Unlocking the timecode of the Dead Sea Scrolls

Using artificial intelligence and radiocarbon dating to date ancient manuscripts.

University of GroningenSince their discovery, the historically and biblically hugely important Dead Sea Scrolls have transformed our understanding of Jewish and Christian origins. However, while the general date of the scrolls is from the third century BCE until the second century CE, individual manuscripts thus far could not be securely dated. Now, by combining radiocarbon dating, palaeography, and artificial intelligence, an international team of researchers led by the University of Groningen has developed a date-prediction model, called Enoch, that provides much more accurate date estimates for individual manuscripts on empirical grounds. Using this model, the researchers demonstrate that many Dead Sea Scrolls are older than previously thought. And for the first time, they establish that two biblical scroll fragments come from the time of their presumed biblical authors. They presented their results in the journal PLOS One on 4 June.

Until now, the dating of individual manuscripts was mostly based on palaeography–the study of ancient handwriting–alone. However, the traditional palaeographic model has no solid empirical foundation. For most Dead Sea Scrolls, a calendar date is not known, and there are no other date-bearing manuscripts from the time period available for palaeographic comparison. Between the few date-bearing manuscripts in Aramaic/Hebrew from the fifth–fourth centuries BCE and the late first and early second century CE a gap is present that prevents accurate dating of the more than one thousand scrolls and fragments from the Dead Sea Scrolls collection.

Digitized manuscripts

This gap is now closed by researchers in the ERC project The Hands That Wrote the Bible by combining radiocarbon dates from 24 scrolls samples, in combination with palaeographic analysis using a machine-learning-based model applying a Bayesian ridge regression method. The new radiocarbon dates are reliable, empirical time markers that bridge the palaeographic gap between the fourth century BCE and the second century CE. They provide an objective date for the writing styles in the tested manuscripts.

Based on this information, the researchers trained the date-prediction model named Enoch. They used a previously in-house developed deep neural network for detection of handwritten ink-trace patterns, BiNet, in digitized manuscripts. This allows for a subsequent geometric shape analysis at the microlevel of the ink trace, such as curvature (called textural), as well as at the level of character shapes (called allographic). It provides a quantitative and empirical basis for the style analysis of handwriting which traditional palaeography cannot deliver. Cross-validation then showed that Enoch can predict radiocarbon-based dates from style with an uncertainty of some 30 years (plus and minus). This is even more precise than direct radiocarbon dating results in the period range of 300–50 BCE.

The first machine-learning-based model

Now that Enoch is ready for use, it becomes possible to date the roughly one thousand Dead Sea manuscripts from this time period. The researchers took a first step in this by feeding Enoch the binarized images of 135 scrolls and having the date predictions evaluated by palaeographers. With Enoch, researchers have a powerful new tool that they can use to support, refine, or modify their own subjective estimates for specific manuscripts, often to an accuracy of only 50 years for manuscripts over 2,000 years old.

Enoch is the first complete machine-learning-based model that employs raw image inputs to deliver probabilistic date predictions for handwritten manuscripts, while ensuring transparency and interpretability through its explainable design. The combination of empirical evidence (radiocarbon from physics and character-shape-based analyses from geometry) brings a degree of quantified objectivity to palaeography never before achieved in the field. And the methods underpinning Enoch can be used for date prediction in other partially-dated manuscript collections.

New chronology

First results from Enoch’s date predictions, presented in the PLOS One paper, demonstrate that many Dead Sea Scrolls are older than previously thought. This also changes how researchers should interpret the development of two ancient Jewish script styles which are called ‘Hasmonaean’ and ‘Herodian’. Specifically, manuscripts in Hasmonaean-type script can be older than the current estimate of ca. 150–50 BCE. And the Herodian-type script emerged earlier than previously thought, suggesting that these scripts existed next to each other since the late second century BCE instead of the mid-first century BCE which is the prevailing view.

This new chronology of the scrolls significantly impacts our understanding of political and intellectual developments in the eastern Mediterranean during the Hellenistic and early Roman periods (late fourth century BCE until second century CE). It allows for new insights to be developed about literacy in ancient Judaea in relation to historical, political, and cultural developments such as urbanization, the rise of the Hasmonaean dynasty, and the rise and development of religious groups such as those behind the Dead Sea Scrolls and the early Christians.

Anonymous authors

Furthermore, this study establishes 4QDanielc (4Q114) and 4QQoheleta (4Q109) to be the first known fragments of a biblical book from the time of their presumed authors. We do not know who exactly finished the Book of Daniel but the common assumption is that this author did that during the early 160s BCE. Likewise, for Ecclesiastes (Qohelet) scholars assume that an anonymous author from the Hellenistic period (third century BCE) was behind this biblical book, instead of the view of tradition that it was King Solomon from the tenth century BCE. Our novel radiocarbon dating for 4Q114 and the Enoch date prediction for 4Q109 place these manuscripts in the same time as these anonymous authors from respectively the second and third centuries BCE. Thus, these results have now created the opportunity to study tangible evidence of hands that wrote the Bible.

______________________________

For this study, writing styles in digitized manuscripts were analyzed using BiNet, a previously developed deep neural network for detection of handwritten ink-trace patterns. By combining writing styles with carbon-14 dates of manuscripts using artificial intelligence, the date-prediction model Enoch is able to produce an accurate date for the manuscript.  Maruf Dhali, University of Groningen

______________________________

This photo shows prof. Mladen Popovic, Professor of Hebrew Bible and Ancient Judaism and Director of the Qumran Institute (front) and his colleauge Dr. Maruf Dhali, Assistant Professor in Artificial Intelligence, working with Enoch to date a manuscript from the Dead Sea scrolls.
This is a still from a short video on the project, which is linked from the PLOS One paper. University of Groningen

______________________________

This is part of Dead Sea Scroll number 109 (4Q109), also known as Qohelet (Ecclesiastes). From Qumran Cave 4. From Qumran (Khirbet Qumran or Wadi Qumran), West Bank of the Jordan River, near the Dead Sea, modern-day State of Israel. The Jordan Museum, Amman, Jordan Hashimite Kingdom. Osama Shukir Muhammed Amin FRCP(Glasg). CC BY-SA 4.0, Wikimedia Commons

______________________________

Article Source: University of Groningen news release.

*Mladen Popović, Maruf A. Dhali, Lambert Schomaker, Johannes van der Plicht, Kaare Lund Rasmussen, Jacopo La Nasa, Ilaria Degano, Maria Perla Colombini, Eibert Tigchelaar, Dating ancient manuscripts using radiocarbon and AI-based writing style analysis. PLOS One, 4 June 2025.

Subscribe to Popular Archaeology Premium. Still the industry's best value at only $9.00 annually.