The basic premise underlying this research is that scientific phenomena are best learned by creating an external representation that complies with the complex and dynamic nature of such phenomena. Effective representations are assumed to incorporate three key characteristics: they are graphical, dynamic, and provide a pre-specified outline of the domain. This study examined the impact of these characteristics on performance and learning. High school students first read an instructional text about glucose--insulin regulation and then created a representation of its content. Representations differed regarding the key characteristics such that the summary (n = 15), concept map (n = 16), model (n = 23), and outlined model (n = 21) all incorporated one additional characteristic compared to their precursor. Main results indicated learning effects in each of these four conditions. Furthermore, creating a model was found to enhance students' learning more than creating a concept map, and students who completed an outlined model were found to learn more than those who created a model from scratch. In conclusion, this study does not univocally verify the necessity of all key characteristics individually, but the results do show that a representational format that combines all key characteristics enhances learning more than other formats.