“Those that Don’t Know History…”
We’re all familiar with Edmund Burke’s well-known line “Those that don’t know history are destined to repeat it.” Burke’s insight applies in spades to the history of data visualization and its practitioners. Each new wave of practitioners insists on rushing in and claiming the laurels of expertise without first studying the lessons and mistakes of the past. We’d be much further along today if this weren’t the case.
I recently read the book Practical Charting Techniques by Mary Spear, which was published in 1969. This was several years before books on graphics by John Tukey, Edward Tufte, or William Cleveland. Spear’s early work is full of practical wisdom and it focuses on best practices as she understood them at the time. Many of the visualizations that I introduce to students in my courses today, which they find new and exciting, can be seen in the pages of this book. Spear included a graph that looks and works almost exactly like Tukey’s box-and-whisker plot years before he introduced it. She even includes graphics that look like an early predecessor of my bullet graph. What I found particularly interesting are two statements that she made about the state of data visualization back in the 1960s.
Here’s the first:
In the mid-1940s and 1950s the three-dimensional chart was popular. Fortunately that vogue seems to be passing. While some were interesting and attractive, they were for the most part practically impossible to read or interpret directly. Comparisons were difficult to make, and scales had to be visually adjusted so that amounts could be read. Such a process required mental and optic gymnastics. (p. 62)
She bemoaned the popularity of 3-D graphs and explained why they fail, but was encouraged by their decreasing popularity. What happened between the decline that she witnessed in the 1960s and the renewed popularity of 3-D graphs today? Software companies got involved in creating tools for the creation of graphs without taking time to learn the lessons of the past. No companies are more to blame than those in the business intelligence industry. Had they built their tools responsibly, we could have avoided another dark ages of graphical display.
Here’s the second of Spear’s statements that I found particularly interesting:
Statistics can be as misleading as the intentionally distorted chart is. Surveys and samples can be biased, correlations too small, factors missing, improper measures taken—not to mention the possible prejudice of the writer or speaker. But fortunately, as the art of graphic presentation advances, more listeners and viewers are becoming aware of these pitfalls and give heed to them; and more designers of charts are being artistic without confounding the true story. (p. 68)
Advancements in data visualization were already, way back then, leading people from a juvenile proclivity to decorate graphs in ways that undermined their stories into a more mature approach to graphical communication. At some point, however, the progress that Spear observed and appreciated was waylaid by a return to silly graphics. Much of this corresponds to the rising interest in infographics, which encouraged students of the graphic arts to venture into the realm of data visualization without bothering to first learn from the past. Graphic design skills can indeed be applied to data visualization, but not without first learning about the field and how it differs in purpose from other uses of graphics. While a few thoughtful infographic designers have honed their craft by studying data visualization best practices and the realms of science that inform these practices, most have taken the fast track born of arrogance and built on ignorance. Infotainment is the result: displays that entertain with flashy tricks while the information on which they’re based becomes unrecognizable and distorted in the glare.
Must every new generation repeat the mistakes of the past? Can’t we use technology for advancement rather than as a streamlined path to dysfunction? Only if we review our history and learn from it.
Take care,
12 Comments on ““Those that Don’t Know History…””
Interesting post Stephen. I was not aware there was a serious modern book on data visualization that predates Tukey and Cleveland Thanks for this post. I just ordered spear’s book.
Indeed an interesting post. I am curious though if the popularity of 3D graphs really decreased in the 60’s, or if this is just false-consensus bias on the part of Mary Spear. Is there any actual evidence for the decline of popularity of these charts? Or maybe just wishful thinking?
Must every new generation repeat the mistakes of the past? Yes, of course, and with an admixture of their own mistakes. Such is the human condition. Technology just helps us to make mistakes faster and bigger than our forefathers. That is why every generation and every discipline needs a prophet, the voice of one crying in the wilderness. Thanks for taking up that mantle ;-)
Dan — The earliest book about designing quantitative data displays that I’m aware of was written by Willard Brinton in 1914, entitled Graphic Methods For Presenting Facts. I included a couple of quotes from his book in the preface to the second edition of my book Show Me the Numbers. A PDF copy of this book can be downloaded from the Web for free.
Barry — I actually don’t know if Spear’s observation about the decline in 3-D graphs in the 1960s was accurate or based on wishful thinking. I don’t find it hard to believe, however, that the practice declined at that time, because ineffective practices such as this usually decline as a field matures. The introduction of software that could be used to create graphs would have invited an entirely new base group of people to produce graphs who had never been trained and were naturally drawn to the flashy stuff. In time, this will decline, but not soon enough.
