Chapter 8 Reading and Text Reconstruction

In: Material and Digital Reconstruction of Fragmentary Dead Sea Scrolls
Jonathan Ben-Dov
Search for other papers by Jonathan Ben-Dov in
Current site
Google Scholar
Asaf Gayer
Search for other papers by Asaf Gayer in
Current site
Google Scholar
, and
Eshbal Ratzon
Search for other papers by Eshbal Ratzon in
Current site
Google Scholar
Open Access

While earlier chapters focused on the fragments and their images, in this chapter we address the text preserved on them, aiming to raise methodological issues which will ultimately lead to creating a fully-fledged digital canvas for each scroll. Imaging technologies raise new issues that should be addressed in this regard, starting with the basic act of reading the signs and reporting them to the reader by means of transliteration. In broken scrolls, it is sometimes important to suggest textual reconstructions to fill the lacunae, or in other cases to fill the space between two fragments that stand in the same column. We evaluate the technique known as letter cloning, in which lacunae in the scroll are filled by means of cut-and pasted letters, to check the validity of the suggested reconstruction. Finally, we assess the act of suggesting text reconstructions, discuss some of its limitations, and offer guidelines for appropriate conduct.

1 Marking Doubtful Letters

A substantial methodological question arises with regard to marking doubtful letters. Sister-professions like classical paleography, where a lot of text is available and therefore the relative significance of individual signs is diminished, employ only one markup of doubt, usually a dot underneath the transcribed letter.1 Editions of DSS usually distinguish three levels of doubt for damaged letters, marking them above the letters according to the certainty of the reading: A dot above (א̇) signals a most probable letter, a circlet (א̊) above an uncertain reading, and an empty circle () in the place of an unidentified letter.2 In his recent edition, Qimron retained these three levels of doubt, while refining the notation by means of new marks for abraded letters, various types of lacunae, etc.3 The use of a circlet (א̊) markup presents a problem, however, because scholars use it to express two separate phenomena. While in some cases it does mark an uncertain letter as required, in other cases, especially when parallel texts are used for the reconstruction or when the composition uses highly formulaic, recurring language which enables easy reconstructions, the circlet is an inadequate marker. In such cases, one encounters very small marks of ink, which could in fact be part of many different letters; had there been no textual parallels or formulas, these letters would have been designated as unreadable marks (). Recording them in the transcription is based solely on other, external indications, while the letters are not there to see.4 In their edition of 4Qpap cryptic A Serekh haEdah, Ben-Dov, Stökl Ben Ezra, and Gayer used hollow letters to mark such cases, in addition to the two usual levels of markup.5 This system is particularly useful in highly fragmentary scrolls, where each letter culled from a parallel or formula carries a great value for the overall understanding of the text. Hollow letters will not be used in the present volume however, due to technical limitations of production.

2 Letter Cloning

Scholarly reconstructions are often made with standard Hebrew computer fonts such as SBL Hebrew or David. Since these fonts reproduce neither the sizes of the original letters nor their exact positions when specific pairs of letters are invoked (e.g., in the sequence כו, when the vav often touches the bottom stroke of kaf), better replacements should be sought. A more reliable technique is employed by skilled paleographers like Émile Puech or the late Ada Yardeni, who draw the suggested letters in the lacuna while perfectly imitating the extant script in the scroll.6 The use of such a technique, however, depends on the rare skills and widely recognized authority of gifted individuals, and cannot be adopted wholesale by non-trained textual scholars. In addition, hand-drawing the letters retains a certain measure of subjectivity and variation on the part of the expert paleographer, one which cannot be achieved in automated letter reconstruction. Such subjectivity may improve the reconstruction, allowing more flexible results, but may also harm it by producing idiosyncrasies or inexact renderings of letters. In this book we suggest a more standardized solution.

The technique of letter-cloning involves copying the shape of the same letter or letters from a proximate position in the scroll, and pasting them on the extant signs, or in the lacuna, as a new layer of the image.7 This technique, first suggested by Armin Lange in 1993, is now much easier to achieve using GIMP or Photoshop.8 Letter cloning is useful when reading fragmentary letters and when completing small-scale lacunae. In the former case, one should validate that the reading fits the actual signs, sometimes no more than minute specks of ink left on the fragment. In the latter case, the suggested reading must fit the space of the lacuna. Letter cloning is efficient and reliable for completing small-scale lacunae. When a reconstruction of longer texts is required, we recommend using custom-designed fonts, as explained in detail in chapter 10.

Adding color to the pasted letters adds transparency to letter cloning, distinguishing the extant text from the artificial one (figure 19). The opacity of the pasted letter can be diminished in order to make the signs of ink on the fragment show more clearly. When completing broken letters, it is advisable to paste the suggested letters only in outline (figure 18).

