The study examined learning while using a linear text, navigational hypertext, or a generative hypertext system. In Experiment One, students were assigned either the linear or navigational hypertexts and expected to learn the information to solve a posed problem, while students in Experiment Two learned the information to pass a reading comprehension test. Results from Experiment One revealed that the Navigational group scored significantly higher on self-determination (SD) and problem-solving (PS) measures, while the Linear group had significantly higher reading comprehension (RC) scores. Results from Experiment Two revealed no significant differences among treatments for either the SD or RC measures. In Experiment Three, we examined the effect of carrying out generative activities on problem-solving and reading comprehension. Results found that the number of generative activities explained a significant amount of the variance in PS and RC scores. Qualitative grouping of log-file data using Pathfinder illuminated patterns of user's navigational processes., HYPERTEXT AND LEARNER CONTROL In these times of rapidly expanding technology, hypermedia instructional programs are becoming commonplace in the educational and business markets. Learners of all ages have increasing access [...]