Designing visualizations of dynamic networks is challenging, both because the data sets tend to be complex and because the tasks associated with them are often cognitively demand­ ing. We introduce the Matrix Cube, a novel visual representa­ tion and navigation model for dynamic networks, inspired by the way people comprehend and manipulate physical cubes. Users can change their perspective on the data by rotating or decomposing the 3D cube. These manipulations can produce a range of different 2D visualizations that emphasize specific aspects of the dynamic network suited to particular analysis tasks. We describe Matrix Cubes and the interactions that can be performed on them in the Cubix system. We then show how two domain experts, an astronomer and a neurologist, used Cubix to explore and report on their own network data.