The following is a guest post by Josh Levy, Historian of Science and Technology in the Library’s Manuscript Division.
What’s a historian to do with a born digital file?
On Christmas Day, 1854, between family gatherings and fretting over the cost of living in Washington, engineer Montgomery Meigs was notating his plans to build a spectacular new dome for the U.S. Capitol, in shorthand. Perhaps seized with a moment of self-awareness, his journal suddenly turned to the shorthand itself. “This phonography,” Meigs wrote, “is a most seductive system of writing. I find myself writing when I have really nothing to write, but seduced by the beauty of the forms and the ease with which the thoughts are put upon paper.”[1]
More than a century later, in 1987, anthropologist Rhoda Métraux and her son Daniel were chatting by phone about her still newish personal computer, an IBM XT. The two agreed, “in no uncertain terms,” that “computing led straight into the sin of prolixity.” That is, of unnecessary wordiness. Typing up a description of the conversation to her friend the following day, Métraux concluded, “don’t you agree?”[2]
Historians know that the medium shapes the message. We wouldn’t expect a handwritten letter delivered by stagecoach in 1800 to resemble the tone or content of a hastily typewritten message in 1950, or an even more hastily typed e-mail today. And historians love sources with great origin stories, especially ones that broadcast how, undaunted by time or space or linguistic boundaries, our research left no stone unturned. But we may not have come to terms with the fact that, increasingly, our sources are in code, and code that most of us can’t understand.
Programming languages aren’t just shaping the content of our sources. They’re creating entirely new kinds of historical evidence. In 2020, Library of Congress Staff Innovators Kathleen O’Neill and Chad Conrady explored methods for providing access to born digital materials. As the Manuscript Division’s historian of science and technology, and without particular training in coding or born digital research, I offered to test these method using the division’s prototype digital workstation.
I focused on two of the division’s collections, partly because they have only the barest resemblance to one another: the papers of anthropologist Rhoda Métraux, and of the mathematician and chaos theorist Edward Lorenz. What I found was exciting, and a little scary.
The computer as typewriter
Rhoda Métraux (1914-2003) was a psychological anthropologist, perhaps best known for her close association with Margaret Mead. She published and traveled widely, and conducted fieldwork in Haiti, Mexico, Argentina, Montserrat, and Papua New Guinea. By the time her correspondence turned digital, in 1987, we find Métraux semi-retired.
But as an anthropologist, her instincts seemed to bend toward reflexivity, of remaining conscious of her own role in shaping her surroundings. So her files, which consist largely of letters to colleagues and friends, make it easy to observe her initiation into computing and to reconstruct much of her original computing environment. Her computer, which Métraux precisely identifies as an IBM XT Model 5151 with a C.ITOH 8510A dot matrix printer attached, sat in the study of a fully restored “largish white” 1840s schoolhouse in Greensboro, Vermont, where she had once summered with Margaret Mead.[3]
The letters Métraux composed on it tend toward the descriptive. In 1987, she wrote:
Here at the computer, I look out – hardly turning my head – at snowy lawn, white fence, birch and dark woods to the S, and (across the unseen road, steeply below) tall pine and spruce woods rising on hills to the E, where darker clouds now shut out the sun… That’s how I work here…[4]
There’s evidence of Métraux’s cautious engagement with the computer as well, and with WordPerfect versions 4.0 and 4.1. “One of the conveniences of the computer,” she wrote to a friend, “you just tap out the code and there is the object, for better or for worse, back on the screen.”[5] Files occasionally went missing, and were found again. Letter drafts were left unfinished. And her computing was often more social than solitary, supported by a “kind friend” named Mary Lee Metcalf who ran a shop in the nearby town of St. Johnsbury. Metcalf offered counsel on computer models and floppy disks, and overhauled what Métraux calls “the machine” when she brought it by for repair.[6]
As born digital collections go, Métraux’s is fairly straightforward, and presents few obvious obstacles to a researcher without prior training. A low-frills file viewer like Quick View Plus can present a readable version of most of the roughly 800 letters in the collection, and most can be made legible with a simple text editor. A deeper forensic analysis using full disc images might reveal some intriguing edits or deletions, but for the most part what we see is what we get.
