Colloquium Methodology of Hebrew Palaeography. Expertise, art or science ?

As Peter Stokes has noted, “[Presenting] palaeographical data… cannot be done using purely computational methods with the computer as a ‘black box’, since this simply replaces the palaeographer with the computer as a source of authoritarian statements. In this situation, how do we define the limits of the method we use as palaeographers? Can a method be questioned, and if so, to what extent and by which means? What constitutes palaeographical expertise, and how is it constructed? How have criteria of authority and validity evolved over time, and what can the history of (Hebrew) palaeography teach us about these changes? What kind of evidence should support claims of expertise?

To what degree must palaeographers defend their conclusions when communicating with non-palaeographers, be they antique dealers, editors of critical editions, digital curators, or historians specialising in other fields? How far can we – or should we – go in defending our views? How is the authority of a palaeographer perceived within and outside the field? Further, what cultural or cognitive biases affect palaeographical analysis? How can palaeographical findings be verified or replicated? Is it possible – or desirable – to formalise palaeographical expertise? Can a computer challenge a palaeographer’s conclusions, and should such computational tools be designed by palaeographers themselves? What does methodological transparency look like in the age of machine-assisted analysis?

In order to address these questions, we are organising a series of thematic sessions that will bring together palaeographers engaging with these questions from different perspectives. We welcome participants working outside the field of Hebrew manuscript studies, as well as palaeographers across the spectrum, from traditional methods to computational approaches. We invite both theoretical reflections on palaeography, and more ‘technical’ or case-based contributions. In all cases, we aim to foster methodological transparency.

Determining the geographical provenance of manuscripts is crucial for any discipline involving the study of texts, from palaeography and textual criticism to linguistics. Typically, this task is entrusted to codicologists and palaeographers, who base their assessments on an analysis of material and graphic features. These assessments are then taken up by philologists and linguists, who build on them when producing critical editions and conducting linguistic research.

However, as Penkower [1–3] has convincingly demonstrated, the reverse path—using text to infer provenance—is sometimes equally viable. Employing a predominantly quantitative method, Penkower was able to determine the “text type” of three specimens by exploiting a characteris tic feature of the medieval Hebrew Bible tradition, namely the strong correlation between ethno geographical area and the number of textual variants — most numerous in Ashkenazic manuscripts, less so in Italian, and least of all in Sephardic and Oriental.

The intuition underlying this method is straightforward: given an unclassified manuscript, its variants are collated and counted against a reference text; this value is then compared with those ob tained from manuscripts from known geographical areas, and the manuscript is ultimately assigned to the most likely area on the basis of its higher or lower number of variants.

We aim to extend Penkower’s intuition to a substantially larger corpus. Using Kennicott’s massive collation of the Hebrew Bible [4, 5], the comprehensive digitization of which has been carried out in the context of the REK project [6, 7], we will attempt to classify all the manuscripts collated in full by Kennicott, simply by looking at the quantity and distribution of their textual variants. To this end, we will test a range of computational techniques, from logistic regression to hierarchical clustering, and possibly phylogenetics as well [8].

It is our position that a quantitative approach of this sort will constitute a valuable additional tool in the arsenal of Hebrew Bible palaeographers: not only is it methodologically sound, but it is also more intuitive and transparent than the black-box methods currently employed in computer vision. Moreover, such an approach is rooted in a terrain that is familiar to both paleographers and philologists, hence rendering it conducive to mutual intelligibility as well as to cross-domain collaboration.

References

[1] J. S. Penkower, “A Tenth-century Pentateuchal MS from Jerusalem (MS C3), Corrected by Mishael ben Uzziel,” [Hebrew], Tarbiz, vol. 58, no. 1, pp. 49–74, 1988, issn: 03343650. Accessed: Apr. 1, 2020. [Online]. Available: www.jstor.org/stable/23597778.

[2] J. S. Penkower, “An Esther Scroll from the 15th Century: Determining its Type among Five Traditions (Oriental, Sefardi, Ashkenazi, Italian, Yemenite),” Textus, vol. 26, no. 1, pp. 209 270, Aug. 2016, Publisher: Brill, issn: 2589-255X, 0082-3767. doi: 10.1163/2589255X 02601011. Accessed: Dec. 21, 2025. [Online]. Available: https://brill.com/view/journals/ text/26/1/article-p209_11.xml.

