Digital Palimpsests: Mark in Trinity College Cambridge MS. O.9.27

Abstract The O.9.27 manuscript of Trinity College Cambridge is a minuscule manuscript of Hesiod’s Opera et Dies. In a 2001 PhD thesis on Greek palimpsests in Cambridge by Natalie Tchernetska, this manuscript is described to contain two distinct lower scripts, one of which identified as a New Testament text. The author read four lines and a partial fifth of the one-leaf palimpsest that contain Mark 1:44, which is remarkable considering that the washing made the lower script virtually the same colour as the page. This note re-examines the Markan lower script in O.9.27 and offers an account of the use of image processing software for the purpose to uncover more text in a difficult palimpsest, a method useful when MSI is not available.

This aim of this short note is to present a method for digitally dealing with palimpsests when multispectral imaging (MSI) is not available which, as will be seen, will lead not only to reading additional text, but also to the identification of the initial manuscript as a lectionary. The following revisits therefore the palimpsest leaf which contains text from the gospel of Mark and offers a discussion of the use of digital image processing software for reading what otherwise is an almost completely washed off lower script. To that end, I first describe the issues that impede the reading of the lower script, followed by a presentation of the digital enhancement process used, and concluding with the results yielded when applied respectively to natural light digital images and UV digital images. Natalie Tchernetska's apt presentation of the palimpsest is a fitting starting point: The lower script I is rotated 180 degrees relative to the upper script and is faint, despite the fact that it is not fully covered by the upper script: only traces of a text in minuscules and of initials in red are visible (Plate 9a). The ruling text is of type Leroy 20E2; the text is written in two columns. The outlines of these letters are blurred and do not let us examine clearly its type of script and hypothesise its date. Nonetheless, with the help of digital enhancement (Plate 9b), I read the bottom line of the left column and several top lines of the right column on the verso of the flyleaf, which contain a fragment of the New Testament, Mark 1:44. The last words on the bottom line of the left column are ]ὅρα μη/, and the first four top lines of the right column read / δενὶ μηδὲν εἴπῃ ἀλλὰ ὕ / παγε σεαυτὸν δεῖξον τῷ / ἱερεῖ καὶ προσένεγκε / περὶ καθαρισμοῦ σου [. [...] It cannot be determined, however, when the flyleaf was added to the main body of O.9.27.6 This parchment flyleaf stands now as the third flyleaf of the manuscript (the other two are left blank and made of paper) and is followed by the first leaf with Hesiod's Opera et Dies (f. 1r). It is slightly smaller than the rest of the manuscript: it measures 16.5 × 25.5 cm, whereas the rest of codex measures around 18.5 × 27 cm. This too shows that it is a displaced leaf that has been added subsequently to the Hesiod manuscript.7 A significant impediment in reading the Markan text today is the result over time of the ruling of the page. On the one hand, it was not followed closely when copying the text which includes Mark 1:44: the ruling is set in a slight angle compared to the baseline of the lower text (which is placed, as said, on two columns, upside down in the newer binding). On the other hand, the ruling has been applied with perhaps too strong a pressure and, as a result, the general aspect of the page today is far from being that of a flat surface: the ruling now forms a grid of parallel arched waves in the parchment, of a rather high relief. Further complicating the matter, the lines of the grid cut through the middle of every other line of the lower script.
Yet the reading of the Markan text is substantially hampered by the fact that the washing off has made the lower script of virtually the same colour as the parchment, as can be seen in the image below, the bottom of the recto of the flyleaf. Edges of the script becomes slightly more visible under an UV lamp but unfortunately not enough for reading additional text.  Cambridge,98. 7 To put this in current codicological terminology-following Andrist, Canart, and Maniaci, La syntaxe du codex-the added overwritten leaf forms now a new 'circulation unit' (p. 61) with the Hesiod manuscript. Of interest here is the 'production unit' (pp. 59-60) with the Markan text, which was broken from its initial codex, effaced, as well as overwritten, and bound in the new codex, O.9.27 (tentatively, transformations D2 + A3 according to Andrist, Canart, and Maniaci, La syntaxe du codex, 81).

