This will be my fifth and final post in this series on stichometry. For the earlier posts, see Part 1, Part 2, Part 3, and Part 4.
One of the things that initially confused me about the stichometric data for the New Testament was the set of evidence for the Pauline letters presented in Bruce Metzger’s classic, The Canon of the New Testament: Its Origin, Development, and Significance (corrected ed., Clarendon, 1989). At page 298, Metzger wrote:
“Scribes of Biblical manuscripts would sometimes indicate the length of each Epistle in terms of number of lines, called stichoi. The statistics are as follows:
To Individuals 1 Timothy 238 stichoi 2 Timothy 182 stichoi Titus 100 stichoi Philemon 44 stichoi”
Metzger cites no source for this data. And if I have understood the evidence of the surviving ancient manuscripts and lists properly, every single one of the numbers reported by Metzger here is incorrect. This strikes me as strange and wholly out of character for Metzger. After making some calculations, I think I see what may have gone wrong. What Metzger reports in this list are not the numbers that “scribes of biblical manuscripts” indicate. Instead, what he seems to be listing are modern estimates based (directly or indirectly) on the work of Charles Graux (1852-1882). In a classic article published in 1878, Graux calculated the number of letters in the editions of several ancient texts, including the Septuagint and the New Testament.1 Metzger (or a predecessor) seems to have divided Graux’s total counts by 36 (the usual number used when a stichos is reckoned as letters rather than syllables) and then rounded either up or down to arrive at these totals. Here are Metzger’s numbers with the calculations from Graux:
Romans 979 stichoi: 35266 letters, divided by 36 ≈ 979.61 1 Corinthians 908 stichoi: 32685 letters, divided by 36 ≈ 907.92 2 Corinthians 607 stichoi: 21851 letters, divided by 36 ≈ 606.97 Galatians 311 stichoi: 11202 letters, divided by 36 ≈ 311.17 Ephesians 331 stichoi: 11932 letters, divided by 36 ≈ 331.44 Philippians 221 stichoi: 7975 letters, divided by 36 ≈ 221.52 Colossians 215 stichoi: 7745 letters, divided by 36 ≈ 215.13 1 Thessalonians 207 stichoi: 7468 letters, divided by 36 ≈ 207.44 2 Thessalonians 111 stichoi: 4011 letters, divided by 36 ≈ 111.42 Hebrews 243 stichoi: 26738 letters, divided by 36 ≈ 742.72 1 Timothy 238 stichoi: 8575 letters, divided by 36 ≈ 238.19 2 Timothy 182 stichoi: 6554 letters, divided by 36 ≈ 182.05 Titus 100 stichoi: 3595 letters, divided by 36 ≈ 99.86 Philemon 44 stichoi: 1567 letters, divided by 36 ≈ 43.53
Metzger’s otherwise strange numbers line up remarkably well with Grauxs calculations, but there are a couple problems with this theory of Metzger’s source. First, Metzger doesn’t cite Graux. Second, to get Metzger’s numbers, it is twice necessary to round down when one would expect to round up (Romans and Philippians). Third, Metzger’s number for Hebrews (243) is impossibly low compared to the calculation from Graux (743).
The third problem is easily handled. Metzger has misprinted 243 for 743. The other two problems are trickier. I’m not sure why the numbers for Romans and Philippians are one digit off, and I don’t know why Metzger doesn’t cite Graux (or some intermediate source) for these numbers.
In any event, the “correct” traditional stichometric totals for the Pauline epistles are those reported by J. Rendel Harris (and reproduced below), although, as we saw in the case of Galatians, the stichometric tallies reported in actual surviving manuscripts can vary substantially.2
It is fitting to close the series on stichometry with portraits of Graux and Harris, who, as relatively young scholars, opened up the field of stichometry of early Christian manuscripts.3
Charles Graux in 1875 and J. Rendel Harris ca. 1885; image sources: Mélanges Graux: Recueil de travaux d’érudition classique dédié à la mémoire de Charles Graux (Paris: Thorin, 1884), frontispiece, and A History of Haverford College for the First Sixty Years of its Existence (Porter & Coates, 1892), plate facing p. 524.
Charles Graux, “Nouvelles recherches sur la stichométrie,” Revue de Philologie de Littérature et d’Histoire Anciennes 2 (1878) 97-143. ↩︎
J. Rendel Harris, Stichometry (London: C.J. Clay and Sons, 1893), 39. ↩︎
Also foundational (though less focused on early Christian material) is Kurt Ohly, Stichometrische Untersuchungen (Leipzig: Harrassowitz, 1928). ↩︎
In my earlier post on the stichometry of Greek manuscripts of Galatians, I was surprised to see that the “traditional” Greek stichometric count for Galatians (293 16-syllable stichoi) matched almost perfectly with the count for the text of the Nestle-Aland (when the latter was adjusted to account for nomina sacra). Earlier Greek manuscripts of Galatians, however, had different counts: Codex Sinaiticus with 312 and P46 with 375 (and the important minuscule 1739 had a count of 392). So, what about the Latin tradition?
