in which to discuss these models. We then examine three particular models, situating them within our framework, and describing some of their implications. Examining these instances allows us to hypothesize more generally about the implications of choices posed by our framework.