These transcriptions use the following transliteration and tagging schemes. An introduction to the transcription system follows. All of this is lifted directly from my PhD dissertation, which I hope to make available in HTML before too long. If I were doing these transcriptions today, I would use TEI Lite or the full TEI. I intend to convert them to TEI Lite one day. I wrote a suite of Perl programs to process the transcriptions into a form suitable for multivariate analysis. The programs can produce a collation as well. Both the programs and transcriptions are free to anyone interested, provided they are not used commercially. Tim Finney October 26, 2001 Transliteration scheme Letter Uncial Minuscule Number alpha A a 1 beta B b 2 gamma G g 3 delta D d 4 epsilon E e 5 waw (digamma) - - 6 zeta Z z 7 eta H h 8 theta Q q 9 iota I i 10 kappa K k 20 lambda L l 30 mu M m 40 nu N n 50 xi X x 60 omicron O o 70 pi P p 80 koppa - - 90 rho R r 100 sigma S s 200 tau T t 300 upsilon U u 400 phi F f 500 chi C c 600 psi Y y 700 omega W w 800 sampi - - 900 This transliteration scheme differs from the Thesaurus Linguae Graecae Project¹s beta code, in which C = xi and X = chi. Specialised tags [ut]...[/ut] uncertain text: could be one of a handful of letters [rt]...[/rt] reconstructed text: missing, or could be any letter [st]...[/st] superscript text: allows isolation for line length counts [it]...[/it] inserted text: often associated with scribal insertion symbols [kc]...[/kc] KAI compendium [sc]...[/sc] scribal contraction, as listed in Gardthausen [ns]...[/ns] nomen sacrum contraction and superscript line [fn]...[/fn] final nu superscript line [di]...[/di] diaeresis [rb]...[/rb] rough breathing [sb]...[/sb] smooth breathing [c0]...[/c0] correction by original scribe [c1]...[/c1] correction by first corrector [c2]...[/c2] correction by second corrector ... [cx]...[/cx] correction by unidentified corrector [d0]...[/d0] deletion by original scribe [d1]...[/d1] deletion by first corrector [d2]...[/d2] deletion by second corrector ... [dx]...[/dx] deletion by unidentified corrector Using the Collate transcription system What follows is an illustration of the Collate transcription system as applied to folio 28 verso of P46 (Heb 7.20-28). If F. G. Kenyon's edition of The Chester Beatty biblical papyri is available, the plates (1937) and his transcription (1936) can be consulted to clarify the process. Complete transcription The manuscripts are transcribed exactly as found. Spelling and punctuation are included because these particulars contain information which may be pertinent to the history of the text. Use of a font which closely approximates the appearance of uncial script reduces the incidence of transcriptional errors such as typing 'P' in place of rho or 'C' in place of sigma. The Greek font used in this section was especially created to fulfil this condition. Basic text Rather than typing in the entire text of the manuscripts, the text of UBS3 (1983) was used as a basis for transcription. After it had been obtained in electronic form, accents and punctuation were removed, final sigmas were converted to normal sigmas, and the entire text was capitalised. Fig. 1 gives the correspondence between Greek and Roman letters employed, as well as numerical equivalents. Location markers The page under scrutiny is folio 28 verso according to the modern editors, or nu delta (i.e., 54) according to one of the manuscript's correctors. The folio numbering is transcribed as: |f 28v||p [c4]ND[/c4]| Whereas the transcriber is free to use whichever set of location symbols is desired, those used here are based on Collate defaults. Lower case characters within location markers (|...|) specify folio, page, section, column, and line number. The assignment of the number nu delta to the fourth corrector is indicated by enclosure in the matched opening and closing tags [c4] and [/c4]. (Corrector assignments will be discussed subsequently.) The specialised font mentioned earlier prints Greek uncials for upper case and Roman letters for lower case. Transcription files which use this font are easier to read. In conjunction with lower case manuscript labels, it enables a collation program to produce printer-ready output using a single font. Diacritics The first line of the folio is transcribed as: MOSIAS EISIN [di]I[/di]EREIS[c3]'[/c3] OI MEN GAR CWRIS OR= [di]I[/di] indicates diaeresis above the iota, while [c3]'[/c3] transcribes a reading mark inserted by the third corrector. Word and line division Words are separated according to the divisions of the standard text unless the positions of features such as diacritics and punctuation suggest that a manuscript divides them differently. Line divisions in the manuscript are marked with a carriage return. (In the examples given below, line-divisions sometimes occur because of the word-processor's word-wrap facility and do not correspond to line-divisions in the manuscript. The actual carriage returns can only be seen in a processor capable of displaying them.) A concatenation marker (=) precedes the carriage return when a line-division does not coincide with a division between