This article reports the results of an ethnographic research about the multimodal science discourse in a sixth-grade sheltered classroom involving English Language Learners (ELLs) only. Drawing from the perspective of multimodality, this study examines how science learning is constructed in science lectures through multiple semiotic resources, including oral language, gestures, and visual products. Specifically, the study asks three research questions: 1. What is the nature of multimodal communication between the teacher and students? 2. How does science learning occur in the classroom from a multimodal perspective? and 3. How does language development occur in the classroom? Data include eight months of classroom observation with video and audio recordings, fieldnotes, formal, and informal interviews with the teacher and students, photographs of textbook pages, students' final products, and teacher products. Results show that the gap between the multimodal representation and communication, as well as the disconnection between teacher discourse and student discourse, provides limited evidence of the students' knowledge reconstruction. The gap between the everyday language and scientific language leads to limited language development. The study expands the current knowledge base about ELLs' multimodal classroom experiences in relating to science learning and language development. The results can be used to help both researchers and educators in the future to examine the multimodal instructional design in terms of its effectiveness for science learning. #