In this paper, we analyze whether structured PhD programs operate at optimal size and whether there are differences between different disciplinary fields. Theoretically, we postulate that the relation between the size of a PhD program and program performance is hump shaped. For our empirical analysis, we use hand-collected data on 86 Research Training Groups (RTGs) funded by the German Research Foundation (DFG). As performance indicators, we use (a) the number of completed PhDs and (b) the number of publications by RTG students (PhD students and postdoctoral researchers). Applying DEA with constant and variable returns to scale, we find that the optimal team size varies between 10 and 16 RTG students in the humanities and social sciences. In contrast, our empirical analysis does not uncover a systematic relation between size and performance for RTGs in the natural and life sciences.