We assess the empirical development of scientific homogenization and differentiation processes in terms of group development comparing U.S. and German sociology from 2004 to 2019. We introduce hierarchical multi-layered Bayesian stochastic block models to the issue of scientific group development. Based on publication data from Scopus, we analyze group formation simultaneously as collaboration networks, reference networks, and topical networks. Employing this method allows us to trace back the temporal dynamics of multidimensional group development and formation for each national scientific field. Overall, the findings indicate clear path dependencies in both fields that have been shaping processes of differentiation and homogenization in U.S. and German sociology and that continue to impact on the formation, stabilization, and dissolution of scientific groups.