Many of todays technology-based resources rely on animated diagrams to explain complex dynamic subject matter. Despite the rapidly growing popularity of such animations, research suggests that their educational effectiveness can leave much to be desired. A major problem that learners face in comprehending animations is in decomposing the presented information into a form that furnishes appropriate raw material for building high quality mental models. This paper proposes an alternative to existing design approaches that shifts the prime focus from the nature of the external representation to the internal composition activity learners engage in during mental model construction.