When Ancient Texts Vanish, Scientists Can Make Them Reappear
To Gregory Heyworth’s naked eye, the coat of arms was nothing more than a smudge. The emblem appeared on the bottom of the epic 14th-century French poem Les Eschez d’Amours; if it could be read, it would reveal to the medieval scholar which family had originally owned it. A firebombing in Dresden during World War II had marred its inscriptions, turning its provenance into a mystery.
“It looked,” he tells mental_floss, “like pigeon poop.”
Heyworth, an associate professor of English at the University of Mississippi, hoped that ultraviolet light might reveal more than what his eye could see. In 2005, he started examining the document with it—but unfortunately, the view didn't improve. So after years of frustrating work, he jumped online and dug up details of the Archimedes Palimpsest, a bundle of 10th century documents that had been erased by a monk so its parchment paper could be reused to write prayers. Imaging scientists had been successful in excavating the “lost” text from the Palimpsest. He wondered if they could do the same for the poem.
In 2010, Heyworth met with Roger L. Easton, Jr., chair of the Rochester Institute of Technology (RIT)'s Chester F. Carlson Center for Imaging Science. Easton had been working on new ways to image and decipher decaying manuscripts since the 1990s. At that point, X-rays (which can identify the iron in certain inks) and ultraviolet light had been in use for decades, but their reach was limited. There are hundreds of pigments, all of them responsive to different wavelengths. To properly exhaust most possibilities, there needed to be more options.
The result of Easton's work was an arsenal of multispectral imaging hardware and software—photographic and analytical techniques that could take faded or erased text and, by reflecting different bands of light, make them visible to the eye for the first time in centuries. A very deliberate, sometimes exhausting practice, multispectral imaging is reviving vanished text and helping historians rewrite world history—a revolutionary new field blending science with the humanities.
Using Easton's equipment, the two photographed Les Eschez d’Amours across a dozen wavelengths, each harboring the possibility of lighting up the pigments on the document. The images were loaded into processing software to further sharpen, enhance, and contrast. And there, viewable for the first time in hundreds of years, was the coat of arms: a unicorn and shield. Within two hours, Heyworth discovered that it was the von Waldenfel family of Bavaria, Germany, that had possession of the document prior to its known whereabouts in the 17th century. It was one missing piece of the poem's chain of ownership.
Les Eschez d'Amours is just one of many documents that can benefit from this process, potentially revealing more than we've ever known about civilization. The downside? There's currently a serious deficit of trained specialists, equipment, and money. "We have a minimum 60,000 manuscripts in Europe alone to image,” Heyworth says, noting that he has the only traveling multispectral system available. “It is, to me, a state of urgency. There is a real danger of some being lost forever.”
A page of the Archimedes Palimpsest, both visible to the eye (L) and after being processed as a multispectral image to reveal "overwritten," hidden text (R). Image Credit: ArchimedesPalimpsest.org
Though it's been refined significantly in the past decade, multispectral imaging isn't an entirely new development. In 1996, Easton and colleague Keith Knox had successfully enhanced faded text from the Dead Sea Scrolls using filtered lenses on a Kodak camera, a process originally developed by the late archeologist Robert Johnston. Easton’s eureka moment came as the team removed two colors of the RGB (red, green, blue) model present in the visible spectrum from the digital image.
“We subtracted pairs of these bands,” he says. “In one of the subtractions, we were able to see some poor-quality, fuzzy characters. I suggested we compare those to the original color image. Upon doing so, we realized that we had not noticed those characters in the original. These characters were new.”
The handwriting had become visible. Later, Easton would introduce multiple wavelengths ranging from ultraviolet to infrared, capturing images as they reacted to a dozen different bands of light.
“One way to think of it is like the black light you see on crime shows,” says Kevin Sacca, a senior undergraduate student who works with Easton analyzing images at RIT. “The pigment has different spectral properties that can absorb, reflect, or transmit light depending on the wavelength.” Hitting the right combination of light and pigment is like having the tumbler in a lock click into place: It can make invisible text glow with new legibility.