words. This enables the collation program to reconstitute words divided across lines before collation. Block markers The next line introduces a block marker identifying the chapter and verse: KWMOSIAS GEGONOTES[c3]'[/c3] O MET ORKWMOSIAS The inclusion of block markers is important because collation programs use them for navigation. All tags must have matching opening and closing sequences within a block, otherwise collation errors may occur. Where tagged text spans a block marker, a closing sequence must be inserted immediately before the block marker and an opening sequence inserted immediately after. (An example of this is given below in the discussion of reconstructed text.) Scribal contractions The following line introduces a nomen sacrum, [ns]KS[/ns]: DIA TOU LEGONTOS PROS AUTON WMOSEN [ns]KS[/ns] The last two letters of the eighth line have a line drawn above them which is scribal shorthand for a final nu: DIA TO QANATW KWLUESQAI PARAMEN[fn]EI[/fn][c3]'[/c3] At the eleventh line a KAI compendium has been inserted above the line by the first corrector: OQEN [c1][st][kc]K[/kc][/st][/c1] SWZEI EIS TO PANTELES[c3]'[/c3] DUNATAI Working from the inner-most tags outwards, [kc]K[/kc] represents a tailed kappa which is a KAI compendium. The tags [st]...[/st] indicate superscript text, while the [c1]...[/c1] tags show that this is an alteration by the first corrector. As the tags are nested, the question of an order of precedence arises. The rule followed here is that tags which are to be 'regularised' first should bind the most tightly to the text being tagged. (Regularisation refers to the replacement of one item with another before collation. In this case, [kc]K[/kc] might be replaced by KAI.) A corollary to this rule is that tags which need to be retained for later stages of processing, including those relating to correctors and text certainty, should be placed outermost. Corrector tags have a higher precedence than uncertain or reconstructed text tags to enable separation of correctional stages without loss of information. Scribal contractions are widespread, especially in the later manuscripts. Many of the corresponding shorthand symbols are listed in Gardthausen's Griechische Palaeographie (1913). Contractions besides those already mentioned are transcribed by enclosing the symbolised letters in [sc]...[/sc] tags. If a symbol is written above the rest of the text it is also enclosed in superscript tags. This allows it to be ignored if an automated count of line lengths is to be performed. It would be more consistent to treat KAI compendia, nomina sacra and final nu superscript lines as particular instances of scribal contraction. This could be achieved by using nested tags in the Collate system (or tags with attributes in SGML). A KAI compendium would then be represented by [sc][kc]K[/kc][/sc]. The present transcriptions do not use this double tagging method, but they could easily be converted if a more general analysis of scribal contraction was desired. Doubtful and obliterated text The next line introduces the uncertain text tags [ut]...[/ut]: TOUS PROSERCOMENOUS DI AUTOU TW [ns][ut]Q[/ut]W[/ns] These enclose any letters which, upon examination, could possibly be one of a number of letters, say two to five. This category is roughly equivalent to letters with sublinear dots in conventional editions. Practically the whole uncial alphabet is prone to confusion once the writing has deteriorated: A L D S E O Q G H P T F I R U Y K C M N Z X The remnant of the uncertain letter in the example is consistent with only a handful of letters. Even though one would expect to find a theta before an omega with a nomen sacrum superscript line at this place, the fact remains that the information content of the letter itself has decreased to the level of ambiguity. If the transcriber resists the temptation to let context influence decisions concerning certainty, it is reasonable to expect that these transcriptions will be similar to those generated by optical character recognition (OCR) scanners, once they are successfully applied to the task. Two lines further down, the tags [rt]...[/rt] appear: KAI EPREPEN ARCIEREUS OSIOS AKAK[ut]O[/ut][rt]S[/rt] These identify reconstructed text, which consists of letters which are either entirely missing or which are so ambiguous that they could be any letters. This class includes text which conventional manuscript editions enclose in square brackets, thus indicating a lacuna. It also includes letters which are very indistinct, but which may still be marked by sublinear dots in printed editions. This scheme has served the purpose of assigning levels of certainty well. However, it has a deficiency. The problem relates to there being no distinction between obliterated letters for which the base is intact, and letters absent because of lacunae. As a result, the transcriptions do not make clear the exact boundaries of the extant writing material. For this reason, future transcription systems should include yet another category for lacunae. The 'reconstructed text' category should be renamed 'speculative text', and be used only for obliterated letters where the