Book Review: Elements of Graph Design
By Gerd Waloszek, SAP AG, Product Design Center – 05/20/2003
This review takes a personal look at Stephen Kosslyn's book Elements of Graph Design.
Book
 |
Stephen M. Kosslyn Information: Information presentation |
Author
|
|
Steven
Kosslyn is Professor of Psychology at Havard University. His research
in cognitive neuroscience includes groundbreaking work on the processes
of mental imagery. His other books include Image and Mind, Ghost's
in the Mind's Machine, and Wet Mind: The New Cognitive Neuroscience.
(From book cover)
Review
We all know the following phenomenon: Data are presented in beautifully looking, tilted 3D graphs, be it pies, bar, or line charts. There are even charts with dancing apples, oranges and lemons. But there is just one small problem with such charts: you cannot find out, which numbers the graphical elements stand for. Often also the relationship between the data is obscured. Areas, for example, may be distorted through tilting, graphs are artificially stretched or enlarged to make data less or more impressive – the list of possible "sins" and misuses goes on. Often, chart users and even designers do not know, which chart is best used for what type of data and/or relationship between the data. Percentages, fractions, or absolute numbers, for example, have all to be presented differently depending on, which aspect of the data the graph wants to emphasize.
Figure: Example of a 3D bar chart with perspective that makes it hard to read and compare actual values.
Stephen Kosslyn is a psychologist whose research includes groundbreaking work on mental imagery. Put simply, mental imagery means that we store images in analog representations in our brains, not in some abstract code, such as list of feature descriptions. Kosslyn's book Elements of Graph Design, however, is not all all academic. It addresses the practitioner as a working reference for the correct preparation, use, and interpretation of graphic information.
In the introduction, Kosslyn's research background shows up; here he tells his readers that the mind is not a camera: It can be fooled by a number of illusions; automatic perceptual effects, such as the Gestalt laws, also play an important role in how we interprete graphs. In chapter 2, Kosslyn provides assistance for the selection of graphs, including when not to use a graph. Here, for example, we learn that pie charts can be used to convey approximate relative amounts. I would, however, have appreciated a table at the end of the chapter that lists the charts together with the recommendations.
The subsequent chapters go into details, starting with the graph framework (including labels and titles), the base graph types plus their variants, and finally, aspects of coloring, shading, and other options. Chapter 8 is especially interesting; it explains "How People Lie with Graphs" – a "must read" for all statisticians. Here, they find all the stretching and distortion tricks that they need in their daily work. The final chapter "Beyond the Graph" offers recommendations for more advanced graphical information displays, such as trees graphs, diagrams, and maps. In the appendices, elementary statistics, checklists for graphs, and – what I like most – a summary of principles and their psychological bases are offered.
All in all, most of this book should be used as a reference. The beginning and the chapters on the coloring and filling and on "cheating with charts" are appropriate for continuous reading. As mostly a reference, however, the book structure and layout could be made more concise for easier scanning. Overview tables of the recommendations would also be helpful.
See also Recommendations for Charts and Graphics
A concise introduction to charts can also be found in the book The GUI Style Guide by Susan Fowler and Victor Stanwick.