I had expected that the Latin counts would be a bit lower just on the basis of Latin lacking definite articles. And a rough count of the Stuttgart Vulgate (3rd edition) for Galatians yielded 4804 syllables (that is, 300 versus of 16 syllables and a last line of 4 syllables), which is indeed a little lower than my rough count of Nestle-Aland 28 (303 stichoi + 11 extra syllables). But again, the subtraction of syllables to account for nomina sacra led to some surprising results (surprising to me, at least). Earlier Latin biblical manuscripts tend to contract a smaller set of words as nomina sacra than do their Greek counterparts, so I only subtracted syllables for this smaller set (forms of Deus, Christus, Iesous, and Dominus). That took away 108 syllables. So, we subtract 108 syllables from the the Vulgate Galatians 4804 syllables: 4804-108=4696. Divide 4696 by 16, and we get: 293.5. This is really quite close to the traditional Byzantine (Greek) count of 293. That seems like a very odd coincidence.
But what do the Latin manuscripts themselves say? Some Latin manuscripts do indeed give this number. Here is St. Gallen, Stiftisbibliothek Cod. Sang. 70, an 8th century codex of Paul’s letters:
St. Gallen, Stiftisbibliothek Cod. Sang. 70 (p. 137), end title of Galatians, argumentum of Ephesians and stichometric count of Galatians; image source: e-codices
Here the stichometric count is added in lighter ink at the bottom of the page after the Argumentum of Ephesians has already commenced: ver(sus) CCXCIII, 293. Other Latin manuscripts give a considerably lower number, such as the 9th century Bible, BnF Latin 1:
Paris, Bibliothèque nationale de France, Latin 1 (fol. 400v), end title of Galatians and stichometric count; image source: Gallica
After the end title: hab(et) vers(us) CCXIII, 213 versus.1 How could the number be so low? I suppose it’s possible that the 293 syllables could be reduced by a count that took into consideration additional nomina sacra contractions and other forms of Latin abbreviation beyond nomina sacra, such as the relatively common dix(it), fec(it), –er(unt), etc. But I doubt this could reduce the count by 80 versus. I think an easier explanation would be an error, CCXIII for CCXCIII. But 213 is not the only other count given in the Latin tradition. The Latin stichometric list in Codex Claromontanus gives a (much) higher number:
Stichometric list in Codex Claromontanus (fol. 468r), showing stichometry for the Pauline letters; image source: Gallica
Here, we get ad galatas ver(sus) CCCL, 350. So, not as high as the counts in P46 (375) and 1739 (392), but considerably higher than 293. I’m not sure how this count of 350 came to be. But there are a lot of strange things about this list.2 For now, I just note that the count for 2 Corinthians seems very low and that it is an odd coincidence that the count for Ephesians is exactly 375 (that is, the count for Galatians in P46). It makes me wonder if these counts simply became jumbled at some point(s) in their history.
Bayerische Staatsbibliothek Clm 6243, a miscellaneous codex of the 8th or 9th century, contains another biblical stichometric list in Latin, and the count for Galatians here is CCCXII, 312 (the same as what is found in Codex Sinaiticus):
Bayerische Staatsbibliothek Clm 6243 (fol. 190r), showing stichometry for the Pauline letters; image source: MDZ
Looking at the Roman numeral, I’m tempted to wonder if this also might be an error for 293: CCCXII for CCXCIII, but this would require both the transposition of the X and C and the loss of an I. Perhaps a better explanation would be to consider this a count based on stichoi of 15 rather than 16 syllables (see below).
At this point, it might be good to summarize the different stichometric counts we have found for Galatians in Greek and Latin witnesses:
293. This count is best attested, and as we have seen, matches up remarkably well with the actual number of 16-syllable stichoi in both Nestle-Aland 28 and the Vulgate.
213. This total may be a mistake due to the similar representation of 213 and 293 both in Roman numerals (CCXIII for CCXCIII) and in Greek numerals (ⲥⲓⲅ for ⲥҁⲅ).
312. This total may reflect a reckoning of stichoi based on 15 rather than 16 syllables (if we divide the Nestle-Aland total syllables by 312: 4684/312≈15.01). It may also be the result of misplacing the traditional count of Ephesians with Galatians.
350. I have no good explanation for this count.
375. I have no good explanation for this count.
392. This total may reflect a reckoning of stichoi based on 12 rather than 16 syllables, an iambic trimeter line rather than a hexameter line (if we divide the Nestle-Aland total syllables by 392: 4684/392≈11.95).3
393. This total is probably a mistake due to the similar representation of 393 and 293 both in Roman numerals (CCCXCIII for CCXCIII) and in Greek numerals (ⲧҁⲅ for ⲥҁⲅ).
So, of the seven reported counts, five have a reasonably good (or at least plausible) explanation. It would be nice to be able to explain the other two counts, especially the 375 stichoi in P46.
