While interdisciplinary courses are regarded as a promising method for students to learn and apply knowledge from other disciplines, there is limited empirical evidence available whether interdisciplinary courses can effectively "create" interdisciplinary students. In this innovative quasi-experimental study amongst 377 Master's students, in the control condition students were randomised by the teacher into groups, while in the experimental condition students were "balanced" by the teacher into groups based upon their initial social network. Using social network analysis, learning ties after 11 weeks were significantly predicted by the friendship and learning ties established at the beginning of the course, as well as (same) discipline and group allocation. The effects were generally greater than group divisions, irrespective of the two conditions, but substantially smaller than initial social networks. These results indicate that interdisciplinary learning does not occur "automatically" in an interdisciplinary module. This study contributes to effective learning in interdisciplinary learning environments.