NEWLY DISCOVERED ILLUSTRATED TEXTS OF ARATUS AND ERATOSTHENES WITHIN CODEX CLIMACI RESCRIPTUS

Abstract This article presents texts recovered by post-processing of multispectral images from the fifth- or sixth-century underwriting of the palimpsest Codex Climaci Rescriptus. Texts identified include the Anonymous II Proemium to Aratus’ Phaenomena, parts of Eratosthenes’ Catasterisms, Aratus’ Phaenomena lines 71–4 and 282–99 and previously unknown text, including some of the earliest astronomical measurements to survive in any Greek manuscript. Codex Climaci Rescriptus also contains at least three astronomical drawings. These appear to form part of an illustrated manuscript, with considerable textual value not merely on the basis of its age but also of its readings. The manuscript undertexts show significant overlap with the Φ Edition, postulated as ancestor of the various Latin Aratea.


INTRODUCTION
This article presents an initial edition of Greek astronomical texts associated with Aratus and Eratosthenes which have been found since 2012 in the underwriting of a palimpsest. The underwriting appears to date from the fifth or sixth century A.D. and extends across nine folios of Codex Climaci Rescriptus (hereafter CCR). CCR consists of 146 folios, and takes its name from John Climacus, Abbot of the Convent of St Catherine on Sinai, whose Scala Paradisi and Liber ad Pastorem, translated into Syriac, constitute the overtext. This palimpsest came into the possession of A.S. Lewis between 1895, when one folio was purchased at Cairo, and April 1906, when forty-eight were purchased at Port Tewfik. 1 Another eighty-eight folios were received by Lewis from an unidentified 'Berlin scholar' in October 1905. 2 One folio is in the Mingana Collection in Birmingham, 3 and eight remain in the Convent of St Catherine. 4 Of the 146 folios, 109 (including the Birmingham and St Catherine's folios) have as their undertext Christian Palestinian Aramaic theological material, especially translations of the Old and New Testaments, while 27 contain as their undertext Greek biblical texts. 5 Folio 55 appears blank. This article is about the remaining nine folios (47-54 and 64), which have long been known or suspected to contain Greek undertext, but which defied decipherment before the advent of more recent multispectral imaging techniques.
The researching the multispectral images. At that time the palimpsest was believed to contain only theological texts in Greek and Christian Palestinian Aramaic. Undergraduate students were engaged in this research during university vacations from 2012 to 2017. In July 2012, Klair, then a University of Cambridge undergraduate, discerned in the underwriting of folio 48 recto, column ii lines 4-6, the sequence of words αὐ|τὸν τεθῆναι ὅτε | εἰϲ Νάξον. He identified the text through the Thesaurus Linguae Graecae as part of Scholium in Aratum 73. 6 This discovery led to the identification of further astronomical material by students over the succeeding years. In March and July 2017 the Early Manuscripts Electronic Library (EMEL) and the Lazarus Project undertook new spectral imaging of the manuscript. 7 In Tyndale House in May 2018 a residential week of collaboration between textual scholars and imaging scholars from EMEL, the Lazarus Project at the University of Rochester, New York, 8 and the Rochester Institute of Technology led to further breakthroughs in identifying text and images, since image processors were guided by textual scholars to develop optimal algorithms for recovery of erased ink from particular parts of specific pages. 9 As the last of the ten groups of undertext within CCR to be deciphered, the astronomical texts are known as CCR10, when it is necessary to distinguish them from other groups of underwriting.

