The Body Politic: A Sociopolitical Look at Anatomical Visualization

“The Agnew Clinic” by Thomas Eakins, 1899. I, like an idiot, already wasted Rembrandt’s The Anatomy Lesson of Dr. Nicolaes Tulp for a blog post about academic conferences, of all things, but that would be just as apropos here.

This blog post accompanies a paper to be presented at CHI 2024, “When the Body Became Data: Historical Data Cultures and Anatomical Illustration” written by Michael Correll and Laura A. Garrison. For more details, read the paper.

I was between gigs throughout most of 2023, so I took the opportunity to get additionally weird and alienating. One of the things I decided to get weird about was John Snow’s famous map of the 1854 London Broad Street cholera epidemic. Snow’s map is often held up as this guiding light of how data visualization can lead to insight, a key turning point in visualization history, so much of a standard bearer that you can buy it on t-shirts and coffee mugs. But pretty much all of the folklore behind it is total horseshit. I did a bunch of research into this but it turns out that Kenneth Field (as he often does) had out-thought me on the subject ages ago. The summary here is that no, the map didn’t stop the outbreak (it was produced after the outbreak had mostly ended), no, it wasn’t convincing evidence (even Snow himself thought it less than unequivocal), it didn’t lead him to his theory that cholera was waterborne (he had written essays on the hypothesis years earlier), it wasn’t the first map of its kind (a similar map by Valentine Seaman beat him to the punch by decades, and Seaman’s map was ironically used as “proof” of the very same miasma theory of disease that Snow was trying to combat), and that Snow didn’t even make the damn map (he paid a guy called Charles Cheffins to do it). “The” map that you usually see isn’t even the only map of that data, and wasn’t even the one he presented! But what I think is equally interesting is how Snow’s map is not really a technological innovation so much as it is a societal one.

What I mean is that there’s nothing particularly technologically complex about “making a map of a local area and making dots where there are sick people.” But it has a number of pretty strong societal prerequisites. For one, you need to have a bureaucracy that is centralized enough and powerful enough and with an interest in public health strong enough for people to go around collecting detailed case information in a way that’s seen as authoritative, and that’s really something that only really starts kicking off in England as a result of post-Reformation political intrigues that I can talk more about on demand if you are bored (it deals with the politics of counting and the complications involved in relying on church parish registers of births and deaths and the Black Death and print culture and a bunch of other things — Jacqueline Wernimont wrote a whole book about this kind of stuff if you’re interested). Then you need to have a very specific set of epistemologies and dueling theories of disease (like, if you believed sickness resulted from displeasing one or more gods, then a map wouldn’t be of much use to you). Snow’s map, in other words, has only a few limited spans of human history in which it was both possible and useful to create. It’s a wider window than a lot of other technological artifacts, your laser discs and betamax and so on, but it’s still a window. But it gets one to thinking: what are the sociotechnical prerequisites in the visualizations we have today? Or, and it’s here where my quasi-Marxist technological determinism sensors start going off, what technologies and political structures create what kinds of visualizations, in the same way that the big sprawling suburbs of America were created by a very particular relationship with the car? Are we designing these things with technology, or are our technologies preselecting or constraining a lot of these design choices for us?

I was chatting about all of this to Laura Garrison while we were waiting in the Leipzig airport after EuroVis 2023. Eightish months later, here we are. Her pitch went something like this: if you thought it was weird how we visualized counts of bodies, wait until you hear how we visualized the bodies themselves. I think anatomy is a particularly interesting case, because it really allows one to isolate how these societal and technological structures inform designs over and above the data themselves. After all, the “data” (the shape and location of our bones and organs and so on) haven’t really changed much over the millennia. But what has changed are the social norms around the body, the technologies we use to collect and represent information, and the cultures into which all of these representations are distributed. Somebody drawing the results of a candlelit autopsy on a (possibly illegally acquired) cadaver for a vellum manuscript is performing a fundamentally different design activity than somebody making an interactive visualization of volumetric imaging data for a virtual reality headset. The fact that both designers might be talking about the same anatomical structures is almost besides the point. The resulting visualizations are just going to look different, be used differently, and be connected to a host of different epistemic and rhetorical projects.

