Hierarchical graphs represent relationships between objects (like computer file systems, family trees etc.). Graph nodes represent the objects and interconnecting lines represent the relationships. In two experiments we investigated what concepts are necessary for understanding hierarchical graphs, what misconceptions evolve when some of the concepts are missing and how misconceptions can be prevented through instruction. Participants were taught different amounts of prior knowledge and then had to respond to a multiple-choice questionnaire with interpretive questions about graphs. In Experiment 1, 72 university students received different amounts of instruction about the concepts necessary to interpret hierarchical graphs. Through detailed analysis of readers' wrong responses to interpretive questions we identified a set of misconceptions. Participants maintained fewer misconceptions and performed better if they had been taught more conceptual knowledge. However, their overall performance was poor. In Experiment 2, 85 students were informed about possible misconceptions, in addition to the instruction of conceptual knowledge. With this intervention they obtained an acceptable level of understanding of hierarchical graphs. The discussion of the results draws on theoretical considerations for the evolvement of misconceptions such as failure to integrate visual and conceptual information and context specificity of the representation.