Visual problem solving with multiple representations is a critical component of chemistry learning and communication. Understanding how students comprehend and utilize visual representations is key to improving chemistry education, and a multimodal approach to understanding how students tackle visual stoichiometry problems offers insight into misconceptions and difficulties that they face. A mixed methods approach was used, employing multimodal data (eye-tracking, drawings, oral responses, and visual problem solving scores) to study participant interaction with representations and develop a framework for understanding college general chemistry students' metavisualization skills. Student performance during a PhET interactive simulation chemistry game was investigated using eye-tracking and qualitative analyses of a talk aloud protocol to isolate key mental blocks contributing to the participants' misconceptions. Cluster analysis and principal component analysis of gaze patterns revealed that participants follow coherent patterns when solving visual problems with multiple representations with respect to the equation, submicroscopic representations, and numbers provided in the question.

Participants were divided into high and low score groups based on quantitative analysis of responses to key questions associated with the conservation of mass in stoichiometric analysis and the groups were further investigated using the of multimodal responses from individuals within each group. Eye-tracking and cluster analysis were found to be valuable tools for framing how students solve chemistry problems with multiple representations.