About a million years ago, I worked in a political consulting firm that specialized in "persuasion mail," which is all those pamphlets you get around election time telling you that Congressman Smith wants to poison your children and molest your house pets, whereas Challenger Jones is the most honest man to ever walk the streets of your district. Early in my time there, as we were designing one mail piece encouraging voters to be more fearful of marauding criminals, my boss said, "I want the cover to be black and yellow stripes." When I expressed my puzzlement, he said, "Yeah, it's ugly, but it's the most arresting color combination there is. You can't not look at it."
I'm not sure if he was right about black and yellow specifically, but the principle was sound: What's visually pleasing need not be what attracts your eye or what you'll remember. The Washington Monthly has an excellent article on info visualization guru Edward Tufte, an interesting character as well as a hugely influential figure who counts Karl Rove as a fan and is currently working to help the Obama administration present its data in ways that don't suck. Here's a tiny portion:
Tufte has coined several terms that have come to define his style, such as "data-ink ratio," the proportion of graphical detail that does not represent statistical information, and "chartjunk," ornamental and often saccharine design flourishes that impede understanding. Jonathan Corum, a former graduate student of Tufte's at Yale who is now the science graphics editor at the New York Times, said that using Tufte's principles is a way of respecting the reader’s intelligence. "I think of it like I'm starting with this rough, unpolished stone," he said. "So how do you cut the perfect gem?"
A lot of Tufte's work is about how to make inscrutable graphics understandable and clear. After you read him, you'll never use Excel defaults again. But what if he's not right about everything? I recently came across a study (described well here) that showed people absurdly overdone charts and simple charts displaying the same information, and found that after a time, people were actually better able to recall the junky charts. Here's one pair they used:
One thing that struck me about this study is that the non-junky charts they used are really bad, just in a different way. They give up opportunities to highlight the important data points or trends, for instance. The choice isn't between something that's totally drab and something that's brimming with chartjunk. Excessive ornamentation is just one of many ways charts can be bad, down to the most basic choices like which type of chart to use (short answer: pie charts are almost always a bad idea). I'd be curious to know how the junky charts perform against well-crafted, non-junky charts. You can make an elegant, attractive, informative chart that doesn't have an animal growling at you, or (in the classic case) a display of oil prices represented by barrels. On the other hand, now that data visualization is such a hot topic, a lot of people are doing it, and I've seen plenty of beautiful visualizations that are absolutely impossible to understand, and therefore fail at their most fundamental task of communicating information.
But what if the best way to make someone remember the content of a chart is by using garish, ugly ornamentation? That would make the world a less beautiful place.