[3] J. S. Penkower, “A Sheet of Parchment from a 10th or 11th Century Torah Scroll: Determining its Type among Four Traditions (Oriental, Sefardi, Ashkenazi, Yemenite),” Textus, vol. 21, no. 1, pp. 235–264, 2002, Place: Leiden, The Netherlands Publisher: Brill. doi: 10.1163/ 2589255X-02101012. [Online]. Available: https://brill.com/view/journals/text/21/1/ article-p235_12.xml.

[4] B. Kennicott, Vetus Testamentum Hebraicum cum variis lectionibus. Oxford: Clarendon, 1776, vol. 1. [5] B. Kennicott, Vetus Testamentum Hebraicum cum variis lectionibus. Oxford: Clarendon, 1780, vol. 2.

[6] L. Bambaci and D. Stökl Ben Ezra, “Enhancing HTR of Historical Texts through Scholarly Editions: A Case Study from an Ancient Collation of the Hebrew Bible,” in Proceedings of the Computational Humanities Research Conference 2023, A. Šeļa, J. Fotis, and I. Romanowska, Eds., ser. CEUR Workshop Proceedings, ISSN: 1613-0073, vol. 3558, Paris, France: CEUR, Dec. 2023, pp. 554–576. Accessed: Dec. 11, 2023. [Online]. Available: https://ceur-ws.org/Vol 3558/#paper6310.

[7] L. Bambaci and D. Stökl Ben Ezra, “Computational Reuse of Printed Critical Apparatuses: The Reverse Engineering Kennicott Project as a Case Study,” in 2025 IEEE 8th Congress on Information Science and Technology (CiSt), ISSN: 2327-1884, Oct. 2025, pp. 500–505. doi: 10.1109/CiSt65886.2025.11224071. Accessed: Dec. 8, 2025. [Online]. Available: https: //ieeexplore.ieee.org/document/11224071.

[8] L. Bambaci, “Is a Stemma Possible for the Hebrew Bible? Towards a Genealogy of Medieval Manuscripts Through Phylogenetic Analysis,” Materia Giudaica– Rivista dell’Associazione Italiana per lo Studio del Giudaismo, vol. XXVI, no. 2, pp. 3–30, 2021

This paper proposes a framework for the palaeographic analysis of substitutes for the Tetragrammaton in manuscripts belonging to the Hebrew Byzantine cultures. It begins with a brief review of what counts as a “divine synbol” on the written surface. It distinguishes between three current understandings of such symbols: (1) lexical forms that function as words (recognizable letters or combinations of letters that function as substitutes for the Tetragrammaton), (2) graphic symbols that operate non-lexically (such as paratextual marks comparable in function to  manicules), and (3) exegetical or traditional markers that encode theological interpretations in writing practices. Within this framework, the paper asks whether conventional markers such as letters (for instance, repeated Yods instead of the Tetragrammaton) retain their function as letters, or become repurposed graphic units.

The discussion is further developed through a palaeographic analysis of divine symbols used to replace the name of God in medieval Hebrew manuscripts. A comparative visual overview across Ashkenazi, Sephardi, Oriental, and Italian scribal cultures highlights the practical, exegetical, and intellectual factors shaping these choices. Selected case studies demonstrate how particular graphic substitutes for the divine name can function as indicators of script type and scribal tradition. At the same time, they pinpoint to the pitfalls of attempts to correlate consistently one type of script with one type of substitute. Attention is given to scribal variation, local preferences, and moments of hybridity between distinct manuscript traditions during key phases in the historical development of Hebrew scripts. These cases reveal how certain substitutes emerged, persisted, or shifted according to scribal training, cultural contact zones, and the transmission histories of particular works.

The study then turns to Byzantine Hebrew script as a focused case study on the relationship between divine symbol and script. Byzantine manuscript cultures have a particularly significant position at the intersection of Italian and Oriental Hebrew scribal traditions. Such a position fostered mixed practices and distinctive graphic alternatives to the Tetragrammaton. The analysis of substitutes draws on both quantitative and qualitative methods. The quantitative component relies on HebrewPal, a structured palaeographic database with predefined descriptive parameters, while the qualitative component focuses on close examination of individual substitutes in specific manuscripts.

