Fragments and Forefathers: An Experiment with the Reconstruction of 4QVisions of Amram

In: Vision, Narrative, and Wisdom in the Aramaic Texts from Qumran
Author: Søren Holst
Open Access

1 Introduction1

Attempting to piece together the original sequence of the fragments of the Qumran Visions of Amram presents an unusually satisfactory experience—up to a certain point: It is certain that extant copies of the text include manuscripts 4Q543–547 which have overlapping sections. Due to these overlaps, two sequences of reasonably continuous text may be reconstructed with full certainty. In between the two, there is a considerable passage, altogether as long as approximately one column in the writing format of 4Q545, in which all that is left is either quite fragmentary, or in the case of the last eight lines preceding the second of the two sequences, literally no text at all.2

In terms of physical reconstruction rather than textual content, though, it would be more precise to speak of one continuous sequence rather than two, since—as is clear from Émile Puech’s and Jean Starcky’s reconstruction work—the passage that begins to grow fragmentary in the middle of 4Q545’s column I, and the passage that picks up around the middle of column II, are quite securely connected by a well-preserved piece of column margin, although not by coherent text.3

The extent of the material lost in the space spanning from 4Q545 I 11 to II 10 may be calculated with considerable exactitude, since the circumstances allow us to determine the precise column height of 4Q545 as well as the width of its first column on the basis of tangible physical evidence, and to estimate with considerable certainty the width of the following column with the help of overlapping material.

The sequence of textual content that may be reconstructed with near-certainty extends beyond what is preserved in 4Q545, since the text at the bottom of 4Q545 II overlaps very convincingly with the content of the largest of all of the other fragments in the five manuscripts identified with certainty as copies of the Visions of Amram, namely 4Q544 1, or—in the terminology of Puech’s edition, which I will employ from here onwards—4Q544 II.

Deplorably, no complete lines are extant in the overlapping sections: In 4Q544 II, the left side is gone from the top part of the fragment, and a smaller bit of the beginning is missing from line 12 and onwards; the case is similar for 4Q545 II, where only the first parts of lines 9–19 are extant on the major fragment, while fragment 1b preserves a few half-words from the last part of lines 15–19. In spite of this, there is an actual overlap that comprises half a dozen completely preserved words in each line of 4Q544 II 1–4 which match material in 4Q545 II 13–19 both in terms of content and relative placement. And in 4Q544, following the section that overlaps with 4Q545 II, we have a further 10 lines, all preserving from 2–3 complete words in a row and up to nine or ten. We thus have a very considerable amount of text directly continuing that of 4Q545 II, which is presumably parallel to what would have been 4Q545 III–IV.

The assuredly interconnected stretches of text that these fragments make up, together comprise a very large percentage of the total extant amount of material from the five manuscripts. Moreover, this stretch of text reconstructed from the largest fragments of 4Q545 and 4Q544 respectively, has partial or complete overlaps with a dozen fragments or combinations of fragments from the remaining three manuscripts.4

An undisputed column numbering can be presupposed for the part of 4Q545 discussed above. This, again, is due not only to considerations of the physical qualities of the remains, but to the evidence provided by overlapping material in another copy. While the beginnings of the lines in 4Q545 I 1–2 are only partially preserved, with the help of a third manuscript, 4Q543 1a–c, we can restore the missing text and arrive at the wording that begins “A copy of the book ‘The Words of the Vision of Amram, son of Kohath’”.5 Thus, there can hardly be any doubt that this forms the beginning of the text. And therefore we can lay out the larger part of extant material in these five manuscripts a) in sequence, b) with certainty as to their position relative to each other, and c) also relative to the beginning of the work. And this evidently will form the basis of any attempts to say anything about the placement of the remaining fragments not assigned a physical location by these observations.

2 Philology, Material and Otherwise

So far, I have simply restated the results of the reconstruction work presented in the editio princeps.6 This has been taken up and elaborated upon by Robert Duke in his electronic edition found on the “Online Critical Pseudepigrapha” web site,7 as well as in a subsequent monograph.8 An independent reconstruction is found in Klaus Beyer’s work.9

Duke’s and Beyer’s editions, however, are eclectic ones, conflating the individual manuscripts into one text, while Puech, true to DJD format, presents diplomatic editions of the individual manuscripts. But all presuppose—as I have been doing above—the feasibility of supplying missing content in one manuscript from other copies of what can supposedly be described as the same work.

