Yardeni dated the charred En-Gedi Leviticus scroll (EGLev) to the second half of the first or early second century CE. Paleographic evidence is often ambiguous and can provide only an imprecise basis for dating EGLev. Nevertheless, a series of important typological developments evident in the hand of EGLev suggests a date somewhat later than the Dead Sea Scrolls of the first–second centuries, but clearly earlier than comparanda from the sixth–eighth centuries. The cumulative supporting evidence from the archeological context, bibliographic/voluminological details (wooden roller and metallic ink), format and layout (tall, narrow columns)—each individually indeterminative—also suggests dating EGLev to the period from the third–sixth centuries CE. I argue that EGLev should be dated to the third–fourth centuries CE, with only a small possibility that it could have been written in the second or fifth centuries, which is possibly supported by radiocarbon dating.
In the early 1970s, a team of archaeological excavators led by Dan Barag, Ehud Netzer, and Yosef Porath discovered the charred remains of at least one scroll from the location of the Torah ark in the burned ancient synagogue of En-Gedi.1 One lump of carbonized parchment2 material was held by the Israel Antiquities Authority (IAA) for many years, until Pnina Shor and Yosef Porath arranged for three-dimensional micro-CT scanning, with the help of David Merkel. The scans were then given to Brent Seales and his research team at the Department of Computer Science at the University of Kentucky, and in 2015 the IAA announced that Seales had succeeded in virtually unwrapping the charred remains, revealing two legible columns from the beginning of Leviticus until Lev 2:11. The beginning of Leviticus was preserved on the innermost layers after a blank space for handling the scroll. The full results and composite images were published in 2016 in two articles describing the techniques used to virtually unwrap the scroll and analyzing the contents of the scroll.3
In the Textus article, Ada Yardeni contributes an appendix on the dating of the En-Gedi Leviticus scroll (hereafter EGLev)4 that warrants further attention. According to Yardeni, the closest parallels to the script of this scroll are found in the late first century or early second century CE, and thus she dates EGLev to this period.5 While Yardeni’s consummate skill in analyzing Hebrew scripts is beyond doubt, I will nevertheless argue in this paper that limitations in data and method render Yardeni’s early date for EGLev improbable, and that the scroll should rather be dated to approximately the third or fourth century CE. This later dating of EGLev in some respects increases the significance of the scroll, because it provides a securely placed script sample on a soft writing support from the land of Israel from a period with almost no attestation. A third- or fourth-century date for EGLev is strongly suggested by a combination of evidence from paleographic considerations, the archeological context of its discovery, radiocarbon dating, material features of the scroll, and its format and layout.
2 Paleographic Analysis
The primary evidence that led Yardeni to propose an early date for EGLev was paleographic in nature. In order to properly appraise the role of paleography in dating EGLev, I will first discuss several limitations inherent in the traditional comparative-typological method and then highlight pertinent paleographic features of the script of EGLev that suggest a later date than Yardeni proposes.6
2.1 The Limitations of the Comparative-Typological Method
The traditional method of paleographically dating ancient manuscripts is fundamentally comparative and typological in nature. It is comparative, in that scripts of unknown dates are compared with script samples of known dates in order to determine their closest parallels. The former are then dated in relation to the latter on the supposition that similarity in script implies temporal proximity. This can work reasonably well with a well-populated dataset that is representative of the various styles, developments, and ranges of variation that characterized ancient scripts. But in poor-quality and poor-quantity datasets with few documented and dated samples, highly fragmentary remains, and countless missing links, the comparative method runs into severe limitations. In particular, I would suggest that in such cases there is a strong potential for bias towards well-documented periods and script styles to the detriment of poorly attested periods and script styles. In other words, if paleographers date manuscripts on the basis of preserved comparanda, they may be more likely to date manuscripts close to these comparanda, rather than to periods for which no comparanda can be presented, even if this does not reflect the actual ancient historical situation.
The mid-second to the sixth centuries CE are precisely such a “dark age” in Hebrew paleography, which makes any attempts at paleographic dating within this period perilous at best. Aside from some roughly datable stone inscriptions and mosaics (mostly from the land of Israel) and a few Jewish fragments on soft supports (mostly papyri from Egypt, but also fragments from Dura-Europos), there is very little evidence for the varieties and development of the Hebrew script in this long period.7 To the best of my knowledge, internally dated materials consist only of a single papyrus from Egypt dated to 417 CE (see comparandum G below), more than fifty explicitly dated Jewish Aramaic tombstone inscriptions from Byzantine Zoora south of the Dead Sea ranging from the fourth to the sixth centuries CE (see comparandum H below),8 a few additional inscriptions,9 and a handful of explicitly dated Aramaic incantation bowls ranging from the mid-sixth century to the early seventh century CE.10 There are absolutely no comparative examples dating between the mid-second and the sixth centuries in a formal script on a soft writing support from the same region as EGLev.
To counter this lack, Yardeni identifies two roughly (paleographically) datable comparanda among the earlier Dead Sea Scrolls from the first century CE or the beginning of the second century, which she considers to be the closest extant parallels to EGLev. Because the script of EGLev is closest to those of these early scrolls, Yardeni proposes a similarly early date for EGLev. And yet, even if we grant for the sake of argument that Yardeni’s parallels are the closest extant, that does not then guarantee that EGLev dates to the same period. Because there are few appropriate comparanda from the third and fourth centuries, Yardeni simply cannot exclude the possibility that now-lost third and fourth century scripts were even closer parallels. In this case, the traditional comparative method employed by Yardeni breaks down due to insufficient evidence and potentially creates a bias (in my view unrealistically) towards an early date for EGLev.
