This study investigated the development of representational competence among organic chemistry students by using 3D (concrete and virtual) models as aids for teaching students to translate between multiple 2D diagrams. In 2 experiments, students translated between different diagrams of molecules and received verbal feedback in 1 of the following 3 intervention conditions: with concrete models, with virtual models, or without models. Following the intervention, diagram translation accuracy was measured in 3 posttests, which were with models, without models, and after a 7-day delay. The virtual models in the 2 experiments differed in the level of congruence between the actions performed with the input device and the resulting movement of the virtual model. Study 1 used a low congruence interface and Study 2 used a high congruence interface. Students learned more when models were available. In terms of learning outcomes, model-based feedback was superior to verbal-feedback alone, models served as a learning scaffold rather than a crutch, and learning with model-based feedback was resilient over a 7-day delay. Finally, concrete and virtual models were equivalent in promoting learning, and action-congruence of the interface did not affect learning. The results are discussed with respect to their implications for instruction in organic chemistry and science, technology, engineering, and mathematics disciplines more generally.