Two or three questions or problems arise, however, when we consider the state of the question summed up so far:

  1. Can more information be wrung out of the lucky coincidence that has supplied us with such a relatively clear view of the placement of the larger fragments of several manuscripts?
  2. On the other hand: Are there perhaps reasons to be cautious and not assume with too great confidence that the remains of one manuscript fit those of another, almost as if they were two copies of an identical jigsaw puzzle of the same picture that just happened to have different pieces missing?
  3. What—if anything—can we say about the placement of the remaining fragments? This last question, which will depend to a large degree not on the physical qualities of the fragments or other hard evidence, but on a consideration of their content, will not be addressed systematically in the present article.

The two first questions are interrelated—or for that matter, they are opposing sides to the same coin, touching as they do upon the question whether we can safely assume that because the manuscripts have clear overlapping passages, therefore they are textual witnesses to exactly the same text, and any information found in one of them can be transplanted to the remaining ones.

The school of manuscript studies referred to as material philology has taught us within the last decade or so, to be careful not to mix up or unwittingly equate three different levels in that complicated mixture of textuality and materiality that is a manuscript. We should be aware of the different levels at which we approach the manuscript, depending on whether we see it as a copy of or textual witness to a literary work, an individual text in and of itself, or an artefact. While the traditional approach to the philology involved in producing a textual edition has tended to focus on manuscripts as sources to be employed in reconstructing the best possible text of an abstractly conceived original work, material philology emphasizes the importance of the individual manuscript both as an artefact and as an individual text and not just as a means of arriving at “the” text of the work in question. In other words, the textual fluidity, which we always knew to be a condition of the transmission of ancient texts, is taken seriously, not just as a problem for textual criticism, but as a quality of the texts.10

While Qumran scholars have in a sense practiced material philology before the term was even coined, routinely producing individual editions of discrete manuscripts, and including meticulous descriptions of their material quality, we have also to a great extent proceeded on the assumption that manuscripts with familiar-looking content were most likely copies of the work that they reminded us of. A systematic application of the principles of material philology is, perhaps not in its infancy, but still growing up.11

3 An Experiment

Having made all these reservations, I intend to make the experiment, while remaining fully conscious of the possible counterarguments, not only to the concrete way it is carried out, but to the endeavour as such: If we assume, on an experimental basis, that the five copies of Visions of Amram are indeed copies of the same work, and proceed to create a single text containing not only all the content available in the overlapping sections, but also all the information that can de deduced about the extent of the missing material, can this “base text” of Visions of Amram be made to fit all the extant manuscripts, or will the project defeat itself by showing that the individual manuscripts are in fact not identical as far as their textual material is concerned?

Proceeding on the assumption that the extant parts of the Amram manuscripts are indeed witnesses to basically the same text, it is relatively simple to produce an electronic version of this “base text,” which we can then pour into the different moulds that are the individual manuscripts with their differing column widths, heights, letter sizes, etc.12

For want of Émile Puech’s talent for taking up the handwriting of an ancient scribe in restoring non-extant passages of text, the ideal solution would be to produce individual Hebrew fonts imitating the script of each Amram manuscript. At this preliminary stage of investigation, I have let it suffice to pick an easily available Qumran-based font imitating a semi-formal Hasmonean script, so as to fit the Amram manuscripts.13 As long as the width of letters relative to each other and the width of spacings relative to letters is reasonably close to that of the manuscripts, this is all we need; the ambition is not to create exact facsimiles of the scrolls, but merely to see how much text will fit into columns, or to put the same thing differently, how many lines and columns in a given manuscript our recreated text will take up.14

An accurate representation of the space that the non-extant parts of the relevant passages would have taken up is essential. The space between extant sections of text therefor needs to be filled up with a suitable number of place-filling dummy ‘words’ made up of hyphens and being approximately equal in length to the average of the extant words. This can be done with considerable accuracy for 4Q545 I, since the endings of most lines, as well as the beginnings of lines 1–7 are clearly preserved. The only source of uncertainty is the possible occurrence of vacats or corrections in unpreserved text.

