I am interested in any academic research or guidelines that suggests the most effective use of color in scientific visualisations. Particularly I am interested in the merits of black on white or white on black color combinations in charts and their effect on readability and comprehension.
Can anyone point me in the right direction?
UPDATE: As a potential side point, I am interested in any advice on what colors to use rather than merely whether to use color or not.
Nasa did a lot of good work on the subject. They made a website a while ago. Unfortunately, it has just been taken down: http://colorusage.arc.nasa.gov/ Perhaps it's only a temporary glitch...
Else I could point you to my own site: http://colorguru.net. Only in Dutch for the time being. But it may perhaps be useful if you use Google to translate it.
Using black on white or white on black is subject of many debates. It all boils down to the background light you have in the room. If you are working in a bright environment, it is better to use black on white. When working mostly in the darker environments with limited light, use white on black. It also depends on how long the users are going to spend watching your program.
Regarding colors, always make sure that colorblind people won't have a problem using your graph. They (mostly) cannot differentiate green from red which are commonly used as a cues for good (green) or bad (red) data.
I prefer light gray backgrounds or soft colors which reduce eyestrain (notice the soft colors in the Plum theme for Windows 98:
What about this guide for using color in charts. It's a practical guide for using colours in charts and presentations. http://www.perceptualedge.com/articles/visual_business_intelligence/rules_for_using_color.pdf
This first link does not contain a recommendation suitable for scientific visualization. Rather, it is a summary of research on visual displays - including the effect of color on performance.
From the US Army's report on helmet-mounted displays.
"A number of studies have expounded on the positive impact of color on performance. In one of the more comprehensive studies, DeMars (1975) concluded that, for certain applications, color enhanced accuracy, decision time, and workload capability. However, Davidoff (1991) and Dudfield (1991) found that the actual significance of color far outweighed its perceived importance. An investigation (Spenkelink and Besuijen, 1996) of whether the use of color, and the resulting available chromatic contrast, could help improve performance in the presence of low luminance contrast concluded that only under special conditions was there an additive effect, and, in general, chromatic contrast cannot be substituted for luminance contrast. Rabin (1996) compared Snellen and vernier acuity, contrast sensitivity, peripheral target detection, and flicker detection for simulated green (x = 0.331, y =0.618) and orange (x = 0.531, y = 0.468) phosphors. For central visual tasks, no differences were found. However, peripheral target detection was found to be enhanced for the green phosphor."
A few things to keep in mind when reading that report. (1) the environment in which one uses an HMD and a scientific visualization are quite different. (2) most likely, performance outweighs preference when identifying the best design for an HMD, (3) visual appeal and performance may be equally important in scientific visualization, (4) colors selected by military/aviation/nuclear power/maritime system designers are invariably highly saturated colors that are not suitable for commercial software.
The importance of visual appeal can be seen in the absence of an achromatic color scheme in the set provided by the Colorbrewer. These are empirically derived color gradients for use as color fills. They are suitable as fills for maps and charts. The research supporting these colors is not relevant to colors for lines or points.
This is an IBM research report with recommendations for gradients in visualization. Take a look at Table 1. Although several of their recommendations include gray as a starting point, all of the gradients include hue variation.
If you have access to a college library database this article could be of some use: http://dl.acm.org/citation.cfm?id=75578
