The supportive arguments for hypermedia and hypertext learning environments typically assert that as the systems permit individuals to self-select content based on needs, to control the pace and content, and/or provide a measure of control otherwise missing in traditional, linear texts and lectures, learning and knowing outcomes increase. There are two difficulties with this argument and some of the resultant empirical literature. The first is that knowing is a complex phenomenon which can be variously defined and measured. However, the literature often focuses on a narrow range of definitions and measures and doesn’t exploit the full range of theory. This, in turn, leads to the second difficulty, that of very limited outcomes. A cognitive model of learning and knowing is presented and contrasted with a sample of empirical research. This model provides a solid theoretical basis for research into cognitively engaging hypermedia systems.