Thumbs Up Viz: Handcrafted
Computers are ubiquitous in contemporary data visualization. Minimalist data-centric graphs and intricate network diagrams rely on computer precision. The inner workings of a cell are animated with cutting-edge rendering techniques and accurate models of molecular wiggling. Interactive visualizations take advantage of advances in human-computer interaction technology. Discussions in the data visualization community often revolve around the latest visualization languages and algorithms. Detailed satellite images and multi-variate maps take advantage of a computer’s ability to manipulate huge datasets.
In contrast to current practice, the history of data visualization is a history of work done by hand. Computer-aided visualization didn't even exist until later stages of the 20th century. Instead, graphs were made with ruled paper, pencils, and drafting tools. Maps were painstakingly sketched and painted. Even early satellite data was captured, transferred, and stored on film or analog video.
Pre-computer data visualizations span an incredible range of forms, styles, and applications. Some uses are decidedly modern: this pair of 1895 maps comparing then-current to proposed flood control infrastructure in New Orleans wouldn’t look out of place in a modern city council meeting, save for the antiquated typography.
Others reflect the aesthetics and scientific practice of the time. Eighteenth- and nineteenth-century biologists used illustrations of plants and animals (think John James Audubon and Ernst Haeckel) to describe and classify new species—eventually leading to the development of a taxonomy that mapped out relationships between species 100 years before the discovery of DNA.
In addition to aiding the production of graphics, computers trivialize the retrieval and distribution of data. In an era when a large dataset might be published in a multi-page table, or even a book, a graph provided a compact and reasonably accurate alternative for communication.
Similarly, physical models allowed extended study of delicate or ephemeral specimens. Harvard’s collection of glass flowers (and other plants and even bees) is a superlative example. Thousands of models, many of them enlarged up to 10 times, enabled generations of scientists to study the intricate anatomy of plants. A glass sculpture may be fragile, but it’s bound to last longer than a flower or a diseased apple.
Some vintage graphics even mixed numerical data visualization with illustration, like this 1968 graph showing the time required for exposure to carbon monoxide to cause illness or death (x-axis, hours), in various concentrations (y-axis, parts per thousand). The sketches of a human figure slumping, then lying passed out on the ground, followed by a levitating skull, gets the point across, no matter your native tongue—I’m pretty sure I know what this means even if the text is in French.
With such a wide variety of styles and applications, pretty much the only commonality in pre-computer data visualization was the difficulty of production. Scott Klein describes the effort required to publish a simple line chart (the first such graph known to appear in a newspaper) in the Sept. 29, 1849 edition of The New York Tribune:
Newspapers were typeset by hand, letter by letter. Illustrations and line art had to be carved by hand into small wooden blocks. Bigger illustrations were made by bolting several blocks together (if you look at the cholera chart closely, you can see some of the seams between the blocks). The process was laborious and hard to pull off on the daily deadline of a newspaper. It took great skill and there were only a handful of craftsmen who could do the work. Few if any newspapers had on-staff engravers, so it’s likely the Tribune would have had to bring in somebody with rare skills who could command a high fee.
Teams of experts were required to create and reproduce even the simplest graphics. This level of time and effort restricted the overall number of visualizations that could be produced, but encouraged a minimal level of quality for each graphic. If cartographic training was required to make maps, maps were usually going to adhere to cartographic principles. Likewise for graphs, infographics, etc. (Although flipping through Graphis Diagrams, first published in 1974, it’s apparent you don’t need a computer to add superfluous 3D elements to a chart.)
As computers evolved from something that occupied a room to a desktop to a pocket, data visualization evolved from something largely done by hand to something largely done on a computer. In many ways, this was great—visualization, especially visualization of big data sets, became much, much easier. Anyone with a PC and a dataset could make an accurate graphic.
If computers make visualization so easy, why bother with the time, effort, and inherent imprecision of drawing something?
One reason was neatly summarized by XKCD (but it’s not only, or even primarily, Microsoft’s fault):
Computers don’t guarantee good graphics, and can even encourage bad graphics. When anyone can make a diagram, the drafters, designers, and printers who formerly collaborated on visualization were no longer absolutely necessary. Many designs were—in effect—dictated by programmers, and no longer meticulously crafted by experts.
Meanwhile, data visualization became its own field of study, and took a decidedly reductionist turn. Graphs lost their ornamentation, sans-serif typefaces proliferated, and visualizers became obsessed with data density and multi-variate displays. Which makes sense—graphics that focus on data tend to efficiently convey information. I personally love a simple line chart (and would be absolutely hopeless without a computer). Computers make these simple, clean graphics extremely well.
After the widespread adoption of the PC and the ascent of minimalist visualization, handmade graphics became few and far between—with one striking exception. Newsrooms worldwide, but especially in Europe and South America, supported teams of skilled infographics designers. These graphics departments steadily produced large format graphics to help explain sports, science, and everything in between.
