The Scholars Deciphering a Lost Writing System

The Atlantic

5 days ago

The heaps of khipus emerged from garbage bags in the back of the tiny, one-room museum—clumps of tangled ropes the size of beach balls. Sabine Hyland smiled as she gazed down at them and said, ' Qué lindo, qué lindo ': how beautiful. Hyland, an anthropologist, had traveled here to the remote mountain village of Jucul in the Peruvian Andes to study them, in the hope of unlocking one of the most important lost writing systems in history, that of the Inca empire.
Instead of writing on clay tablets or papyrus, as other ancient societies did, the Incas recorded information by tying knots into long cords they called khipus. Only a few Andean villages have preserved their khipus through the centuries; those that have survived are revered, and village elders have sometimes kept their existence secret even from other community members. Yet beyond scraps of lore, most villagers have no idea what their khipus say: Knowledge of how to read them has all but vanished in the 500 years since the Spanish conquered and destroyed the Inca empire in the 1500s.
Jucul sits at an altitude of 11,800 feet, six hours north of Lima on axle-rattling mountain roads. The village is surrounded by green-brown slopes streaked with rocks, like waterfalls frozen in place. Most of its roughly 150 inhabitants live in mud-brick homes with tin roofs, and dogs roam freely. The gradients are steep; you can walk one block and ascend two stories. Nearly everyone wears a Stetson or sun hat or ball cap—L.A. Lakers, Miami Heat, KEEP AMERICA GREAT.
The people of Jucul kept their khipus locked away for centuries; Hyland and I were among the first outsiders ever to see them. New khipus rarely turn up anywhere in the Andes, so these cords could amount to a major breakthrough for Hyland, a professor at the University of St Andrews, in Scotland.
Although some scholars doubt that they'll ever be able to read khipus fully, even a partial reading of the undeciphered cords would help illuminate the history of the Andean people who began recording information on them more than a millennium ago. Hyland has already published a proposed decoding of a few syllables on khipus from other villages. If the Jucul ones provide additional clues, she and her colleagues might one day be able to use them to crack open the lost history of the Inca empire, which was, at its peak, the largest civilization in the Americas.
The Incas began conquering nearby kingdoms in the mid-1400s, and in less than a century they had subdued a population of 12 million. The nearly 25,000 miles of roads they built, many through punishing mountain terrain, facilitated communication between far-flung areas, as did the numerous rope bridges they suspended over dizzying gorges. The Incas had advanced calendars and ceramics as well, and perfected a type of neurosurgery, likely to treat skull wounds suffered in battle, among other ailments. But most traces of the empire have vanished. The hilltop complex of Machu Picchu is one of its few enduring relics.
From the November 1967 issue: Peru's Inca renaissance
Khipus are another. Approximately 1,400 khipus have survived, but hundreds of thousands were likely in use in the 1400s. Most khipus are made primarily of cotton or animal hair (llama, alpaca) and have a similar structure: a long, thick 'primary' cord from which up to 1,000 tasseled or knotted 'pendant' cords dangle. The majority consist of plain beige, brown, or white cords, but others display a wide range of colors; Hyland has studied one that contains strands of 'crimson, gold, indigo, green, cream, pink, and shades of brown from fawn to chocolate.' Some also have objects knotted into them; Hyland has heard that a few khipus in Jucul might contain locks of human hair, bags of coca leaves, and a doll that might represent a god or supernatural being.
'I am not leaving this village without seeing that doll,' she told me.
Beneath her excitement, though, Hyland confessed that she was nervous. The bundles in the museum were so snarled—real-life Gordian knots—that unraveling them seemed hopeless. The khipus' centuries-old fibers also looked fragile, as though one errant tug could snap the strands and destroy the information encoded there. That's to say nothing of the task of actually working out what they might mean.