Others among the “Tours Bibles” also give this count (see Köln, Erzbischöfliche Diözesan- und Dombibliothek, Cod. 1, fol. 365r and Bibliothèque nationale de France, Latin 3, fol. 375r). ↩︎
See the interesting recent article by Kelsie G. Rodenbiker, “The Claromontanus Stichometry and its Canonical Implications,” JSNT44 (2021) 240-253 https://doi.org/10.1177/0142064X211055647↩︎
The tally of 392 that is found for Galatians in 1739 also appears in GA 436 (11th-12th century, Vat. Gr. 367, fol. 124v); GA 209 (14th century, Venezia Biblioteca Nazionale Marciana, gr. Z. 010, coll. 0394, fol. 146r; GA 205 (15th century, Venezia Biblioteca Nazionale Marciana, gr. Z. 5, coll. 0420, fol. 429r). ↩︎
In an earlier post, I set out the evidence used to establish that for ancient Greek and Latin prose writing, a stichos (or versus) was generally equal to 16 syllables. In a subsequent post, I drew attention to the discussion of the payment to writers based on number of stichoi in Diocletian’s Edict on Maximum Prices. So, stichometric counting served to measure works in a standardized way and as a basis for payment of writers. This seems simple enough. The odd thing is that, at least when it comes to Christian texts in the early period, these “notional” or “standard” line numbers don’t actually appear to be very standardized.
A spot check of some later Greek manuscripts of Galatians does give a regular stichometric count. Here is the end of Galatians in Codex Angelicus (GA 020, assigned to the 9th century):
Rome, Biblioteca Angelica, Manoscritti greci, Ang. gr. 39, folio 134 recto, end title of Galatians with stichometric count; image source: Biblioteca Angelica
The last datum given is the stichometric count: ⲥⲧⲓⲭⲱⲛ ⲥҁⲅ, 293 stichoi (using a cursively formed koppa to represent the number 90). This number, 293, is generally accepted as the “standard” stichometric number for the Greek text of Galatians. The problem is that earlier manuscripts give different numbers. For instance, Codex Sinaiticus:
Codex Sinaiticus, end title of Galatians with stichometric notation; image source: codexsinaiticus.org
Written in lighter ink and smaller letters below the closing title ⲡⲣⲟⲥ ⲅⲁⲗⲁⲧⲁⲥ is the notation ⲥⲧⲓⲭ ⲧⲓⲃ, 312 stichoi.1 This count differs again from the tally provided in the Beatty-Michigan Pauline Epistles codex (P46):
Beatty-Michigan Pauline epistles codex (P46), end of Galatians with stichometric count; image source: Chester Beatty Online Collections
Here the notation (in a hand decidedly different from that of the main text) is ⲥⲧⲓⲭ´ ⲧⲟⲉ, 375 stichoi. This is considerably more that the number of lines given in Codex Sinaiticus and the later manuscripts, so much so that I almost suspect that the number could be an error for ⲧⲓⲉ, 315.
There are a few later Greek manuscripts with high stichometric counts for Galatians, but these are mostly explicable as errors for 293 (for instance, some manuscripts have ⲧҁⲅ for ⲥҁⲅ, so 393).2 It is less easy to explain the stichometric total for Galatians in 1739 (assigned to the 10th century):
Monastery of the Lavra B.64 (GA 1739), fol. 75 recto, end title of Galatians showing stichometric count; image source: Library of Congress
Here we get ⲥⲧⲓⲭ´ ⲧҁⲃ, 392. It’s more difficult to see that as the result of an error of copying ⲥҁⲅ. We would have to posit mistakes in copying two out of three digits. (I should add that it’s not clear to me that the stichometric notes were produced by the copyist of the main text; I’ve never seen this manuscript in person.)
I didn’t have the time to count the syllables in each of these manuscripts to see what the “real” number of stichoi is in each of them. But I did do a rough count of the number of 16-syllable stichoi in the text of Galatians printed in Nestle-Aland 28, and I came up with 304 stichoi (or, to be more precise, 303 stichoi + 11 extra syllables, or 4859 total syllables).This is a little closer to the 312 stichoi reported in Codex Sinaiticus than it is to the Byzantine count of 293. But here’s where it gets interesting: My count of Nestle-Aland included unabbreviated nomina sacra. If we contracted the standard nomina sacra (forms of θεός, κύριος, Ἰησοῦς, Χριστός, υἱός, πνεῦμα, πατήρ, and σταυρός/σταυρόω) and count again, we would need to subtract about 175 from the total number of syllables (I say “about” because the forms of contraction can of course vary a little). So 4859-175=4684. Divide that by 16: 4684/16=292.75. That’s shockingly close to the Byzantine standard of 293.3
It makes me curious to know what the “real” stichometric count for the Byzantine text of Galatians is. And I am also at a loss to explain the count of 375 stichoi reported in P46, if it is not indeed an error as suggested above. And overall, I’m not sure what to make of this variety of stichometric counts for Galatians. In another post, I will take a look at the Latin evidence.