CODICOLOGY
The non-biblical Greek content in the undertext of CCR occupies folios 47-54 (which are currently bound as a single quire) and folio 64. Conjugate folios are 47 with 54, 48 with 53, 49 with 52, and 50 with 51. Together these form quire 7 of CCR. Folio 55 (seemingly without underwriting, though probably ruled) along with folio 64 functioned as the outermost leaves or fifth bifolium of quire 8, which otherwise consists of folios 56-63, all of which have Christian Palestinian Aramaic undertext. Folios 55 and 64 are isolated leaves which probably did not form a single bifolium of the undertext. 10 The layout is shown in Diagram 1 below. The central sheet of quire 7, namely folios 50 and 51, contains a continuous Greek text of the anonymous Proemium II to Aratus sections 1-6 in the sequence 51r, 51v, 50r and 50v. Therefore, this sheet must also have been the central sheet of a quire in a codex that was recycled to make CCR, and it is clear that the sheet not only was rotated prior to reuse but was also turned over. The hair and flesh sides were identified through autopsy by P.M. Head: the hair sides are (working from the centre of quire 7) 51r with 50v, 52v with 49r, 53r with 48v, and 54r with 47v, along with 55r and 64v from quire 8.
Folios 49, 50, 51 and 52 were rotated 180°for reuse such that the Greek undertext is upside down in relation to the Syriac overtext.
Moir reported that 'The average size of the sheets is now about 23 or 24 cm. × 18 or 19 cm. or 9¼ × 7½ in., but there is evidence that at least some parts of the MS. were upwards of an inch longer and also broader and that they have been trimmed to suit the present format.' 11 Several folios have lost not only their outer margin but even some of their undertext as a result of this trimming (for instance both 48v and 52r have lost the ends of their hexameters).
Other than Aratus' hexameters, which are laid out in a single column, texts are laid out in two columns of twenty-six lines, most commonly with thirteen or fourteen letters per line. Though the sloping ogival majuscule hand of the undertext permits a wider range of dates, a date from the mid fifth to the mid sixth centuries is not in conflict with any obvious palaeographical data. 16 CONTENTS Nine folios contain or share a bifolium with astronomical material, while a tenth (folio 55) is merely physically associated with these by its current placement in the manuscript. Five pages (47r, 48r, 54r, 54v, 64r) contain several of the Catasterisms that were attributed to Eratosthenes, and a further page (64v) contains the same type of material, albeit previously unknown in Greek. Three pages (48v, 49v, 64r) contain diagrams or drawings. Two folios certainly (48r and 53v) and probably a third (54v) contain astronomical measurements. Four hair sides (47v, 49r, 52v, 53r) as well as the isolated but codicologically associated folio 55 have not yet revealed either identifiable word sequences or indications that diagrams were once present. 17 Folio 48 shows that the manuscript integrates the text of the Phaenomena with Catasterisms and illustrations. The manuscript thus appears to belong to the family of the Φ Edition which Martin posited. 18 The contents in their current order are shown in the following table. Our provisional reconstruction of the quires of the original astronomical manuscript is given in Diagram 2.

TRANSCRIPTIONS
The erasure of the undertext of this part of the codex was generally more thorough than for other parts. 19 Much of the undertext of this palimpsest is not legible, either being obscured by the overtext or, more frequently, because the undertext was erased so thoroughly. Since the transcriptions are based on study of various multispectral images, letters that are legible in one image may be invisible in another which in turn does not show the letters visible in the first image. Underdots have been reserved for letters that are uncertain, substantially obscured or lost; square brackets [ ] are used for breaks in the parchment and angle brackets < > for modern editorial additions. Round brackets () enclose text which has been editorially restored using space calculations, and not in response to visible remains of letters. Text within such brackets must be regarded as particularly uncertain when it covers more than several letters, since within this Round brackets also enclose modern section and line numbers. Where text is recorded as undeciphered, individual letters and words may have been read, but extended sequences have not been read with confidence. Except where specifically noted, all breathings, accents, apostrophes and most spaces between words are editorial, while any punctuation outside of round brackets is from the manuscript-though judgements about the presence of punctuation are often uncertain within a palimpsest. Absence of punctuation in the transcription should not be taken as an indication of its absence in the manuscript. Diaeresis, other than in restored sections, is from the manuscript, and its representation is given priority over accents and breathings. CCR often uses ε for αι-for example 64r λέγοντε (i 7), καλῖτε (ii 4), ὁρῶντε (ii 4-5), λέγετε (ii 9-10) and μιγῆνε (ii 11 and 13-4)-and these and other non-standard spellings are preserved in our transcriptions. Iota adscript is generally used in CCR, but only occurs at the end of words. Capitalization represents enlarged letters in the manuscript, and has otherwise DIAGRAM 2: Proposed Quire Structure of Original Astronomical Codex been avoided in the transcriptions, in order better to highlight the enlarged letters. Opening letters of paragraphs or columns are generally enlarged, though the degree of enlargement varies. Columns mostly begin with ekthesis. In the manuscript Greek letters used as numbers have overlines, and overlines have been editorially restored in some cases of presumed numbers. Discussion has been ordered by bifolia in the presumed order in which their first text occurred in the original manuscript. Select images of the relevant pages of the manuscript are available at motb.me/ codex-climaci-rescriptus and https://tyndalehouse.com/CCRimages.

