Most chemistry teaching operates at the macro (or laboratory) level and the symbolic level, but we know that many misconceptions in chemistry stem from an inability to visualise structures and processes at the sub-micro (or molecular) level. However, one cannot change a student’s mental model of this level by simply showing them a different, albeit better, model in an animation. Molecular-level animations can be compelling and effective learning resources, but they must be designed and presented with great care to encourage students to focus on the intended ‘key features’, and to avoid generating or reinforcing misconceptions. One misconception often generated is the perception of ‘directed intent’ in processes at the molecular level, resulting from the technical imperative to minimise file size for web delivery of animations. An audiovisual information-processing model – based on a combination of evidence-based models developed by Johnstone and Mayer, cognitive load theory, and dual-coding theory – has been used to inform teaching practice with animations, and seed questions for research on student attributes affecting development of mental models using animations. Based on this model, the constructivist VisChem Learning Design probes students’ mental models of a substance or reaction at the molecular level before showing animations portraying the phenomenon. Opportunities to apply their refined models to new situations are critical. [Chem. Educ. Res. Pract., 2006, 7 (2), 141-159]