A superlative (and award-winning) example is Adolfo Arranz’s City of anarchy (left). Published 20 years after the Kowloon Walled City was demolished, the graphic features intricate drawings of daily life inside the densely-populated city block. The informal style of the central illustration emphasizes the ramshackle architecture. A timeline along the bottom indicates increasing population density, accompanied by a series of miniature sketches that detail the area’s metamorphosis from fortress to claustrophobic high-rises to urban park. Callouts highlight interesting features, and precision typography complements the otherwise organic feel of the work.
(This style of infographic is by no means computer-free: John Grimwade showcased a spectacular collection of draft sketches and work-in-progress for City of Anarchy, including screenshots of Arranz coloring apartment interiors with Corel Painter.)
Another modern source of handcrafted visualization is comics. (Or, if you prefer, sequential art.) As the industry matured in the 1980s and ’90s, comics increasingly covered serious topics, both fiction and non-fiction. Among my favorite graphic novels is Clan Apis, the story of a honeybee written and illustrated by Jay Hosler, a bee biologist who happens to be allergic to bee stings.
Comics employ their own visual language, one which is well suited to information visualization. A particular strength is the ability to show change over time in a static graphic. Check out the illustration of a honeybee’s larval stages in the right-hand panel above—it shows how a tiny larva grows into a fat pupa (the inquisitive Nyuki), and how the pupa subsequently surrounds itself with a cocoon. All within a single page.
Handcrafting can also help solve a particularly difficult visualization challenge—how do you visualize the invisible? David S. Goodsell, a professor of molecular biology at the Scripps Institute, illustrates the interiors of cells. But the individual bits that make up a cell (proteins, RNA, DNA, etc.)—even though large by molecular standards—are tiny on a human scale. For example, the width of a hemoglobin molecule is about one percent of the wavelength of violet light. We simply can’t see them optically—the shape of these molecules is not directly observed, but inferred from techniques like x-ray diffraction.
Goodsell solved this problem by developing a visual vocabulary—conventions of color and shape—to depict these molecules, and employs them in renderings and exquisitely detailed watercolors. In the painting of a red blood cell below, hemoglobin proteins are red, elements of the cell wall are purple, and other proteins in the blood are yellow, green and orange. The painting helps viewers picture otherwise abstract crystal structures.
Sometimes handcrafted data visualization is an aesthetic choice. Profiling the Parks is a short animation by RJ Andrews that puts the topography of some of the United States’ most popular parks in context. It could have been rendered in a realistic, minimalistic, or polygonal style, but the use of colored pencils and informal lettering lends the piece a casual or even whimsical tone. This invites viewers to relax and enjoy the experience, rather than asking them to analyze it.
Finally, handcrafted illustrations can literally give a visualization a sense of perspective. The satellite images below show a glacier in Tibet that collapsed catastrophically in July 2016.
Compare the satellite data to the drawing by James Tuttle Keane below. The nadir (downward-looking) view of a satellite flattens out the topography, making it difficult to interpret the landscape. In contrast, the illustration shows how the glaciers were nestled high in steep alpine valleys. Keane also added a cutaway showing the temperature gradient in the interior of the glacier, and the water between the base of the glacier and bedrock that allowed the avalanches to break free of the valley floor and speed down the mountainside. Annotations highlighting important features round out the graphic.
I deliberately titled this post Handcrafted, not Hand-drawn: not only are many stylized figures done with the aid of a computer, but pre-computer data visualization could be just as precise as a contemporary vector graphic. In my mind the common element in these examples is that they were designed with intention by a skilled individual or team that was versed in the data, the available tools, and the principles of effective visual communication. Each graphic is specifically suited to the data it is representing. This level of effort takes time and expertise, but helps engage and inform a viewer.
While I was trying to figure out exactly what I was going to write about, I asked Twitter for additional examples of visualizaers who work by hand, and received some wonderful suggestions in return:
Illustrators & Data Artists
- Mara Averick
- Alex Cagan
- Mona Chalabi
- Jen Christiansen
- Patterson Clark
- Jonathan Corum
- Nirja Desai
- Sean Vidal Edgerton
- Kenneth Field
- Katherine Haugh
- Nigel Hawtin
- Kristin Henry
- Wes Jones
- Giorgia Lupi
- Catherine Madden
- Shayle B Matsuda
- Jamie Molaro
- John Nelson
- Jill Pelto
- Stephanie Posavec
- Jane Pong
- Leila Qışın
- Mesa Schumacher
- Cristophe Viau
- Jeremey Wood
Special thanks to: RJ Andrews, Jen Christiansen, James Tuttle Kean, Jenna Mukuno, Jon Schwabish, Will at 50watts.com, and Daniel Wolfe.