There is a widespread assumption that the addition of still images, animation and sound to text will enhance any information product. The research reported in this paper is investigating such claims for multimedia in an educational context and for a specific user group: grade-six primary school students. The role of text complexity and text type (descriptive or procedural) in influencing the impact of animation was explored. Design criteria such as the level of semantic integration between text and animation, and the importance of including captions with animations, was also investigated. The initial stages of the research used a commercial CD-ROM multimedia product - Compton's Multimedia Encyclopedia -while the later stages have employed multimedia sequences generated for the purposes of the experiments. The findings to date suggest that the impact of multimedia is subtle. Multimedia produced the greatest relative improvement in recall and inference levels in the case of the simple procedural article, but this was also the article which exhibited the greatest integration of text and animation. Experimentation is continuing to establish whether similarly good results can be obtained with a descriptive article when animation integration is improved. Students who have seen only the text are the most successful at recalling what they have read. The addition of animation to a text produces more impact on inference than on recall, and the multimedia group of students scored more highly than the others on inferencing. At a still higher level of comprehension - the ability to identify the main themes in an article or to enact a procedure described in an article -multimedia demonstrated a clear advantage over text alone.