So, almost 900 words in, that’s how we get to this specific paper. The idea here is to look at anatomical visualizations not as some grand teleological chain of increasing sophistication and complexity from the days of Galen to the MRI machine, but instead data-driven views of the body as these almost conversational artifacts, the designers of which were shaped by their technological and social circumstances, sure, but ultimately trying to accomplish similar tasks to the best of their ability. I like this framing not just because I let me poke around medical museums on a couple of continents but also because it means that it’s very self-destructively silly to pretend that all of these centuries of prior work are nothing but atavistic artifacts with nothing of value for the contemporary designer. Because, it turns out, we are likewise shaped by our technological and social circumstances, and it might be beneficial to learn how all of the thousands if not hundreds of thousands of past designers dealt with the problems that are really not that dissimilar to the ones we have today.

Here’s an example of a so-called “Five Figure Series,” a visual genre that was already well over a millennia old at the time of its reproduction here in a late 13th-century manuscript:

The five figures are views of the same idealized body, but cribbing from the Greeks (and especially the philosopher/writer/physician/polymath Galen) the body is divided into five systems: veins, arteries, nerves, bones, and muscles. To the modern eye these images may look simplified and even a little silly, but it’s important to keep the context surrounding them in mind. We live in a society that takes it for granted that you can go around and perform autopsies or otherwise consult some authoritative source. In most places on the planet, for most of recorded history, autopsies were taboo. And even they weren’t, you didn’t really have the technology to keep a cadaver well-preserved for very long, or the imaging technology to really dig deep into what you saw. Your main options were to sort of do things by analogy to the other animals that you could dissect and just hope things were close enough to humans, or to wait for somebody to get injured in an (un-)lucky enough place that you could get a glimpse of what was going on inside. So you might have known in general what the major systems of the body are and roughly how they are arranged internally or connected to each other. And your main method of circulating what you knew was through manuscripts that were hand copied, so even if you did know a lot of that stuff, you had to present it in a robust way that would survive a few centuries of the telephone game.

So, okay, if all of that is your technological base, what does the superstructure look like? I think the five figure series is actually a pretty reasonable solution to all of these design problems, given the milieu in which it was immersed. Simple diagrams showing connections, because those are the data where you are the most confident. Major systems laid out more in a network connection way (we draw an analogy to these diagrams and modern subway maps in the paper: you want to know what’s connected to what, and are less troubled by specifically where this connection exactly happens in space). Iconic rather than super detailed representations of organs because that’s what you can reliably communicate and you don’t have a lot of detail about what those structures look like anyway.

Okay, you might say, glad we’re out of that era and into our era of autopsies and graphics cards and so on. Not so fast. Because what we also learned is that the constraints and hard design problems that show up in one (data-)cultural context don’t go away. They just repeat themselves, sometimes in new guises. For instance, here’s the diagram by OpenStax College that (at the time of writing this post) shows up right near the top of the wikipedia article for the circulatory system:

Rough locations of body parts, separated by system, represented iconically rather than in detail and with a focus on connection rather than exact topography? This is one of the five figure series with some better labels and and .svg file on the back end, but it’s not a radically different species. I like to think that a 13th century monk would be able to recognize what was going on in this diagram and nod approvingly. Radically different constraints (here, on pedagogy and focus and filtering; then, on data scarcity and uncertainty) resulting in similar visual design choices, even as everything else about the way that these visualizations were generated (like the philosophies around the body, the ways and types of anatomical data collected, the tools for image generation and reproduction) totally shift.

Or, here’s another example, from Gray’s Anatomy:

Notice how the labels are placed on this thing: when there’s room, they run along the blood vessels and muscles of the head, allowing not just easy identification but also communicating some depth and 3D structure that would otherwise be missing from the 2D drawing. This might seem easy, “oh, label things in situ when you can, otherwise throw the labels off to the side”, but there’s a real art to getting this right. And it’s an art that we’re still trying to master, especially once you have interactive 3D models to worry about. It turns out to be a difficult thing to automate! You have to make sure you’re not misleading people about the actual underlying structure of the parts in question, and be willing to sacrifice some of that fine detail just to reliably fit the labels on the surface. The same design challenges and techniques, over a century apart.