312 is also the stichometric number Codex Sinaiticus gives for Ephesians. But 312 also happens to be the traditional stichometric number for Ephesians. ↩︎
GA 1891 (Jerusalem, Orthodox Patriarchate, Hagios Sabas 107, folio 173r) 10th-11th century, has ⲧҁⲅ, 393. ↩︎
Rendel Harris did this exercise for the whole New Testament using the text of Westcott and Hort, and for Galatians, he also got a stichos count of 304. When he reduced that to account for nomina sacra contractions, he got a slightly higher number than me (296), but he only considered four words as nomina sacra (θεός, κύριος, Ἰησοῦς, and Χριστός). See J. Rendel Harris, Stichometry (London, 1893), 38-41. ↩︎
The auction of several items from the collection of Martin Schøyen took place yesterday in London. The highlight of the sale was the so-called Crosby-Schøyen codex, which sold for just over the high end of the estimated price range at £ 3,065,000. Other items, such as the cover of Nag Hammadi Codex I, seem not have sold.
I do not know who purchased the Crosby-Schøyen codex. According to the BBC, “A spokesperson [for Christie’s] said they could not reveal who bought the book due to client confidentiality.”
The book seems to have been a sound financial investment. When it appeared on the market in 1955, it was sold to the University of Mississippi (together with a substantial part of another codex and loose papyri) for $5,000. When the University of Mississippi sold this codex and the portion of the other codex to the dealer H.P. Kraus in 1981, the reported cost was $250,000. When just our codex alone was sold to Martin Schøyen in 1988, the reported sale price was £ 200,000 (about $350,000 at the time).1 And now the book has sold for £ 3,065,000 (about $3,900,000). So, even accounting for cumulative inflation over time, the cost of the book has soared.
In any event, I do hope the new custodian of the codex will keep it available to scholars for study.
These earlier purchase prices are reported in J.M. Robinson, “The Manuscript’s History and Codicology,” in J.E. Goehring (ed.), The Crosby-Schøyen Codex MS 193 in the Schøyen Collection (Peeters, 1990), xvii-xlvii. ↩︎
In an earlier post, I reviewed the evidence for a stichos or versus being equal (in theory) to 16 prose syllables. The count of stichoi in a work provided a standardized way of describing its length. A stichometric count could also be used as a means of calculating the cost of writing or copying a text. An important piece of evidence for this is the so-called Edict on Maximum Prices of Diocletian issued in the year 301. The complete text of the edict does not survive, but epigraphic remains, some quite substantial, preserve much of it.
Drawing of the Corone fragment of the Edict on Maximum Prices, including the Greek text of the lines on the cost of writing; image source: M.N. Tod, “A New Fragment of the Edictum Diocletiani,” Journal of Hellenic Studies 24 (1904) 195-202.
Both Greek and Latin versions survive for the portion of the edict that is relevant to the discussion of stichometry.
CIL edition of the lines concerning prices of writing from the Stratonikeia inscription of the Edict on Maximum Prices
There are a few lacunae, but between the two versions, the sense is reasonably clear. The text given here is that of Lauffer.1
39 scriptori in sc<ri>ptura optima versus n. centum D̸ XXV 40 sequ[enti]s scripturae bersuum no. centum D̸ XX 41 tabellanioni in scriptura libelli bel tabularum [in ver]sibus no. centum [D̸] X
After a word about payment for the parchment maker, the inscription gives maximum prices to be paid to a scriptor or καλλιγράφος (copyist) for two types of writing, “the best writing” and “second quality writing.” In each case the basis for price is the sum of στίχων ρʹ, versus centum, 100 stichoi. A third type of writing is attributed to a different type of writer. The ἀγοραῖος or tabellio,2 figures who execute the writing used in different kinds of documents (the terms libelli and tabulae cover a variety of documentary texts). Again, the unit for determining pay is 100 stichoi.
So accuracy of stichometric counts would seem to be important not just for knowing the length of texts but also for paying those who copied texts of all kinds. One would thus expect that stichometric counts for a given text would be basically stable, but this doesn’t really seem to be the case.3 I will explore this issue a bit in another post.
Siegfried Lauffer, Diokletians Preisedikt (Berlin: de Gruyter, 1971). ↩︎
Presuming that tabellanioni is synonymous with (or an error for) tabellioni. ↩︎
The statement in the Cheltenham List mentioned in an earlier post already suggested that getting such a count might not be an entirely straightforward matter: “Because the index of verses (indiculum versuum) in Rome is not clearly given, and because in other places too, as a result of greed, they do not preserve it in full, I have gone through the books one by one, counting sixteen syllables per line, and have appended to each book the number of Virgilian hexameters it contains.” Edition and translation of Rouse and McNelis, “North African Literary Activity: A Cyprian Fragment, the Stichometric Lists and a Donatist Compendium,” Revue d’histoire des textes 30 (2000) 189-238: Quoniam indiculum versuum in urbe Roma non ad liquidum sed et alibi avariciae causa non habent integrum per singulos libros computatis syllabis posui numero XVI versum Virgilianum omnibus libris numerum adscribsi. ↩︎
I’ve been thinking a bit about stichometry lately. As some of the primary sources did not seem to be easily accessible online, I’ve decided to write up a couple posts on the topic. We can begin with the opening of the entry for “stichometry” in the Oxford Classical Dictionary (2012):
“stichometry, the modern name for an ancient system of numbering lines in literary texts. In Greek papyri, this numbering takes two forms. (1) Marginal: each hundredth line marked with a letter of the alphabet (A = 100 up to Ω = 2400, then again from A). (2) Final: the sum total of lines in the work (roll) stated at the end, often introduced by ἀριθμός, ‘number,’ and most often in acrophonic numerals. Any individual copy may exhibit both, one, or neither; a few copies show lines checked off in fives, tens, or twenties. In verse, the ‘line’ defines itself. In prose, the numbering assumes a notional or standard line (the actual lines would differ in length from copy to copy): apparently15-16 syllables.”
