Global comparisons of learning from hypertext/hypermedia and traditional presentation formats like text have yet failed to show major advantages concerning the effectiveness of hypermedia learning. Thus, it is proposed in the current paper to evaluate hypermedia environments more specifically with regard to their potential to implement and support well-defined learning approaches. According to this view, an effective hypermedia design needs to be based on thorough cognitive task analyses with regard to structures, processes, and resources that are required to benefit from a specific learning approach. This claim is illustrated by two experiments in which we explored the instructional potential of hypermedia environments for improving schema acquisition from worked-out examples. A cognitive task analysis was used to identify specific types of information comparisons that are crucial for successful schema induction and that might be effectively supported by suitably designed hypermedia environments. The experiments investigated two methods for enhancing comparison processes in hypermedia-assisted learning from worked-out examples, namely, elaboration prompts and an interactive comparison tool. Both methods improved performance for near-transfer problems. Ways of extending this task-analytical approach to facilitating far transfer are also discussed.