A book I read with a reading group in 2012.

About the book

I bought this famous book long ago, but had only dipped into it until the study group finally prompted me to read it carefully from cover to cover.

For a book about information visualization, its meticulous layout was impressive. Almost all illustrations and notes sit beside or immediately below the sentences they concern, avoiding the need to flip back and forth. If an illustration’s explanation continues onto the next page, the same image sometimes appears again (the Salyut 6 cyclogram, for instance). Important images are reproduced as faithfully as possible: John Snow’s cholera map, for example, is printed at its original size.

It was not easy reading, both because the examples span so many fields—engineering, medicine, architecture, magic, astronomy, astrophysics, computer science, typography, art, seventeenth-century English literature—and because, unlike ordinary technical or popular-science books, it frequently uses literary language.

Throughout, the author cites inappropriate visualizations and criticizes them with extraordinary sharpness, thoroughness, exactingness, acerbity, cynicism, and humor. It feels as though his personality comes through.

Preface

The preface explains the book’s overall subject and organization. Its most important theme, it says, is “verbs”: how to visualize change, movement, and causal relationships effectively.

Chapter 1. Images and Quantities

The first part briefly introduces the history of data visualization. I was struck by early attempts to fit diverse data onto the two-dimensional plane of paper.

It was also interesting that it took 5,000 years to progress from maps—simplified, scaled-down drawings of the real world—to discovering that “arbitrary” numbers could be plotted on the X and Y coordinates of an abstract Cartesian plane.

The latter part discusses losses and distortions when quantities or real objects become abstract or representational images: loss of absolute or relative scale when reproducing artworks in books, for instance, or dequantification, in which scientific visualizations fail to convey quantitative information present in the original data. Dequantification during visualization of numerical data is one of the author’s chief concerns.

Conversely, suitable distortion or transformation can sometimes reveal patterns better. Compressing the Y axis of an ordinary line chart, for example, can make a trend easier to see.

Chapter 2. Visual and Statistical Thinking: Displays of Evidence for Making Decisions

Two examples explain what makes visualization effective in supporting decisions.

The first is the London cholera outbreak of 1854. Instead of a conventional graph of daily deaths, John Snow mapped the locations of the dead and public water pumps suspected as the source, persuasively displaying evidence for his hypothesis that cholera spreads through contaminated drinking water. Today we would call this geovisualization.

The author identifies four particularly important approaches.

First, the visualization (a map showing pumps and deaths) revealed causal relationships effectively.
Second, evidence was presented to answer “Many compared with what?” rather than simply “There are many deaths.” Quantitative comparison is central to scientific/statistical thinking, and visualizations must contain it too.
Third, Snow considered alternative explanations and carefully investigated apparent counterexamples rather than overlooking them, strengthening his hypothesis.
Fourth, he did not assume the data were accurate, but carefully evaluated possible errors (people moving house, for example). Today we might call this data cleansing.

The second example is the Space Shuttle Challenger disaster on January 28, 1986. Although sufficient evidence existed to predict the disaster and the engineers warned of it beforehand, inadequate explanations and visualizations (and bureaucracy, among other things) led to the explosion. (Naturally, an exaggerated account like “because of the visualization” should be understood in light of this being a book about visualization.)

The criticisms include:

Losing credibility by failing to identify the author or organization responsible.
Increasing confusion by including unnecessary data irrelevant to the decision.
Obstructing sound decisions by showing only some, rather than all, relevant data.
Distracting from the issue with unnecessarily detailed data.

The author also criticizes materials prepared for the subsequent hearings: unnecessary decoration obscured essential information, and the reports still included only a tiny fraction of the decision-relevant data.

He emphasizes that information visualization must follow the principles of scientific/statistical reasoning and be evaluated with equal rigor. The following passage probably captures the chapter’s conclusion:

Visual representations of evidence should be governed by principles of reasoning about quantitive evidence. For information displays, design reasoning must correspond to scientific reasoning. Clear and precise seeing becomes as one with clear and precise thinking. —p53

Chapter 3. Explaining Magic: Pictorial Instructions and Disinformation Design

Magic books relate to the book’s subject in two ways.

