A key goal of recent reforms in science education has been to support students' collective sensemaking, in which they work together to understand a natural phenomenon. This work is often done during class discussions, but facilitating those discussions is challenging. In this study, we used a framework of three discussion types to investigate how classroom discussions differ throughout a storyline curricular unit: "initial ideas" discussions aim to surface students' ideas about a phenomenon; "building understandings" discussions aim to construct claims based on an investigation or data; and consensus discussions aim to bring together multiple claims over time to explain a phenomenon or event. We used this framework to analyze the idea work and turns taken in classroom discussions from one highly skilled middle school teacher. This analysis resulted in three takeaways. First, we found the teacher consistently worked to surface and clarify his students' ideas. This is as an important step in collective sensemaking that comes before the construction and critique of those ideas and can support a sense of belonging for students who typically do not identify with science. Second, we characterized a distinct talk pattern closely associated with surfacing and clarifying ideas across all three discussion types: the Propose-Probe-Clarify-Restate (PPCR) sequence. PPCR is distinct from previously identified talk patterns in the teacher's use of clarifying and restating to honor the ideas students bring to a discussion and to restate them to ensure they are accepted into a public record. Finally, we found that each discussion type required a distinct set of teacher moves to achieve its goal. Surfacing and clarifying ideas is necessary but insufficient to support collective sensemaking. Therefore, different talk moves, such as asking for evidence and evaluating claims for their connection to the phenomenon of interest, are required in building understandings and consensus discussions.