Method and results when using normal light digital images
Natalie Tchernetska's methodological solution for rendering readable the lower scripts of the palimpsests in the thesis is "digital enhancement," which indicates the computerized manipulation of the digital image of a manuscript page. This approach is used throughout the thesis, the latter including a descriptive chapter on the matter.8 Tchernetska lists a number of ways to proceed in using a specialized software-Adobe Photoshop-and rightfully draws attention to the fact that digital enhancement would work even better when one starts from an UV digital photography, instead of a digital image in natural light.
However, none of the Tchernetska's described methods were effective on the O.9.27 palimpsest, since the colour of the lower script is too close to that of the parchment. As seen in the quotation above, digital enhancement was used in the case of this particular manuscript, but the exact method applied in its case is not described.9 In any event, she transcribed the following text from the bottom of the left column (last two words of the last line), respectively from the top of the right column (four lines) on the verso of the flyleaf, from Mark 1:44, which are the areas in which the lower script is the most visible:10 I had the chance to work with the manuscript on several sessions in the Wren Library of the Trinity College in 2010, with subsequent study visits in 2011 and 2014. Unfortunately, at that time it was not possible to produce in Cambridge UV digital images of the palimpsest leaf, or MSI. However, I did receive digital images with which I could work, through the kindness of the Wren Library custodians. In the following I will describe briefly my own solution for manipulating the digital image of the flyleaf-with the help of the Adobe Photoshop software-in order to read further text of the lower script.
The starting point is the fact that there are several possible formats for encoding a digital picture. The one which includes the highest number of possible colours and which is also the most common is the RGB colour space.11 The RGB mode decomposes virtually any visible colour in three channels: Red, Green and Blue, whose initials form the name of this particular mode of image encoding. The three channels are overlapping, and from their combination over each pixel results in the colour we perceive on the screen. And conversely, when scanning an image, the visible colour is decomposed by the scanner's software in these three basic channels.
Since in the initial digital image the lower script has almost the same colour as the parchment itself, I proceeded to verify whether any of the three channels displays a more pronounced contrast between the script and page. I found that while the Red channel has almost no contrast at all, the Green and particularly the Blue channels offer at least some contrast between the script and the colour of the page, as can be seen in the image below. 8 Tchernetska,Greek Palimpsests in Cambridge,[133][134][135][136][137][138][139][140][141][142][143][144] Ibid., 143. 10 Ibid., 98. 11 For a discussion of colour spaces in digital context see Plataniotis and Venetsanopoulos, Color Image Processing and Applications, 1-50. They classify the RGB space among the "physiologically inspired color models, which are based on the three primaries, the three types of cones in the human retina," of which "the Red-Green-Blue (RGB) color space used in computer hardware is the best known example" (p. 3). Consequently, I digitally manipulated in various ways the two relevant channels, taking into account the peculiarities of this manuscript. For instance, since the lower script appears in these channels among the midtones and in a tone which is very close to that of the page, sharpening the overall contrast of the image would have just render the extant traces invisible. Using the Curves adjustment tool instead, however, one can add control points to the curve as in the image to the right in order to keep both highlights and shadows virtually unchanged (and therefore avoiding an overall sharpened contrast) which allow the user to only modify the middle tonal areas. This can result in an increased contrast between the lower script and its background, which is the desired effect, even though it does not result in an easily readable image. Additionally, in order to tone down the visual impact of the upper script, I have changed its colour into a neutral colour. For that, I produced a selection of all darker ink areas and created a layer with a grey filling within that selection, places above the layer containing the image of the manuscript.
Using these features and constantly comparing magnified areas in normal light and various manipulations creates an impression of the lower text in a number of places on the flyleaf. As a result, I was able to read most of the text that Tchernetska read, though departing from her transcription on one occasion. In addition, it was possible to ascertain the fact that the verso has 28 lines on each column and that the recto has a corresponding two-column text. More importantly, this allowed me to partially read further six lines-Mark 2:2-3-as well as several other disparate words and groups of letters in the two columns. Below there is the relevant fragment from the second column of the verso, lines 12-18, as shown on the digitally manipulated Blue channel. Reading more text from Mark in the column opened new possibilities with regard to the question of grasping the initial state or purpose of the leaf, before it was washed off, overwritten and rebound as the third flyleaf in Trinity College Cambridge MS. O.9.27. The fact that this column starts with Mark 1:44 on its first four lines, reading then Mark 2:2-3 in lines 12-18, seemed to suggest a continuous text of Mark or at least a text which contains a larger, continuous, quotation from Mark. It also became apparent that between the 6th and the 12th line of the second column-which separate the two sections which were now transcribed, containing respectively Mark 1:44 and Mark 2:2-3-there is not enough space to accommodate both 1:45 and Mark 2:1.12 Although no such omission is signalled in NA2813 (or in the editions of Tischendorf14 and von Soden15), the apparatus in Swanson's edition shows that a number of continuous-text Markan manuscripts have lectionary markings at Mark 1:44 and 2:1.16 This suggests therefore that the flyleaf, in as much as it omits Mark 1:45, is a lectionary and not a continuous-text manuscript of Mark.17 While this has produced new results, they were still limited in the absence of UV digital images, which would have provided a starting point with a sharper contrast between the ink of the lower script and the colour of the parchment than the digital images produced in normal light. Presumably, the further enhancement of this initial contrast could have resulted in a more legible image.

