An alarming level of the “what-would-Tufte-say” syndrome can be found in this post and some of its comments discussing a New York Times’s infographic. This syndrome has some recognizable features like the extensive use of “chart junk”, “lie factor” or other terms and expressions coined by Tufte that reveal a somewhat misunderstanding or abusive usage for legitimacy purposes (argument by authority).
You can spot the syndrome as soon as you land on the page. In spite of what the author claims, there is no “Chart Junk in the New York Times”: there is no grid, there is no moiré, there are no “decorative forms or computer debris” (Tufte, The Visual Display…), there is no Duck.
A never-ending story: the glamorous scale break discussion
The real problem in the chart is the scale break, but that is not “chart junk”. It could be “the data-ink ratio”. It could be “the lie factor”. But is it? The never-ending discussion over the use of scale breaks is almost as absurd as the one over pie charts. And the answer is, as always, the definitive “it depends”.
A scale break should be obvious, explicit, prominent. You must be sure that the reader will notice it. And it should not be used lightly. But if zero doesn’t make sense with your data or the variation is so small that you get an empty chart with an almost straight line at the top you should use a scale break to improve resolution.
As a rule of thumb, you can use scale breaks in line charts but not in column charts. Or at least that’s what Stephen Few believes:
“You should generally avoid starting your graph with a value greater than zero, but when you need to provide a close look at small differences between large values, it is appropriate to do so.
Never eliminate zero from the quantitative scale, however, when bars are used to encode the values. Why? Because a bar encodes quantitative value primarily through its length, and, without zero as the base, the length will not correspond to its value.”
But Kosslyn expresses a different opinion:
“Unless the zero value is inherently important, make the visible scale begin at a value slightly lower than the smallest value in the data, and the upper value slightly larger than the largest value.”
(His example uses column charts…)
Tufte would be proud
Hermeneutics enjoys a long and noble history, so it comes as no surprise those 3.230 results that you get if you google for “‘Tufte would be proud'”.
I would be proud of myself if I could fully understand and practice his design ideas and understand also at what point they stop working. Because they do. And sometimes…
- pies can be used;
- you are not lying just because you broke the scale;
- even the “above all, show the data” principle can be disputed;
- an excessive dosage of a minimalist design can be perceptually wrong;
- the audience doesn’t get your cleverly designed chart;
- people just don’t like it;