When the Archimedes Palimpsest was rediscovered in the late 1990s, Easton saw an opportunity to put his techniques to a considerable test. Archimedes was a mathematician born in 287 BCE who had his elaborate formulas copied on dried animal skin known as parchment. In the 13th century, a monk had used an abrasive liquid—likely orange juice—to scrape off the ink describing Archimedes’ work. (At the time, parchment was difficult to find and often reused.) This recycling is known as palimpsesting. In this case, the monk took seven of Archimedes’ scrubbed manuscripts, tied them together, and used them as a canvas for his own writing.
“Archie,” as the book is known to scholars, started out in rough shape and spent the next 700 years getting worse. Mold, age, and some ill-advised glue had all conspired to create a book that looked to be on the verge of crumbling. Imaging would not only provide a possible key to unlock the text, but a way of preserving it for future researchers to examine.
Though it had been photographed before Easton’s digital excavation in the 2000s, the scientist used multiple bands of light to create the best opportunity for the “undertext,” or the remains of the erased pigment, to be seen. A cell phone camera, for example, might take a picture in the three RGB bands visible to the eye; Easton photographed in a dozen bands, then blended the layers to form multispectral images. From there, the files would be examined in a software program called ENVI that can work to bring out faded or obscured writing by utilizing the different wavelength-specific bands used during photography and manipulating pixels for contrast.
“The chances are, the ink written over it is different from the ink below,” Sacca says. “The spectral properties will be different, and we can separate them.”
The initial approach was to blend the “overtext,” or the monk’s writing, together with the parchment to isolate the undertext. But it was too blurry—and if the overtext was written directly over the faded ink, it would all disappear. Instead, Easton essentially turned the pages into three distinct layers, “lifting” the undertext off, using ENVI to sharpen and darken the text for visibility, and sending the results to scholars. Figuring out which wavelength the pigment responds to can take days. Since ink and damage can vary even on the same page, the process has to be repeated constantly; ENVI can take hours to run a single software process on an image, whether it's a whole page or just a portion.
A page of the monk's work in normal light (L), imaged (M), and with the undertext made visible (R). The hidden text was written vertically on the page. Image Credit: RIT/Center for Imaging Science.
The results, however, were nothing short of stunning. Archimedes, it turns out, was on his way to discovering calculus and was pondering the concept of infinity well over a thousand years before scholars believed anyone had. The discoveries that trickled out beginning in 2000 essentially rewrote what historians had believed about math.
After much of the Archimedes work had been completed—some passages that had been painted over and resisted all attempts under multispectral responded to a Stanford X-ray examination—Easton began helping Heyworth with his studies in 2010. Heyworth’s model for a portable imaging system, a key part of what he dubbed the Lazarus Project, would bring Easton’s abilities to a wider audience. They’d also entertain proposals from scholars eager to unlock the hidden knowledge of their own work. A request to examine some charred pages written by William Faulkner revealed never-before-seen poetry; the Library of Congress employed similar techniques to discover that Thomas Jefferson had erased “subjects” and written “citizens” in the Declaration of Independence.
While manuscripts were a foremost consideration, one historian was intrigued by a map likely used by Christopher Columbus that was slowly being lost to time. Easton had performed his document archaeology for manuscripts. Could he do the same for a massive canvas rendered in multiple kinds of paint?
A segment of the Martellus map before processing, viewed under an (unsuccessful) wavelength, and finally showing the faded text. Image Credit: courtesy of Chet Van Duzer.
The Martellus map warned of monsters. Four feet high by 6 feet long, the geographical guide was crafted by cartographer Henricus Martellus in 1491. Scholars believe it almost certainly informed Christopher Columbus about the shape of Asia and the (erroneous) location of Japan before he set about discovering the New World. It had fascinated scholar Chet Van Duzer ever since he had first seen images of the map taken under ultraviolet in the 1960s. The light had illuminated spores of ink.
“It proved there was text on the map,” he says. “But you couldn’t see most of it.”
Van Duzer reached out to Heyworth and Easton in 2012, who were collaborating to steer the Lazarus Project into new directions. Heyworth knew that many universities didn’t have the finances to install expensive imaging rooms with just a handful of historical documents, making his portable equipment (which was provided free of charge) attractive.