Yet even if Métraux used her XT as a kind of glorified typewriter, computer code still set the terms within which her correspondence took shape. The result isn’t all that flashy: just a more intimate, roughly-hewn version of what we might have found if Métraux had typewritten her letters instead. Her vigorous defense of Mead’s legacy against anthropologist Derek Freeman’s attacks. Scheming to get back into the field. Encouragement of younger colleagues. Pride in her son. Health scares. And reminiscences of an academic life in the company of luminaries like Ruth Benedict and Bronislaw Malinowski.
Even so, Métraux’s files do offer us an interesting data point in the history of digital non-natives adapting to emerging computing technologies. Digital literature scholar Mark Marino has observed that “if code governs so much of our lives, then to understand its operations is to get some sense of the systems that operate on us.”[7] That argument is meant as a pitch for humanities scholars to learn how to read computer code as closely and critically as they read any other source, and to seek an understanding not only of what code does but what it means.
Read alongside Métraux’s correspondence, though, it also implies a recent past when code didn’t yet exert such power over us. And maybe one of the most striking remarks in these files is one simply tossed off, in a letter dated July 18, 1987, when Métraux writes, “My first go at the computer since June!”[8] Only a few decades later, that degree of distance between human and machine is growing harder for some of us to imagine. I’m a little jealous.
Computing and chaos
By contrast, the digital files in the papers of Edward Lorenz (1917-2008) speak to a life lived in close company with computers. In the 1950s, Lorenz was a meteorologist with a foundation in mathematics and a suspicion that the linear equations used to model weather predictions were failing to capture some fundamental force, limiting their success.[9] Testing that suspicion led him to his great insight, that minor shifts in initial conditions can have major impacts on complex systems over time. Or, in more practical terms, that forecasting the weather more than two weeks out might be impossible.
This work put Lorenz at the forefront of chaos theory. It also represented a break with both statistical weather modeling and classical understandings of the predictability of nature, what Isaac Newton had called the “clockwork universe.”[10] Better described as sensitive dependence on initial conditions, Lorenz’s ideas were popularly shorthanded as “the butterfly effect” after a conference where he asked, provocatively, whether a butterfly flapping its wings in Brazil could set off a tornado in Texas.[11] As it happened, when he tried to demonstrate the limitations of chaotic phenomena by plotting equations, some of them looked like butterflies too.
But this wasn’t just clever research, it was computer research. In the late 1950s, Lorenz had become the sole user of a whirring, clunking Royal McBee LGP-30, a desk-sized personal computer installed in his office at MIT. The machine could perform 60 multiplications per second, and had 16KB of memory. Few scientists at the time had that much access to that much computing power, and Lorenz used it to tinker. Without knowing which equations to test or parameters to use, he later recalled, he was simply “trying out an enormous number of things, more than I ever could possibly have gone through” with hand calculations.[12] That freedom led to an accident: the rerun of a simulation with a small rounding difference in just one variable which, to his surprise, transformed an entire long-range weather forecast. Lorenz and his computer had discovered chaos.
Most of the digital files in the Library’s Lorenz Papers were created long after these early revelations, in the 1990s. But while they lack Métraux’s descriptive flair and her wonder at computing’s novelty, Lorenz’s files do hold plenty of clues to his relationship with computers. Alongside the e-mails, documents, PowerPoints, audio files, and images, there’s a FORTRAN compiler, and a handful of DOS-based programs that help users create striking animated graphs of equations associated with chaos theory.
I find these graphs fascinating for two reasons. The first is simple: it’s fun to imagine what viewers may have felt when they first saw animated equations rendered so easily. And even though those stories are still fresh, they do raise the question of how programs like these may have influenced the way we now visualize data. Historians have already pointed out that people of the past sometimes “saw” things in paintings or diagrams that would entirely escape our notice today. A diagram in the 1543 Nicolaus Copernicus treatise De revolutionibus, for instance, looks for all the world like a group of static geometric shapes. But Copernicus actually expected his readers to animate those shapes mentally, and failing to do so properly would have rendered his argument incomprehensible.[13] We probably wouldn’t find such dramatic shifts in observation over just a few decades. But maybe the Lorenz programs can provide some similar evidence for visual histories of science education, or histories of data visualization itself.
