Molecular visualizations have been widely endorsed by many chemical educators as an efficient way to convey the dynamic and atomic-level details of chemistry events. Research indicates that students who use molecular visualizations are able to incorporate most of the intended features of the animations into their explanations. However, studies also suggest that learning from visualizations is imperfect. In this study a new theoretical framework, variation theory, was used to examine learning and understanding based on students' reported experience of variation while viewing visualizations. Student metacognition was examined to gain insight into how students recognized variation between their mental models and the visualization models. Results from this study provide visual evidence of a transitional state of understanding in which students' previous conceptions merge with new conceptions learned from the visualizations. However, limitations in students' ability to monitor their understanding made unpacking what students understood challenging, as students tend to communicate only what they believe to be the most essential details.