Deciphering a lost writing system requires a rare combination of linguistic flair, statistical savvy, and deep cultural knowledge of the region in question. Some scholars have spent their whole lives toiling on lost scripts and died with nothing to show for their efforts. The most famous decipherment ever, that of Egyptian hieroglyphs, required the discovery of the Rosetta stone, which contained near-identical texts written in ancient Egyptian and ancient Greek. Even with that enormous head start, decoding the script still took two decades.
Yet khipu scholars seem optimistic these days. 'Everybody feels like we're close,' says Jon Clindaniel, an anthropologist and computer programmer at the University of Chicago. There's a new collaborative spirit in the field; key data are being shared more widely than even a few years ago. At the same time, sophisticated radiocarbon dating methods and novel approaches involving AI are being employed. As Hyland put it, 'We're in a whole new Renaissance of khipu studies.' Further progress could open up new tracts of knowledge about the origins of writing, as well as the rise—and fall—of one of the greatest lost empires in history.
Hyland and I arrived in Jucul on a sunny day in June. We were greeted with offal-and-corn soup and presented with necklaces threaded with carnations, roses, toffee, lollipops, and circular knots of bread, a sort of Peruvian bagel. That night in the village museum—a room featuring Jucul's most prized possessions, including ancient skulls and youth-volleyball trophies—we took part in a ceremony she called a chacchadero, meant to bless Hyland with good luck in deciphering the khipus.
A table was spread with coca leaves, liquor made from raw sugar cane, and rolled cigarettes. Hyland had advised me not to refuse anything I was offered—people might get offended—so after an opening prayer, I wadded some crinkly coca leaves into my cheek and swigged the cane liquor, then dutifully choked my way through my first cigarette. A score of speeches followed. At one point, someone passed around a gourd with white powder inside. I was alarmed to think I might be snorting my first cocaine that night as well, until Hyland explained that it was lime, a calcium-based mineral that, when dissolved in the mouth, draws more stimulants out of the coca. The trick worked. Despite the 40-degree cold outside, I was flushed warm when we emerged from the museum, and I spent a few restless hours on my cot before the buzz from the coca wore off.
The next morning, we swept coca dust and cigarette ash off the table, opened a garbage bag, and plopped down the first of four khipu bundles, which weighed about 20 pounds and supposedly contained the goddess doll. I'd volunteered to help unravel, although I was suddenly regretting it. Imagine a snarl of Christmas lights so big you need two arms to carry it. Wary of the brittle strands, I hunted around as delicately as possible for loose ends and wriggled my way elbow-deep into the rat's nest, palpitating every loop and twist. Unfortunately, disturbing the ropes like this caused them to shed, and before long, a cloud of dander was tickling my nose; some settled on my tongue.
One stretch of cord looked particularly fragile. It was dark yellow and Hyland said it looked like maguey, a vegetable fiber. I spent 20 minutes teasing it free, centimeter by centimeter, and exhaled with relief when it emerged intact. (I later learned that this section wasn't maguey but animal hair that had suffered damage from rodent urine.) Still, it was just one liberated foot amid seeming miles of khipus.
Luckily we had help. Victor Margarito, who runs the Jucul museum, had a knack for untangling the snarls: Like a magician pulling handkerchiefs from inexplicable places, he kept wriggling his fingers into the bundle and emerging with entire yards of free rope. Thanks mostly to him, we eventually extracted 20 separate khipus and khipu fragments from the bundle—including a black-and-white one with a barber-pole-swirl primary cord that sent Hyland into another chorus of ¡Qué lindo! We didn't find a doll, but we did uncover small tassels that resembled ghosts, as well as shriveled scraps of rawhide bound to hair that looked uncannily human. It turned out to be llama or alpaca fibers. The whole thing stretched 74 feet—longer, she said, than any other khipu ever recorded to date.
Most exciting of all, a coca pouch the size of a wallet was sewn onto the primary cord. It was dyed pink, and had blue tufts on each corner. Hyland was thrilled: Coca is a quintessential ritual item in Andean culture, and she said the bag provided good evidence that the khipus were used in rituals as well. Sure enough, Margarito worked open the bag's knot and found, amid desiccated coca leaves, a pair of ancient cigarettes rolled in centuries-old paper—an echo of the previous night's ceremony.
