Almost all research evaluation exercises, by construction, deliver data at the level of departments and universities, not at the level of individuals. Yet, the aggregate performance is the average of the performance of individual researchers. This paper explores the issue of the relative magnitude of variability in performance within departments and between departments. It exploits anonymized data at the individual level from one of the largest research evaluation exercises, the Italian VQR 2004–2010 (Valutazione della Qualità della Ricerca). If the variability between departments were much larger than variability within departments, we would see evidence of a process of stratification, or vertical differentiation, arguably driven by competition and researcher mobility. The data show that the opposite pattern is at play.