Using UV digital images
Fortunately, in October 2018 I was provided with UV digital images of the flyleaf of O.9.27 through the kindness of Mr. Sandy Paul, sub-librarian at Trinity College Library Cambridge, causing new impetus for this study. As expected, the initial image indeed contained a contrast similar to that obtained by processing the normal digital image as described above. To begin with, it contained more information and contrast on each of the three colour channels, as can be seen below. Incidentally, the Blue channel proved again to be the most useful. Again, by constantly comparing magnified areas in normal light, UV, and the various 12 While in lines 1-6 there are about 110 letters, there are about 250 letters in NA28 in Mark 1:45-2:1, to be distributed in six lines (7-12). While there is little that can be read in the palimpsest to verify this, it seems very unlikely to have had all this text squeezed in this space.   Digital UV images proved helpful. As a first result of the method described above, it was possible to read in lines 5 and 6 further text from the remainder of 1:44, including its last word-αυτοι[ς-at the end of line, ascertaining that the whole verse was there. More importantly, in line 7 εκ κα(τα) μαρκ(ον) became readable. This is a common abbreviation for a liturgical incipit which introduces a liturgical reading from Mark, whose presence between Mark 1:44 and the beginning of Mark 2 confirms that the leaf comes from a lectionary.18 In Trinity College Cambridge MS. O.9.27 verse 1:44 is certainly included up to αυτοις, and it can also be verified that verse 45 is left out entirely. The introduction to 2:1 is still not legible, but there are traces of decoration in red ink in the margin (noted by Tchernetska) as well as further traces of a spiralled decoration in the same ink as the lower text. However, there is no general sense of what symbol they may have formed before being washed. Riddle, who has studied the lectionary text of Mark, does not include the ending of Mark 1 in his discussion,19 yet Colwell notes that "Mark's Gospel supplies the lections for the Saturdays and the Sundays of Lent."20 Both their collation21 and Gregory's study22 concur in placing the reading of Mark 1:35-44 in the second Saturday of Lent,23 and Mark 2:1-12 in the second Sunday of Lent, preceded by Mark 2:28-3:5 (on the first Saturday of Lent) and John 1:44-52 (first Sunday), and followed by Mark 2:14-17 (third Saturday), and Matt 25:1-13 (third Sunday). Consequently, it was possible to read most of the first nine lines of the two columns on the other side, with passages from Mark 2:5-6 and 2:11-12 respectively.
18 Biblioteca Vallicelliana MS D 63 offers a more accessible parallel: it is a 12 th century lectionary-l. 137-of a similar dimension (18.5 × 23.7, compared to 16.5 × 25.5 of O.9.27), also written in two columns. On the bottom half of folio 12r there is virtually the same situation: the reading from Mark 1:44 ends in α προσταξεν Μωσης εις μαρτυριον αυτοις and does not include verse 45 at all. This is followed by the abbreviated indication, written in red ink, that another reading from Mark follows, and then an incipit introduces Mark 2:1 and goes up to 2:5 on the second column. Digital images of the manuscript are available online at http://www.internetculturale.it/jmms/iccuviewer/iccu.jsp?id=oai%3Awww.internetculturale.sbn.it%2FTeca%3A20%3AN T0000%3ARM0281_Vall_D_63&mode=all&teca=MagTeca+-+ICCU. The Pinakes entry is https://pinakes.irht.cnrs.fr/notices/ cote/56319/. However, the method used in this short note was not successful for reading the flyleaf entirely. In this sense, this a provisional and partial transcription of the manuscript. Most of the text remains unreadable because it is very well washed out and for most areas also covered by the upper script. What I was able to read was enough to ascertain that this is a lectionary, and also that the recto of the flyleaf contains the continuation of the lectionary reading started on the second column of the verso. What follows is the updated transcription of the flyleaf (using Vallicelliana MS D 63 as the base text), with no attempt at a full edition or textual commentary, which will have to await the perusal of multi-spectral imaging (MSI). Until then, the advances possible through this method are establishing this is a lectionary and reading additional text on both sides. It is hoped that this note will succeeds in drawing attention to this curious, misplaced, lectionary leaf.24 Until then, the lectionary text of Mark on the flyleaf has received the Gregory-Aland designation L2484 from the Institut für Neutestamentliche Textforschung in Münster.

Concluding remarks
The advances presented above in reading the text of Trinity College Cambridge MS. O.9.27 do not, unfortunately, overcome one of the initial problems posed by the flyleaf palimpsest: the blurriness of the erased lower script continues to hinder any sensible approximation of the dating, even after the digital manipulation of UV images. It still stands that it was written before the late 13th century, a terminus ante quem given by Tchernetska's dating of the upper script.25 Other than that, it is rather difficult to describe the script, since so little of it can be read, and in a precarious way, even if one can get an impression of it during examination. The script seems to have a squared manner in the writing of the letters, to write και with the κ written separately from αι, and the two angled strokes of the former jotted down separately from the vertical strokes. There is virtually only one ligature (ει).
As to the initial function and state of the Markan text before being washed and overwritten, Natalie Tchernetska suggested a number of possibilities: "It is plausible that the original manuscript contained either a liturgical text of the kind in which scriptural readings are frequent (for example, a prayer book or a lectionary); or some other Christian text where NT fragments might be cited, such as a commentary or a homily; finally, it could have been a modest NT manuscript copied for private use."26 Following the use of digital UV images, we can now better determine its purpose, and show that the palimpsest which has been washed and repurposed as a flyleaf in Trinity College Cambridge MS. O.9.27 was initially, or has been copied from, a lectionary.