This paper explores the link between the composition and the performance of junior research groups. The authors argue that the heterogeneity-performance link depends on the type of heterogeneity (cultural vs. study field) and on the disciplinary area. The authors test their hypotheses on a data set of 45 junior research groups and find a U-shaped relation between cultural heterogeneity and performance in the humanities and social sciences, but no link between the two in the natural sciences. The link between study field heterogeneity and performance in the natural sciences is negative, and in the humanities and social sciences study field heterogeneity and performance are not related. Interaction within the group helps reap the benefits of heterogeneity. The study results are derived in the context of junior research groups in Germany, but are generalizable to other countries and contexts where PhD education is taking part in groups.