Bertin - Special topic
Bertin’s Sémiologie Graphique @ 50 – II
Color can be one of the most satisfying and most frustrating design experience, as color has various perceptual properties, and semantic connotations that are difficult to control for a designer. This challenges relates to any information visualization, across all of the sciences. Color features prominently amongst Bertin’s (1967/1983) list of visual variables, the only one with 3 dimensions that can be varied (Bertin, 1983: 87). Some of his color principles have been tested empirically (i.e., Brewer, 1994, Garlandini & Fabrikant 2009), and some of this research has led to widely used software to support users in selecting appropriate color schemes for maps, statistical graphics, and other multivariate information visualizations.
We report on an experiment with commonly employed schemes in neuroimaging for brain activity maps (i.e., spectral/rainbow and heated body schemes) and in cartographic displays (i.e., univariate and bi-variate schemes), following Bertin’s (1967/1983) color principles. Experts in neuroimaging (N=134) and geovisualization (N=197), and also participants (N=486) sampled from the general public (via Amazon Mechanical Turk) were recruited worldwide to participate in an online web-based study. Participants were asked to interpret shown data and provide trust assessments with 1) brain maps showing various brain activity states (i.e., from normal brains to brain death), and 2) corresponding cartographic displays, depicting the outcomes of an environmental sustainability model for a particular country that matched the employed brain state descriptions.
We found significant differences in data interpretations and trust ratings, across color schemes and expertise. Contrary to our hypothesis that domain experts’ interpretations and ratings would be least affected by a particular color scheme in their own field, we did find that neuroimaging experts were most strongly influenced in their assessments, due to differing color schemes.
Geovisualization experts exhibit largest differences between color preference and data trust ratings, particularly for both the rainbow and heated body color schemes, commonly employed in neuroscience. The heated body scheme is diametrically opposed to Bertin’s (1967/83) principle to use color shades that progress in parallel to data values, that is, the higher the data value, the darker the shade. This principle has even emerged as one of the few standards that statisticians seemed to have agreed upon since their first few international congresses by the end of the 19th century (Palsky, 1999).
Geovisualization experts trusted displays with the rainbow and heated body schemes more than their stated preferences, that is, when rating the suitability of this color scheme for visualization. This is especially noticeable for the geovisualization experts’ ratings with brain maps.
Our empirical results suggest to further raise awareness for the potential influence of visualization design decisions on scientific data interpretation in changing contexts, and across domain expertise, and to provide empirically validated design defaults in widely used visualization tools across the sciences.
Christen, M., Brugger, P., Fabrikant, S.I. (2015). The influence of color-manipulation on data interpretation in neuroimaging and geographic information visualization. 2015 INS Annual Meeting, Chicago, Oct. 15-16, 2015.
Brewer, C. A. (1994). Color Use Guidelines for Mapping and Visualization. In: MacEachren, A.M., and Fraser Taylor, D.R., (eds.), Visualization in Modern Cartography, Elsevier, New York, NY: 269-278.
Garlandini, S. & Fabrikant, S. I. (2009). Evaluating the Effectiveness and Efficiency of Visual Variables for Geographic Information Visualization. In: Stewart Hornsby, K., Claramunt, C., Denis, M. and Ligozat, G., (ed.), LNCS 5756: Spatial Information Theory: 9th International Conference (COSIT 2009): 195-211.
Palsky, G. (1999). The debate on the standardization of statistical maps and diagrams (1857-1901). Elements for the history of graphical language. Cybergeo: European Journal of Geography, 85: online.