FOLIOS 51 AND 50: ANONYMOUS II PROEMIUM TO ARATUS
Folios 51 and 50 contain sections 1-6 of the Anonymous Proemium II to Aratus, deciphered in reliance on the edition of Maass, 20 who dated this Proemium to the sixth century A.D. or earlier. 21 Based on the radiocarbon date for folio 52, the sixth century should be taken as the latest possible date for this material. Folios 51 and 50 present the longest sections of continuous text recovered from CCR. A number of readings particularly agree with witness group Λ, or with a significant number of its members: 51r i 18-19 κάτω ἐπί; ii 20 καί; ii 22-4 omission of ἐπεὶ προείρηται μὲν ἔχων; 51v ii 6-7 omission of κατηϲτέριϲται before κρατήρ; ii 14 ἐκτόϲ; ii 21 omission of τε καί; ii 22 omission of ϋάδεϲ; 50r i 18 τούτων ἐξήγηϲιν; i 21-2 περι|ελθών; 50v ii 2 ἀτάκτωϲ; ii 8 presumed omission of μέϲον before ἀγκῶνα on grounds of space.

Folio 51 rectο
Some text from the right-hand side of column ii may have been lost through trimming but, because legibility decreases at the right margin, it has not always been possible to identify precisely where this has occurred. The title ending φαινόμενα ἀράτου, which other words may have preceded, was added in a later, less elegant, hand. Table 3 i The reconstructed line is unusually long, and could be shortened by omitting δέ. ii 2-14 The reconstruction of these lines is tentative. The assignment of text to lines often only illustrates a possibility. ii 26 The final three letters of the column are clearly τοϲ, which we take to be in error. The ekthesis presumably is to mark the following content about the smaller circles beginning with ἐλαχίϲτο<υ>ϲ as a new unit.
Folio 51 verso Table 4 i The spellings ζώδια and ζωδιακόϲ without iota after the omega are consistently used in CCR, which only uses iota adscript at word endings. i 13-14 λύρα is omitted between ἐνγόναϲι and καϲϲιέπεια. i 20 κατηϲτέριϲται or κατεϲτήρικται. i 23 CCR's θηρίον appears haplographic alongside the traditional θηρίον, θυτήριον. ii 15 The reading ἐν rather than ἐπ' agrees with the correction in L II . ii 16-17 CCR exhibits an unparalleled word order: κατηϲτεριϲμένων ἀϲτέρων. This phrase is not followed by eta (= 8) as in Maass's edition.
Greek and Latin stories of the marriage of Dionysus and Ariadne give the location either as Naxos or as Dia, 30 which is either another name for Naxos or the name of an island near to Crete. 31 i 13 The rough breathing is in the manuscript. i 17 φθῖρε, that is, φθεῖραι. ii 1 The first letter of this column seems to be both a theta and a nu, one being a correction of the other, though it is hard to tell which. Presumably, in i 26-ii 1 ϲωθῆναι is intended and some confusion has occurred during the movement between columns. ii 1-2 The spelling λαβιρύνθου is clear. ii 25 The abbreviation μ̊(mu with circlet over it) here and on 53v stands for μοῖραι 'degrees'. 32 The restoration is based on the number suggested by Victor Gysembergh. Though the initial letter is not enlarged, there appears to be ekthesis and the initial letter of ii 1 also appears slightly enlarged. We have therefore taken this as the first line. The word ϲκορπίου was thus spread between 48r and 53v. The number has also been read as ι ̅ β̅ . The number is thus 14½ or 12½, or, if μέϲηϲ is taken as subtractive, 33 13½ or 11½. A possibility is that ι ̅ δ̅ μέϲηϲ is corrupted from ι ̅ <καὶ> δ̅ ʹ μ̊, which allows the addition to work. 34 i 4 here and in i 7 and ii 7 represents the manuscript's symbol, which we take to mean ½. We infer its meaning not based on the shape of analogies in other manuscripts but from its position within sequences of numbers from the greater to the lesser. The number in this line is 6, ½ and ¼, that is, 6¾. i 13 The number δ̅ is also possible. Table 9 i i 14-18 In their extant form these lines are problematic since i 14-15 cannot go with the preceding lines which have dealt with the most westerly star. The translation above is thus of a text which is presumed corrupt. Textual confusion has probably occurred owing to the resemblance of i 14-15 with i 16-17 (δέ is in second position in both, ἀνατολάϲ in the same position in both, and ἔϲχατοϲ shares much visually with ἐχόμενοϲ). We expect here a description of the last star to appear, namely ε CrB, rather than the easternmost star ι CrB. On the analogy of νοτιώτατοϲ δὲ ὁ γ ̅ ἀπὸ τοῦ λάμπρου πρὸϲ ἀνατολὰϲ ἀριθμούμενοϲ in ii 2-6, we may conjecturally restore ἔϲχατοϲ δὲ πρὸϲ ἀνατολὰϲ κεῖται ὁ δ̅ ἐχόμενοϲ ὁ ἐπ' ἀνατολὰϲ τοῦ λάμπρου ἀϲτέροϲ, 35 'In last place towards the East lies the one which is fourth to the East of the bright star.' ii 8 Underneath the text is a scribal flourish, about three lines in height. Sketch by Zaman.