I’m also fascinated by the interactivity of these programs, which makes them fairly unique as historical sources. That’s especially evident in a program we can’t quite get to run, Dynamics. It was developed by chaos theorist James A. Yorke in 1985, as a way to collect all the tools the Maryland Chaos Group was using to visualize the properties of dynamical systems. The program was aimed at both researchers and students, and circulated freely, in constantly improving iterations. Dynamics was meant to be interactive and entertaining, and disks included the full source code, written in C.[14] In fact, the program was so interactive that performing certain functions, like adding a new map, required the user to make changes to the code and then recompile the program. In other words, this code wasn’t just meant to be read, it was meant to be tinkered with. So historians who can’t code really can’t fully engage with this source. And if they can code? Maybe they can get our versions of Dynamics to run.
Historians who code
The shorthand passage at the beginning of this post was written in 1854, but it wasn’t decoded until the 1990s. The reason was simple: no one still fluent in the Pitman shorthand system was willing to undertake such a tedious, time-consuming project. Finally, William D. Mohr, a retired stenographer for the U.S. Senate, spent four years of his time translating the Meigs journals.[15] And now they’re online, and accessible to anyone.
I’m a bit intimidated by the increasing codification of primary sources. But we can’t wait 140 years before we start learning to read them. In parallel with the Library’s work to make these born digital sources available, humanities graduate students are increasingly engaging with programming. And, while all historians won’t need to learn code, there is a lot of untapped potential. As scholar Mark Marino suggests, code isn’t just a mechanical tool. It’s a “means of expression between humans through communication with machines.”[16] I think historians will find that computer code can tell us a lot about ourselves.
[1] Montgomery C. Meigs and Wendy Wolff. Capitol Builder: The Shorthand Journals of Montgomery C. Meigs, 1853-1859, 1861. (Washington: U.S. GPO, 2001), 177.
[2] Rhoda Métraux to Dorothy, June 13, 1987. LETR102.!BK. Digital ID: mss83228_164_001, Rhoda Métraux Papers, Manuscript Division, Library of Congress, Washington, DC. (LC Métraux Papers)
[3] Rhoda Métraux to Dorothy, July 30, 1988. LETR288. Digital ID: mss83228_164_002. LC Métraux Papers.
[4] Rhoda Métraux to Toby, January 13, 1988. LETR187. Digital ID: mss83228_164_001. LC Métraux Papers.
[5] Rhoda Métraux to Dorothy, July 31, 1987. LETR135. Digital ID: mss83228_164_001. LC Métraux Papers.
[6] Rhoda Métraux to Martha, February 24, 1988. LETR207. Digital ID: mss83228_164_002. LC Métraux Papers; Rhoda Métraux to Mary Lee Metcalf, September 16, 1988. LETR319. Digital ID: mss83228_164_003. LC Métraux Papers; Rhoda Métraux to Dorothy, February 22, 1991. LETR626. Digital ID: mss83228_164_006. LC Métraux Papers.
[7] Marino, Mark C. Critical Code Studies: Initial Methods. Software Studies. (Cambridge, Massachusetts: The MIT Press, 2020), 4.
[8] Rhoda Métraux to Judy, July 18, 1987. LETR119. Digital ID: mss83228_164_001. LC Métraux Papers.
[9] Kathleen T. Alligood, Tim Sauer, and James A. Yorke. Chaos: An Introduction to Dynamical Systems. (New York: Springer, 1996), 359-360.
[10] Peter Dizikes. “When the Butterfly Effect Took Flight.” MIT News Magazine, February 22, 2011.
[11] Lorenz, Edward N. “Predictability: Does the Flap of a Butterfly’s Wings in Brazil Set Off a Tornado in Texas?” In AAAS Section on Environmental Sciences, New Approaches to Global Weather: GARP. Boston, 1972.
[12] Robert W. Reeves. “Edward Lorenz Revisiting the Limits of Predictability and Their Implications: An Interview from 2007.” Bulletin of the American Meteorological Society 95, no. 5 (May 2014), 684.
[13] Kathleen M. Crowther and Peter Barker. “Training the Intelligent Eye: Understanding Illustrations in Early Modern Astronomy Texts.” Isis 104, no. 3 (September 2013), 437-438, 448-449.
[14] Helena E. Nusse and James A. Yorke. Dynamics: Numerical Explorations. (New York: Springer-Verlag, 1994), v, 2.
[15] Meigs and Wolff, x.
[16] Marino, Critical Code Studies, 8.
Comments
Interesting take but I think machine learning can do this faster.