Ultimately, the paper demonstrates that palaeographic methods offer more than technical classification. They provide insight into broader questions in the history of the book, including scribal identity, cultural exchange, and the ways in which communities visually negotiated the sanctity of the divine name.

The National Library of Israel has digitized almost all of the extant 100,000 medieval manuscripts in Hebrew script. State-of-the-art computational tools are currently being developed to paleographically date and provenance a significant fraction of them, and to find joins of fragmentary manuscripts, including those from the Cairo Geniza. The regions of images that contribute most strongly to the algorithmic decisions are highlighted, so that experts can judge their significance. Additionally, linguistic and intertextual tools will be used to help analyze the texts therein.

The Egrapsa Project (University of Basel) studies the paleography of ancient Greek papyri (3rd century BCE–7th century CE) through the reconstruction of the pen movements that produced the traces we see today. We argue that handwriting must be understood as a dynamic process rather than a static result. This methodological shift raises two central questions. First, what does objectivity mean in paleographic research, and how can it be achieved? We compare the capacities of human observers and computational systems—such as machine learning and neural networks—to approach objectivity through replicable and transparent procedures. Second, we examine the crucial problem of measuring similarity. Can computers genuinely quantify resemblance, or does their metric logic differ from human perception? By addressing these issues, our study aims to move paleography from art toward science—from subjective interpretation to analytical reconstruction grounded in movement and comparison.

Meticulous distinction between different types of the Hebrew script in the Middle East and Inner Asia still remains desiderata in the field of medieval palaeography. However the Persian type usually appears as a clearly separate variation among the Oriental Hebrew scripts, especially towards the end of the Middle Ages. The present paper deals with the still narrower area and its peculiar sub-type within the Persian writing, namely the script of the Timuridic Transoxiana. Although not numerous, the manuscripts copied according to their colophons in the cities of Bukhara and Samarkand share specific forms of the letters that can be seen as a further development of the general Persian script. The features of this local type are examined here with intention to formalise their palaeographical description. We should hope it can help us to localise more precisely some manuscripts without an indication of place of copying that came down to us from the Greater Iran region. All the manuscripts under consideration were written in the second half of the 15th century and are described in the Sfar-Data codicological database. Most of them came from the Moussaieff collection, now in the Bar Ilan University library (Ramat Gan, Israel), they were easily available in the original form. Two additional manuscripts are kept in the AIU library in Paris and the JTSA library in New York, they are available in microfilms or black-and-white scans. The examination of the palaeographical characteristics is largely based on the HebrewPal Album and its detailed description of graphic elements.

The Ancient Egyptian writing culture existed for over 3,000 years. It included more than just the pictorial and monumental hieroglyphs; there were also more or less simplified forms of characters mainly used in handwriting and inked on papyrus or other materials. Depending on the degree of cursivity and for reasons related to the history of science, the script variations are referred to as cursive hieroglyphs. hieratic, abnormal or cursive hieratic, and demotic.

The long-term project “Altägyptische Kursivschriften. Digitale Paläographie und systematische Analyse des Hieratischen und der Kursivhieroglyphen” (AKU project), based at the Academy of Sciences and Literature | Mainz (Germany), is dedicated to the paleographic analysis of hieratic and cursive hieroglyphs. 

The relationship between hieratic and hieroglyphic writing is intricate. Both script variants developed concurrently, exerted mutual influence, and produced partially different repertoires of signs. Since hieratic, unlike hieroglyphs, was the daily handwriting in ancient Egypt, there exists considerable variation in the forms writing a sign. The sign forms of a hieratic or cursive manuscript are referred to as “hieratograms” or “allographs” of a single grapheme. The diversity of hieratic forms is not solely attributable to the idiosyncrasies of individual scribes. It also depends on factors such as the historical development, education and training, writing tools, the physical characteristics of the writing surface, and the function of the text being produced, among others. Digitizing hieratic signs is crucial for the modern paleographic study and systematic processing of the vast number and variety of hieratic allographs. Some analyses have already been conducted using digitized material, paving the way for further research.

After a presentation of the AKU project’s aims and methods, this contribution centers on two key aspects: the complex relationship between hieroglyphs and hieratic, and the challenges and opportunities digital analysis can add to the field of ancient Egyptian paleography.

