Learner control is an important instructional design in video-based learning. This study assessed two conditions: a full learner control where learners direct their learning path, and a hybrid learner control where learners follow the instructor-set path but still enjoy certain aspects of control. Two groups of university students participated in this study by learning statistics through online video courses. The findings show that the full learner control condition attained higher learning performance than the hybrid learner control condition, but spent more time than the latter. The full learner control condition scored higher than the hybrid condition in the difficult sections of video-based learning; but no significant difference was found in the easy section. There was a significant difference between the two conditions in learning agency, but no differences in cognitive load and affective and cognitive engagement. Hierarchical regression analysis indicated differences between the full and the hybrid learner control conditions in the factors predicting overall scores. The findings carry important contributions and implications for the research and practice of instructional designs in online video-based learning such as MOOCs.