Fully understanding the khipus' meaning, though, would take far more time and study. The tassels were different colors, different fiber types, different thicknesses—all variables that could encode meaning in different ways. As Hyland put it, 'Where do you even start deciphering that?'
The first breakthrough in khipu decipherment took place in the 1910s. An American math teacher and amateur historian named Leland Locke had been studying the history of counting devices, and he turned his attention to a cache of khipus at the American Museum of Natural History in New York. He determined that most khipus record numbers, functioning like textile abaci, a theory later confirmed by a khipu unearthed in an ancient Inca cemetery. The hanging pendant cords are divided into 'decimal zones' of different values. To log the number 237, for example, a khipukamayuq, or 'khipu animator,' would first make two overhand knots in the 'hundreds zone' near the primary cord. Then they'd scoot down an inch or two and make three more knots in the 'tens zone.' Finally, after scooting down another inch, they'd tie a special knot with seven circular loops. Some khipus encode numbers that reach into the tens of thousands.
Scholars now believe that the Incas often used these numerical records to count goods. In 2013 and 2014, for instance, archaeologists excavated an Inca storehouse and found several khipus alongside caches of peanuts and chili peppers; a 2015 paper argued that the cords helped track how much food was on hand. Close examination of the numbers also suggested that storehouse officials would subtract a fixed amount from each cache and set it aside, probably either as taxes or as seeds for the next year's planting.
But not all khipus served as ledgers. Spanish chronicles from the 1500s state that the Incas used khipus as letters, calendars, legal documents, biographies, historical texts, and possibly even poems.
The Jucul khipus, Hyland said, almost certainly contain some linguistic information: The hanging pendant cords contain no knots, so if they did record numbers, it was by some other means. Hyland believes that they may encode words instead, through variables such as color; fiber type (cotton, animal); and the left- or right-hand twist of the strands. The Jucul khipus also resemble another, badly damaged set from the Andean village of Rapaz, where locals say their khipus functioned as religious calendars, documenting items offered for sacrifice during festivals. Both the Jucul and Rapaz khipus likely originated in Spanish colonial times, perhaps as early as the 1500s, after the Inca empire disintegrated but before the Andean people stopped using the medium for record keeping.
It's an open question whether contact with the Spanish changed the nature of khipu writing, and whether Inca-era khipus (pre-1530s) and colonial-era khipus record information in similar ways.
A bigger question is at stake here too. Over time, writing arose in multiple locations in Asia and Africa. Yet because those continents were in constant contact with one another, exchanging goods and ideas, scholars have debated whether writing sprung up independently in each spot or first appeared in one place before spreading elsewhere. By contrast, scientists are certain that new-world civilizations developed their writing systems independently from those of other continents, because these systems originated before any contact with the Old World. These writing systems, then—including, possibly, Inca khipus—could illuminate how and why our ancestors first adopted written language: a record of one of the most consequential changes in human history.
Hyland, now 60, first fell in love with Andean culture at age 16. In 1980, her father, an agricultural scientist, took a year-long post in Lima, at a crop-research station called the International Potato Center, and brought his family with him from their home in New York State. He studied seed-storage techniques, and whenever he visited rural areas, she would tag along. The Andean landscape and lifestyle thrilled her. She recalls a field trip with a church group to a museum of ancient artifacts in Lima. In addition to seeing some erotic pottery that scandalized her chaperones, she glimpsed a magnificent khipu with blue and brown cords hanging on a wall.
During college at Cornell, Hyland studied anthropology and learned Quechua, the dominant language of the Inca empire. She earned her Ph.D. at Yale in 1994; one of her professors was Michael Coe, who helped decipher the hieroglyphs of the Mayan empire. His success inspired Hyland to believe that deciphering narrative, nonnumerical khipus might be possible.