When copying and pasting letters one should choose the most complete and reliable letter forms.9 Important methodological points to take into account are:

  1. Letters should preferably be copied from a close region on the fragment.

  2. Letters should preferably be copied from the same constellation in the word (beginning of the word, position with regard to the neighboring letter, connected letters etc.).

    Figure 18
    Figure 18

    Reading the words ד̇ב̊ק̇י פ̇לא in 4Q405 19a 5. The suggested letters are pasted in outline on top of the extant signs. Left: the fragment in its original state. Right: the letters ב and ק, cloned from the same fragment and pasted on the extant ink signs. The cloned letters validated Carol Newsom’s reading of this fragment (“MasShirot ʽOlat HaSabbat” in Qumran Cave 4 VI: Poetical and Liturgical Texts, Part 1, DJD XI, Esther Eshel et al. [Oxford: Clarendon Press, 1998], 239–52), contra Qimron’s later suggestion. For a complete discussion and references see Noam Mizrahi, “Eleventh Song of the Sabbath Sacrifice: Literary form and Exegetical Content,” [Hebrew] Tarbiz 87.1 (2020): 5–36.

    © IAA, LLDSSDL, Shai Halevi. Graphics: Asaf Gayer

  3. It is preferable to copy-paste the same sequence of letters rather than a single letter.10

    Figure 19
    Figure 19

    A section from 1QSa I 18–21, with suggested completion of the lacunae. The image reproduced (colored blue) is a graphic reconstruction of the scroll based on a photograph taken by Bruce Zuckerman and Kenneth Zuckerman, West Semitic Research, in collaboration with Princeton Theological Seminary.

    Courtesy Jordan Museum. Graphics: Einat Tamir

  4. Letters should retain the original scaling of the copied letters.

3 Reconstruction Based on Parallels

While scholars often suggest completions of words based on the context or on biblical allusions, lacunae and broken letters are better reconstructed based on a parallel text. A convenient list of overlaps between copies in the non-biblical Qumran corpus was composed by Tigchelaar, but other lists have arisen since then.11 Caution should be practiced when using the text of other copies of the same composition in textual reconstructions, because copies may reflect different forms of the source composition, and significant variations are ubiquitous in Second Temple compositions.12 In contrast, quite a few other texts – whether sectarian or not – display a reasonable amount of stability, with copies differing only in minor scribal mistakes and corrections.13 Reconstructions based on textual parallels seem more justified in these cases. For example, a reconstruction of the highly fragmentary cave 4 copy of Serekh haEdah was made possible by means of a digital reconstruction of its text in parallel to the well-preserved 1QSa, despite some recensional differences between the two.14 It seems that Instruction belongs to the group of scrolls with a relatively stable text. In this composition, the long overlaps preserved between different copies attest to meager textual changes, amounting to several letters or – at a maximum – one word.15

Potential pitfalls of this procedure must be kept in mind.16 Orthography may be different from copy to copy; for that purpose, the scholar must first establish the orthographical profile of each manuscript, which will then allow them to verify the readings by means of letter cloning. Second, there is the problem of the frequency of vacats, as some scribes insert more or longer vacats in their text than others. For example, in the parallel text appearing in 4Q418 9 and 4Q416 2 iii, the scribe of 4Q418 clearly uses longer vacats, as seen at the end of lines 12 and 16. This problem will be addressed in chapters 9 and 10, as part of the procedure of textual reconstruction.

In general, we are wary of suggesting hypothetical completions of lacunae when a parallel is absent.17 However, textual completions have often proved useful when a new join or a new arrangement of fragments is suggested. Scholars who suggest the new join can significantly strengthen their case by positing words to bridge the gap between the respective fragments while maintaining valid syntax and reasonable content. Arguably, a new distant join can only be put forward if accompanied by a feasible completion. In such cases, some measure of textual speculation and creativity is indeed warranted, as is employed in the present volume. Of course, scholars may be content with the separate fragments as they are, not suggesting joins and not having to invent text completions. Such minimalism, we fear, is not to the benefit of our profession.


This is the rule in the Leiden Conventions. See:


These definitions are our translation from Émile Puech, Qumran Grotte 4. XXII: Textes araméens, première partie: 4Q529–549, DJD XXXI (Oxford: Clarendon Press, 2001), xviii. Most DJD authors use the same categories. Barthélemy and Milik, Qumran Cave 1, 48, define three levels of doubt, the lowest one being a letter that is paleographically improbable but required by the context; this category appears with a question mark above the letter. The same annotation of a raised question mark had been initially used by Strugnell and Harrington to denote a third level of doubt, but it was removed as part of the preparation for the DJD XXXIV volume (see Tigchelaar, Increase Learning, 20).


Qimron, The Dead Sea Scrolls, 1.כח, lists the various markups on an inserted bookmark for use next to the edition. The rules for these new signs are not indicated in the actual book.


