In this paper we discuss the use and challenges of identifying communities with shared semantics in Enterprise Modeling. People tend to understand modeling meta-concepts (i.e., a modeling language’s constructs or types) in a certain way and can be grouped by this understanding. Having an insight into the typical communities and their composition (e.g., what kind of people constitute a semantic community) would make it easier to predict how a conceptual modeler with a certain background will generally understand the meta-concepts he uses, which is useful for e.g., validating model semantics and improving the efficiency of the modeling process itself. We demonstrate the use of psychometric data from two studies involving experienced (enterprise) modeling practitioners and computing science students to find such communities, discuss the challenge that arises in finding common real-world factors shared between their members to identify them by and conclude that the common (often implicit) grouping properties such as similar background, focus and modeling language are not supported by empirical data.