While existing toolkits offer users a means for creating custom data visualizations, building complex, interactive systems remains a tedious task. The most notable shortcoming is the inability to compose interactive visualization components and marks into a cohesive whole without a great deal of rework to scale and place marks manually. We introduce CIViL, a domain specific language for creating composable and interactive information visualizations. Focusing on concise representations and consistent semantics, CIViL allows users to specify interactive visualizations by composing simple visual components. By leveraging declarative principles, users are able to specify visualizations in terms of both the visual elements and the behavior that they desire, without needing to consider how the visualization will be generated, scaled, or displayed. We present a prototype implementation in Haskell along with several examples to demonstrate the potential of the language.