Abstract Assuming that all theoretical texts are at least intentionally aimed at introducing, defining, or relating concepts to one another, we propose a system for categorizing the components of such texts: we argue that on an abstract level, all parts of a theoretical text can be assigned to the categories theses, arguments, and notions. In the article, we explain how these categories were derived and how sthey can be used to create concept maps of theoretical texts, facilitating the analysis of the relation of such concepts in and across texts. We also report on an annotation workflow in which our theoretical concept is subjected to empirical testing and which serves as the basis for (semi-)automatic recognition of the three text parts and the creation of concept maps.