FOLIO 64: NEW MYTHOLOGICAL TEXT AND CATASTERISMS
Based on purely physical considerations it is not possible to determine which of 64r and 64v is to be read first, since it is probably a single leaf, not part of a bifolium. Whereas most CCR bifolios were recycled intact, with the trimming process producing a wide inner margin and a narrow or absent outer margin, folio 64 has narrow margins on both sides of the text intact. It may therefore have been trimmed on both the inner and the outer edge before reuse. However, the content of 64v does not readily follow on from 64r, and therefore 64v should be judged as coming first. Comment i 3-4 ἀγανόν is highly uncertain, as it is generally found in poetry and is awkward as an adjective with an accusative of respect, though it is elsewhere used in the context of mortality in a hunting context. Another possibility is to read αθανον by parablepsis for ἀθάνατον. υἵαν 'Hyas' is a reading suggested by Zingg. i 8

Folio 64 recto: [Cat.] 14 and 23, Taurus and Pleiades
The collocation of Taurus and Pleiades is found in the Latin tradition of Hyginus but not in other Greek manuscripts of Eratosthenes. The wording here shows both significant alignments with and differences from existing traditions, including additional material, for example text with analogies to Anonymous II at i 19 and the longer explanation for the presence of seven stars in ii 3-7. 37 Since this section is already dealing with the positions of the stars in Taurus, we must assume that the mythology of Taurus preceded 64v and also that the treatment of Taurus and its associated constellations (Pleiades and Hyades) continued onto the page following 64r. Thus Taurus and associated material occupied at least four sides. Table 11 i εἶν]α̣ ι ̣ τ ̣ ̣ ῶ̣ ν ̣ ἄ̣ τ ̣λ ̣α̣ ν ̣τ ̣ο ̣ϲ ̣ μερόπη δὲ ϲιϲύφωι θυγ](ατέ)ρ ̣ ω̣ (ν) possible text ϲὺν ̣ θ ̣ν ̣η ̣ τῶι διόπερ i 2 The punctuation appears mistaken. i 6-7 Possibly ϋάδαϲ was corrected to ϋάδεϲ. i 12 Preposition appears to be missing. i 15 This line is read as having atypically few letters, but further text may be present. i 21 It is uncertain whether the article is present. i 22 Letters are lost in a tear in the manuscript. ii 3 The presence of the final supralineal stroke on ἑπτάϲτερο ̅ is uncertain, but this appears more likely than ἑπτάϲτεροϲ. ii 5 The syntax is awkward. ζ̅ may have been omitted between δέ and ἀλλά. ii 11 CCR most probably reads φηϲιν with L, not the plural φαϲιν of E, but it is possible that it has both readings with alpha or eta overlayed on the other as a correction. ii 19 Reading ποϲειδόνι for ποϲειδῶνι. ii 22 CCR supports Heyne's conjecture λύκοϲ against λεῦκοϲ in E. ii 26 CCR supports Kiesslingius's conjecture διόπερ. In the lower margin, set to the left, there appears to be a diagram consisting of lines and small triangles, which, if contemporary with the rest of the underwriting, might be the horns of Taurus. The diagram, however, is surprisingly flattened relative to the other two known diagrams in CCR and this, combined with its position in the margin, may suggest that it is not part of the initial astronomical work. Sketch by Zaman.
FOLIOS 52 AND 49: ARATUS, PHAENOMENA, A DIAGRAM AND UNDECIPHERED TEXT We do not know the distance between these folios in the original codex, but we note that both known sides feature aquatic entities. We tentatively order them 52r, 52v, 49r, 49v, so that the creature we identify as Delphinus would be nearer to the part of Aratus' poem which follows the section on 52r. In our provisional reconstruction (see Diagram 2 above) folio 49 comes after folio 54, but it is treated here in order to keep its treatment next to that of folio 52, with which it is physically connected.
Folio 52 recto: Aratus, Phaen. 282-99 The top and left margins are preserved (in relation to the undertext), but the right-hand side of the folio has been trimmed with the result that the ends of the lines have been lost. In addition, in the upper lines the writing becomes less legible towards the righthand side. The text ends about three quarters of the way down this page and its ending appears to be marked with a paragraphos. Above the top line, there is a fingerprint which we presume to be ancient.

