This article addresses, conceptually and empirically, the classification of public research organizations (PROs) understood as non-university and non-enterprise research-focused organizations that are public by nature or in which the government has an influence. The construction of archetypes of research performing organizations has been a standard method of analysis, as reflected in the Frascati Manual that guides national statistical offices to delineate the perimeter of the institutional sector of PROs. However, this practice has often overlooked the emergence of new types because traditional approaches to classification tend to characterize previously defined mutually exclusive categories, rather than allow evidence to reveal categories "ex-post." This gives rise to a number of concerns related to the scientific validity of the classification of entities in the organizational field of research. The present article discusses conceptual and methodological issues associated with different classificatory strategies. It also presents the empirical results of a taxonomical exploration that allows the identification of categories not determined "ex-ante." Our empirical strategy consists in applying clustering techniques on a number of organizational dimensions, chosen based on theoretical grounds and proxied by variables determined by data availability. We implement it on a pilot dataset of 197 research-focused organizations from eight different European countries.