First, they must visually explain a magician’s performance over time, and that performance has two aspects: what the audience should see and what must remain hidden from them, known only to the magician. They offer much to learn about conveying highly complex information: the passage of time, visible movements, and movements that must be concealed.

Second, magicians must hide some of what actually happens to deceive the audience. Their techniques let us infer what not to do when we design. We should study how information is effectively hidden and actively avoid it (called disinformation design).

After magic, the chapter considers examples of effectively displaying incorrect information. The intended point seems to be that before asking whether information is effectively presented, we must ask whether it is correct.

Beyond illustrations in magic books, it also discusses magicians’ onstage actions, which offer many lessons for presentations and lectures.

Chapter 3 emphasizes:

Is the information correct in the first place?
Is it appropriately represented?
Are all the data fully described?
Does the presentation prevent misreading?
Are suitable context and comparisons provided?

Chapter 4. The Smallest Effective Difference

The smallest effective difference is the minimum difference that can be perceived effectively. For example, do not use red and blue when gray and black distinguish things adequately. It is a principle of visualizing with the smallest discernible difference. The author compares it to Occam’s razor in science. Perceptual fields (perceptual psychology, psychophysics, etc.) use a similar term, just-noticeable differences. Of course, the author’s smallest effective difference is greater than a JND; he slightly alters the phrase to say that visualization needs just notable differences.

His definition:

Make all visual distinctions as subtle as possible, but still clear and effective. —p73

(In some respects, this resembles Do the simplest thing that could possibly work. —ak)

Put another way, conserve scarce visual resources rather than wasting them. Humans can easily distinguish and remember fewer than ten colors, so when different hues are unnecessary, vary luminance within the same hue. If line weight alone can distinguish layers, do not add other visual devices, and so on.

What does this conservation gain? Unused materials (visual resources) increase freedom. Designers have more room to maneuver when adding a dimension to a chart or emphasizing something that truly matters. It also leaves more flexibility when the medium changes—for instance, to monochrome newspaper printing or a small mobile screen.

Chapter 5. Parallelism: Repetition and Change, Comparison and Surprise

Chapter 5 introduces ways to go beyond the constraints of a two-dimensional plane and increase the number of representable dimensions by juxtaposing similar visualizations, mostly in space. Its title’s words—repetition, change, comparison, surprise—are key constituents of “information” (probably as distinct from mere “data”).

Placing two or more images close together invites viewers to compare their similarities and differences actively. This need not be limited to image–image juxtaposition. For example, cubes seen from different angles can be arranged on the left, with corresponding numerical values in the same arrangement on the right. This supports both visual recognition and precise numerical reading for each cube. Another interesting example is a UI showing German lyrics and English translations as a vocal performance progresses. Sound (music) and two sets of text (German and English) unfold in parallel over time.

The latter part presents and criticizes inappropriate juxtapositions.

Chapter 6. Multiples in Space and Time

“Multiples” arrange several images, much like the previous chapter’s juxtapositions, to reveal repetition, change, patterns, and unexpected surprises. They are especially useful for spreading motion over time across two-dimensional space. This follows naturally from chapter 5 (the opening sentence is almost identical to that chapter’s title).

Examples include effectively explaining why Saturn’s rings appear differently depending on its position relative to Earth (C. Huygens, 1659), depicting Pangaea’s transformation into today’s continents through continental drift (A. Wegener, 1929), and densely depicting dynamic human movements in successive steps like film frames (E. Muybridge, 1887). One point these examples emphasize is that the required density varies with speed and regularity of movement. The highly regular motions of planets can be understood even with much wider intervals than dynamic human movements, through a kind of interpolation or extrapolation.

Beyond movement over time, the chapter shows a patient’s medical records, comprising thousands of numbers, condensed into a single page of multiples.

