This article presents "concept landscapes"--a novel way of investigating the state and development of knowledge structures in groups of persons using concept maps. Instead of focusing on the assessment and evaluation of single maps, the data of many persons is aggregated and data mining approaches are used in analysis. New insights into the "shared" knowledge of groups of learners are possible in this way. Electronic collection of concept maps makes it feasible to aggregate the data of a large group of persons, which in turn favors a data mining approach to the analysis. The educational theories underlying the approach, the definition of concept landscapes, and the accompanying analysis methods are presented. Cluster analysis and Pathfinder networks are used on the aggregated data, allowing new insights into the structural configuration of the knowledge of learners. Two real-world research projects serve as case studies for experimental results. The data structures and analysis methods necessary for working with concept landscapes have been implemented as a GNU-R package that is freely available.