The traditional method of paleographic dating is also typological in nature, in that scholars must “connect the dots” between dated and datable documents to identify typological developments that take place over time and in different script styles.11 In lieu of near-perfect parallels, undated manuscripts are often placed in relation to a constructed (and unverifiable) typology pegged at points to a calendar, i.e., where a given script fits within perceived trajectories of broader script developments. These typologies always lack a certain level of resolution, and factors such as long transition periods, gaps in the typology, the idiosyncrasies of individual scribes, conservatism and archaism (particularly in formal hands like that of EGLev),12 parallel trajectories of script development, interference from different script traditions and styles, and a host of unknowable and unquantifiable sociological and material factors mean that it is only possible to suggest approximate ranges of dates for manuscripts dated typologically.13 The formal Hebrew (proto-)square script is particularly notorious for its long-term stability, greatly exasperating the problems, though it too, of course, can be shown to have undergone evolutionary development.14
In best-case scenarios, bold paleographers may even attempt to give date ranges narrower than the working life of an individual scribe. In most cases, however, I would suggest that this is unrealistically precise, and paleographers should allow for wider ranges of possible dates.15 Either way, we must never forget that—given the uncertainties and unpredictability of historical processes—these ranges are in fact probability distributions (similar to the below-mentioned radiocarbon dating results), even if humanities scholars are less adept at recognizing and reporting their error. In other words, though the error is difficult to quantify in reality, a paleographer might suggest a date of 75 CE for a hypothetical manuscript, with an (unverifiable) margin of error of ± 25 years to produce a date range of 50–100 CE. But if we (somewhat more realistically) view this as a confidence interval of one standard deviation, that means that there is a 68.2 % probability that the actual date falls within that range. In order to achieve a higher degree of confidence—say 95.4 %—we would then need to expand the proposed range to 25–125 CE. This level of precision may perhaps be possible for Hebrew scripts in a relatively well-documented period such as the first century CE, but such precision is almost universally recognized to be impossible in the poorly documented period from the second to the sixth centuries CE.16 Since the possible dates of EGLev bridge these two periods—and because of the inherent imprecision in any form of typological dating—we must be prepared to accept broader ranges of possible dates as the result of typological paleographic dating methods than Yardeni allows in her proposed range of approximately fifty years. Thus, not only is Yardeni’s date improbably early, but I would suggest that it is also improbably precise, bringing her into direct conflict with other evidence for dating the scroll. When paleographic dating is understood as making probabilistic statements rather than absolute-looking date ranges, this better highlights the possibility of a productive synthesis between traditional paleographic methods and other indicators of date (especially radiocarbon dating), which is quickly and rightly becoming common practice for dating ancient manuscripts.
2.2 Paleographic Analysis of the En-Gedi Leviticus Scroll
In order to facilitate a paleographical analysis of EGLev, let us propose a number of illustrative writing samples for comparison from the period between the first and the eighth centuries. We will first reconsider the two comparanda proposed by Yardeni, as well as several samples from across this period. These have been selected from the preserved texts based on four criteria: 1) stylistic similarity to EGLev; 2) the confidence and specificity with which they can be dated; 3) similarity of writing material; and 4) geographic spread.
A = Israel Museum, Shrine of the Book, 11QTemplea, Scribe B (late first century BCE or first century CE).
Yardeni’s first comparandum, 11QTemplea (or 11Q19)—the famous Temple Scroll—was discovered in cave 11 in the vicinity of Qumran.17 The bulk of the scroll is written in a hand paleographically attributable to the late first century BCE or first century CE, and it was almost certainly finally deposited in the cave during the First Jewish War against Rome (66–73 CE). In 1990–1991, 11QTemplea was dated in Zürich to a radiocarbon age of 2030 ± 40 BP (calibrated to a 1σ range of 97 BCE–1 CE, according to the 1986 dataset).18 11QTemplea is written in an elegant formal hand with even lines falling neatly below dry ruling lines. Its strokes are generally straight, angular, and of a homogenous thickness. Many letters have decorative adornments, but the serifs are often subtle. The script leans to the left.
B = Israel Antiquities Authority, 5/6ḤevPsalms (second half of first century CE or early second century).
Yardeni’s second comparandum, 5/6ḤevPsalms (or 5/6Ḥev 1b), was discovered in the Cave of Letters in Nahal Ḥever, and it was finally deposited in the cave during the Second Jewish War with Rome (132–135 CE).19 It is paleographically attributable to approximately the second half of the first century CE or early second century. 5/6ḤevPsalms is written in a well-executed, calligraphic formal hand in neat, straight lines hanging just below ruled guidelines. The strokes are mostly straight and of homogenous thickness, and the resulting angular script generally leans to the left. The scribe consistently includes additional decorative strokes and subtle serifs.
C = Israel Antiquities Authority, SdeirGenesis (late first century CE or early second century).
SdeirGenesis (or Sdeir 1) was claimed to have been discovered in an unidentified cave in Wadi Sdeir, where it was apparently finally deposited during the Second Jewish War with Rome (132–135 CE).20 It is paleographically attributable to approximately the late first century CE or early second century. The script is an elegant, formal hand, written in neat, straight lines hanging well below the ruled guidelines. The straight but soft strokes betray some minimal (inconsistent) differentiation between thick horizontal and thin vertical strokes, and the relatively thick strokes give the script a somewhat heavy appearance. The vertical strokes generally lean to the left and often end in a sharp tip. The scribe makes regular use of additional ornamental elements and subtle serifs.