It also includes a kind of fish identification guide for divers to consult underwater. What impressed me was that it exploited the magnifying effect of diving masks: the air inside causes light to refract, so information could be packed more densely onto each page. Turning pages underwater is inconvenient; using the mask’s magnification reduced the need to do so. Generalizing the lesson: good design requires understanding and making use of the context in which information will be used.

Chapter 7. Visual Confection: Juxtapositions from the Ocean of the Streams of Story

The last chapter concerns confections (the dictionary offered “a work of fine or elaborate craftsmanship”). A confection weaves together nouns, verbs, and narrative streams scattered across space and time into a single display, expressing a “richer story.” The opening quotes a lovely fairy tale. Let’s read it together (such beautiful writing):

…the Water Genie told Haroun about the Ocean of the Streams of Story, and even though he was full of a sense of hopelessness and failure the magic of the Ocean began to have an effect on Haroun. He looked into the water and saw that it was made up of a thousand thousand thousand and one different currents, each one a different color, weaving in and out of one another like a liquid tapestry of breathtaking complexity; and the Water Genie explained that these were the Streams of Story, that each colored strand represented and contained a single tale. Different parts of the Ocean contained different sort of stories, and as all the stories that had ever been told and many that were still in the process of being invented could be found here, the Ocean of the Streams of Story was in fact the biggest library in the universe. And because the stories were held here in fluid form, they retained the ability to change, to become new versions of themselves, to join up with other stories and so become yet other stories… —p120

The discussion of computer interfaces later on is also interesting. It makes several claims I find engaging (and mostly agree with). For instance, the claim that in applications dealing with content, the information itself should become the interface (“the information becomes interface”) resembles Jef Raskin’s The Humane Interface. The author calls this information-sensitive design, contrasting it with generic (interface) approaches. For content itself to act as the interface, he argues, one needs a deep understanding of it. (Think of the UI in Deadspace: such an interface could not emerge without deep understanding of the overall design.)

The chapter introduces quantitative approaches to evaluating the information efficiency of computer interfaces: measuring the space devoted to content against the space devoted to administrative controls (similar to the data-ink ratio); measuring text density against paper or other screens (earlier, the book compares telephone-directory typography with computer user interface typography); and counting the commands available on one screen (more is better if presented minimally; it discusses Hick’s law and related ideas).

Here is the book’s final paragraph. The last sentence in particular summarizes confections, and perhaps the whole book:

The development of perspective by Florentine architect during the 15th-century Italian Renaissance was a special gift to the world of visual thinking, for now people could see diverse objects located in geometrically correct context. Confectionary designs are a similar gift to understanding. Like perspective, confections give the mind an eye. Confections place selected, diverse images into the narrative context of a coherent argument. And, by virtue of the architecture of their arguments, confections make reading and seeing and thinking identical. —p151

Brief summary

When creating information visualizations or evaluating their results, adopt the same attitude as when conducting and assessing scientific/statistical reasoning. Do not omit or distort important information; exclude unnecessary information; provide comparisons or context; and always question the information’s truthfulness (effectively or efficiently conveying incorrect information is a problem).

Principles and techniques:

The smallest effective difference: conserve visual materials, using the minimum necessary.
Parallels, multiples: reveal information through juxtaposition.
Confections: carefully “compose” images and text to convey a rich story.

Reservations

I also have a few reservations.

First, the approach seems insufficiently grounded in human vision and cognition. This could be supplemented by Colin Ware’s work (such as Visual Thinking for Design) or more foundational writings (David Marr, J. J. Gibson, etc.).

Second, apart from the final chapter on confections, it feels too focused on the mechanical efficiency of conveying information. Something like Donald Norman before Emotional design, perhaps? Indeed, at least some visualizations he dismisses as rubbish (chartjunk) are contested. See Efficiency and beauty in data visualization.

For details, see Reservations about Edward Tufte’s work.