Investigated the application of a generative theory of multimedia learning to the design of second-language multimedia learning environments and possibilities of supporting visual and verbal learning preferences to improve learning outcomes. 103 United States college students enrolled in second-year German language courses participated in the study as a regular class activity. Verbalizers and visualizers were identified using the Visualizer/Verbalizer Questionnaire and the Edmonds Learning Style Identification Exercise. Participants read a 762-word German language story presented by an interactive multimedia computer program in their own time. For keywords, a verbal translation and a picture or video clip (visual annotation) were available. A vocabulary posttest confirmed the hypothesis that students remembered word translations best when they made use of both visual and verbal annotations rather than only one or neither. A comprehension test indicated that propositions are recalled better when annotations in the preferred mode are available. The results are consistent with the generative theory of multimedia learning and emphasize the importance of providing both visual and verbal learning options in the design of multimedia instruction.