You are here

Artificial Intelligence Helps Identify Authors of Qumran Caves Scrolls

Among the biggest open questions about the Dead Sea Scrolls (also the Qumran Caves Scrolls) are who wrote them, and where. Were these 2,000-year-old manuscripts penned by a single group in the Judean desert, perhaps the enigmatic sect known as the Essenes? Or did they originate in different places and within various Jewish religious streams?

The simple answer is that we don’t know, since the biblical scribes of antiquity didn’t sign their work, cite their allegiance, or give us many clues about their identity. But now Dutch researchers have enlisted artificial intelligence to analyze the handwriting on the scrolls and determine how many different scribes were behind each text.

The team of biblical scholars and computer scientists from the University of Groningen tested their analysis on the so-called Great Isaiah Scroll. The study, published Wednesday in the journal PLOS ONE,  confirmed that the scroll, which contains the text of the biblical Book of Isaiah, was penned by two different hands.

Suddenly, a different handwriting

What difference does it make how many scribes wrote an ancient manuscript? Actually it opens the door for a completely different approach to studying the Dead Sea Scrolls, explains Mladen Popovic, a professor of Hebrew Bible and Ancient Judaism and the lead researcher on the team.

Ever since their discovery, experts have been focusing on the content of the manuscripts, looking for patterns in style, word usage or ideology to glean some knowledge about the people behind the scrolls, Popovic explains. While content analysis will continue to be important, the new AI-supported method will give researchers physical evidence to connect different manuscripts that were written by the same hand, he says.

The method that was road-tested on the Isaiah scroll involved algorithms designed by PhD candidate Maruf Dhali and Lambert Schomaker, professor of Computer Science and Artificial Intelligence at Groningen. Their artificial neural networks were taught to first distinguish the ink traces from the background parchment and then to find statistically significant patterns in the styling of the characters, including the changing curvature and shapes of different elements within single letters.


The advantage of using computers for the job is that, unlike the human eye, a machine can compare hundreds or thousands of different character features at the same time. For example it can look at the 5,000 occurrences of the letter “aleph” (the first letter in the Hebrew alphabet) in the Great Isaiah Scroll and find patterns of similarities and difference at the microlevel between those thousands of repetitions.

“The human eye is amazing and presumably takes these levels into account too, but that is often not a transparent process,” Popovic says. “And we also find it difficult to process when there is much data.”

The data from the Isaiah Scroll show that roughly midway through the manuscript there is a subtle but significant change in the handwriting, which is best explained by postulating the existence of two different scribes.

The fact that the styles are so close that only a computer could distinguish them is already an important piece of information for researchers. It suggests that the two scribes had learned how to mimic each other’s handwriting, possibly pointing to a common origin or shared training, Popovic says. “That is so exciting, because this opens a new window on the ancient world that can reveal much more intricate connections between the scribes that produced the scrolls,” he says.

Essenes? We hardly knew ye

The first seven scrolls, including the Isaiah manuscript that was the focus of the new study, were found in 1947 by Bedouin shepherds in a cave near the ancient settlement of Qumran, located on the shores of the Dead Sea in the modern-day West Bank. Thousands more fragmentary texts emerged in the following decades near Qumran and at other sites across the surrounding Judean desert – with the latest find announced by Israeli archaeologists in March. 

The Qumran scrolls range in date from the third century B.C.E. to the first century C.E., before the destruction of the Second Temple by the Romans in 70 C.E. at the end of the First Jewish Revolt. They include not just biblical manuscripts but also additional religious texts such as commentaries, apocalyptic prophecies, prayers, rules on community life and much more.

The scrolls have revolutionized our knowledge of Second Temple Judaism and the many groups and sects from which rabbinical Judaism and Christianity eventually emerged. But who exactly wrote them remains a mystery.

Traditionally, scholars have linked the scrolls and the isolated settlement of Qumran to the Essene movement, described by some ancient historians like Josephus Flavius. This ascetic sect, which embraced voluntary poverty and communal life, was one of the three main streams in which Judaism was divided in the late Second Temple Period.

Recent scholarship however has moved away from linking Qumran and the scrolls found in its vicinity exclusively to the Essenes. In part this is because, as mentioned, the scrolls are never signed by a specific person or group, and because there are significant ideological differences between what we know about the Essenes from ancient historians and what emerges in some of the sectarian texts from Qumran.

“Today most scholars speak more generally of the ‘Qumran community,’” says Joe Uziel, head of the Dead Sea Scrolls Unit at the Israel Antiquities Authority. “But beyond that you have other scrolls that may have arrived in the caves after being written elsewhere.”

That’s why the possibility of identifying different writing styles and tracing the hand of a specific scribe in multiple texts is so important, as it promises to highlight new connections between scrolls and give insight into their authors.

Puzzle pieces

Another advantage of the AI-driven approach is that it makes it easier to figure out when a text was written, Popovic tells Haaretz. The Great Isaiah Scroll, for example, has been radiocarbon dated to the second century B.C.E. Using their algorithms, the researchers can now identify other manuscripts from the same period by matching the writing styles, without need to sacrifice precious ancient parchment to carbon dating, he says.

The study of the Isaiah scroll is merely the first major breakthrough for the project by Popovic and colleagues, titled “The Hands that Wrote the Bible” and financed to the tune of 1.5 million euros (US$ 1.8 million) by the European Research Council.

It should be noted that the Isaiah scroll, measuring some seven meters in length, is one of the largest and most complete of the Dead Sea manuscripts. Most of the other 900 or so known scrolls have reached us in the form of around 25,000, often confetti-sized, fragments that have had to be painstakingly pieced together. Fortunately, the handwriting identification analysis is already showing good results even if it is run on short texts of just around 100 characters, Popovic tells Haaretz.

This suggests another application of the new method, in that the handwriting analysis may tell us whether certain fragments should indeed be attributed to the same scroll or whether they belong to two different manuscripts, says the IAA’s Uziel. “This is a great example of how new avenues of research can be used to understand new things about the scrolls, well beyond the specific discovery of the two writers of the Isaiah manuscript,” Uziel tells Haaretz.

Space tech and cow DNA

The Dutch project is by far not the only recent application of cutting-edge scientific methods to the study of the scrolls. Aided by NASA technology, the IAA has produced high-resolution multispectral images of the scrolls, which make the texts more readable and have even revealed sections of script that are invisible to the naked eye.  (In fact, these highly detailed images were also used to train the neural networks used in the Dutch study.)

Meanwhile, an international team of scientists has been perfecting a method to use particle accelerators and advanced imaging techniques to read scrolls that are too damaged to be unrolled. And finally, recent research at Tel Aviv University has succeeded in extracting from some scroll fragments the DNA of the animals whose skin was used to make the parchment.

This project has the closest synergies with the handwriting analysis developed by Popovic and colleagues, because it too promises to find patterns and connections between different texts and fragments of scrolls. Instead of matching or differentiating manuscripts according to the handwriting, this method accomplishes the goal by looking at DNA.

So, for example, the presence of fragments written on parchment genetically traced to a cow, an animal not usually raised in the Judean desert, has already given more weight to the idea that at least some of the scrolls did not originate within the Qumran community.


In fact, in the future it should be possible to correlate the data gathered through the handwriting identification and the genetic analysis to see if they match and to glean even more knowledge about the anonymous hands that wrote the Dead Sea Scrolls, Popovic notes.

“We will never know their names,” he says. “But after seventy years of study, this feels as if we can finally shake hands with them through their handwriting.”

Ariel David