Colin — I do at times feel as if I’m a lone voice crying out in the wilderness. It’s always encouraging to find that there are many others, such as you, who share my concerns about the harm that is done by poorly designed graphs.
I think part of the problem is that the interests of software companies are not always aligned with effective presentation. I remember in the 90’s, advertisements for spreadsheets would always show off the most ‘impressive’ graphics. And for the purposes of showing how clever your software is a 3D chart will do a good job. But the impression that the graphic makes is so effective that users think that ‘what the software can do is what I should do’. After all, it’s what I see in all the pictures, so it must be what everyone else is doing.
They still show their most technically advanced and/or beautiful graphics in advertising. It makes sense if you’re selling a product. It doesn’t necessarily make sense to use it as an information provider, but who will tell that to the user? The impression has been made.
Really interesting that this was said in 1969. Thanks for the post.
I think there is a need to distinguish between 3D graphics that are needlessly 3D (don’t show any data in the third dimension (to pinch from your website (http://www.perceptualedge.com/example12.php))), those that show data in a third dimension that could be better encoded by other means (http://www.perceptualedge.com/example3.php) and those where a third dimension is actually useful. In the latter category I’d suggest two dimensional mathematical functions and (some) two-dimensional hit patterns. Some times contour plots work fine but often you don’t get the real sense of how high the peaks and how low the troughs are. Often there is a difference between what works best for EDA and what works best for a printed publication (for example being able to spin a 3D surface around on my computer has been enlightening but I can’t do that on a printed page).
Interesting links
It Used to Take Three Highly-Trained Professionals to Make a Presentation
http://blog.duarte.com/2009/09/it-used-to-take-three-highly-trained-professionals-to-make-a-presentation/
I think that 3D surface like rendering is very limited in usability in visual BI. I didn’t see yet a convincing one. But as Tim said before, there are situations where the third “dimension” proves very useful. Here is a link on Edward Tufte forum with a picture from Science magazine where the third “dimension” is used with creativity: http://www.edwardtufte.com/bboard/images/0002Sj-3347.jpg
That is effective for several reasons: similar variability for adjacent curves, use of discrete values with distinct color code for readability, right “angle” for third axis avoiding “realistic” and unnecessary 3D objects projection techniques.
In my opinion, the 3D effect of bar or pie charts has still a psychological impact on audience. It actually detours the meaning of the graph to (arguable) aesthetically skills of the designer. From my point of view is just a matter of time till this approach will prove insufficient for quality analysis. As it happened in 60s.
Here is a nice article on the differences between Infographics and Data Visualization in case anyone is interested. I thought it was very informational.
http://www.business2community.com/content-marketing/what-is-the-difference-between-infographics-and-data-visualization-0389249
In response to the article Scott posted the link to on 2/11/13.
I prefer Alberto Cairo’s definition of the difference, “An infographic is an edited, summarized presentation of data selected by a designer to tell a story. A visualization is a display designed to explore data so every reader will be able to extract his or her own stories.”
I thought it was interesting that the example of an infographic in the article was an example of a poorly designed one.
How does the new Microsoft GeoFlow rate? At first glance it appears to be a terrible data visualization practice but it certainly is getting “hyped” by their MVP community that usually is respected as thought leaders.
John,
I spent some time with the creators of GeoFlow last September while providing consulting services for Microsoft. The software is impressive, but it would rarely be useful for quantitative data visualization. A 3-D representation of geography is only useful when you need to get down into the terrain or city to explore the physical world from various spatial perspectives. Flying around the landscape and viewing the mountains or buildings towering above you is not useful when you want to see the geographical location of quantitative values, however. A 2-D display works better for this. Pairing 2-D geospatial displays with the ability to switch to a 3-D display can be useful if you need to explore values in a 2-D overview and then dive down into the landscape to explore the actual physical appearance of the landscape or cityscape that is associated with the location of the values. For example, imagine a 2-D map that shows the locations of restaurants in space, perhaps with a bubble for each restaurant sized by popularity, with the ability to then switch to a 3-D display to fly down into the neighborhood to see the restaurant and its surroundings to getter a better sense of it. You can also use 2-D and 3-D displays in collaboration sometimes to tell stories about geographically related quantitative information. The 3-D display isn’t all that informative, but it provides a sense of the place that is concrete rather than abstract, which is sometimes useful.