For column II, the matter is slightly less obvious: Part of the text restored in lines 11–19 in DJD comes from parallel passages (primarily from 4Q544 and 4Q546), but a certain amount is the inspired work of Puech. The column width suggested by the reconstruction in DJD is at least one third in excess of the first column, and one might be tempted to ponder, therefore, if it could reasonably be conceived of as being any narrower.15 The placement of fragment 1b relative to the main fragment 1a, however, seems to be secure, based on the overlap of 4Q544 II 3 as well as 4Q546 2 3, with both 4Q545 1a and 1b in line 17 of the column. Only reconstructed text to the left of fragment 1b comes into consideration, therefore, and seeing that some of this in 4Q545 II 15–16 is paralleled by 4Q546 2 1–2, even if it were possible to propose different text forms for the relevant manuscripts, the column in question could at the very most have been half a dozen letter spaces narrower.

A reconstruction like the following, therefore, seems certain. For the lines where no text is preserved at all, the length of dummy lines should possibly be kept to slightly below the average of the fully extant lines, since we have no way of knowing how often the last part of a line was left blank or a vacat inserted. This allows us to arrive at a version of 4Q545 III that looks like this:

For the large piece of text that makes up 4Q544 II, the matter is slightly less straightforward. The complete text of lines 2–3 (consisting of extant text in the fragment itself plus parallel text from 4Q545 and 4Q546 plus a certain amount of restoration by Puech to fit these together into a coherent text) makes up an average of 121 letter spaces. This is not impossible. Emanuel Tov mentions as an extreme example that 4QJeremiahb may have had one column that was 130 letter spaces wide.16 This is an extreme case, though, and it seems unlikely that Puech would have reconstructed a column this wide unless the material available from fragments presumably containing a parallel text had been available. It is illustrative to notice the amount of text that he finds it necessary or reasonable to posit for the remainder of the column. All of the following lines have sections where neither parallel material nor restored text is offered, although it is not easy to see from the edition how Puech envisioned the original layout of the fragment; the endings of lines 9–13 are reconstructed (the first parts of the lines are extant, except for a very few letters at the beginning of lines 12 and 13), but the Aramaic transcription itself does not clearly indicate to what extent we are to assume an amount of non-extant and non-reconstructable text between the extant beginnings and reconstructed endings of these lines. The triple-dot markings in the French translation of all lines in the bottom half of the fragment do indicate, however, that what is there is not coherent text with no lacunae in it.

To put it another way: Puech reconstructs a column which may seem a bit top-heavy, since it has very long lines at the top, and lines of uncertain length at the bottom. For our experimental purpose, then, the question is, should we really put in enough dummy words to inflate the rest of the column to the width that Puech posits for lines 2–3. The answer seems to be yes: If we mark in 4Q544 what corresponds to the beginnings of 4Q545 II 13, 15, 17 and 19—cf. fig. 2 below—we see that they are almost directly underneath each other, i.e. one line in this column corresponds in length almost exactly to two entire lines of 4Q545 II, if the manuscripts did in fact have the same text.

If the initial lines of the column can be shown almost conclusively to be of this unusual length, then evidently we must assume that this represents the width of the entire column, and that we must fill in dummy text to correspond to this. As mentioned above, though, it makes sense to keep slightly below the average of the fully reconstructed lines; the most excessively long line might for instance have had a few words left out by mistake which were later added above the line.

4 Visions of Amramc (4Q545) III–IV

By means of the overlap between the bottom of 4Q545 II and the top of 4Q544 II described above, we may splice these two rather certainly reconstructed passages of text-and-dummy-markings together and use the resulting “base text” to test the assumptions that can be made about the physical layout of the different copies. Thus, we could posit that columns III–IV of 4Q545 might have looked like this, assuming a column width somewhere between the narrower column I and the necessarily broader column II:

Figure 3
Figure 3

4Q545 III–IV (reconstructed)

This fits Puech’s suggestion that the small fragment 2, which he places—based on its shape similar to fragment 1b—at the bottom of column IV, refers back to the material preserved in 4Q544 2,17 as there would be ample room for the content of this fragment—one of the largest ones without any direct overlap with the main stretch of text reconstructed—in the lower half of 4Q545 IV.18

5 Visions of Amramb (4Q544) I

Going back to 4Q544, since we have a dependable estimate of the extent of the material in its first column (namely everything that went before the preserved top of the extant column II), we can say something about its possible dimensions. While we do know the amount of material, however, we know neither the width nor the height of the column. If what we have of column II reflects almost the entire height of the column (Puech presupposes a column height of 16 lines),19 then the width of column I must have been approximately 75–80% of the width of column II, which still makes it an unusually wide one:

On the one hand, this makes for an unusually wide column, relative to its height. On average, in the typical Qumran scroll, the height would exceed the width, whereas here the opposite is the case.20 On the other hand, while no tangible facts keep us from envisioning a taller and narrower column, this would further increase the contrast between the width of this column and the subsequent, inordinately wide, one. And with an increased column height, which would of necessity apply also to the following columns, would come an increase in the amount of material we would have to posit between 4Q544 1 and 2, thus rendering Puech’s suggestion of locating the description of angels or watchers in 4Q544 2 prior to 4Q545 2 at the bottom of 4Q545 IV increasingly difficult. And, as Puech points out, irregularities in the line rulings seem to allow us to align fragment 2, preserving a bottom margin, with fragment 1, to demonstrate a total height of 16 lines.21

6 Visions of Amrama (4Q543)

Since the beginnings of the first six lines at the top of column I in 4Q543 are preserved, we can say with certainty how wide the first column will be if the manuscript contained the same text found in 4Q545. Furthermore, since fragments 2a and 2b, which contain text parallel to 4Q545 I 14–19, preserve a top margin, we can also estimate the total amount of text of the preceding column and thus compute the approximate column height and say that the scroll must have had 14 lines to a column:

Figure 5
Figure 5

4Q543 III (reconstructed)

The rather unequal length of lines makes it tempting to suggest that perhaps this manuscript had a shorter text in column I line 6 than the parallel passage in 4Q545. And similarly in column II, a quite undulating left margin results from the attempt to make the text actually preserved in 4Q543 2a–b align more or less like it does in the fragment. This may reflect on the inaccuracies inherent in the attempt to represent lost text by dummy markers in the first place, or flaws in my actual estimation of the amount of dummy markers needed to represent the amount of text presumably lost from 4Q545—or, again, it might actually tell us something new about the manuscripts, namely that they very likely did not have exactly the same text. The two first explanations are certainly likely ones, but they do not rule out the possibility that the third one might apply as well.

Proceeding to column V of the same manuscript, where fragment 5 would have belonged, we encounter an even clearer example of the same phenomenon. Fragment 5 has five lines of clearly legible words, corresponding to lines 4–8 in the passage of text assembled by Puech out of fragments 5–9. Puech suggests that the passage made up lines 10–14 of column V. If we attempt to suggest a reconstruction of the relevant context on the basis of the “base text” employed above, something like this is the result (with the approximate outline of fragment 5 drawn in):

In the top lines of the reconstructed column, we have assumed approximately the same column width that our work on column I suggested. This presents no problem, since nothing remains of these lines in 4Q543 itself, and we can suit ourselves in distributing the “base text”. Once we reach the extant fragments of the column, a problem arises. In order to get the text preserved in the five lines of fragment 5 to align under each other so as to correspond to that fragment, we are forced to let every second line of our reconstructed column V extend quite dramatically beyond the left hand margin demarcated by the remaining lines. This extra material represents the dummy text which the dimensions of 4Q544 led us to assume as having been lost from that manuscript. In other words, if our reconstruction of the amount of text originally contained in 4Q544 fragment 1 was correct, this seems to indicate that the relevant part of that manuscript held a much longer (or altogether different) version of the text than does the 4Q543 fragment.

If we leave the experiment aside for a moment and check this observation against Puech’s edition of 4Q544 itself, we see that the text preserved in the five lines of 4Q543 5 overlaps with extant text of 4Q544 at the beginning of lines 13, 14, and 15 (corresponding to 4Q543 5, lines 1, 3, and 5) and a few centimetres into lines 13 and 14 (parallel to lines 2 and 4 of the fragment).22 As we noted above in discussing the amount of dummy text to be used in filling out the bottom part of 4Q544 II in our experiment, there can be no doubt that a certain amount of text must be assumed in addition to the extant beginnings and reconstructed endings to the lines at the bottom of that column.

In the light of the comparison between 4Q543 5 and 4Q544 II, therefore, although it is conceivable that we have slightly overestimated the amount of dummy text needed to fill out the relevant lines of 5Q544 in the experiment carried out in this article, it seems certain that the lines of text in 5Q544 II 11–15 must have held too much material for them to be basically the same passage that 4Q543 5 comes from. Most plausibly 4Q543 had a substantially shorter version of the same text, or—less likely, I presume—this fragment of 4Q543 does not belong to the passage found at the bottom of 4Q544 II at all.

