Background: There is growing interest in using argumentative discourse in educational settings. However, in a previous study, we found that discourse goals (persuasion vs. consensus) while arguing can affect student outcomes in both content learning and reasoning. Aims: In this study, we look at argumentative discourse data from a previous study to ask how differences in discourse might account for the differences we observed in learning and reasoning outcomes. Sample: One hundred and five dialogues (57 disputative, 48 consensus) between 7th grade science students attending a public high school near Tarragona, Spain. Methods: Participants were randomly assigned to conditions and paired with peers who disagreed with them on three topics related to renewable energy sources. After instruction on each topic, they were asked to either "argue to convince" (persuasion condition) or "argue to reach consensus" (consensus condition) on that topic. Conversations were audio-recorded and transcribed for analysis. Results: Students in the persuasion condition engaged in shorter conversational exchanges around argumentative claims and were more likely to use moves that foreclosed discussion, whereas students in the consensus condition were more likely to use moves that elicited, elaborated on, and integrated their partners' ideas. Conclusions: When arguing to reach--rather than defend--a conclusion, students are more likely to coconstruct knowledge by exchanging and integrating arguments. These findings are consistent with predictions about the potential of argumentation for knowledge building and suggest that teachers must attend to discourse goals when using argumentation to support learning and reasoning.