The three would eventually sit on the Lazarus Project's board; for now, Van Duzer was explaining how badly he wanted to resurrect Martellus’ old legends.
In August 2014, team members traveled to Yale University, where the map is kept in the school’s library behind a protective enclosure. Their in-house archivists freed it from the wall and balanced it on an easel. (The map had been backed to help preserve it.) Easton used a quartz lens made by MegaVision to take 50-megapixel images of overlapping sections—55 in all—while an LED light source loomed over the canvas. Because the map’s surface is uneven and painted, varying the distance to the stationary lens, Easton had to refocus the camera as they made their way across.
That fall, Easton and Sacca worked in Rochester to pull the faded text from the map, sending digital files to Van Duzer in California to translate Martellus’ Latin. Sometimes words would trail off, leaving him to infer meaning; other times, he’d squint and try to decide whether he was seeing a “V” or “LI.”
Courtesy of Chet Van Duzer.
Like a developing negative in a dark room, the words of Martellus slowly appeared. He warned of sea dangers, and how some cultures fished for sharks. "A sea monster that is like the sun when it shines,” he wrote of the orca, “whose form can hardly be described, except that its skin is soft and its body huge."
Text in specific regions told Van Duzer which sources Martellus had used. Citing the work of Marco Polo, for example, came from one of the early manuscripts and not a published edition. (Details can vary between the two.)
“We know almost nothing about Martellus,” Van Duzer says, “so whenever we can generate or verify his sources, it’s exciting.” Martellus was himself a source for later mapmakers like Martin Waldseemuller, the first cartographer to name America. Knowing how Martellus crafted his topography would increase our understanding of how other important maps were created.
Because of Van Duzer’s knowledge of the map, he was able to request Easton and Sacca focus on specific areas. “He’d email and say, ‘Can you check there? I think there’s text but I can’t see it,’” Sacca says. “I spent four or five days running data on that one area. Sometimes you get single words, sometimes entire paragraphs.”
The Martellus map, Sacca says, is mostly imaged, with roughly 90 percent of the faded text now visible. Other technicians could go over it and possibly find data he’s missed, but that requires time and resources RIT doesn’t have. Despite pleas from many scholars and universities to examine their holdings, Easton only has two students working full-time to unravel documents.
“People will ask me to image their grandfather’s diary,” Sacca says. They don't realize the thousands of documents already in the queue, or that there’s only so much expertise to go around.
An overwritten illustration of a 5th century medicinal herb becomes wholly visible after being imaged. SinaiPalimpsests.org.
At any given time, Easton, Heyworth, and other advocates for the burgeoning field of textual science are traveling the world. Part of their mission is to image delicate relics that their owners wouldn’t dare think of transporting. (RIT is currently assisting in imaging the library at St. Catherine’s Monastery, home to thousands of ancient folios written in 11 languages and left behind by visiting monks as far back as the 4th century.) Another is to train students and other scholars how to use the technology so more manuscripts can be preserved and better understood.
“These students are the ones who will be doing the real work that will follow up on our efforts,” Easton says. “It is only by collaborations by people whose loyalties are to the objects and not to personal recognition or financial gain can the need be addressed.”
The rising tide of skilled image specialists face a danger beyond decaying pages: In 2012, Islamist extremists attacked one of the famed libraries of Timbuktu and burned its books. Fortunately, scholars had switched out their rare manuscripts, preserving the African writings, which date from the 10th century to 14th century.
“It’s the only record of scholarship of the continent from that period,” Heyworth says. “They’re endangered objects.”
The more work that can be done, the more documents can be excavated, making interest in the field as much of a priority as imaging itself. Heyworth recalls a day not long ago when he invited a first-year student to sit down and interact with the ENVI software. A page from an ancient Vatican manuscript was onscreen. With a few mouse strokes, the text revealed underwriting. The student began to read the Greek out loud.
"It was the first time anyone had heard that in over a thousand years," Heyworth says. "That moment made him a scholar. I want other people to have that experience.”