D = Israel Antiquities Authority, Mur 24 (134 CE).
Mur 24 consists of a series of farming contracts on papyrus found in Wadi Murabbaʿat and internally dated to the 20th of Shebaṭ, year 2 of the redemption of Israel by Shimʿon b. Kosiba (134 CE).21 The bulk of Mur 24 (fragments A–I), though a documentary text, was uncharacteristically inscribed in the skilled formal bookhand of a professional scribe, albeit not as neatly and carefully as might be expected for a literary scroll. This rare example of an explicitly dated formal hand from the second century is thus an important comparandum for our purposes.
The text of Mur 24 is written parallel to the fibers in somewhat uneven lines. The letters generally respect a virtual baseline and are upright or leaning slightly to the left. The scribe incorporates ornamental additions and distinct serifs, and the script is clearly legible, though the letters are not consistently well executed. The strokes are typically homogeneous and somewhat straight, but also frequently slightly rounded.
E = Israel Museum, Shrine of the Book?, P. Yadin 44—5/6Ḥev 44 (134 CE).
5/6Ḥev 44 is one of a series of three related documents (5/6Ḥev 44–46) written by the scribe Joseph b. Simon concerning the division of leased land in En-Gedi.22 It was written in En-Gedi on 28 Marḥeshvan in the third year of the Bar Kokhba revolt (134 CE).23 Unlike most documentary texts (but like Mur 24), 5/6Ḥev 44 was written in an elegant, calligraphic, formal (proto-)square script of a professional scribe undoubtedly accustomed to copying literary scrolls.24 As an explicitly dated bookhand from second-century En-Gedi, 5/6Ḥev 44 is an invaluable comparandum for dating EGLev.
The text is written in neat, straight, closely packed lines parallel to the papyrus fibers. The script is small, with a rough sense for a baseline and generally leaning to the left. The letters were carefully executed, paying attention to the ornamental additions prevalent in formal hands of the period. The scribe does not consciously differentiate between the relative thickness of vertical and horizontal strokes. A distinctive feature of his scribal practice is the use of large X’s as space fillers at the ends of lines.
F = Yale University, Beinecke Library Inv. DPg 25 = P. Dura 11 (c. 165?—c. 256 CE) (see Fig. 1 above).
Yale University, Beinecke Library Inv. DPg 25 (P. Dura 11) is a liturgical poem written in Hebrew.25 It was discovered in a fill near the synagogue in Dura-Europos,26 belonging to a Jewish community which may have had connections to the Persian Orient.27 This Roman frontier fortress-city was destroyed by the Sasanians c. 256 CE and never reoccupied, so the manuscript cannot postdate this date.28 If the Jewish presence in Dura-Europos is indeed related to the establishment of the Roman garrison,29 then the poem was likely written down after the Roman conquest of the city from the Parthians c. 165 CE and the construction of the synagogue soon thereafter, but we cannot be certain.
P. Dura 11 is written in a noncalligraphic, semiformal hand, characterized by a lack of precision.30 Its cramped, uneven lines are not guided by ruling lines, and the somewhat large letters vary in size (avg. about 4 mm bilinear height). Its relatively thick but inconsistent strokes vary in thickness from letter to letter, but the writer does not consciously distinguish between the relative thickness of the horizontal or vertical strokes. The script is mostly unadorned, except when decorative elements have become integral parts of the letters. The handwriting has a slight general slant counterclockwise from vertical. Letter formation varies greatly within the small fragments, with several peculiar features (e.g., the almost horizontal left diagonal of mem and the protrusion of the roof of qoph to the right of its right downstroke).31
G = Universität zu Köln, Papyrussammlung Inv. 5853—the Cologne Ketubah (417 CE) (see Fig. 2 below).
This famous Cologne Ketubah was written in Aramaic and Greek in Hebrew script in Antinoopolis, Egypt, and it is internally dated to November 15, 417 CE.32 It is written on papyrus in a fairly neat, elegant, semiformal script, though the lines are not entirely even. The strokes are characteristically thin and frequently curved. While lacking many of the adornments of more formal hands, it frequently has the exaggerated serifs so common in the semiformal scripts of this period.
H = Zoora Aramaic Inscriptions 34 and 36 (466 and 468 CE) (see Figs. 3 and 4 below).
Zoora 34 and 36 are two Jewish Aramaic tombstones from Byzantine Zoora (south of the Dead Sea), which share almost identical iconography and were almost certainly written by the same hand.33 Zoora 34 is an epitaph for a person who died in 466 CE,34 and Zoora 36 for a person who died in 468. In both, the text is painted with red paint on white sandstone tombstones in a professional, calligraphic hand very different from most of the crude hands reflected in the Zoora inscriptions. Average letter height on Zoora 34 is about 2 cm. The texts are written without guidelines in neat, but not perfectly straight lines. The strokes are of even thickness and frequently have some rounding, which softens the harsh angularity of the formal script. Most letters lean slightly to the left. The scribe includes ornamental additions and subtle serifs on many letters. In many respects, Zoora 34 and 36 may be the most similar dated comparanda we have, supporting the possibility of a relatively late dating of EGLev.
I = En-Gedi Synagogue Mosaic Floor (mid- to late fifth century CE).