(282)
Reading μεταϲκαίροντα with the direct manuscript tradition. Since the text in the codex is not accented, it does not provide evidence in relation to Martin's division as μετὰ ϲκαίροντα.

(284)
CCR has τάνυται with E in ras S, against τετάνυϲται in M and τετάνυται in P. The reading in CCR requires hiatus at the caesura (cf. 287, though not at the caesura, 292, and 296, which involves an initial digamma). The reading of M requires elision of -αι, which is possible in Aratus (for example 293).

(286)
With ἲϲ τρέπετ' ἠε[λίοιο] CCR supports the earlier position of ἴϲ, as conjectured by Grotius and adopted by Kidd, against the readings transmitted by the manuscripts: ἵνα τρέπετ' ἠελίοιο M* S; ἵνα τρέπετ' ἠελίοιο ἴϲ M 2 ; ἵνα τρέπετ' ἠελίου ἴϲ E; ἵνα τε τρέπετ' ἠελίου ἴϲ A. 10 (291) ἔλθοι with E S, against ἔλθοιϲ M. 11 (292) ἐπιρήϲϲουϲι displays simplification of geminate -ρρ-, but metre shows that -ρρ-is required. 12 (293) ϲυμφέρετ' with the direct manuscript tradition against ἐμφέρετ' in Stobaeus. τότε δὲ with M E, against τότε δὴ S, and τὸ δέ τοι V. 13 (294) μαλκῑόντι with Stobaeus, against μαλκιόωντι of the codices. κακώτατον, again with Stobaeus, against κακώτερον of the codices. 15 (296) κολυμβίϲιν uniquely, and problematically, as an adjective, with the direct manuscript tradition, against the adjective κολυμβάϲιν of Stobaeus. 16 (297) πολλάκιϲ with S, against πολλάκι δ' in M and E. Since the text on the conjugate folio (52r) is also upside down in relation to the Syriac overtext, we know that the bifolium was rotated 180°prior to reuse and thus that the drawing originally represented an upright aquatic creature. When rotated to its original position it is on the left-hand half of the page with its tail rising to the right. The creature faces away from the manuscript's binding with its mouth in the outer margin and preserved intact. No traces of a label or other text have been recovered. The seven or more groups of four dots which show up clearly in the image are not part of the same layer, but are punctuation in red ink accompanying the Syriac upper writing. This may suggest that the aquatic creature, which is only visible after digital processing, is also in red ink. This drawing bears a remarkable resemblance, both in its form and in its orientation, to a drawing labelled δελφῒν on folio 306v of the fourteenth-century Codex Vaticanus Graecus 1087, the earliest manuscript hitherto known of the Fragmenta Vaticana of the Catasterisms. Space above the image is sufficient for the text of Phaen. 316-18 concerning Delphinus, or possibly for 316-21, which would follow Phaen. 282-99 which was identified on 52r, the conjugate folio of this bifolium.  It is noteworthy to find the Catasterism of Capricorn in CCR, because Capricorn is mentioned three times in the passage of the Phaenomena preserved on 52r (lines 284, 286 and 292). These are, of course, not the only references to Capricorn in Aratus, but they are a cluster of references and the first in the Phaenomena. This confirms the close connection between elements of the Phaenomena and the Catasterisms also seen on folio 48. The left margin has been trimmed but without loss of Greek undertext. i 1-2 Reconstruction of dotted letters is particularly uncertain. i 3 There is probably insufficient space for δέ after ἐκείνου, though its absence is not certain. i 10 The spelling κρητηκά for κρητικά seems likely, as the crossbar of the second eta appears visible. i 19-22 Traces of letters are visible and it seems likely that the text here was longer than has otherwise been preserved in the Greek tradition. See, for instance, Hyg. Poet. astr. 2.28. i 26 There is no space between ἔθηκεν and τήν and therefore the corrector has added καί immediately above them. ii 17 The reading οὐρᾶϲ rather than οὐραίου is possible, but fits the space less well. ii 19 κ ̅ β̅ is also possible.