The so-called Italian Rotunda occupies a distinctive yet methodologically challenging place in the classification of Hebrew scripts. Traditionally described as a formal, rounded book-hand characteristic of the Italian peninsula between the late medieval and early modern periods, the Rotunda has served as a key marker for dating and localisation. Yet its internal variability, its intersections with different scribal practices in Italy, and the evolving criteria used to define it raise crucial questions about the nature of palaeographical expertise and the methodological transparency of our field.

This paper proposes a critical re-examination of the Italian Rotunda as a palaeographical category by integrating close visual analysis, historical contextualisation, and insights drawn from ongoing computational approaches developed within the MiDRASH framework. Taking as a starting point a corpus of manuscripts produced in Italy (13th-16th centuries), the study explores three main axes:

1.         Terminology and Definition – How has the term “Rotunda” been employed and defined by scholars of Hebrew script from the 19th century to the present? Which implicit assumptions about script evolution, scribal identity, and regional coherence underlie current definitions?

2.         Expert Judgment vs. Formalised Criteria – Which features are considered diagnostic (ductus, letter modulation, curvature patterns, ligatures), and how consistently are they applied across manuscript descriptions? Can these features be formalised without flattening the nuance essential to palaeographical expertise?

3.         Intersections and Hybridities – Many manuscripts traditionally labelled “Rotunda” bear traces of different graphic habits, reflecting the mobility of Jewish scribes within the peninsula. How should palaeographers handle hybrid or transitional scripts: reinforce typologies, or question their boundaries altogether?

By presenting a series of case studies the paper demonstrates how classificatory choices are influenced by both cultural assumptions and cognitive biases. Finally, it reflects on how computational tools can support (but not replace) expert analysis: can machine-detected curvature metrics or ductus approximations make our reasoning more transparent, or do they risk introducing new opaque “black boxes”?

Ultimately, this contribution aims to show that studying the Italian Rotunda is not merely a technical exercise but a methodological lens through which to interrogate authority, expertise, and the future of Hebrew palaeography.

This paper offers a methodological overview of one of the earliest, most documented stages of the Hebrew script, focusing on how Iron Age Hebrew palaeography is studied today. Central themes include the challenges of palaeographic dating, complications dependent on the writing medium, and the question of whether—and in what sense— we may speak of the “evolution” of early Hebrew script. What constitutes script evolution? Which analytical procedures allow us to identify developmental trends, and where do their limits lie? The scripts of Hebrew seals and ostraca will serve as the main case studies, illustrating both the diversity of palaeo-Hebrew writing practices and the methodological issues involved in classifying and dating them.  In addition, the lecture will reflect on the challenges of teaching West Semitic epigraphy and palaeography to today’s students—highlighting both the advantages afforded by new technologies and the difficulties posed by shifting learning habits, limited language backgrounds, and the high level of interpretive nuance required.

In the late 1980s, Colette Sirat, a pioneer in the modern study of palaeography and codicology of manuscripts in Hebrew characters, authored L’Edition des textes philosophiques médiévaux: Questions de methodologie. In this article, Sirat expressed concerns about the limited availability of scholarly editions of Hebrew philosophical texts. She briefly identified some of the methodological issues that could constrain researchers in undertaking such fundamental scholarly endeavors. Her analysis thereby illustrated the close relationship between the study of manuscripts, textual traditions, and editorial processes involving Hebrew texts. Although these processes are interconnected, their interaction can lead to methodological conflicts. Similar observations were noted by Saverio Campanini in his article: Problemi metodologici e testuali nell’edizione del Sefer ha-Bahir. In his examination of methodological challenges in editing kabbalistic texts, Campanini offers a theoretical framework grounded in two seminal works of modern classical philology: Paul Maas’ Textkritik and Giorgio Pasquali’s Storia della Tradizione e Critica del Testo. Drawing upon these two articles concerning Hebrew text editions, I intend to extend this investigation by applying contemporary philological methodologies to practical research cases, focusing on kabbalistic texts. I will begin my intervention by expanding upon the theoretical framework established by Campanini, both retrospectively, with a brief reference to the debates between Friedrich Nietzsche and Ulrich von Willamowitz Moellendorf, and prospectively, emphasizing the foundational contributions of material philology and the concept of textual vitality during the Middle Ages. Following the establishment of these methodological foundations, I will proceed to present several intriguing cases of unique interactions between the characteristics of manuscripts and kabbalistic texts. These cases will offer insights and depth to the foundational question: what do we edit when we edit?

