Mental models are internal representations of external phenomena. During their interaction with visualizations, the users construct mental models to represent these visualizations internally, to visually reason on them and solve problems with them. This chapter synthesizes existing theories on mental models and visualization to discuss their role and relevance for the design of visualization systems. From a mental models perspective, we discuss two challenges of visualization design: (a) supporting the initial construction of mental models and (b) supporting the integration of information from multiple views by synchronous or sequential coherence techniques. We argue that the theory of mental models allows to understand visualization research and practice in a more unified fashion as an advanced model-building endeavor, operating on human computer ensembles engaged in “distributed cognition.”