Such markup is frequent in formulaic texts like the Astronomical Enoch, as amply attested in Henryk Drawnel, The Aramaic Astronomical Book (4Q208–4Q211) from Qumran: Text, Translation and Commentary (Oxford: Oxford University Press, 2011).


Jonathan Ben-Dov, Daniel Stökl Ben Ezra, and Asaf Gayer, “Reconstruction of a Single Copy of the Qumran Cave 4 Cryptic-Script Serekh haEdah,” RevQ 29.1 (2017): 21–77, here 38–39.


See for example Bezalel Porten and Ada Yardeni, Textbook of Aramaic Documents from Ancient Egypt, Volume 1: Letters (Jerusalem: Hebrew University Department of the History of the Jewish People, 1986), 125; Émile Puech, “11QPsApa: Un rituel d’exorcismes. Essai de reconstruction,” RevQ 14,3 (1990): 377–408, especially 404–8. An extreme application of this technique was carried out by Andrew Fincke, The Samuel Scroll from Qumran. 4QSama Restored and Compared to the Septuagint and 4QSamc (Leiden: Brill, 2001).


The term “letter-cloning” was suggested by Bruce Zuckerman, “Every Dot and Tiddle”; Zuckerman, “The Dynamics of Change.” See 13–19 in the online version:


Lange, Computer-Aided Text-Reconstruction; Lange, “Computer Aided Text-Reconstruction.”


These restrictions were formulated by Zuckerman, “The Dynamics of Change,” 13–19.


See the considerations in Yigal Bloch, Jonathan Ben-Dov, and Daniel Stökl Ben Ezra, “The Rule of the Congregation from Cave 1 of Qumran: A New Edition,” REJ 178.1–2 (2019): 1–46, here 6. Some earlier studies are: Michael Langlois, “Les manuscrits araméens d’Hénoch,” 115–119; David Hamidović, Les traditions du jubilé à Qumran (Paris: Geuthner, 2007); Hamidović, “In Quest of the Lost Text”; Bronson Brown-deVost, “4QEnc (4Q204) Column I: A New Reconstruction,” in From Enoch to Montréal and Back: New Vistas on Early Judaism and Christianity. Papers from the Fifth Enoch Graduate Seminar, Montréal, 20–24 May 2014, ed. Lorenzo DiTommaso and Gerbern S. Oegema (London/New York: Bloomsbury T&T Clark, 2016), 60–84; Tucker and Porzig, “Between Artefacts.”


Eibert Tigchelaar, “Annotated Lists of Overlaps and Parallels in the non-Biblical Texts from Qumran and Masada,” in The Texts from the Judaean Desert. Indices and an Introduction to the Discoveries in the Judaean Desert Series, ed. Emanuel Tov. DJD XXXIX (Oxford: Clarendon Press, 2002), 285–322; For new overlaps see for example Ariel Feldman, “An Unknown Prayer from 4Q160 and 4Q382,” [Hebrew] Meghillot 11–12 (2014–15): 99–109.


Famous examples are the differences between the MT and LXX versions of Jeremiah. Among the non-biblical texts, a well-known example is the various texts of S (1QS, 4Q255–264, 5Q11, and 5Q13), see e.g., Philip Alexander, “The Redaction History of ‘Serekh Ha-Yaḥad’: A Proposal,” RevQ 17 (1996): 437–56; Charlotte Hempel, “Shifting Paradigms Concerning the Literary Development of the Serekh,” in The Qumran Rule Texts in Context (Tübingen: Mohr Siebeck, 2013), 109–19; Qimron, The Dead Sea Scrolls, 1.209–10.


A notable example where only minor variants are attested is Shirot Olat HaShabbat, whose manuscripts were found both in Qumran and in Masada (4Q400–407, 11Q17, Mas1k). Noam Mizrahi stated the textual stability of Shirot Olat HaShabbat in an oral presentation: “Textual Pluriformity and Literary Development in the Qumran Scrolls,” lecture at the conference “Textual Plurality beyond the Biblical Texts,” Université de Lorraine, 2017.


See Ben-Dov, Stökl Ben Ezra, and Gayer, “Reconstruction of a Single Copy.” In turn, this reconstruction called for modifications in the text and reconstruction of 1QSa, see Bloch, Ben-Dov, and Stökl Ben Ezra, “The Rule of the Congregation.”


This matter will be discussed further in chapter 14. Tigchelaar, Increase Learning, 64, mentions “instability of the text” of Instruction. This is relative, however, as he refers to instability in terms of spelling and other minor variants, not of large textual differences.


See Herbert, Reconstructing Biblical Dead Sea Scrolls, 5–6.


See Qimron’s cautious methodological remark in The Dead Sea Scrolls, 1.יב.

  • Collapse
  • Expand