A series of mosaics from the floor of the same synagogue in which EGLev was discovered serve as a natural point of comparison.35 The older synagogue of stratum III was renovated in the mid- to late fifth century CE, at which time these new mosaics in stratum II were apparently laid.36 These mosaics reflect multiple different, but similar hands, and can be considered together. The medium of inlaid tile decreases the precision of the script, but the inscriptions are clearly intended to reproduce the contemporary semiformal script, sometimes displaying its distinctive serifs. The additional decorative strokes and shading of more formal hands are lacking.
J = Oxford, Sackler Library Papyrology Room, Antinoopolis P. 47 + P. 48 (sixth to eighth century CE?) (see Fig. 5 below).
P. Ant. 47 + 48 are two parchment fragments of a large scroll preserving parts of 1–2 Kings.37 It was discovered during excavations in Antinoopolis, Egypt, a site which also yielded numerous Greco-Roman papyri from the third to the sixth centuries CE (as well as smaller numbers from the second and seventh to eighth centuries). As such, McHardy suggests that the Kings fragments may likewise date from the third to the sixth centuries,38 and Birnbaum argues alternatively for a date in the fifth century39 or sixth century.40 Sirat dates it to the fifth or sixth century,41 and Dukan seems to prefer a date in the sixth or seventh century.42 Olszowy-Schlanger argues for the possibility of a later date for P. Ant. 47 + 48 based on the fact that Coptic materials were discovered at Antinoopolis up to and beyond the Muslim conquest and its paleographic similarity to early scripts from the Cairo Genizah.43 Yardeni and Engel also suggest a later date, perhaps in the seventh or eighth century.44 The dating of the scroll is complicated by a dearth of suitable comparanda at the beginning of the range of possible dates, the reverse situation of what obtains in our attempt to date EGLev.
P. Ant. 47 + 48 is written in a neat, calligraphic formal hand characterized by a high degree of precision and consistency. The lines fall slightly below the ruled guidelines. Its script inclines slightly to the left, and base strokes descend to the left. Most strokes are gently curved, but the letters are generally square in shape. Downstrokes tend to end with a sharp rounded tip. The scribe consciously differentiated between thick horizontal and thin vertical strokes, but the contrast is not always obvious.
K = Staatliche Museen zu Berlin—Ägyptisches Museum und Papyrussammlung, Inv. No. 8492 (seventh or eighth century CE?) (see Fig. 6 below).
P. Berlin 8492 is a papyrus fragment containing Hebrew liturgical poetry, which was found somewhere in the Fayum.45 Given the use of papyrus and the fact that it was found along with similar Arabic papyri dated by Sachau to the eighth century CE, Steinschneider suggests that it can hardly be dated later than the eighth century.46 Chwolson concurs, placing the earliest of the Berlin fragments in the seventh century and the latest in the eighth.47 The Academy of Hebrew Language website entry suggests a date in the eighth century,48 whereas the LDAB/Trismegistos database suggests a more conservative date from the fifth to the ninth centuries.49
P. Berlin 8492 is written in an elegant formal hand in many respects very similar to that of EGLev. The lines are straight and neat, but somewhat cramped. The letters are frequently adorned with additional decorative strokes, but the serifs are more subdued than in many semiformal hands. Horizontal strokes are regularly thicker than vertical strokes, though the differentiation is not consistently observed. The script inclines slightly to the left, and base strokes descend to the left.
L = MS London-Ashkar (seventh or eighth century CE) (see Fig. 7 below).
MS London-Ashkar consists of two recently reunited large fragments of Exodus, one in the possession of the Duke University Library (the Ashkar fragment), and the other the property of Stephan Loewentheil of the 19th Century Rare Book and Photograph Shop in New York (the London fragment, having previously been in the possession of Jews’ College in London). It has been dated on the basis of radiocarbon dating and paleographic analysis to the seventh or eighth century CE.50
MS London-Ashkar is written in a neat formal hand in clean columns and straight lines written below the ruled guidelines. The script leans ever so slightly to the left, and there is little or no differentiation between the thickness of horizontal and vertical strokes. It is characterized by regular use of additional ornamental, diamond-shaped strokes on certain letters, as well as subtle serifs on others.
2.2.2 Comparative Description of EGLev
EGLev (see Figs. 8 and 9 below), is written in a formal, elegant, and well-executed calligraphic (proto-)square hand. There is a general impression of a virtual baseline, though not consistently respected. The lines are straight and neat, with letters hanging directly from roughly evenly spaced ruled dry lines an average of about 4 mm apart (visible in the form of horizontal cracks in the parchment running across the entire preserved part of the first sheet of the scroll, even though burning and digital unrolling of the scroll have created some distortions). The script is quite small, with letters averaging between 1.5–2.0 mm in bilinear height. Even accounting for shrinkage due to carbonization, the small, dense writing is clearly remarkable, being more typical of the early periods than that of later Torah scrolls, with their increasing preference for larger formats and scripts.51
One of the most distinctive features of EGLev is the conscious differentiation between thick horizontal and thin vertical strokes.52 Such a degree of shading is uncommon in the Judean Desert scrolls and in hands from the third to the eighth centuries, and is usually only found in some of the most self-consciously calligraphic of scribal hands (=CJK; ≠ABDEFGHIL; for earlier examples, see 4QSamb, 4QJosha, and 4QPse; for an example roughly contemporary with EGLev, see Oxford, Bodleian Library, Ms. Heb. d.89 [P] i).53 As such, it strongly indicates the formal character of the script of EGLev. The script is oriented with a general left-leaning slant at an inclination of about 20° counter-clockwise from vertical, though the left leg of aleph leans strongly to the right (=ABCDEGHIJKL; ≠F; typical of formal and semiformal hands throughout the early periods). The letters aleph, gimel(?), zayin, ṭet, lamed, nun, ʿayin, tsade(?), and shin all have additional decorative strokes (=BCDEJL; ≈AHK; ≠FGI; typical of formal hands from around the first century CE onwards). Bet, dalet, he, kaph, final mem, samekh, qoph, resh, and taw are adorned with subtle serifs on their upper horizontal strokes (=ABCDEHJKL; ≈FG; ≠I; characteristic of formal hands throughout the early periods, with the exception of taw). The letters bet, kaph, nun, pe, and tsade are unusually tall relative to their width, creating large interior spaces (≈BG; ≠ACDEFHIJKL).
