Traditionally, content assessments have been carried out through written language. However, the latest standards in U.S. K-12 education expect all students, including English learners (ELs), to demonstrate their content learning using multiple modalities. This study examined the performance of fifth-grade students at varying levels of English proficiency on four science tasks that elicited responses in visual, written, and oral modalities. Findings revealed that approximately half of students performed differently in visual versus written modalities on each task. However, performance did not consistently favor the visual modality for ELs, likely due to challenges related to visual representation in some areas of science. Additionally, triangulating students' visual and written responses with their oral responses yielded more accurate interpretations of their science understanding. Collectively, these findings indicate the potential of multimodal assessment for providing more complete and accurate information about what ELs and their peers know and can do in the content areas.