The ability to interpret graphs and diagrams requires observers to map perceptual input to abstract concepts. For example, colormaps of neuroimaging data require mapping colors to different amounts of brain activation. Our lab is interested in how observers’ pre-existing biases in color associations influence their interpretations of visual representations. We are currently developing a perceptual-cognitive mapping (PCM) framework to address this problem, which is motivated by the structure-mapping theory of analogy (Gentner, 1983). The prediction that visualizations will be easier to interpret when internal mappings between percepts and concepts match the stimulus mappings as specified by legends or labels. The problem is understanding what determines internal mappings between percepts and abstract concepts. We are addressing this question in a variety of domains (e.g., data visualizations, navigational maps, and recycling). We are working towards formalizing a computational theory that optimizes color-concept mappings, which requires identifying a set of colors that have a natural correspondence to a set of domain items and which serve to discriminate among the items.