While it is impossible to give an exhaustive paleographic analysis of each individual letter form here (see Figs. 10 and 11 below), in her detailed paleographic analysis of the script of EGLev, Yardeni rightly points out numerous close parallels with her early comparanda. In many respects, EGLev certainly is similar to the latest phases of the formal Hebrew script attested in the Dead Sea Scrolls.54 Most of the letter forms are quite typical of the first- and second- century scripts, and virtually all can find parallels somewhere in the voluminous early materials (in inscriptions, if not scrolls). As discussed above, however, that in and of itself is insufficient to ensure a date for EGLev in the second half of the first to the beginning of the second century CE, given the lack of resolution in our paleographic typologies for the periods in question. There are no prominent features from scripts from the second and first centuries BCE to compel an early dating. Nor are there any features that can be demonstrated to be incompatible with a third- or fourth-century CE dating.
On the other hand, there are significant features in the script of EGLev that suggest a date typologically later than the first- and second-century parallels. Some of these advanced developments were seen already by Yardeni, who nevertheless suggests that EGLev does not postdate the chosen comparanda.55 Viewed in isolation, these features do not necessarily preclude a first- or second-century date, but these late features never occur in combination in any scrolls from the first and second centuries, suggesting that EGLev should be dated somewhat later.
First, the left leg of the he of EGLev intentionally and consistently begins below the roof (=HIK; ≈C; ≠ABDEFGJL; mixed forms in Zoora inscriptions, but the separated form clearly predominates in the fifth and sixth centuries),56 a feature which can occasionally be found in early (usually ossuary) inscriptions and cursives (e.g., Mur 18, an Aramaic loan bill dated to 55 CE), but appears usually only in the form of accidental and exceptional cases in scrolls from the first and second centuries,57 suggesting a date from the third century or later.58
Second, final mem is particularly important for dating EGLev, because the left downstroke frequently (almost always?) starts below the roof,59 creating an opening in the top-left of the letter that is not a feature in scrolls from the first or second century (though it can be found in some inscriptions), but is common in later periods (=GHIK; ≈CJ; ≠ABDEFL; closed form predominant throughout in Zoora, but open becomes more common in the 5th century).60 Furthermore, final mem barely drops below the baseline established by most other letters, whereas earlier forms tend to be larger and drop further down below the baseline (=EFGHIJKL; ≈BCD; ≠A).61
Third, the left leg of taw in EGLev has shifted to the right and downward, such that there is either a clear four-way “+” intersection between the left leg and roof or the left leg starts to descend from a point on the roof to the right of its left tip in a sort of fully developed “T” pattern (=DGIKL; ≈EHJ; ≠ABCF; Zoora evidences a period of transition, with mostly the older form in the fourth century and mostly transitional forms later, with the “T” form becoming more prominent from the fifth century onwards). The transition to this form from the earlier “⊢” form where the roof moves right from below the top of the left leg has hardly begun in scrolls from the first and second centuries,62 but seems to have taken place primarily sometime around the third to the fifth centuries.63 Concurrent with this change, the left end of the roof of taw in EGLev appears already to be at an early stage of developing a subtle serif (=DGKL; ≈EH?; ≠ABCFIJ).64
Other less prominent features may also slightly favor a later date. In many hands from the first and second centuries waw and yod are virtually indistinguishable, but they are clearly distinguished in length (rather than shape) in EGLev, a feature which is characteristic of later hands (=CDFGHIJKL; ≈ABE).65 The kaph of EGLev is normally narrow, but sometimes has a longer roof stroke (e.g., the kaph at i.8.4.2)66 than is characteristic of the first and second centuries (=FGIJL; ≠ABCHK). Final nun is short and relatively straight, which tends to be a late typological development (=BHI; ≈CJL; ≠ADEFGK). ʿayin is large and rounded on the right, whereas examples from the first and second centuries tend to be smaller and more angular (=GHIJKL; ≠ABCDE). Yardeni also notes that nonfinal nun sometimes has a fully developed “roof” stroke (e.g., the nuns at i.2.5.2 and i.8.2.2), which is a relatively late decorative feature.67
Thus, while Yardeni is certainly correct that in many ways the hand of EGLev is very close to those of the first and second centuries,68 these significant typological developments beyond documented bookhands in scrolls from the first and second centuries CE would seem to suggest dating EGLev later than the early second century. Early precursors for some of these features can be found (particularly in inscriptions or cursives),69 but the combination of these late features does not occur in extant scrolls from the first and second centuries.70 Such belated introduction into the calligraphic, formal bookhand of features already evident sporadically in early inscriptions and cursives is not unexpected. This later dating is further supported by many close similarities with Zoora 34 and 36 (comparandum H) of the fifth century, clearly demonstrating the long-term stability of the formal script.
