Rainbows Revisited: Modeling Effective Colormap Design for Graphical Inference

Color mapping is a foundational technique for visualizing scalar data. Prior literature offers guidelines for effective colormap design, such as emphasizing luminance variation while limiting changes in hue. However, empirical studies of color are largely focused on perceptual tasks. This narrow focus inhibits our understanding of how generalizable these guidelines are, particularly to tasks like visual inference that require synthesis and judgement across multiple percepts. Furthermore, the emphasis on traditional ramp designs (e.g., sequential or diverging) may sideline other key metrics or design strategies. We study how a cognitive metric-color name variation-impacts people's ability to make model-based judgments. In two graphical inference experiments, participants saw a series of color-coded scalar fields sampled from different models and assessed the relationships between these models. Contrary to conventional guidelines, participants were more accurate when viewing colormaps that cross a variety of uniquely nameable colors. We modeled participants' performance using this metric and found that it provides a better fit to the experimental data than do existing design principles. Our findings indicate cognitive advantages for colorful maps like rainbow, which exhibit high color categorization, despite their traditionally undesirable perceptual properties. We also found no evidence that color categorization would lead observers to infer false data features. Our results provide empirically grounded metrics for predicting a colormap's performance and suggest alternative guidelines for designing new quantitative colormaps to support inference. The data and materials for this paper are available at: https://osf.io/tck2r/.