This dissertation study was intended to verify whether the positive effects of the roll-over annotation presentation format on reducing cognitive load and enhancing vocabulary and comprehension of 5th-grade children (Morrison, 2004) can be extended to college students' learning from a web-delivered text. In order to answer this research question, relevant constructs, such as cognitive load, learner control and learning, have been examined in this experimental study of 149 undergraduate students in a state university at College of Education. No single effect of annotation presentation format on cognitive load was found, but an interaction effect on cognitive load was revealed between annotation presentation formats and pre-existing knowledge in this study of online education, similar to the interaction effect between annotation presentation formats and learners' reading experience found in traditional textbook learning (Yeung, Jin and Sweller, 1998; Yeung, 1999). Besides, students' computer experience also had a significant impact on their perceived cognitive load. One more key finding from this study was that the embedded annotation presentation format generated the least learner control, significantly different from other annotation presentation formats. In conclusion, an adaptive approach to the design of annotation presentation formats is recommended, for example, individual differences including learners' familiarity with content should be considered along with different annotation presentation formats so as to reduce learners' overall cognitive load. Additionally, learners' computer experience should be examined when hypertext annotations are used. Finally, choices of annotation presentation formats should be well-conceived to balance cognitive load, learning, and learner control.