On the other end of the spectrum, EGLev consistently has earlier forms than comparanda from around the 6th–8th centuries. In conjunction with the extensive similarity with late hands of the Dead Sea Scrolls, this undoubtedly means that EGLev cannot be dated paleographically later than perhaps the fourth or fifth century (the period of Zoora 34 and 36). This is not the place to trace all of the later developments in the Hebrew script, but a representative example is in order. The aleph of EGLev, with the relatively straight and simple left leg leaning to the right (=D; ≈ABCEHIJ; ≠FGKL; tends to become more vertical in later periods) and the intersection between the central diagonal and left leg at the top tips of each stroke (=ABCDHI; ≈EF; ≠GJKL; in time the intersection tends to move down the left leg; Zoora inscriptions show both forms, indicating a period of transition), probably indicates a date before the sixth or seventh century at the latest.
In sum, most of the features of the script of EGLev are equally compatible with dates within the range of the first to the fourth centuries CE. A pattern of significant features fit better in the third or fourth century, whereas there are no features that clearly suggest the first or second century over and against the third or fourth century. Significant differences from later comparanda make it unlikely that EGLev was written after the fifth century. Thus, based on examination of the script, I would suggest dating EGLev to around the third or fourth century CE, with only a small probability of it having been written in the second or fifth century.
3 Supporting Arguments
As observed above, there are compelling paleographic reasons to suggest a date for EGLev in the third or fourth century CE, but much of the evidence is ambiguous and imprecise. Indeed, the very nature of paleographic evidence and method often precludes definitive answers, especially at the level of precision necessary for dating EGLev. For this reason, it is crucial that we also consider several independent, nonpaleographic supporting arguments that can provide important additional evidence for dating the scroll.
3.1 Archeological Context
According to archeological evidence, the synagogue at En-Gedi was burned to the ground somewhere between the middle of the sixth century CE and the beginning of the seventh century. A horde of coins was discovered in a house adjacent to the synagogue, the latest of which dates to the early years of Justinian I (reigned 527–565 CE; coins datable up to about 540 CE), which led Barag et al. to conclude that the synagogue was destroyed in religious persecution under Justinian in the middle of the sixth century.71 Another horde consisting of thousands of low-value bronze coins was found in the niche for the Torah ark in the synagogue itself, and an initial screening by Bijovsky suggests that the latest date to the period between 498–538 CE.72 Two later coins from Justin II (565–578 CE) and a gold tremissis of Maurice (582–602 CE)—the latter of which was found in a thick layer of ash from the destruction—were discovered in the village of En-Gedi in undisturbed stratigraphic contexts.73 This fact leads Yizhar Hirschfeld to date the destruction to the end of the sixth or early seventh century during a Saracen raid,74 but Bijovsky considers this evidence insufficient to move the date back that far.75 The destruction of the village and synagogue between the mid-sixth and early seventh centuries necessarily puts an upper limit on how late EGLev could have been copied.
The apparent location of EGLev in the synagogue’s Torah ark suggests that it was still in active use within the community that maintained it.76 If Yardeni’s proposed early date were accepted, that would imply that the scroll was nearly 500 years old when the synagogue was destroyed, which I suggest is highly unlikely.77 On the one hand, in recent years scholars have become increasingly aware that ancient books often had long useful lifespans. The evidence of Roman book collections collected by George Houston suggests an average useful lifespan of around 150–200 years for papyrus bookrolls, though on rare occasion books may have been preserved for 300–500 years.78 Manuscript evidence from Judean Desert find sites in close geographical proximity to En-Gedi suggest similar distributions, as the large majority of scrolls are generally thought to date around 50–150 years prior to their final deposition in their respective caves during the First and Second Jewish Wars against the Romans.79 Rare exceptions from Qumran cave 4 are generally dated about 200–300 years before the destruction of the site. Thus, while it is not in principle impossible that a scroll would survive intact for 500 years, on average it is intrinsically highly improbable that any given scroll in active use by the ancient Jewish community of En-Gedi would be close to 500 years old.80 If dating ancient manuscripts is indeed a matter of weighing probabilities, this fact surely argues strongly against Yardeni’s proposal for an early date for EGLev. A date for EGLev in the third to the fifth centuries CE, however, would fit well with expectations based on the archeological context, since the scroll would have been approximately 50–350 years old when the synagogue was destroyed.
3.2 Radiocarbon Dating
As radiocarbon dating methods and measurements continue to improve, paleographers have begun to appreciate the value of this content-independent measurement of time. Indeed, in many cases, radiocarbon dating may provide reliable ranges of probable dates even more precise than traditional paleographic typologies can reasonably be supposed to allow, especially in cases where paleographic typologies are based on low-quality and low-quantity datasets. As noted above, this is precisely the case with the study of the development of the formal Hebrew (proto-)square script from the second to the sixth centuries CE, exponentially increasing the relative weight that should be given to the radiocarbon dates.
A sample presumed to be from EGLev was radiocarbon-dated in conjunction with the publication of the scroll, and the results are published as supplementary material in the Science Advances article. EGLev was delivered by its excavators to the Israel Antiquities Authority (IAA) in a box (box 198) containing charred material (the intact part of the scroll had not by that point been isolated from the surrounding material for further study). The sample to be dated was taken from the carbonized material surrounding the chunk we now know to be EGLev before the material was transferred to archival boxes. Thus, while the published, intact portion of EGLev was not itself sampled, the IAA is fairly confident that the sample was taken from the same material.81 Nevertheless, given the fact that the sample was not taken from a contiguous portion of the scroll and that the original contents of the box are not entirely clear to me, we must admit some degree of uncertainty as to whether the sampled material was in fact originally from EGLev or another source, such as perhaps another scroll from the Torah ark or perhaps even the wood of the Torah ark itself. Only a further sample from part of the contiguous remains of EGLev (ideally the top margin) is likely to clear up the ambiguity entailed in the sampling.