At present, I am mostly interested in stichometric counts of prose texts. The basis for this standard length of a prose line is an incidental remark in Galen (On the Doctrines of Hippocrates and Plato 8.1.22-25). While engaged in a lively medical dispute, Galen repeatedly apologizes for writing so much when the matter at hand is actually, in his view, quite simple. In the course of one of these apologies, he provides some interesting equivalences. I provide the reference here in full in the translation of Phillip De Lacy.1
“Thus the true account is as short as I shall demonstrate to you; it reaches its conclusion in a few syllables (δι᾽ ὀλίγων συλλαβῶν), as follows: ‘Where the beginning of the nerves is, there is the governing part. The beginning of the nerves is in the brain. Therefore, the governing part is there.’ This one argument has thirty-nine syllables (ἐννέα καὶ τριάκοντα συλλαβῶν), equivalent to two and one-half hexameters (δυοῖν καὶ ἡμίσεος ἐπῶν ἑξαμέτρων). A second argument is in all five lines long (πέντε τῶν πάντων ἐπῶν): ‘Where the affections of the soul more visibly move the parts of the body, there the affective part of the soul is. The heart is observed to undergo a great change of motion in anger and fear. Therefore, the affective part of the soul is in the heart.’ If you thus join these two arguments together, the combined total will be no more than eight hexameter lines (ἐπῶν ἑξαμέτρων ὀκτώ). Who, then, is to blame for my having written five books dealing with these matters that could have been scientifically demonstrated in eight heroic lines (διὰ ὀκτὼ στίχων ἡρωϊκῶν)?”2
On Galen’s reckoning, then, two and a half hexameters lines is 39 syllables of prose. If we do the math: 39/2.5=15.6 At the same time, eight ἡρωϊκοὶ στίχοι (or ἑξάμετρα ἔπη) are greater than 122 syllables. If we again do the math: 122/8=15.25
So, it seems then that an epos or a stichos in prose writing is 16 syllables. This finds support in (later) Latin tradition, in which a versus was also equivalent to 16 prose syllables. A Latin Christian sitchometric list generally called the Cheltenham List is preserved in two medieval manuscripts.3 The maker of the list appended the following note (given here in the text and translation of Rouse and McNelis):4
“Because the index of verses (indiculum versuum) in Rome is not clearly given, and because in other places too, as a result of greed, they do not preserve it in full, I have gone through the books one by one, counting sixteen syllables per line, and have appended to each book the number of Virgilian hexameters it contains.”5
The point of stichometric counts (or at least one of the points) seems to be to provide a standard way of measuring the length of a work. But as the statement above from the Cheltenham List indicates, it could be challenging to get an accurate stichometric count for a book. And in fact surviving lists and stichometric notations for biblical books often disagree with each other, sometimes significantly. But that will be the topic of another post.
St. Gallen Stiftsbibliothek Cod. Sang. 133, p. 490; image source: e-codices
Phillip De Lacy, Galen: On the Doctrines of Hippocrates and Plato. Second Part: Books VI-IX, 2nd ed. (Berlin: Akademie Verlag, 2005). ↩︎
The list itself is usually assigned to the second half of the fourth century, but the manuscripts that preserve them are later Biblioteca Nazionale Centrale di Roma, Vitt.Em.1325 (tenth or early eleventh century) and St. Gallen Stiftsbibliothek Cod. Sang. 133 (late eighth or early ninth century). ↩︎
Richard Rouse and Charles McNelis, “North African Literary Activity: A Cyprian Fragment, the Stichometric Lists and a Donatist Compendium,” Revue d’histoire des textes 30 (2000) 189-238. It’s not totally clear whether this statement goes with the material that precedes it (a list of the stichoi of biblical books) or the material that follows it (a list of the stichoi of the works of Cyprian). ↩︎
Rouse and McNelis’s edition of the Latin: Quoniam indiculum versuum in urbe Roma non ad liquidum sed et alibi avariciae causa non habent integrum per singulos libros computatis syllabis posui numero XVI versum Virgilianum omnibus libris numerum adscribsi. ↩︎
After being down for several days due to a cyber attack, the Christie’s website appears to be back up and running. A more detailed description of the items being auctioned from the Schøyen Collection is now available. Among them is the leather cover of Nag Hammadi Codex I along with the separately framed waste papyrus extracted from the cover.
