Evidence is beautiful if it's artfully displayed

June 22, 2002

Edward Tufte has made an art of the visual display of information. And he's made something of an industry of it too, with books, posters and a one-day course on information display that he presents some 30 times a year around the country.

In his brochures advertising the course, he uses a graphic (you can see it on his Web site, edwardtufte.com), drawn by Charles Joseph Minard in 1861, displaying Napoleon's disastrous invasion of Russia in 1812. The number of troops, originally 422,000 men, is represented by a thick band that narrows inexorably as the army sweeps east. Napoleon reached Moscow with only 100,000 men.

On the retreat westward, the steadily narrowing band is linked to the date and the temperature as the onset of a bitterly cold winter harrows the remnant. Just 10,000 men returned to the Polish border where the invasion began. It's quite possibly the best statistical graphic ever drawn, Tufte says, and in its own way as powerful an anti-war statement as Picasso's Guernica.

It's also possibly the most complicated advertisement ever published, and I've had a copy of it on my desk for years. So I just had to attend the seminar when Tufte presented it this week in Denver.

One of his themes is the need for information displays to get "out of flatland" -- that is, to show multiple variables on the two-dimensional surface of the paper or computer screen. His prop is a copy (it was once in the library of Ben Jonson) of Euclid's Elements published in 1570. To represent three-dimensional figures, the publisher glued 80 tiny paper cutouts on the page that fold up into geometrical shapes. In every copy.

After Tufte has shown the book, his assistant puts on her white gloves and carries it around the room so everyone -- some 350 people -- can see it up close.

He also has a first edition of Galileo's 1613 book on sunspots, and shows a brief video that one of his students made of Galileo's careful drawings of 40 consecutive days of sunspot observations.

He shows a slide of the page on which Galileo wrote about the "yearly movement of the Earth," with the famous phrase highlighted in yellow.

"Don't worry," he says serenely when everyone gasps in unison at the desecration of a precious artifact. "We Photoshopped the highlighting."

Lest you think this lecture is all about old books, let me add that he has a lot to say about Power Point -- absolutely none of it good -- and the design of effective Web pages, as illustrated by some really awful ones. And then I'll proceed directly to the graph that could have saved the space shuttle Challenger.

Many people remember physicist Richard Feynman, who as a member of the commission that investigated the accident dramatically illustrated its cause by dropping a section of O-ring into a glass of ice water. Cold, it was not resilient enough to seal the joints in the shuttle's booster rockets.

But the engineers at Morton Thiokol had that information, at least implicitly. Most of it was included in the charts they sent to NASA in the hours before launch, when the decision on whether to delay it was being debated. In the 24 previous launches, they had several times found damage to the O-rings in recovered boosters, some of it quite severe.

But the information was buried in the tables. Though the engineers recommended against launching that morning, when temperatures were predicted to be in the high 20s, their charts and tables never explicitly showed the relationship between damage and temperature. NASA was not convinced.

A simple graph makes the connection unmistakably clear. The horizontal axis shows temperature at launch for all previous flights; the vertical scale shows the severity of damage.

The lowest temperature at which a damage-free launch occurred was 66 degrees. All four launches at lower temperatures had damage, and to the extent one can judge a curve by four points, the damage was progressively worse at lower temperatures. At the lowest, 53 degrees, the shuttle probably barely survived. On both the booster rockets, the primary O-ring had burned through and leaked, the secondary O-ring was badly eroded and there was soot in places in the joints where no soot should have been. To launch at 27 to 29 degrees was insanely risky.

"Had the correct scatterplot or data table been constructed," Tufte said, "no one would have dared to risk the Challenger in such cold weather."

Tufte believes clear design both demonstrates and promotes clear thinking. His books are masterpieces of clear design.