Envisioning Information - Tufte

Color and Information

While there are more than a million colors discernable by the trained colorist, only about 20,000 colors are typically perceivable, and only about a maximum of 20 to 30 colors are impactful when used in visualizations. In Color and Information, Tufte breaks down the use of color in envisioning data, and how color can be used most efficiently, fluently, and coherently.

He references Imhof’s rules rules that he refers to regarding color in maps:

  1. Use color on muted/neutral fields to highlight data, and also to achieve harmony.
  2. Large patches of color should be muted.
  3. Colors should be themed and sprinkled around a single background color to ensure unity.

Tufte breaks down the purpose of color in data visualizations as either to 1) label, 2) represent, 3) measure, or 4) quantify. In his example, he says color is used to label different objects (e.g. ground vs ocean), represents reality with accurate colors or shading, measures altitude with shades, and finally, increases the aesthetic value through the map.

He goes on to describe how color should be used in representing, what colors should be used, and the effects of using color. For example, his “grand strategy is to use colors found in nature” because they are more familiar and coherent– because they are “natural,” they are less “colorjunk.”

He also gets technical with colors, breaking them down to hue (what we think of as “color”), saturation (amount of gray), and value (brightness/intensity). Here, he also starts discussing the impact of perception and how colors interact with each other, referring to the work of Josef Albers.

Finally, he talks about the redundancy of “signals” and representation, becasue that helps reduce ambiguity and confusion. This is important because it can emphasize certain properties and delineate differences even more.

The Chartjunk Debate

Tufte calls non-data and repeated data elements “chartjunk,” and criticizes it for being harmful, uninformative, and mostly ornamentation. Stephen Few, however, defends this ornamentation and tries to reconcile both sides and arguments.

Through a study that examined “chartjunk” vs plain charts, examining recall and comprehension, there is a qualification of ornamentation. Through the study, it was discovered that the ornamented charts were about the same amount of legible as the plain ones, but participants had a much easier time remembering the embellished charts later.

However, there were issues that seemed to cloud the study’s conclusion that embellishments are actually fine– namely, the charts were actually very easy to understand and had simple messages to begin with, there was only a small set of values, and embellishments were created by a trained designer. Stephen Few then goes on to talk about the extreme embellishments that constitute chartjunk, and shows the amount of junk that can be put into charts is actually quite impressive.

The paper ultimately seems to point at the term “chartjunk” and how there needs to be a beter definition. It is not that there needs to be a completely minimal, precise representation, but rather, a more curated, designed version in order to engage the reader, draw attention, and make the information memorable.