Constrained interactivity for relating multiple representations in science: When virtual is better than real

• We compared virtual and concrete models during a spatial chemistry task.
• Both model types supported accurate performance.
• Task performance was more efficient with virtual models.
• Virtual models can constrain users to make task-relevant manipulations.

Virtual models are increasingly used in science education, especially in spatially demanding domains. However, few studies have directly compared the effectiveness of virtual and concrete models, or systematically characterized differences between them. Here, we compared students' accuracy and efficiency using virtual and concrete models to align and produce different representations in the domain of organic chemistry. Naïve undergraduate students learned the conventions of different molecular representations (diagrams and models) and then performed tasks that involved matching models to diagrams and using models to complete diagrams. The results indicated similar levels of accuracy for virtual and concrete models and greater efficiency for virtual models. Students preferred virtual models, but rated the usability of the two model types about equally. The efficiency benefit associated with using virtual models can be explained by their constrained interactivity, which prevented students from making task-irrelevant manipulations and increased the salience of the task-relevant information in the models.