But the field of khipu studies had grown stagnant, in part because some prominent scholars argued that nonnumerical khipus were merely personal mnemonic devices. That is, they believed that each khipu maker would record information using an idiosyncratic pattern of colors, knots, and fibers—a code that no one else could understand. Khipu makers could read their own cords, the theory went, but no one else could, and after they died, their khipus became indecipherable.
An anthropologist named Gary Urton challenged that idea. In a series of papers and books he wrote while teaching at Colgate University in the 1990s, he argued that the Inca empire was highly centralized, and that officials wouldn't have left the recording of vital information to the whims of individual scribes. There had to be a standardized system. Urton's theory eventually won over his colleagues and revitalized the field.
Urton had a charming backstory: He told reporters that he'd quit Boy Scouts after failing knot tying, and that he was inspired to decipher the 'trapped' words inside khipus because a childhood stutter had left his own voice trapped inside him. In 2000, he won a MacArthur genius grant, then jumped from Colgate to Harvard and quickly became the field's star scholar.
Shortly after arriving at Harvard, Urton, then 56, decided to create a database to promote the systematic study of khipus, with information on their length, number of cords, and other attributes. He hired 32-year-old Carrie Brezine to build it. She was uniquely suited for the role: She had a degree in mathematics and an interest in the subject. She was also an enthusiastic amateur weaver; she'd even made khipus herself.
The two began an affair; Brezine later said she felt pressured by Urton to acquiesce and to continue the affair because of his power and status in the field. She also enrolled in a Ph.D. program at Harvard, with Urton as her adviser. Although she'd built the database, she says that sex was a condition of her ability to use it. 'Gary made it clear that he could and would revoke my access at any time if I did not perform adequately,' she told Science magazine in 2020. (In an email to The Atlantic, Urton disputed that he pressured Brezine and denied using sex as a condition for access to the database, calling Brezine's description 'a complete misrepresentation of anything I ever said.')
Harvard eventually launched an investigation into sexual-misconduct allegations from Brezine and other former Harvard students, and in June 2021 the university stripped Urton of his emeritus status and banned him from campus. (Urton told me that Harvard's investigation was 'profoundly unfair and unjust,' and said he had not had relationships with any women while they were students.)
Urton had dominated the field of khipu studies for years. After he was forced out, Jon Clindaniel, who had been his graduate student at Harvard, took over administration of the database and, along with several other scholars, made the site easier to access and search. Since then, the field of khipu decipherment has flourished. Some of this work involves using computers to analyze khipu data in creative ways. For her part, Hyland is focusing on linguistics. In fact, she's drawing on a classic strategy for deciphering lost languages—one that relies on, of all things, the power of puns.
Puns have historically played an important role in written language. In ancient Egypt, the words for vulture and mother sounded alike ('mwt'), so whenever scribes needed to mention someone's mother on a sarcophagus or temple wall, they chiseled in a vulture hieroglyph. (An equivalent in English would be drawing a circle with rays—☼—to mean son.) A related tool for early writing systems was the rebus, in which a series of pictures and letters stand for sounds, such as 👁️🥫CU for 'I can see you.'
Hyland already knows of a few potential Quechua puns on khipus. One appeared on a khipu believed to be from a family named Yakapar. Hyland reasoned that the khipu's last few cords probably were a signature. As she explained in an article in the journal Current Anthropology, the very last cord was a rich yellow color, like ripening corn. The word for this color in Quechua is paru —a near-perfect match for the last syllable of Yakapar. Another pun involved a type of modified khipu that had cords dangling from a wooden board. The board featured carvings of monkeys, and it recorded the consumption of a corn beverage on feast days. In Quechua, 'monkey' is k 'usillu, and kulli refers to this beverage.
Scholars are also working to decipher khipus by sorting them into genres. Although Spanish chroniclers documented many different types of khipus, given the lack of archaeological context for most surviving ones, we have no way of knowing which khipus belong to which category. Manny Medrano, a graduate student at Harvard who formerly studied under both Urton and Hyland, explains the dilemma with an analogy. 'It's as if someone raided a bookstore overnight and flung all the books on the floor,' he told me. 'We don't know which are detective novels, which are accounting books. So for me … the decipherment problem first and foremost is a reshelving problem.'
