A few thoughts on E. Tufte's Visual Display of Quantitative Information

March 22, 2021

Edward R. Tufte is probably the most controversial data visualisation practitioner. I'd heard data professionals argue about his ideas dozens of times, even before I acquired one of his books myself. Now that I've read the Visual Display, I'd describe it as inspiring, but also a little arrogant and cold.

Let me explain.

Inspiring design and examples

I love the design and the layout of the book. It's carefully thought-through and minimalistic, just like the author wanted. It's rather rare to see a dataviz book designed by the author themselves, as the publishers usually take care of that. This is the reason Tufte self-published and the results speak for themselves.

The content in the first chapters of the book is inspiring, too. The author discusses visualisation examples that are quite design-heavy, like this line chart by William Playfair:

Or this rather complex graphic, also by Playfair:

Seeing these designs surprised me, as Tufte's praise for them didn't align with his reputation of dataviz best practices dictator. Well, I didn't have to wait long for the tone of the book to change.

Arrogant critique

A couple of chapters in, the ideas become more extreme. Tufte criticises many existing graphics. Some of them deserve it, like the one below with doubtful area calculations (Time, 1979):

However, I don’t appreciate his tone and approach. Tufte makes a list of journalistic and scientific publications and ranks them based on (his own arbitrary measure of) the sophistication of their graphics.

And it gets worse. After critiquing the lack of sophistication in the press, the author introduces the build a duck syndrome to describe computer-made decorative graphics. To me, this feels like an unnecessarily obnoxious way to express one's disagreement with the designs other people made.

Cold data-to-ink ratio

The chapters that follow dive deep into suggestions for good graphical design (no ducks!). I strive for simplicity in my own designs, but Tufte takes his (in)famous concept of the data-to-ink ratio to the extreme. Consider, for instance, the traditional representation of the box plot below:

In Tufte's suggested redesign, the box plot becomes this:

I don't know about you, but I would never guess on my own that the tiny lines above show a distribution.

Tufte adheres to a very strict rule of how much ink there should be on a page, and forgoes the idea of a middle ground. We probably don’t need dark distracting gridlines or borders, but there’s also the option to make them gray and hence push them to the background. Perhaps the data-to-ink equation should account for the saturation of the ink too, not only for its mere existence.

***

In the epilogue of the book, Edward Tufte writes: "<...> most principles of design should be greeted with some scepticism." I would apply this to his own book, too. I suggest you read it, as it has become a classic in the big library of writings on data visualisation. But make sure this is not the only book you read on the topic. There is more to graphics than the cold data-to-ink ratio. You also want your readers to be engaged, moved, and even to marvel at your work! ✨