Background: Dialogic teaching, specifically the use of argumentation in teaching, is seen as promoting scientific literacy and understanding. However, international evidence consistently shows that the prevailing modes of classroom talk are monologic. The problem is how to transform science education into dialogic and argumentative spaces. One way to do this is through the use of curriculum materials that support and scaffold dialogic practices. Although there is growing interest in promoting argumentation through curriculum materials, their effect on students' learning and the relationship between curriculum-supported argumentation and learning are still not clear. Purpose: The aim of this study was to contribute to this knowledge gap. We replicated research conducted in the UK as part of the epiSTEMe project, in which curriculum materials supporting dialogic classroom talk were developed and evaluated through a randomised experimental study. Sample: A total of 220 students (aged 10-11 years) from 18 classrooms participated in the study, with the classrooms spread across 18 public schools all located in Santiago, Chile. Design and method: We conducted an experimental study randomised at school level. Eleven teachers delivered science lessons following a teaching programme especially developed to foster dialogic and argumentative classroom talk (the intervention group), and seven teachers delivered lessons in their usual way (the control group). Students were assessed individually using pre- and post-measures of science content knowledge and argumentative skills. Results: The results showed that the intervention group obtained significantly higher pre- to post-gains on science content knowledge. Although the control group engaged in significantly more whole-class argumentative dialogue than the intervention group, this did not predict content knowledge learning in the control group. By contrast, in the intervention group the frequency of whole-class argumentative dialogue had a strong and positive effect on the delayed content knowledge post-test, controlling for initial measures.