7 Visions of Amrame (4Q547)

One last example: In fragments 1 and 2 of 4Q547 are preserved the endings of one and ten lines of text respectively. The text in the third to seventh lines of fragment 2 (reconstructed by Puech as column III 6–10)23 clearly parallels material in 4Q544 II 7–11, and the one-and-a-half legible word in 4Q547 1 fits a well-preserved passage in 4Q545 II 16 that may be restored with near-certainty as part of 4Q544 II 2 on the basis of the overlap between 4Q545 and 4Q544.

Reconstructing 4Q547 from our experimental base text on this background yields a very neat result, except for line 12 of the column (the approximate placement of what corresponds to the preserved parts of 4Q547 in fragments 1–2 are marked in fig. 7):

This supplies us with yet another example of the fascinating phenomenon that scribes who copy Visions of Amram have a preference for inordinately wide but rather short columns (Puech sees both a top and a bottom margin in the two fragments). But more pertinently, it is clear that for the most part this corroborates Puech’s very precise reconstruction: With the one exception of the second-to-last line, the words preserved in 4Q547 2 do fit a column of the proper width containing the text found in 4Q544 II. The same observation could in fact have been arrived at simply by noticing where the words that correspond to 4Q547 2 3–6 (4Q547 III 6–9) are located in 4Q544 II: They are neatly spaced with approximately 90–95% of a full line between them, corresponding to the column width of the reconstructed column 4Q547 III (the equivalents of 4Q547 III 9–10 are at a distance of exactly one complete line, reflecting possibly a slightly longer text or a vacat in the space between them).

The one exception to this neat correspondence between our reconstructed columns 4Q544 II and 4Q547 III, too, is evident both in the above reconstruction and by means of noticing the distances between the corresponding words in 4Q544: The word בעה (4Q547 III 12) occurs in 4Q544 II 12 less than one-and-a-half lines after the words דילוהי ואחדין which are equivalent to 4Q547 III 10 (the text of 4Q547 III 11 is not extant at all in 4Q544), rather than at the expected distance of approximately 1.9 lines. In other words: 4Q547 III must have had a considerably longer text in line 12 than what is the case in the parallel passage partly preserved and partly reconstructed in 4Q544 II 12. Again, we seem to come up against the fact that parallel passages take us a long way towards reconstructing the individual manuscripts, but when we look more closely, the result also tells us that most likely there were considerable divergences between the individual manuscripts that we find it convenient to regard as copies of the same work.

8 Conclusion

It seems indisputable that for practical purposes of reconstruction, 4Q543–547 can conveniently be regarded as copies of the same work, at least up to a certain point. It would be meaningless to abstain from getting the maximum information out of the evident parallels identified by Puech, and there can be no doubt that the manuscripts are so closely related that one justifiable way of reading them is indeed as textual witnesses to one work. At the same time, exactly the effort to combine the information they give, leads at several points to the conclusion that the text of the manuscripts must occasionally have differed considerably.

Furthermore: While the philosophy of text editing is evidently open to debate, and proponents of material philology would possibly want to apply editorial principles that emphasize the abstract work less and the individual manuscripts more, my considerable respect for Puech’s practical application of the principles that happen to be his, have by no means been diminished by this little exercise.


  • Beyer, Klaus. Die aramäischen Texte vom Toten Meer samt den Inschriften aus Palästina, dem Testament Levis aus der Kairoer Genisa, der Fastenrolle und den alten talmudischen Zitaten. Göttingen: Vandenhoeck & Ruprecht, 1984.

    • Search Google Scholar
    • Export Citation
  • Beyer, Klaus. Die aramäischen Texte vom Toten Meer. Ergänzungsband. Göttingen: Vandenhoeck & Ruprecht, 1994.

  • Beyer, Klaus. Die aramäischen Texte vom Toten Meer. Band 2. Göttingen: Vandenhoeck & Ruprecht, 2004.

  • Cerquiglini, Bernard. Éloge de la variante: Histoire critique de la philologie. Paris: Seuil, 1989.

  • Cerquiglini, Bernard. In Praise of the Variant: A Critical History of Philology. Translated by Betsy Wing. London: Johns Hopkins University Press 1999.

    • Search Google Scholar
    • Export Citation
  • Duke, Robert R., ed. “Visions of Amram.” The Online Critical Pseudepigrapha,

  • Duke, Robert R. The Social Location of the Visions of Amram (4Q543–547). StBibLit 135. New York: Peter Lang, 2010.

