By New research applying artificial intelligence to ancient religious texts is providing fresh statistical support for one of biblical scholarship’s longest-running debates: whether the Hebrew Bible was written by multiple authors or editorial traditions rather than a single unified voice.
An interdisciplinary research team that includes scholars affiliated with Duke University has used machine-learning techniques to analyze subtle linguistic patterns in the earliest sections of the Hebrew Bible. The study focused on the first nine books, commonly referred to as the Enneateuch, examining word frequency and sentence structure at a granular level.
According to the researchers, the AI model identified three distinct stylistic clusters that closely align with categories long recognized by historians and textual critics: the Priestly source, the Deuteronomistic History, and the Book of Deuteronomy.
“We found that each group of authors has a different style – surprisingly, even regarding simple and common words such as ‘no,’ ‘which,’ or ‘king.’ Our method accurately identifies these differences,” said Thomas Römer, a professor at the Collège de France.
The project grew out of earlier research on ancient inscriptions. Mathematician Shira Faigenbaum-Golovin began studying handwriting on pottery fragments dating to around 600 BC more than a decade ago, using computational tools to distinguish individual scribes.
“We concluded that the findings in those inscriptions could offer valuable clues for dating texts from the Old Testament,” Faigenbaum-Golovin explained. “That’s when we started putting together our current team, who could help us analyze these biblical texts.”
While most chapters fit cleanly into one of the three identified styles, the analysis also revealed anomalies. Certain passages — including parts of the Ark Narrative in 1 Samuel — did not align with any of the major clusters, suggesting the possibility of additional contributors or editorial layers.
Researchers say the implications of the work extend beyond biblical studies. “If you’re looking at document fragments to find out if they were written by Abraham Lincoln, for example, this method can help determine if they are real or just a forgery,” Faigenbaum-Golovin explained.
The team described its conclusions as offering “interpretable, statistically significant evidence” of distinct linguistic traits among biblical writers or editors, arguing that the approach provides a new empirical tool for addressing long-disputed questions in the field.
“Our methodology thus provides a new tool to address disputed matters in biblical studies,” the researchers said.
Faigenbaum-Golovin emphasized the broader significance of the collaboration behind the project. “It’s such a unique collaboration between science and the humanities,” she said. “It’s a surprising symbiosis, and I’m lucky to work with people who use innovative research to push boundaries.”
