This article reports the use of concept mapping to reveal patterns of student learning (or non-learning) in the course of master's level teaching for research methods. The work was done with a group of 12 postgraduate students, and the concept maps of four individuals produced "before" and "after" a single teaching intervention are shown in detail. The data are presented as case studies that document the incidence of deep learning, surface learning and non-learning. These are terms that are widely used in the educational research literature, but most evidence for these learning approaches comes from students' conceptions of learning, not from empirical measures of changes in knowledge structure. Here precise criteria for defining change in terms of deep, surface and non-learning are developed, and concept mapping is used for assessment of learning "quality" using these criteria. The results show that deep, surface and non-learning are tangible measures of learning that can be observed directly as a consequence of concept mapping. Concept mapping has considerable utility for tracking change in the course of learning, and has the capacity to distinguish between changes that are meaningful, and those that are not. This is discussed in the wider context of learning, and teaching and research. (Contains 2 tables and 9 figures.)