Another Harvard graduate student, Mackinley FitzPatrick, is leading an effort to sort khipus into genres based on the colorful patterns woven into their primary cords. In the past, many scholars neglected primary cords, as if they were mere scaffolding. But there's renewed interest in these cords: FitzPatrick thinks that, like the spine of a book, they might signal a khipu's subject matter.
From the September 2024 issue: Will the mystery of the Voynich Manuscript ever be solved?
Artificial intelligence can help determine genre too. A few years ago, Clindaniel trained an AI system to analyze the colors of 37,645 cords on 629 khipus, as well as the colors of the cords that surround them, which may indicate context and genre. Clindaniel's program found that rare khipu colors—red, certain blues, orange, yellow, certain grays, greens—were all clustered together, indicating that they were probably used in highly similar contexts. Based on Spanish chronicles and other clues, Clindaniel suggests that this context might have involved religion or Inca royalty. In the future, scholars could analyze fiber type and other variables to search for more clusters.
A better understanding of the materials used to make khipus might also help with decipherment—another area where Urton's diminished presence has created openings for new techniques and theories. Urton told me that when scrutinizing khipus to determine what kind of fibers they were made of, he just eyeballed the fibers and guessed; most, it was assumed, were made of cotton. More recently, a graduate student of Hyland's named Lucrezia Milillo has been using microscopes to examine khipus, and has found animal hair and non-cotton vegetal fibers. On some khipus that Milillo has studied, those fibers appear at regular intervals, systematically, suggesting that its use encoded meaning somehow.
One obstacle to deciphering khipus is a lack of firm dates for them; knowing which ones were made prior to the Spanish conquest is especially important. Given that khipus are made from organic material, scientists should in theory be able to date them by measuring the amount of radioactive carbon-14 they contain. But a 2014 paper co-authored by Urton argued that carbon-14 tests cannot cleanly distinguish pre-1530s khipus from post-1530s khipus. (He told me that this is due to a bombardment of cosmic rays in the 1500s and a subsequent jump in carbon-14 levels in the atmosphere.) Ivan Ghezzi, an archaeologist in Peru who has done extensive work on carbon dating, says this pronouncement discouraged other scholars, and only several dozen of the approximately 1,400 known khipus worldwide have been carbon dated today. More recently, though, Ghezzi and other experts have devised potential work-arounds for the complications because of atmospheric fluctuations. With firmer dates in hand, Hyland and others will have a better grasp on whether studying colonial khipus can help crack Inca ones.
Still, some scholars remain pessimistic about the odds of deciphering khipus with any certainty. Galen Brokaw, a khipu expert at Montana State University, cites one concern above all: his belief that khipus are 'not a single code.' Instead, he suggests, they may be 'multiple codes that work together.' Just as brown and white cords might have different meanings in different genres, other variables could shift too: a llama-hair cord might mean one thing in a census and something else entirely in a tribute record. And if that's the case, even the complete decipherment of one genre of khipus wouldn't necessarily help scholars read another; each would become its own laborious puzzle.
Other lost writing systems did not face this obstacle. Once Egyptologists determined what, say, a lion or hippopotamus hieroglyph meant on a temple wall, those hieroglyphs meant the same basic thing in prophecies, medical documents, and recipes. If an archaeological team today discovered a new site with Mayan or Egyptian hieroglyphs, it could call in an expert to read them and get a translation in short order. 'I'm not sure that'll ever be possible' for khipus, Brokaw told me. 'I hope I'm wrong.'
