Inscrutability and data visualization: the lost language of the khipu

While it has long been understood that the Inca khipu was an advanced thread/cord-based numerical accounting system, there are tantalizing shreds of evidence suggesting that it was much more — a singular kind of writing, capable of containing myth and recording history. Unfortunately, in 1583, “Peru’s nascent Roman Catholic church decreed that khipus were the devil’s work and ordered the destruction of every khipu in the former Inca empire.” This decree spelled the end of the khipumayoc, the Incan caste of writers and readers who alone could decipher the meanings of the khipu’s intricate knots and finely-woven colored strings.

It’s possible . . . that khipus were actually examples of semasiography, a system of representative symbols—such as numerals or musical notation—that conveys information but isn’t tied to the speech sounds of a single language, in this instance Quechua. (By contrast, logographic languages such as Chinese and Japanese are phonetic as well as character-based.) The Incas conquered a huge number of neighboring peoples in a short time span, between 1438 and 1532; each of these groups had its own language or dialect, and the Incas wanted to integrate those new territories into their hyperefficient organizational network quickly.

If khipus are examples of semasiography, the obvious next step is to break their code. Nearly a decade ago, Gary Urton, a professor of pre-Columbian studies at Harvard, began the Khipu Database project (KDB), a digitized repository of 520 khipus. (831 khipus are known to exist worldwide.) Urton has argued that khipus contain vastly more information than once believed—a rich trove of data encoded in each cord’s colors, materials, and type of knot. The KDB may have already decoded the first word from a khipu—the name of a village, Puruchuco, which Urton believes was represented by a three-number sequence much like an Inca ZIP code. If he’s correct, the system employed to encode information in the khipus is the only known example of a complex language recorded in a 3-D system.  Khipus may turn out to be something like bar codes that could be “scanned” by anyone with the proper training. (

Separated from the meanings that gave rise to their intricate structures, khipus appear to us not as texts but as abstract textile artworks. However, because we know that their origin is in the representation of data, khipus have a powerful and mysterious presence. A similar experience can be had when we look upon an abstract data visualization: simply knowing that a given visualization is somehow derived from real flows of sensor data or network traffic imbues it with mystery and a patina of inscrutable meaning. No “legend” is required to appreciate such works — indeed, a legend or key can have the effect of lessening the mystery and robbing the viewer of the imaginative act of speculating on the “meaning” of the visualization. Perhaps this makes the loss of the khipu’s secrets a little more tolerable: so long as we cannot read the tangles of their knots and cords, we are free to imagine what they must mean.

2 Replies to “Inscrutability and data visualization: the lost language of the khipu”

  1. This is one of the most beautiful photographs of a khipu I’ve seen: where is it from?

Comments are closed.