One observation about the catalog entry for this lot: The buyer will also get the bowl that James Robinson often claimed was used to seal the jar in which the Nag Hammadi codices were said to have been found. I want to point out that this bowl has a provenance history that differs from that of the leather cover (and cartonnage) of Codex I. While the quires and leather cover of Codex I took different paths to their current repositories (the quires to Cairo and the cover eventually to the Schøyen Collection), the bowl had a more direct path: an Egyptian family to James Robinson (in 1976) and then to the Schøyen Collection in 1994.
Now, was this in fact “the” bowl used to seal the jar with the codices? For reasons I have outlined in detail elsewhere, I regard the find story or–more accurately–the multiple, conflicting find stories of the Nag Hammadi codices as unreliable. Nevertheless, it’s worth teasing out the tale of the bowl a bit. To the best of my knowledge, Robinson’s fullest account of the bowl is in his book, The Nag Hammadi Story (2014). While summarizing his story of the discovery of the codices, Robinson gave a detailed description of the jar in which the books were said to have been found (a jar that had allegedly been smashed and that Robinson had never seen). He eventually moved on to describe the way this jar had allegedly been sealed and what happened to the bowl used to seal the jar:
“The jar had been closed by laying into its mouth a pottery bowl, which Khalīfah took with him to al-Qaṣr. There he was employed as a servant and camel driver for a Copt, Sami ʿAbd al-Malāk, who took the bowl as a talisman that would bring a blessing. Through the mediation of his wife Umm Nadya, a relative Salib ʿAbd al-Masīḥ, for a modest consideration, made it available for study.”1
In other words, Robinson bought the bowl and took it back to Claremont. What is interesting about the story is that the connection of the bowl to the alleged discovery of the books is the figure of Khalīfah, that is, the brother of Muhammad ʿAli al-Samman, the man most closely associated with the discovery of the codices, whose story (or rather, stories) Robinson spent years investigating. One part of Muhammad ʿAli’s story was a denial that Khalīfah was present at the time of the discovery of the codices.2 Robinson did not believe this part of Muhammad ʿAli’s story.
James Robinson’s Egyptian bowl, now MS1804/1 (3) in the Schøyen Collection; image source: Christie’s
Another account of Robinson’s purchase of the bowl is provided in James Goehring’s publication of a similar bowl found during excavations of a church at Pbow. Goehring says the following about the bowl Robinson bought:
“Robinson notes that Khalifah took it [the bowl] to the home of the Coptic family of Sami Abd al-Malak and his wife Umm Nadia. When Robinson located the bowl, it was Salib Abd al-Masih, Sami’s nephew, who sold it to him. While Robinson did not know for sure the whereabouts of Sami at that time, one may assume, since Umm Nadia was still present and Salib acted as the one with authority to sell the bowl, that Sami had passed away. Robinson concurred with this assessment.”3
I think it’s important to emphasize that this bowl took a path in the antiquities trade that is different from that of the cover of Nag Hammadi Codex I. We might assume that they had a common origin, but that really is just an assumption. The first time the bowl and the cover of the codex actually have a documented history together is in Claremont, California in 1976 (the cover of Codex I had arrived in Claremont in 1972).4 Treating the cover and the bowl as a unit tends to create the impression of unity and coherence in a “find story” that is actually quite messy when we focus on the details.
Inscription on the bottom of the bowl, MS1804/1 (3) in the Schøyen Collection; image source: Christie’s
James M. Robinson, The Nag Hammadi Story (Leiden: Brill, 2014), 1.35-36. ↩︎
James E. Goehring, “An Early Roman Bowl from the Monastery of Pachomius at Pbow and the Milieu of the Nag Hammadi Codices,” in L. Painchaud and P.-H. Poirier (eds.), Coptica – Gnostica – Manichaica: Mélanges offerts à Wolf-Peter Funk (Louvain: Peeters, 2006), 57-371, quotation at 361, n. 15. ↩︎
See the account in Robinson, The Nag Hammadi Story, 1.383-390. ↩︎
The upcoming auction of the Crosby-Schøyen Codex prompted me to revisit the edition of the codex and think a bit about the book’s construction. One datum that didn’t really register with me before I started making papyrus rolls is the length of the kollēmata, the individual sheets of papyrus pasted together to make up a papyrus roll. Pliny the Elder provides descriptions of different classes of these sheets, and the largest that he mentions is the macrocolum, which is 1 cubit long (about 44.4 cm).
Papyrus codices that are sufficiently well preserved sometimes allow us to reconstruct the rolls from which their bifolia were cut. The kollēmata of the rolls reconstructed from some codices do fall below the maximum width reported by Pliny. For example, in 1956 Jean Scherer reconstructed the rolls from which the bifolia of one of the Tura codices of Origen (TM 62347) were cut.
