For designing effective and tailored instruction, valid instruments that measure the level of expertise are necessary. We propose a graph-oriented approach for in-depth analyses of knowledge structures. Therefore, four measures of integration and encapsulation of knowledge structures were validated in an experiment. Participants (six experts and six intermediate students) recalled and explained the symptoms and laboratory data of a medical case description in the domain of cardiology. The results showed that the graph-oriented measures were more discriminative towards expertise-related differences than classic measures. Thus, our graph-oriented measures offer a more adequate and a more fine-grained analysis of knowledge structures.