Our paper is a case study centered around Neirab I (KAI 225), a 6th century BCE Old Aramaic epitaph for Sin-zer-ibni the priest of Sahar at Neirab. Recently, we have identified two groups of partially erased letters along the border of the stele using the RTI sequence produced by the West Semitic Research Project. These groups match portions of the main text on the stele, and we suggest they represent two unsuccessful attempts to inscribe Sin-zer-ibni’s epitaph. These observations illuminate an imperfect process of inscription and highlight two cultural and cognitive biases in the study of epigraphy. Our conclusions challenge these biases by taking the evidence of poor planning in Neirab I as an impetus to re-evaluate the role, and literacy, of the stone carver as opposed to the scribe, re-envision the production of letter forms as a more subjective task than generally presumed, and displace the reed pen as the measuring stick for the production of letters on stone. On the basis of these challenges, we propose two modifications to epigraphic practice: increased engagement with the theory of craft literacy as put forth by Alice Mandell and improved methodological sophistication, and thus transparency, in the evaluation of ductus. Regarding theory, we promote a broader critical awareness of the impact of writing material on the production of letters and further advocate for increased nuance regarding the centrality of the “scribe” in our theorizations of epigraphic production. Regarding methodology, Neirab I exemplifies the need for quantitative analysis for evaluating ductus in order to combat the subjectivity of visual observation, and description, to improve methodological transparency. Using tools such as Adobe Illustrator to measure letter orientation and stroke angles in relation to an object’s surface and performing statistical analysis on this data can provide a consistent basis for comparison of letter formation regardless of their orientation. Proper utilization of these tools, coupled with quantitative analysis, can offset the subjectivity of visual inspection across variant materials and object shapes. Ultimately, it promotes both analytic accountability and methodological transparency in the evaluation of ductus, and consequently, epigraphic analysis more broadly.

Ductus is a fundamental concept in palaeographic analysis, and its definition affects every analytical level, from the morphology of individual letters (and other signs) to the identification of script modes. Building on the methodology of HebrewPal, this study proposes an extended definition of ductus that takes into account micromovements and examines its usage as a new methodological tool in several areas that continue to present methodological challenges.

The first area concerns letter morphology, specifically the analysis of junctions between letter components. Rather than highlighting the shape of the juncture, the extended definition of ductus prioritizes the change in direction and speed of writing as primary analytical parameters.

The second implication affects the global level of analysis, namely, the determination of writing mode. Currently, the term “non-square” or “semi-cursive” covers a broad range of scripts that differ in visual characteristics, execution techniques, and function. The proposed extended definition of ductus provides a more precise basis for the classification of bookhand non-square scripts.

The third area of application addresses the relationship between individual handwriting and script models. Here, the extended definition of ductus makes it possible to define the borders of the script model, potentially enabling the application of computational methods for its reconstruction. At the same time, it clarifies the mechanisms through which individual hands diverge from shared models, resulting in, along with other factors, the graphic variability of each individual handwriting.  

The study is illustrated through a corpus of 13^th to 16^th century manuscripts written in non-square bookhand, analysed within the framework of HebrewPal. The case studies encompass Sephardic and Ashkenazic manuscripts, supplemented by selected examples from Oriental and Italian manuscripts.

dating

handwriting identification

evolution of script

localisation

terminology and its definition

the universality (or lack thereof) of palaeographical terminology

Paper title and abstract (250-350 words).

Short CV (maximum 2 pages).

Current institutional affiliation.
All materials must be submitted in English.
Please note that the number of papers we can accommodate is limited; proposals will be selected
based on their relevance to the colloquium’s theme.
Please send your application to Evgeniya Zarubina [evgeniya.zarubina@ephe.psl.eu] using
“Hebrew Palaeography Colloquium Application” as the subject line.
Application deadline: 21 November 2025
Notification of acceptance: 20 December 2025
If you have any questions regarding the colloquium’s organization, please do not hesitate to
contact the organizer using the email above.
Organizer: Dr Evgeniya Zarubina, EPHE-PSL, Paris, France
Organised within the framework of the ERC project MiDRASH (Grant N° 101071829).