Reconstruction of the kollēmata of the Tura codex of Origen’s Contra Celsum; images sources: CSAD and J. Scherer, Extraits des Livres I et II du Contre Celse d’Origène (Cairo: IFAO, 1956), p. 3
Each individual kollēma was about 20 cm long. The Beatty-Michigan codex of the Pauline epistles (P46, TM 61855) has kollēmata that average about 20 cm long (though they range in length from about 10 cm long to about 33 cm).1 There are usually two kollēseis (pasted joins) on each surviving bifolium.
A bifolium of the Beatty-Michigan codex of the Pauline epistles (P46) with two kollēseis indicated by white lines; image source: Chester Beatty Online Collections
But a number of codices were constructed from rolls that have much longer kollēmata. The conservator Hugo Ibscher (1874-1943) reported that some of the kollēmata used to make the rolls used for the Berlin “Gnostic” codex (P.Berol. 8502, TM 107765) were over 1.6 meters long. Rodolphe Kasser reported that the Bodmer Menander codex (TM 61594) was made up of cut from a single very long kollēma measuring at least 4.20 meters!2 [[Correction 28 Sept. 2025: Kasser claimed the length of the roll was 4.20 meters, but he points out the roll had two kolleseis, so the individual kollemata were shorter, but still quite long.]] This seems implausible given the usable length of a stalk of papyrus. I have not been able to verify or refute Kasser’s claim, but from the digital images now available online, one can reconstruct quite long kollēmata. Here is a set of bifolia certainly from a single kollēma that was at least 75 cm long:
Bifolia of the Bodmer Menander Codex showing a single long kollēma stretching across three bifolia, white arrows highlight especially prominent horizontal fiber patterns; image composed from material from the Bodmer Lab
When the rolls used to make the Nag Hammadi codices were reconstructed, it was found that these kollēmata were also quite long, with lengths of up to 1.62 meters.3 So, there is good evidence for rolls composed of kollēmata much longer than Pliny’s macrocollum. This suggests that the mode of manufacturing papyrus must have undergone a change from Pliny’s day. Eric Turner drew out the implications:
“A roll without kolleseis of a length of about 2 m. would have been constructed using the full height (about seven feet) of the papyrus stem. If such rolls were made, they would appear to be a novelty designed to meet the requirements of codex-makers. It is to be noted that most of these supposed rolls with kollemata 2 m. wide are of very coarse, thick papyrus.”4
Turner located this shift in the fourth century. It is thus possible that the presence of these long kollēmata may be an indicator that a codex was produced no earlier than the fourth century.
The kollēmata of the Crosby-Schøyen codex fall into this later, longer group. The editors of the codex reported kollēma lengths between 1 and 1.5 meters.5 Radiocarbon analysis of the papyrus of this codex would allow for a date anywhere from the middle of the third century to the middle of the fourth century. It may be that the presence of the long kollēmata in the Crosby-Schøyen codex suggest a date in the later part of that range. But some caution is necessary. The number of samples we have is small. And the existence of the relatively short kollēmata in the Tura Origen papyrus noted above (which is assigned to the sixth or seventh century) suggest that shorter kollēmata continued to be used throughout late antiquity. So, shorter kollēmata may not necessarily be indicative of an earlier date, but longer kollēmata (over about 45 cm) may perhaps be an indicator of production in the fourth century or later.
See the data gathered in E.B. Ebojo, A Scribe and his Manuscript: An Investigation into the Scribal Habits of Papyrus 46 (P.Chester Beatty II – P.Mich. inv. 6238) (Ph.D. diss., University of Birmingham, 2014), 79-86. ↩︎
R. Kasser, Papyrus Bodmer XXV, Ménandre: La Samienne (Bibliotheca Bodmeriana, 1969), at 11-12: “En effet, un examen attentif des pages du P. Bodmer XXV-IV-XXVI nous a convaincu du fait suivant: tous les folios de ce codex ont été découpés dans un même rouleau. …Ce volumen devait avoir une longueur de 4,20 m au moins.” ↩︎
J.M. Robinson, The Facsimile Edition of the Nag Hammadi Codices: Introduction (Brill, 1984), 67-70. ↩︎
E.G. Turner, The Typology of the Early Codex (University of Pennsylvania Press, 1977), 53. ↩︎
J.M. Robinson, “The Manuscript’s History and Codicology,” in James E. Goehring (ed.), The Crosby-Schøyen Codex Ms 193 in the Schøyen Collection (Peeters, 1990), xix-xlvii, at xlv-xlvi. ↩︎
In 2022, I published a short paper on the construction of the Beatty-Michigan codex of the Pauline epistles (P46, TM 61855). I suggested that the surviving page numbers in the codex might not be an entirely reliable guide to the original size of the single quire that makes up the codex (I had noticed that some other better preserved single-quire papyrus codices were not symmetrical in terms of numbered pages in the two halves of the codex).
I realized after publication that I had not noted an additional fact relevant to the discussion: Early codices, even single-quire codices, can have multiple sequences of page numbers. I pointed to the Crosby-Schøyen codex (TM 107771) as an example in a subsequent post.
