On chart visualizations and chartjunk

Tufte, “Color and Information”

With progress in computer technology and increased familiarity among users since this 1990 edition, it seems worth revisiting Tufte’s opening claim for this chapter that any number of colors over 20-30 produces negative returns. We’re now used to attributing number values to hues (in the RGB24 system, for example), so it makes sense to create visualizations that associate color values directly with data values. Color can serve as very precise indicator of measurement, if a range of data is made to fit 256 shades, for instance. Tufte gestures at this in his analysis of Ware and Beatty’s work on p. 88, though the paragraph on color’s “multidimensional quality” stops short of saying whether viewers could parse these clusters easily or not. My hunch is that these clusters worked fairly well in the 90s (an “obvious improvement over black-only dots,” as Tufte says), but also that they’d work even better now: partly because we’d improve the visualization with higher resolution and more precise customization, and partly because viewers are now so much more acquainted with these computer-based gradations.

Few “The Chartjunk Debate”

The poles of the debate are represented by Tufte (who promotes a minimalist approach to chart design) and Holmes (an advocate of artistic, metaphorical additions to emphasize meaning, and Tufte’s victim in the “Diamonds” critique). Few and Bateman et al. fall somewhere in between: Few is sympathetic to Tufte, but wants to define chartjunk a little more precisely as anything that does not contribute at all to chart information; Bateman et al. concluded from their small study that chartjunk didn’t contribute to comprehension but contributed slightly to memorability. The lesson I take from Few’s middle road, perhaps primed by my reading and commentary on Tufte above, is oriented around color: the striking feature of Tufte’s improved charts is that they make use of 1-2 simple colors to make charts more engaging while not making them overwhelming (that is, between the dull simplicity of the study charts and the chaos of the Powerpoint/Excel creation… but definitely closer to the dull end). Bar charts are a particular favorite of Few’s, it seems, and there’s no attempt to respond to the kinds of attacks Drucker made on bar charts; but, even if Few’s basic formats and 1-2 accent colors don’t account for humanistic ambiguities, that’s not the task at hand. They’re undeniably clearer and friendlier.

Enter text in Markdown. Use the toolbar above, or click the ? button for formatting help.