Elisabetta Boaretto at the Weizmann Institute D-REAMS Radiocarbon Laboratory dated the selected “charcoal” sample using the Accelerator Mass Spectrometry technique within a single standard deviation to 1754 ± 40 BP. Calibrated according to the OxCal v.4.3.2 program, that yields the following probability distribution. There is a 68.2 % probability that the parchment (if indeed it was that) came from an animal that was killed between 235–340 CE. Expanding our sights to include a 95.4 % confidence interval, the probability for the date of the skin is distributed as follows: 140–160 CE (2.0 %), 165–200 CE (4.2 %), and 205–390 CE (89.2 %). In other words, there is a high probability that the sample should be dated somewhere between 235–340 CE (68.2 %), and an even higher probability that it should be dated somewhere more broadly in the third or fourth century (approximately 90 %). There is a small probability that it could be dated to the mid- to late second century (just over 6 %). The combined probability of the sample dating from the period between 50–125 CE as suggested by Yardeni, however, is negligible. Thus, if the sample was indeed from EGLev, radiocarbon dating strongly supports our suspicion that the proposed early date is improbable. Instead, it indicates a probable date in the third or fourth century at 90 % confidence, with the most likely interval from about 235–340 CE.
Of course, this radiocarbon dating has additional complicating factors, which must be taken into account. For example, of the five internally dated documents from the Second Jewish War (dated between 128–135 CE) that were subjected to radiocarbon dating in the 1990–1991 and 1994–1995 series, four yielded good dates (at or near 1σ), but XḤev/Ṣe 8a (internally dated to 135 CE) was dated to 1758 ± 36 BP,82 which calibrates according to the 2013 dataset on the OxCal v.4.3.2 program to 235–335 CE (68.2 %), with a 2σ range covering a period from 140–385 CE. This distribution is strikingly similar to that of EGLev, and yet the actual date of writing unexpectedly falls somewhat outside the 2σ range.83 Nevertheless, XḤev/Ṣe 8a is dated somewhat later than Yardeni’s proposed early date for EGLev, so the results from EGLev would have to be even further removed from the actual date of writing than the anomalous (and as yet unexplained), problematic XḤev/Ṣe 8a results for Yardeni’s dating to hold up. Yet in the archeological reports, treatment records of the IAA, and lab results from the Weitzmann Institute, I see no reason to suspect either contamination or laboratory error in the results for EGLev. They conform well to the expected age of the material based on the archeological context and other circumstantial evidence discussed in this article. EGLev was burned in antiquity, remained buried in situ for well over 1000 years, and was never the object of chemical treatments in the hands of conservators, making it an ideal candidate for radiocarbon dating. There is very little opportunity in the known afterlife of the charred remains for any contamination to have occurred, and the normal δ13C value (–21.2 ‰ PDB) and carbon content of 45.5 % are consistent with a clean parchment sample.84 Any possible contaminants were likely removed in the pretreatment cleaning process, though the details of this process are not reported in the article. Thus, if the sampled material was indeed originally from EGLev, all indications suggest that the reported radiocarbon dates are likely to correspond well to the actual date of writing.85 This range of dates, however, is inconsistent with Yardeni’s early dating of EGLev.
3.3 Bibliographic Analysis
In the study of the development of the Hebrew book it is essential to consider not only the script, but also all the material features that make these textual artifacts “books.” Several such bibliographic/voluminological details of EGLev may be significant for its proper dating. One of the most important is that the scroll was apparently originally rolled around a wooden roller.86 The micro-CT scan shows a well-preserved solid inner core in the most protected parts of the scroll where complete revolutions have been well preserved.87 This core cannot be parchment, but looks very similar to what might be expected from the distorted cellular structure of burned wood.88 Wooden rollers of this type are well documented in Greco-Roman book production,89 but Haran argues that they were adopted relatively late in Palestinian Jewish book production.90 As far as we can tell, such rollers were infrequent in scrolls known from the first and second centuries CE,91 but are well known in later rabbinic literature (e.g., m. Yad. 3.4; b. B. Batra 14a; y. Meg. 1.71d; Sof. 2.5) and medieval scrolls. Most of the Dead Sea Scrolls were not discovered rolled up in situ or with sufficient remains from their beginnings or ends so as to exclude the possibility of their having originally been attached to a wooden roller. In most cases where the evidence is sufficient to make a determination, however, such rollers were lacking. The only confirmed example among the Dead Sea Scrolls is 4Q82 (4QXIIg), where the innermost layers at the end of the scroll are still preserved wadded together around part of a wooden stick.92 11Q11 (11QapPs) was originally thought to have been rolled around a wooden stick, but the conservationist Elena Libman has since shown that the inner core was in fact tightly folded and tied parchment from the end of the scroll.93 11Q4 (11QEzekiel) has also been informally suggested as a possible example of a scroll with a wooden roller, but I have not been able to verify the validity of this suggestion.94 A few sticks have been found in the caves around Qumran, but I am not aware of any indications for their use as rollers for scrolls.95 The scroll being read in the paintings from the synagogue at Dura-Europos lacks wooden rollers.96 Thus, as far as the evidence permits evaluation, it appears that wooden rollers were uncharacteristic of scrolls from the first and second centuries CE (though not entirely unattested) and only became a normal feature somewhat later. Furthermore, even if a few of the Qumran scrolls did have wooden rollers at their ends (as in 4Q82), it is interesting to note that EGLev has a roller preserved at the beginning of the scroll, possibly in agreement with rabbinic prescriptions for adding rollers at both the beginnings and ends only of Torah scrolls (e.g., b. B. Batra 14a; y. Meg. 1.71d). Overall, the presence of such a roller at the beginning of EGLev suggests a date from the third century or later.