base remains. In my opinion, it is important to retain a separate category for such text. To remove it results in the loss of distinction between letters of which a transcriber is reasonably confident, and those where almost complete uncertainty prevails. Returning to the example, if reconstructed text is ignored, it implies that the word here is AKAKO rather than the word AKAKOS with an indefinite final letter. This could be transcribed as AKAKO_, where the underscore represents obliterated text. A problem with using a single character such as an underscore to represent all reconstructed text is that collation programs can become confused where large portions of a reconstructed block are represented in this way. One possible solution would be to construct a character set which gives corresponding certain, uncertain and reconstructed letters different appearances. The collation program would then be instructed to treat the corresponding letters as equivalent for the purpose of collation, while the program's output would give a visual indication of the text's certainty. A warning is necessary with respect to reconstructed text. Tags required for later stages of processing (such as corrector tags) should not be placed within text that is tagged as reconstructed. This is because a preprocessing step, such as replacing everything within reconstructed text tags by an underscore (_), can eliminate the required tags unintentionally. Experience has shown that it is always possible to rearrange the tagging so that this does not happen. Where tagged text encompasses a block marker, a closing tag should be put at the end of one block and an opening tag placed at the beginning of the next block. To illustrate, the block marker is preceded by a closing tag [/rt] and followed by an opening tag [rt] in the transcription of the last line of folio 27 recto: [rt]TW [ns]QW[/ns][/rt] [rt]KAI KAQ OSON OU CWR[/rt][ut]IS OR[/ut][rt]KW[/rt]= Where only part of a block's text is preserved, the rest should be supplied as reconstructed text to help collation programs stay on track. This requirement may change as collation routines become more sophisticated. Fragmentary manuscripts sometimes contain enough information to narrow the range of possibilities for reconstructed text. The surviving fragments may indicate that a lacuna is more likely to hold one reconstruction than another. For large lacunae, where educated guesses about the most likely text are precluded, a single standard text should always be used to supply the missing text. In this way, variants within reconstructed text will always be due to a conscious choice that one possible reading is more likely than another. It is tempting to use more than one standard text in certain situations. For example, it may be known that a particular manuscript conforms to one standard rather than another, as in the case of manuscripts with a substantially Byzantine text conforming to the Textus Receptus. However, to use both the UBS text and the Textus Receptus results in at least two undesirable outcomes. The first is the result of differing verse division positions which sometimes occur. A collation program which uses verse divisions to mark collation block divisions will interpret two texts with different division positions as being variant even though no variation exists in the texts themselves. The second occurs where different standards are used for reconstructed text regardless of what is likely to fit the gaps. Given two manuscripts with lacunae covering the same extensive section of text, apparent variations will be found where the standard texts happen to differ, but where the actual manuscript texts are completely unknown. Spaces, punctuation, breathings, and accents Among the early papyri, there are sometimes spaces between words which coincide with pauses in sense. An example of the inclusion of an ellipsis (...) to indicate such a space occurs in the seventeenth line. AMARTWLWN... K[ut]A[/ut]I [ut]U[/ut]YHLOTEROS TWN It is difficult to find a satisfying way to define these spaces. For convenience and consistency, the simple criterion that they must exceed a certain dimension may be used, although this can result in authentic spaces being omitted. These transcriptions include three classes of punctuation: upper, medial, and lower. Colons and apostrophes are transcribed as such, while paragraph divisions are marked with a paragraph symbol (¶). Breathing and accent marks are not usually transcribed as they are not within the scope of this study. Their inclusion would entail a great deal of extra work for little return with respect to discovering the early history of the text. Therefore, they are only transcribed if they were written by the earlier scribes or correctors. Where they are transcribed, the present system labels these marks with tags such as [rb]O[/rb] for a rough breathing. As an alternative, it would probably be better to use SGML-type entity references for highly accented texts. They consist of a standard abbreviation for the combination of letter and diacritic inserted between an ampersand and semi-colon. Accordingly, an omicron with a rough