The Visual Display of Quantitative Information is an absolute classic on the creation and use of graphs. Done correctly, a good graph can make complex information instantly comprehensible, reveal relationships and patterns, and both delight and inform. Done poorly, a bad graph causes eyestrain, confusion, and the deliberate obfuscation of the truth. And in a world where graphs are ordinary, Tufte provides a quick history of how they came to be, and the cognitive leaps required.

Tufte rails against the sins of bad graphics: scaling and axes that lie about trends in the data; the use of unnecessary ink to convey redundant information; visual clutter and bad aesthetics. He advocates for a kind of elegant minimalism, conveying the most information with a few well-chosen lines of varying weights, and cleverly using edges and white space to mark boundaries, while supporting information with text. The advice is for a pre-computer graphics era (at least in my signed 1983 edition), but the aesthetics still hold, even if we aren't drawing graphs with a marker and straight-edge.

The problem is that Tufte turned out to be a voice crying in the wilderness. There are the majors flaws, like the use of flashy cluttered "infographics" that combine the worst features of text-heavy articles and data graphics. But then there is the minor things. I have at my fingertips about a half-dozen data visualizations packages, from Excel (boo!) to ggplot and bokeh. And not a single one, by default, does everything that Tufte says. They get close, but the defaults are not quite minimalist enough. And truly great graphs, like Minard's plot of Napoleon's invasion of Russia, with his army vanishing into the snows, still require an artist's touch.