  • Lundhaug, Hugo and Liv Ingeborg Lied. “Studying Snapshots: On Manuscript Culture, Textual Fluidity, and New Philology.” Pages 119 in Snapshots of Evolving Traditions: Jewish and Christian Manuscript Culture, Textual Fluidity, and New Philology. TUGAL 175. Edited by Liv Ingeborg Lied and Hugo Lundhaug. Berlin: Walter de Gruyter, 2017.

    • Search Google Scholar
    • Export Citation
  • Monger, Matthew P. “4Q216: Rethinking Jubilees in the First Century BCE.” PhD diss., The Norwegian School of Theology, Religion and Society, 2018.

  • Parry, Donald W., and Emanuel Tov, eds. Parabiblical Texts. DSSR 3. Leiden.: Brill, 2005.

  • Puech, Émile. Qumrân Grotte 4.XXII: Textes araméens, première partie: 4Q529–549. DJD 31. Oxford: Clarendon Press, 2001.

  • Tov, Emanuel. “71. 4QJerb.” Pages 171176 in Qumran Cave 4.X: The Prophets. Edited by Eugene C. Ulrich et al. . DJD 15. Oxford: Clarendon Press, 1997.

    • Search Google Scholar
    • Export Citation
  • Tov, Emanuel. Scribal Practices and Approaches Reflected in the Texts Found in the Judean Desert. STDJ 54. Leiden: Brill, 2004.


I would like, with this article, to pay tribute to the unparalleled scholarship of Émile Puech, who was not able to come to the symposium chronicled in the present volume, but whose work forms an indispensable part of the basis for much that was discussed at the symposium. In the article, I use Puech’s work on one group of texts as a testing ground for asking questions about our methodology in reconstructing texts. It should be evident from what follows, that even if my attempts to suggest improvements should prove successful, the conditions for working on the texts at all, inevitably belong under the well-known metaphor of a dwarf enjoying the privileged view afforded by standing on the shoulders of a giant.


Émile Puech, Qumrân Grotte 4.XXII: Textes araméens, première partie: 4Q529–549, DJD 31 (Oxford: Clarendon Press, 2001), 333–38: The top half of the first column in 4Q545 as well as the bottom half of the second one are extant to a large extent. Column I, lines 11–19 preserve at most a word and a half (mostly less) at the end of each line, and column II, lines 1–8 preserve no text at all, while only a few letters are extant at the beginning of lines 9–10, before continuous text picks up again in line 11. A little additional text, but no coherent sentences, may be reconstructed on the basis of 4Q543 2a–b which overlap convincingly with 4Q545 I 14–II 3.


Puech, DJD 31: plate XIX.


4Q543 1–2 as well as 4Q546 1 overlap with 4Q545 I. Three sets of fragments, 4Q543 3, 4Q546 2 and 4Q547 1–2, have overlaps with both 4Q544 II and 4Q545 II. And finally, 4Q543 4 and 5–9 overlap with the part of 4Q544 II that has no parallel in 4Q545.


Following the translation of Edward Cook in Donald W. Parry and Emanuel Tov, eds., Parabiblical Texts, DSSR 3 (Leiden: Brill, 2005), 413.


Puech, DJD 31: 283–405.


Robert R. Duke, ed., “Visions of Amram,” The Online Critical Pseudepigrapha,


Robert R. Duke, The Social Location of the Visions of Amram (4Q543–547), StBibLit 135 (New York: Peter Lang, 2010), 12–35. Duke presents a composite text which he subdivides into four “chapters”, supplemented by some of the larger unplaced fragments: His chapters 1 and 2 make up the first of the reconstructed passages, chapters 3 and 4 the second. The material basis for the reconstruction is as follows: Duke chapter 1 = 4Q545 1a i 1–12 // 4Q543 1a–c // 4Q546 1. Chapter 2 = 4Q545 1a i 13–19 // 4Q543 2a–b. Chapter 3 = 4Q545 1a–b ii 9–19 // 4Q543 3 + 4 // 4Q544 1 1–9 // 4Q546 2 // 4Q547 1–2 1–9. Chapter 4 = 4Q543 5–9 // 4Q544 1 10–15 // 4Q547 1–2 9–13.


