Two experiments are presented that examined the role of individual differences on cognitive load in multimedia learning. In Experiment 1, 103 students using a German multimedia software were allowed to look up visual and verbal annotations for unknown words. In Experiment 2, 152 students were assigned to one of four treatments in which they could use either verbal or visual annotations, both, or none. Providing a choice of annotation type resulted in increased learning, while assigning an annotation type led to higher cognitive load, resulting in reduced learning for low-ability learners. Results are in line with a Generative Theory of Multimedia Learning and Cognitive Load Theory, and have implications for the design of Web-based and multimedia learning environments.