In this paper we present our vision of how the practice of ontology analysis and engineering could be used to support the integration of (conceptual) models capturing various different aspects of an enterprise, originating from different stakeholders. Creating an overall enterprise model (or architecture) requires a deep understanding not only of what the words used in models mean to their creators, but also how the various modelers interpret the meta-conceptual constructs they use (i.e., categories in the modeling language). We argue this cannot be achieved by assuming a priori consensus on the used semantics (as is common practice), but that we need to incorporate methods that let us discover the existing categorization tendencies of specialized discourse communities and their individual members. Thus, we propose that conceptual integration in enterprise modeling, and perhaps also in other fields, would strongly benefit from an approach to Ontology that is more concerned with figuring out how people actually categorize than with stipulating how they ought to categorize. To show the possibilities such an approach offers we describe how the semantic differential can be adapted to fit this need and demonstrate its application by investigating the categorization tendencies of some computing science students in regard to modeling notation constructs.