Conceptual modeling is an essential activity during information systems development and, accordingly, a learning task faced by students of Information Systems. Presently, surprisingly little is known about how learning processes of conceptual modeling proceed, and about modeling difficulties learners experience. In this study, we integrate complementary modes of observation of learners' modeling processes to identify modeling difficulties these learners face while performing a data modeling task using a modeling tool. We use the concept of cognitive breakdowns to analyze verbal protocols, recordings of learner-tool interactions and video recordings of learners' modeling processes and survey learners about modeling difficulties. Our study identifies five types of modeling difficulties relating to different aspects of constructing conceptual data models, i.e., entity types, relationship types, attributes, and cardinalities. The identified types of modeling difficulties motivate a taxonomic theory of modeling difficulties intended to inform design science research on tool support for learners of conceptual modeling.