What’s In A (Color) Name?
The role of color names in graphical perception.
A very important dimension in visualizations today is color. Selecting the right colors for spatial representations — such as heat maps and choropleths — is therefore essential for greater inference. How then, can a designer evaluate which colors might evoke higher perceptual accuracy? Given the large spectrum of colors to choose from, colormaps can come to the rescue. But how can a designer go about selecting the right colormap?
Reda et al. investigate and suggest in an award-winning study that linguistic associations of colors — of all parameters — are effective and have high interpretability. In fact, color nameability may even outweigh existing perceptual systems!
The authors study two linguistic metrics of colormaps — name variation and name salience. Variation is a “measure of the number of nameable colors”, while salience is the “degree to which those colors have unique names”. A second goal is to test these findings across a variety of representations — choropleths and two-dimensional geographical maps, for instance — and evaluate their performance.
The authors mathematically formalize the concepts of salience, variation and perceptual discriminability — all three of which are functions of a continuous colormap C. They select 6 colormaps, or rather, three pairs: <teal-beige and purple-pink>, <blue-brown and grey-red>, and <turbo and jet>. Each pair consists of a low salient and highly salient colormap, as can be seen below.
The colormaps are evaluated in two experiments conducted with participants from Amazon Mechanical Turk. The participants are given a graphical inference task in which they are shown representations from different models, and have to pick the odd one out. In the first experiment, the representations are choropleths and smooth maps; in the second, they are heat maps and scalar fields. Both experiments result in jet to be the most accurate colormap, and better performing than turbo, an equivalent rainbow scale. The explanation for jet’s effectiveness lies in its “variety and salience of its color names (prototypical blue, green, yellow, orange, and red)”. The paper thus establishes that color nameability affects people’s interpreting ability when deriving insights from spatial visualizations.
But how can color names be of any relevance, when color is simply representing some attribute of quantitative information? Especially when there is certainly no correlation between color names and data?
In an attempt to answer this burning question, the authors speculate that we are better at graphical inference when the representations afford “visual grouping (or ‘chunking’)”. They also suggest that nameability improves perceptual discriminability, and therefore, salient names are effective in communication, and also in one’s own abilities in problem solving.
While the paper excels with its contributions in color interpretability and visual representation, it has its shortcoming in not establishing the performance of color nameability beyond graphical representations. Additionally, given the importance of accessibility in our systems today, the paper could have also gone deeper in exploring this study for the visually impaired.
In conclusion, the paper presents some excellent findings that would be insightful to designers in their choice with color encodings. It also gives scope to explore nameability by language and region, and create new colormaps that combine nameability with established principles of perceptual ability.
References
- Reda, K., Salvi, A. A., Gray, J., & Papka, M. E. (2021). Color Nameability Predicts Inference Accuracy in Spatial Visualizations. Computer Graphics Forum, 40(3), 49–60. https://doi.org/10.1111/cgf.14288
