Revisiting Tufte, Pt 1: Visual Display of Quantitative Information

[note: I have this to say about pneumonia: Bleh! I'm feeling much better, but let me just say that as weight-loss programs go, I do not recommend this one]

In response to Katie Benjamin's question, I decided to go back and read Edward Tufte. I read him years ago, but it's been years.

I've started with the first of his "big books," The Visual Display of Quantitative Information.

A few thoughts:

The book starts out so well. Really good observations and history of statistical graphics. And then the basic concept of graphical integrity, one in a lineage of "graphic parades of horribles," in a line with How to Lie With Statistics and How to Lie With Maps. It's an effective and entertaining technique, and Tufte does a good job with it, in particular excoriating those who make infographics that in their pictoriality distort the data.

In particular, the idea of the ink-to-data ratio is a useful one: if a mark doesn't mean something, leave it out. It's really a Strunk-and-White kind of practical model towards lean and effective graphics.

After that, I fear things go downhill. First, his analysis of why we have bad statistical graphics. He ascribes this to (1) bad training on the part of graphic folk who haven't spent much time with data analysis: they are drawing about things they know little about; (2) the idea that statistics are boring and need "livening up" if readers are going to pay attention; and (3) the idea that readers won't understand charts if they're too plain.

Basically we have charts that lie because the people who make graphs are ignorant, disrespectful and bored. If you make bad graphics, it's a character fault. I think it's a cheap shot.

He then goes down the road towards graphic language reform in the name of "data-ink" maximization, proposing specific new techniques that feel to me a little like the spelling reforms of Melvil (aka Melville) Dewey: based on sound theory, but ignoring the power and centrality of cultural habit. They assume because something makes sense and fits an ideal of simplicity that people will either do that something or be dunderheads.

Maybe people like complexity for its own sake. Sometimes at least. Coming soon, a blog entry: in defense of the graphically baroque.

The following chapter I think illuminates what's really going on: It's called "Multifunctioning Graphic Elements" and it's about graphics that can be read fully from a variety of different view points. The graphics in this chapter are like poems: they are multi-faceted, complex, sometimes ambiguous in the sense that they play different sorts of information off each other in the same expression.

They really are things of beauty.

So here's what I think: sometimes you want a poem, but sometimes you you want a straight answer to the question, "how much does that tomato cost?"

The former, you want to pack in multiple layers of meaning in a way that may miraculously land in your lap or that may take years of gathering and pondering. The latter, you want a clear social mind operating in the moment, which can respond in the same language to a conversational question.

The presumtion that all communications ought to strive for a single ideal oversimplifies the variety of human communication. I think Tufte actually has interesting reads on multiple pulses; I'm just not convinced that in his quest for Overarching Rules he hasn't Overreached.

In any case, more from the Tufte trail soon...