breathing would be transcribed as &Orougr; (TEI classical Greek supplemental, 1992). Other combinations and ligatures could also be transcribed by this method. Implied original disposition The last few lines of the folio are transcribed as: EAUTON ANENEGK[ut]A[/ut]S[c3]'[/c3] O NOMO[rt]S GAR KA[/rt]= [rt]QISTHSIN ANQRW[/rt][ut]P[/ut]OUS [di]I[/di][rt]EREIS ECONTAS[/rt] [rt]ASQENEIAN O LOGOS DE THS ORKWMOSIAS[/rt] This varies from Kenyon's transcription (1936, 33): eauton anenegkasæ o nomoªs gar kaqi sthsin arciereis anqrwºpous ªecontas ªasqeneian o logos de ths orkwmosiasº The difference lies mostly in the word order of the reconstructed text. Given that not too large a reconstruction is attempted, a particular original disposition is often implied. After removing characters such as punctuation and spaces, the range of possible reconstructions that are compatible with the manuscript evidence can be narrowed by considering line-lengths, likely locations of line-divisions, and possible textual variations. To demonstrate that this is so, the relevant parts of my reconstruction can be compared with Kenyon's reconstruction: EAUTONANENEGK ASONOMOSGARKA QISTHSINANQRW POUSIEREISECONTAS EAUTONANENEGK ASONOMOSGARKAQI STHSINARCIEREISANQRW POUSECONTAS The spaces correspond to a vertical split in the papyrus. As may be seen, the first reconstruction conforms better to the disposition of the writing on the papyrus. This procedure of rearranging the text with the assistance of a word processor using a font that is similar to the manuscript's writing, contributed to the discovery of a previously unpublished variant in P46 at this location (Finney, 1994). Scribal alteration Through transcribing a number of manuscripts of Hebrews, a system for assigning and tagging scribal corrections has been developed. The heart of the system is a decision table which lists criteria used to discriminate between correctors: Corrector 0 1 2 3 4 5 Ambiguous Style x Size x Pen-width x Shade x Kenyon x Corrector '0' is the original scribe correcting his own work, corrector '1' is the first corrector, and so on up to corrector 'x'. The last classification is for all corrections which cannot be assigned to a specific corrector with confidence. Style relates to the handwriting, size to the dimensions of the letters, pen-width to the width of the stylus, and shade or colour, as the case may be, to ink characteristics. None of these characteristics is an infallible guide to identity. However, when taken together they do give a basis for consistent assignments. The opinions of various editors and commentators can be included, and each criterion may be weighted to reflect its importance or reliability. Where the decision table gives an evenly divided or indeterminate result, the correction can be assigned to the unidentified corrector 'x'. In this hypothetical example, the assignment falls quite strongly to the first corrector. Corrections which involve changes to existing words are transcribed as deletion of the original and replacement with the corrected word, even if only part of the word is affected. The eighth line of folio 28 recto provides an example of this procedure. The first corrector has crossed out the iota in ANQRWPOIS: [d1]ANQRWPOIS[/d1] [c1]ANQRWPOS[/c1] Nested deletion and correction tags are used where more than one corrector has been at work on the same words. Whereas individual letters can be marked as added or deleted, such an approach makes it more difficult for a reader to comprehend the stages of correction. It is necessary to know the number of scribes and correctors of a manuscript and their distinguishing features before a decision table can be constructed. The requisite knowledge is gained by examining numerous alterations. The notes of others who have edited the manuscript can also contribute valuable instruction. Even so, subtle differences between correctors sometimes go unnoticed until a transcription is well advanced, and might call for a complete revision of former assignments. References Aland, Kurt, Matthew Black, Carlo M. Martini, Bruce M. Metzger, and Allen Wikgren (eds.). 1983. The Greek New Testament. United Bible Societies, 3rd corr. ed. New York: United Bible Societies. Finney, T. J. 1994. 'A proposed reconstruction of Hebrews 7.28a in P46'. New Testament studies 40, 472-473. Gardthausen, V. 1913. Griechische Palaeographie. Vol. 2. Die Schrift, Unterschriften und Chronologie im Alterum und im Byzantinischen Mittelalter. 2nd ed. Leipzig: Veit. Repr. 1978. Berlin: Nationales Druckhaus. Kenyon, Frederic G. (ed.). 1936. The Chester Beatty biblical papyri: descriptions and texts of twelve manuscripts on papyrus of the Greek Bible. Fasciculus 3 supplement. Pauline epistles: text. London: Emery Walker. Kenyon, Frederic G. (ed.). 1937. The Chester Beatty biblical papyri: descriptions and texts of twelve manuscripts on papyrus of the Greek Bible. Fasciculus 3 supplement. Pauline epistles: plates. London: Emery Walker. Robinson, Peter M. W. 1994b. Collate: interactive collation of large textual traditions, version 2. Computer program. Oxford: Oxford University Centre for Humanities Computing.