AbstractEmbodied learning tools (ELTs), such as building and manipulating models, bolster students’ learning of spatially demanding science content, such as stereochemistry. However, studies comparing the effectiveness of virtual to physical ELTs on student learning are limited. This study compares online instruction using a virtual ELT, virtual models, to a physical ELT, physical models, on organic chemistry students' stereochemistry understanding. Furthermore, it identifies which tool promotes the transfer and application of spatial science content to higher order concepts, such as reaction mechanisms. Lastly, the role of students' spatial skills is considered. To assess the magnitude of the effect of each tool, two control groups were included, a text‐based instruction and a no instruction group. Results revealed that the physical model group showed greater stereochemistry understanding compared to the no instruction group, while there was no difference between the virtual model and the no instruction groups. When asked to apply their knowledge to reasoning about reaction mechanisms, the physical model group outperformed the virtual model group and both control groups. Students' spatial skills did not influence their understanding of either concept in any learning condition. These results suggest that regardless of spatial skills, physical ELTs may provide the necessary scaffolding for students to learn spatially demanding science content, whereas virtual ELTs are no more effective than receiving no formal instruction. Furthermore, physical ELTs may be more effective than virtual ELTs in facilitating the application of that knowledge to more complex concepts. These findings have broad implications for online science pedagogy.