Scientific discourse comprises a specialized, lexically dense and abstract language, with a need for students to handle multiple resources for meaning making in parallel, e.g. verbal language, graphs, gestures, models, and symbolic language (Halliday & Martin, 1993; Lemke, 1998). This puts high demands on the learner, in particular for multilingual students as they learn science in their second language. We will present results from an interdisciplinary project on interaction in science classrooms, focusing on how science content is elaborated and negotiated through various semiotic resources. The project draws on i) social semiotics (Kress, 2010), including systemic functional linguistics (SFL) (Halliday & Matthiessen, 2004), and ii) practical epistemology (Wickman & Östman, 2002). Data (video/audio recordings, digital photographs and texts collected in the classrooms) is analyzed through SFL-analyses, multimodal analyses and practical epistemology. We will concentrate on findings from a primary school classroom working with "time". Here, the teacher uses a variety of semiotic resources in different semiotic modes, and students are engaged in small group work and whole class discussions. Our analyses, made from different perspectives, reveal a need for interdisciplinary cooperation for researchers to be able to give a broader picture of what learning opportunities are given the students.