Furthermore, Barag claims that EGLev was found next to the remains of “a wooden disc set at the base of a rod on which a scroll was rolled,” as well as “a small seven-branched silver lamp, possibly one of the decorations of the Torah ark curtain or one of the ornaments of the mantle of the Torah scroll.”97 While it is not certain that the disk and decoration were ever attached to EGLev or that either they or the wooden roller were attached during the initial production of the scroll (rather than being added later), the possibility is suggestive. Circular bases and ornaments on Torah scrolls are all unattested in Hebrew scrolls from the first and second centuries CE.
Another factor potentially important for dating EGLev is the composition of its ink, which the editors claim is likely to have contained metals, based on the higher density of the ink in the micro-CT scan in comparison with the surrounding material.98 Iron gall inks began to replace carbon-based inks in late antiquity (perhaps around the fourth century CE?),99 but the precise history of this transition remains unknown. Nevertheless, recent studies have identified earlier mixed inks that are carbon-based, but contain metallic elements, though it remains controversial whether these are accidental contaminants or intentional additives. Some papyri from Herculaneum (before 79 CE) contained quantities of lead,100 and some papyri from Egypt (second century BCE to third century CE) contained copper.101 Some Dead Sea Scrolls likewise may have had corrosive, metallic, mixed inks, though again it is debated whether or not these were intentional additives.102 The elemental compositions observed in Dead Sea Scrolls by Nir-El and Broshi, however, would be unlikely to produce the same degree of visible contrast between ink and substrate evident in the imaging of EGLev.103 Given current knowledge of ink compositions, the probable presence of metal in the ink of EGLev is possibly suggestive of a relatively late typological development, but a date in the first or second century CE cannot be precluded.
3.4 Format and Layout
In a 2016 survey of the column layouts of late Herodian and post-Herodian scrolls from the Judean Desert, Kipp Davis has suggested a diachronic development within high-quality literary scrolls moving from wide to narrow columns.104 According to Davis, there are few (if any) good examples of late or post-Herodian scrolls from Qumran with tall, narrow columns, and most scrolls from Qumran have columns between a 1:1 and 2:1 height-width ratio. In contrast, approximately half of the literary scrolls from sites from the Second Jewish War exhibit high columns of over 30 lines and narrow columns of less than 10 cm in width.105 Tall, narrow columns became standard for quality biblical scrolls in later periods (e.g., comparanda J and L, as well as Oxford, Bodleian Library, Ms. Heb. d.89 [P] i, which would have had about 25–30 letter spaces per line). This profile suggests a gradual transition towards tall, narrow columns that was only beginning in the first and second centuries CE. EGLev is reconstructed with columns of 35 lines at a height of about 15.6 cm (just under half preserved) by its editors, who also calculate that column I averages 31–35 letter spaces per line and column II averages 34–37.106 Column I measures 4.3 cm in width on the ABF master view, and column II measures 5.2 cm, measured from their straight right margins to their notional left margins. The columns of EGLev, therefore, are laid out between a 3:1 and 4:1 height-width ratio, which is certainly in keeping with the trajectory identified by Davis. The original measurements are more difficult to evaluate due to shrinkage from the carbonization of the scroll, but even if we factor in shrinkage of 20–30 %, the columns would still be significantly narrower than 10 cm. Further study is necessary to write a full history of column layouts in the periods in question, but the tall, narrow columns of EGLev appear to reflect a relatively late typological development, and a date in the third or fourth century CE is somewhat more probable than one in the first or second century, though we cannot preclude an earlier date.107 Consideration of the margins, initial blank space, and overall dimensions are less relevant for our purposes.108
Dating ancient manuscripts based on script alone is always a bit of a tenuous endeavor, but this is especially the case in poorly documented periods and styles, which advises caution in paleographically dating EGLev.109 Much of the evidence is ambiguous and imprecise. Nevertheless, a series of important typological developments evident in the hand of EGLev suggests a date slightly later than the Dead Sea Scrolls of the first and second centuries, but clearly earlier than comparanda from the sixth to the eighth centuries. The supporting evidence from archeological context, bibliographic/voluminological details (wooden roller and metallic ink), and format and layout (tall, narrow columns)—each individually indeterminative—reinforces the suggestion that EGLev should probably be dated within the period from the third to the fifth centuries CE. As such, I would suggest that EGLev should be dated to the third or fourth century CE, with only a slight possibility that it could have been written in the second or fifth century.110 If the sample subjected to radiocarbon dating was indeed originally part of EGLev, a date in the third or fourth century would be particularly secure. This argument for a later dating is not based on the absolute exclusion of a first- or second-century date on the basis of any individual piece of evidence, but rather a pattern of script- and nonscript-related features in EGLev that are uncharacteristic of scrolls from the first and second centuries, but appear to fit well in the third and fourth centuries. The cumulative evidence makes for a compelling probabilistic case for such a later dating. If this dating is correct, EGLev is an unprecedented example of a formal Hebrew script on a soft writing support from third- or fourth-century Palestine. And as Engel and Mishor note, “any written material that can be assigned to that ‘silent era’ is of inestimable value.”111, 112