From left to right: Scholars Eli Piasetzky, Shira Faigenbaum-Golovin, and Israel Finkelstein—co-authors of a PLOS ONE paper published on June 3, 2025—featured in the documentary Ark of the Covenant: The Bible’s Origins, directed by Thierry Ragobert. (Screenshot)
A revolutionary study published in the PLOS ONE Journal has harnessed the power of artificial intelligence to unravel the mysteries of biblical authorship. Researchers from Duke University, Reichman University, the Protestant Faculty of Theology of Paris, Tel Aviv University, Collège de France, and the University of Haifa have employed advanced word frequency analysis to identify unique linguistic fingerprints within the Bible, offering scholars a powerful new tool for textual analysis.
The study, titled “Critical biblical studies via word frequency analysis: unveiling text authorship,” addresses one of the most enduring questions in biblical studies: who wrote the Old Testament? By analyzing subtle variations in word usage, the team distinguished between three distinct scribal traditions spanning the first nine books of the Hebrew Bible, known as the Enneateuch.
“The Bible reflects centuries of oral and written transmission, with layers of revision and redaction,” explains Axel Bühler, Senior Lecturer at the Protestant Institute of Theology in Paris. “Our method provides a quantitative approach to differentiate these authors, offering interpretable and statistically significant evidence of authorship,” adds Shira Faigenbaum-Golovin, Assistant Research Professor at Duke University.
Using statistical hypothesis testing, the researchers evaluated the likelihood of authorship, determining whether a text was more likely written by a particular author. This framework not only enhances accuracy but also provides interpretable results, deepening scholarly understanding of biblical composition.
In the first part of the study, the team focused on identifying linguistic characteristics across three major corpora: the oldest layer in Deuteronomy (D), the Deuteronomistic History (DtrH) in the Books of Joshua to Kings, and the Priestly writings in the Pentateuch (P). Their analysis confirmed that Deuteronomy and Deuteronomistic texts exhibit greater similarity to each other than to the Priestly writings—a finding consistent with established biblical scholarship.
From biblical texts to word distributions – and all the way to identifying the three corpora in the 50 examined texts (D, DtrH, and P). Graphs are taken from our PLOS ONE paper.
In the second phase, the researchers applied their statistical model to disputed biblical texts, generating new insights into their origins. These results offer fresh perspectives on the formation of biblical texts, potentially reshaping our understanding of ancient literacy and authorship practices.
“This methodology opens exciting possibilities for the field of biblical studies,” says co-author Thomas Römer of the Collège de France. “It complements traditional critical scholarship by providing objective, reproducible results.”
Israel Finkelstein from the University of Haifa highlights the broader implications: “Beyond biblical studies, this research exemplifies the growing potential of AI in the humanities. By bridging the gap between computational analysis and historical scholarship, the study introduces a new paradigm for analyzing ancient texts.”
Looking ahead, the team plans to expand their methodology to investigate additional texts of uncertain authorship and further refine their analytical models, paving the way for future discoveries in the field.
Contact: Shira Faigenbaum-Golovin
Department of Mathematics and Rhodes Information Initiative
Duke University
Email: alexandra.golovin@duke.edu