Klaus Beyer, Die aramäischen Texte vom Toten Meer samt den Inschriften aus Palästina, dem Testament Levis aus der Kairoer Genisa, der Fastenrolle und den alten talmudischen Zitaten (Göttingen: Vandenhoeck & Ruprecht, 1984), 210–14; Die aramäischen Texte vom Toten Meer. Ergänzungsband (Göttingen: Vandenhoeck & Ruprecht, 1994), 85–92; Die aramäischen Texte vom Toten Meer. Band 2 (Göttingen: Vandenhoeck & Ruprecht, 2004), 117–25.


Hugo Lundhaug and Liv Ingeborg Lied, “Studying Snapshots: On Manuscript Culture, Textual Fluidity, and New Philology,” in Snapshots of Evolving Traditions: Jewish and Christian Manuscript Culture, Textual Fluidity, and New Philology, ed. Liv Ingeborg Lied and Hugo Lundhaug, TUGAL 175 (Berlin: Walter de Gruyter, 2017), 1–19 (9–10). Cf. the title of the groundbreaking work of the field, Bernard Cerquiglini, Éloge de la variante: Histoire critique de la philologie (Paris: Seuil, 1989); English translation, In Praise of the Variant: A Critical History of Philology, trans. Betsy Wing (London: Johns Hopkins University Press, 1999).


Norwegian scholars have been at the forefront of this. In addition to the work of Liv Ingeborg Lied (cf. the preceding note), a full-blown example from Qumran studies is Matthew P. Monger’s dissertation “4Q216: Rethinking Jubilees in the First Century BCE”, defended at the Norwegian School of Theology, Religion and Society in March 2018.


I speak of “basically” the same text, as there are some minor textual variants visible in the parallel passages: 4Q544 II 2 has עבדתנא שגי over against עבידתנא [ ש]גיאין in 4Q545 II 15; in 4Q543 I 5 we must assume that the text read ושל[ח וקרא whereas 4Q545 I 4–5 has וקרא, but, judging from line length, most likely did not have ושלח; and the beginning of 4Q544 II 10 is read by Puech as בחזוי חזוה and the corresponding ending of 4Q547 III 9 as חזית בחזות.


The font used is ‘Hebrew Square Isaiah’, based on 1QIsaa, created by Yoram Gnat and available at The Open Siddur Project,


The tools being developed by the Scripta Qumranica Electronica project ( will be able to do a task like this with more precision. The present study is merely a tentative experiment.


Emanuel Tov, Scribal Practices and Approaches Reflected in the Texts Found in the Judean Desert, STDJ 54 (Leiden: Brill, 2004), 82–83, remarks that, generally speaking, the individual columns of a sheet of parchment are either equally wide, or “columns that are unusually wide or narrow are generally found at the beginning or end of sheets”.


Tov, Scribal Practices, 83. Elsewhere, though, he cautions that this reconstruction (for it is a reconstruction only) “would be much beyond the length of the known Qumran scrolls,” Emanuel Tov, “71. 4QJerb,” in Qumran Cave 4.X: The Prophets, ed. Eugene C. Ulrich et al., DJD 15 (Oxford: Clarendon Press, 1997), 171–76 (176).


Puech, DJD 31:340–41, cf. plate XIX.


On the other hand, this reconstruction places the content of 4Q544 2 rather close to where the material preserved in 4Q544 II leaves off. This runs counter to the obvious suggestion that in 4Q544 itself, fragment 2 makes up the bottom part of the column following that preserved in fragment 1, which is to say that approximately 10 full lines of text—and possibly very long ones, if the column in question bore any resemblance to the preceding one—divided fragment 2 from fragment 1 (= column II). We could accommodate the reconstruction of 4Q545 III–IV to this observation by having a wider column III, creating more room for the content of 4Q544 2 between that of 5Q544 1 and the bottom margin preserved in 4Q545 2. An exact suggestion for the dimensions involves an equation with two variables: Neither the width of 4Q545 III nor of the column that ends in 4Q544 2 (column III according to Puech’s very likely reconstruction, DJD 31:326–27) are known to us. The larger the width that we posit for the latter, the more material will have to go between the content of 4Q544 II and fragment 2, and the wider we will also have to assume 4Q545 III and/or IV to have been.


Puech, DJD 31:319.


Tov, Scribal Practice, 83, while stressing that rabbinic rules prescribing for Torah scrolls a column width half of the height (Maseket Soferim 2:15) are not adhered to in Qumran, still points to “a positive correlation between the height and width of columns: the higher the column, the wider the lines, and the longer the scroll”.


Puech, DJD 31:319.


Puech, DJD 31:322–23.


Puech, DJD 31:379.