Some scholars also question whether khipus represent 'true' writing. In true writing systems, symbols (i, x) map directly to sounds ('eye,' 'eks'). Although keen on decipherment, Hyland admits that Inca khipus might represent more of a 'proto–writing system' still coming into being when the Spanish invasion disrupted its development. Carrie Brezine won't even go that far. To explain her theory of how khipus work, she invokes Homer. Imagine if Homer had encoded The Odyssey in knots that signified ideas such as 'hero/geographical obstacle/challenge/opponent.' However useful to ancient bards, such a spare description would mean little to us today.
Brezine believes that focusing so much on decipherment can obscure what's truly special about khipus: that a large empire 'functioned with textile data as its core bureaucratic tool.' Cords made from the hair of different animals can look identical, especially when dyed, and distinguishing one fiber type from another requires running your fingers along the strands to feel how coarse or silken they are. Certain khipus, then, require both sight and touch to make sense of them. As Hyland notes, even if we never read a single Inca word, they provide a whole new understanding of what written language can be.
On our third day in Jucul, the town lost power, as happens often there. In the little museum, we hauled our table over to the sunlit doorway and, with the mountains framed before us, continued untangling the snarled bundles.
Eventually, 96 khipus and khipu fragments emerged from the garbage bags. The bundle in the last bag was in the worst shape yet, riddled with rodent droppings and giving off a pungent odor. But it also produced another new record, Hyland said, for the longest khipu ever discovered, which stretches an astounding 224 and a half feet. Among its brown, white, and black tassels, it contained tufts of human hair, perhaps as 'signatures' for whoever made different portions of the khipu. When disentangling the cords, Hyland also saw a flash of green silk deep inside the bundle and felt her heart leap into her throat—could it be the doll? She had to wait more than two hours while Victor Margarito separated everything.
Sadly, the doll itself, probably made of wax, had disappeared; it was likely devoured by mice. But its green silk skirt remained. Hyland ran her fingers over the stitching, marveling at how delicate it was. Based on the skirt's style, she suspected that it dated from around 1700.
As we were working, Rubén Susanibar, a tanned, wiry farmer in a dusty Stetson, walked into the museum and sat down. At first he just watched us, saying nothing. Then Hyland invited him to help untangle, and drew him out with a few questions. Some townspeople, it turns out, had more information about the khipus than they'd let on.
Several khipus contained cords with matted tangles of animal hair attached. Susanibar explained that these mats, which he called tancash, can form naturally on vicuñas, llamas, and alpacas if their fur gets soaked. Tancash is useless for anything practical, its fibers too snarled to be spun into cloth or rope. People must therefore have collected it solely to add to khipus, to encode meaning somehow. Hyland wondered whether tancash might pun on some important name or concept in Quechua.
Later, after I left Jucul, an elderly man named Lenin Margarito wandered into the museum and told Hyland about another possible pun, as well an old village ritual that required hauling coca, rum, cigarettes, and food up a mountaintop in the middle of the night. Lenin is the father of Victor Margarito, the museum caretaker, but Victor had never heard the stories his father was telling. He ended up scribbling notes as fast as Hyland—heritage passing down in real time to the next generation.
To protect the khipus, Hyland had planned to wrap them in acid-free paper for storage. But she'd forgotten to pack any in her rush to leave Scotland, so it was on to Plan B. She rooted through her luggage and selected two clean cotton T-shirts to sacrifice, one green, one red. She swaddled some of the longer khipus in those and packed them into cardboard boxes for storage in the museum, to await her return next year. She was already excited to get back.
Since leaving Jucul, Hyland has purchased an antique wax doll, and she plans to commission a tiny silk gown so the town can display a replica of the goddess alongside the skulls and volleyball trophies in the museum. Even now, months later, she still can't believe the luck she had in uncovering so many new khipus. 'It feels like finding a cave with the Dead Sea Scrolls,' she said, 'or entering an untouched ancient Egyptian tomb filled with hieroglyphic inscriptions.'
Hyland and her colleagues might never decipher khipus fully, much less resurrect the Inca equivalent of the psalms or Homer. But even a few spare lines would be invaluable. Such words would give the people of Jucul and throughout the Andes something that many of us today simply take for granted—a chance to hear their ancestors speak.