Government agencies and other authorities often communicate earthquake risks using maps derived from geographic information systems. Yet, little is known about the effects of these maps on risk perceptions. While mental models research and other approaches are available to inform risk communication text design, similar empirically derived guidance is lacking for visual risk communications, such as maps, which are likely to trump text in their impact and appeal. This paper reviews the empirical research that might inform such guidance. Research on graphs, spatial and visual perception, and map design suggests that graphics increase risk avoidance over numerical risk representations, and countable visuals, like dots, can increase the accuracy of perceived risks, but not always. Cartographic design features, such as color, animation, interactivity, and depth cues, are all candidates to represent risk and uncertainty and to influence risk perception. While there are robust known effects of color (e.g., red = danger), with some cultural variability, animation can increase the salience of otherwise obscure features but is not uniformly effective. Depth cues, dimensionality, and the extent to which a representation depicts versus symbolizes a scene will influence the viewer's perspective and perception, depending on the viewer's familiarity with the scene; their effects on risk perception remain unclear. The translation and representation of technical information about risk and uncertainty is critical to risk communication effectiveness. Our review suggests a handful of candidate criteria for evaluating the effects of risk visualizations, short of changes in behavior: accuracy, accessibility, retention, and perceived risk and usefulness.