Contrary to system-controlled multimedia learning environments, hypermedia systems are characterized by a high level of interactivity. This interactivity is referred to as learner control in the respective literature. For several reasons this learner control is seen as a major advantage of hypermedia for learning and instruction. For instance, learner control might increase students' interest and motivation, facilitate adaptive instruction, or provide affordances for active and constructive information processing. In this paper we analyze the instructional potentials of learner-controlled hypermedia environments as well as possible reasons for the ambiguous results of studies that have aimed at determining the effectiveness of hypermedia learning. According to our analysis, the potential effectiveness of self-controlled learning with hypermedia might be difficult to demonstrate due to (1) usability problems (i.e., disorientation, distraction, cognitive overload), (2) moderating learner characteristics (i.e., prior knowledge, self-regulatory skills, cognitive styles, and attitudes towards learning), (3) a lack of conceptual foundations, and (4) methodological shortcomings of many hypermedia studies. The findings reviewed in this paper and the corresponding claim that hypermedia may be effective only if used in a sensible way are used to derive a couple of guidelines for further research on hypermedia learning. This paper on hypermedia learning is part of a special issue on interactive learning environments. Broadly speaking, interactive learning environments allow learners manipulating the presentation of information that they contain. In the literature on computer-based instruction, the term interactivity is most often used for describing the malleability of single