One Lexeme at a Time

Can the spelling of a Hebrew word really tell us when a biblical text was written?

This joint presentation challenges the traditional assumption that "full" spelling in Biblical Hebrew simply increases uniformly through time and can therefore function as a straightforward chronological marker. Instead, Naaijer and Ehrensvärd investigate whether orthographic change spreads one word at a time through the vocabulary — a process known as lexical diffusion.

Drawing on evidence from the Masoretic Text, the Samaritan Pentateuch, and the biblical Dead Sea Scrolls, and combining traditional Hebrew philology with computational analysis and hierarchical Bayesian regression, the presentation asks whether the familiar contrast between defective and full spelling is better understood as a lexically conditioned process rather than as a single chronological clock for dating biblical texts. For additional scholarly perspective, see the Institute's research on comparative textual evidence from the Masoretic Text, Samaritan Pentateuch and Dead Sea Scrolls.

Scholarship on the matres lectionis has long treated "plene spelling" as a single quantity that increases over time, and has used that rise as an indicator for dating biblical texts. This paper argues instead that orthographic fullness advances by lexical diffusion: it spreads through the vocabulary, governed by properties of individual lexemes rather than by a single corpus-wide clock. We make the case through two complementary computational studies drawn from the Masoretic Text, the Samaritan Pentateuch, and the biblical Dead Sea Scrolls.

The first is the name דוד/דויד. In our data, it is the only lexeme in the biblical lexicon to show near-total replacement of an earlier defective variant by a later full one in the MT. Mapping its distribution across the MT and DSS reveals the shape of a diffusion process that has run to completion, and exposes how heavily conventional "late" orthographic profiles depend on this one word.

The second is diffusion in progress. Within a single syllable type (qatōl), phonologically identical words behave oppositely: גדול is often full, קטן almost always defective. Using hierarchical Bayesian regression, we test which lexeme-level factors, token frequency, morphological environment, and semantic domain, predict a word's position along the defective-to-full continuum.

Together the two cases show that spelling change in Biblical Hebrew is lexically gradient, not uniform, with consequences for any argument that treats aggregate spelling counts as evidence of a text's date.

Dr. Martijn Naaijer is a Guest Researcher in the Section of Biblical Exegesis at the Faculty of Theology, University of Copenhagen. His research lies at the intersection of Biblical Hebrew linguistics, textual scholarship, corpus linguistics, statistics, and machine learning. He received his PhD from Vrije Universiteit Amsterdam in 2020 with the dissertation "Clause Structure Variation in Biblical Hebrew: A Quantitative Approach," in which he applied quantitative methods to linguistic variation in the Hebrew Bible.

From 2021 to 2024 he was a postdoctoral researcher at the University of Copenhagen, where his work brought together the study of the Dead Sea Scrolls, Biblical Hebrew, textual development, and machine-learning methods. He is currently also a postdoctoral researcher at the University of Zurich (2025–2027) in the project "Towards an Integrated Methodology of Dating Biblical Texts." His research investigates how statistical and computational approaches can contribute to questions of linguistic variation, textual transmission, historical linguistics, and the dating and development of biblical texts.

Naaijer is also involved in the creation and analysis of digitally annotated corpora of ancient Hebrew and other Semitic texts, including research on the Samaritan Pentateuch, Biblical Hebrew, Syriac, and the Dead Sea Scrolls.

Martin Gustaf Ehrensvärd, PhD, is Associate Professor in the Section of Biblical Exegesis at the Faculty of Theology, University of Copenhagen. He holds a PhD in Semitic Philology from Aarhus University and is a specialist in Biblical Hebrew, Semitic philology, historical linguistics, the linguistic history of Hebrew, and the methodological problems involved in using linguistic evidence to date biblical texts.

A major strand of Ehrensvärd's scholarship has challenged overly simple chronological models of Biblical Hebrew and emphasized the complexity of linguistic variation, scribal transmission, textual development, and the distinction between linguistic history and the date of individual biblical compositions. Together with Ian Young and Robert Rezetko, he co-authored the major study "Linguistic Dating of Biblical Texts."

At the University of Copenhagen, Ehrensvärd has also worked at the intersection of traditional Hebrew philology and computational research. Together with computer scientist Anders Søgaard, he has been a principal investigator in the project "Artificial Intelligence and Ancient Hebrew Texts," which uses machine-learning methods to investigate the linguistic and textual history of the Hebrew Bible and the biblical Dead Sea Scrolls.

His broader teaching and research background includes Biblical and Rabbinic Hebrew, Classical Arabic, Classical Syriac, other Aramaic dialects, Biblical translation, and the historical study of Semitic languages.