So that’s one class of observations, how anatomical visualization is not just this steady march of progress where we discard the old once we have new technologies, but that various forms seem to recur and repeat and remix across time because, well, many of the underlying problems remain the same, even as the data cultures behind them shift. Another observation is just how much rhetorical work these anatomical images do.

For one, these diagrams are usually not pictures of actual people. Even the anatomical figures driven by “modern” data are still combinations of observations, idealized and simplified and smoothed over for pedagogical or practical purposes. Just like the standard stats class anecdote that the U.S. Air Force tried to design cockpits for the “average” pilot and found that the resulting design fit nobody, because no pilot was “average” in every respect, so too is it the case that nobody’s bones or organs actually match everything you’d see in an anatomical textbook. So there’s a sleight of hand here in how these idealized bodies are put together (like, how many people that are proportioned just like the Vitruvian Man do you see walking around in your day-to-day life?), and that sleight of hand gets connected to all of the racism and classism and what have you that feeds into the rest of our cultures (a fun game to play for the generative AI-inclined is to generically ask for anatomy textbook covers until you get one that’s not a super built, usually white, dude). And don’t get us started on The Anatomical Basis of Medical Practice, the 1970s attempt to try to make anatomy more “interesting” for the (presumed) audience by using Playboy-style centerfold nude models.

So that’s one rhetorical task, to communicate what the ideal or idealized body looks like. But another rhetorical task we saw was doing the work of actually convincing people that all of this otherwise taboo or morbid stuff was of actual scientific or pedagogical utility, which often looked a lot more like advertising than data visualization. Here’s one of my favorite examples from the paper, from 1747:

Again, there’s some idealization going on here. Per the authors, it’s “a skeleton of the male sex, of a middle stature and very well proportioned; of the most perfect kind, without any blemish or deformity,” which is already doing some sociopolitical work. But the other important characteristic is that there’s a huge rhinoceros in the background. That is not just any old rhinoceros, that’s Clara, practically the Taylor Swift of 18th century European rhinoceroses. Why is she there? Well, because that’s the way to get people interested in the work, to sell it as not just some sort of morbid or lurid curiosity but as a popular, acceptable thing to look at or circulate. We saw these same sort of rhetorical moves being played, this sort of tacit back and forth across the line of social acceptibility between the “oh this is just for pedagogy, don’t mind us” versus the “come one, come all, gawk at the body!” Part of the work of an anatomist is convincing one’s audience that it’s acceptable and useful to even do anatomy, in the face of all of the potential pressures to the contrary.

This kind of marketing happens for new technologies as well. Some of the first efforts by people who discovered new imaging technologies like the microscope and the x-ray were not just to “objectively” depict what they saw, but to sell those technologies, to show that they were useful or credible, in ways that bear not a little bit of similarity to how today’s tech companies adding “AI” or “smart” to every product under the sun.

But I have gone on long enough; if you want to know more, check out the paper. But the general idea of my involvement with the work was to unite several of my passions in a way that I think is useful. The first is my insistence that the way we’ve made data visualization so focused on computers and computer science is a mistake, and really we’ve got quite a few centuries of work on how to design and structure information to draw on that we really should be consulting if we want to make useful progress. The second is a perhaps perverse delight in getting to use my philosophy degree while nominally in a technical field (there is a whole section on the ethics of all of this body-related data that I haven’t even touched on in this post). In general, though, I want readers to realize that we chose anatomical visualization not because it’s a crazy edge case, but because it is an easy example of how there’s no such thing as a straightforward relation between “raw” data (whatever that means) and visualization, but that these things are mediated, and situated, and just generally part of larger cultural and societal projects than some obvious or objective process of turning a row in an Excel spreadsheet into pixels.