Now, I follow up with another addendum. In the original article I used the Berlin Akhmimic Proverbs codex (TM 107968) as an example of a single-quire codex in which the outer folia of the quire were incorporated into the cover, relying upon the conservator Hugo Ibscher’s description of the book:
The original cover of Berlin Ms. or. oct. 987, now on a paper quire; image source: Paola Buzi, The Manuscripts of the Staatsbibliothek zu Berlin preussischer Kulturbesitz, Part 4 (Steiner, 2014), plate 5
“The Berlin Akhmimic Proverbs codex presents further interesting evidence. Hugo Ibscher noted that the outermost folia of this single-quire codex were incorporated into the leather cover of the codex along with waste papyrus to help stiffen the cover.”
I added a footnote to this sentence expressing some caution about this interpretation of Ibscher:
This is at least how I understand Ibscher’s somewhat confusing description: “The cover of the book consists of 6 to 8 layers of papyrus, which, when glued on top of each other, resulted in a strong cardboard. The outer sheets of this cardboard contain remains of Greek writing, while the inner ones are apparently blank. It fits well with this that from the first papyrus roll, before the surviving sheet 1 of the codex, four more sheets have been taken from the first papyrus roll [from which the bifolia were cut] that are not found in the codex and were probably used to strengthen the book cover and were bound through at the same time” (Hugo Ibscher, “Von der Papyrusrolle zum Kodex,” Archiv für Buchbinderei 20 [1920] 21–40, at 39: “Der Buchdeckel besteht aus 6 bis 8 Papyruslagen, die übereinandergeklebt eine starke Pappe ergaben. Die äußeren Blattlagen dieser Pappe enthalten griechische Schriftreste, während die inneren anscheinend unbeschrieben sind. Hierzu paßt es gut, daß aus der ersten Papyrusrolle, vor dem erhaltenen Blatt 1 des Codex, noch vier Blätter entnommen sind, die sich im Kodex nicht vorfinden und wahrscheinlich zur Verstärkung des Buchdeckels verwendet und gleich mit durchgeheftet wurden”). I take heart from the fact that Theodore C. Petersen showed a similar understanding: “The papyrus boards of the binding were found by Dr. Ibscher to have been built up of six to eight papyrus leaves pasted one over the other. Several of the innermost of these were found to have been sewn as part of the original codex, i.e., as its outermost sheets.” See Theodore C. Petersen, Coptic Bookbindings in the Pierpont Morgan Library (ed. Francisco H. Trujillo; Ann Arbor: Legacy, 2021) 430. It is unclear how common this type of construction was among single-quire codices. I know of no other documented examples. J.A. Szirmai, citing Ibscher’s “Der Kodex,” mentions another alleged example of this phenomenon, but he has misread Ibscher’s reference, which is almost certainly to the Berlin Proverbs codex (Szirmai, The Archaeology of Medieval Bookbinding [Aldershot: Ashgate, 1999] 12).
I was a little unclear on what Ibscher meant by “und gleich mit durchgeheftet wurden.” And I probably should have given Ibscher’s reference from “Der Kodex” in full. The context there is a discussion of single-quire codices. Ibscher does not specifically name the Proverbs codex (which is why Szirmai, I think, misinterpreted him), but it must be the Proverbs codex that Ibscher had in view:
“In one of these early codices I was able to make the interesting observation that the six outer bifolia were left blank. They were firmly attached to the quire by stitching, then glued together and now also served as a book cover, which was covered with leather.”
“Bei einem dieser frühen Kodizes konnte ich noch die interessante Beobachtung machen, daß die sechs äußeren Doppelblätter unbeschrieben gelassen wurden. Sie wurden mit der Lage durch die Heftung fest verbunden, dann zusammengeklebt und dienten nun zugleich als Buchdeckel, der mit Leder überzogen war.”
Now, I am happy to report that another passage by Ibscher confirms this understanding of the construction of the Proverbs codex in no uncertain terms. While looking up some information about the Medinet Madi codices, I came across this note by Ibscher specifically in reference to the Proverbs codex (Ms. or. oct. 987):
“The Coptic papyrus codex of the Berliner Staatsbibliothek, which was still bound when I received it for conservation, consists of 40 bifolia and some stubbed bifolia, while the 6 outermost bifolia were glued together to form the book cover.”
“[D]er koptische Papyruskodex der Berliner Staatsbibliothek, der noch im Einband saß, als ich ihn zur Konservierung erhielt, umfaßt 40 Doppelblätter und einige Einzelblätter, während die äußeren 6 Doppelblätter zusammengeklebt den Buchdeckel ergaben.”
It’s good to have Ibscher’s view of this matter fully clarified.
Bibliography:
Ibscher, Hugo. “Die Handschrift.” Pages v–xiv in Hans Jakob Polotsky and Alexander Böhlig, Manichäische Handschriften der Staatlichen Museen Berlin: Kephalaia, 1. Hälfte (Lieferung 1–10). Stuttgart: Kohlhammer, 1940.
Ibscher, Hugo. “Der Kodex.” Jahrbuch der Einbandkunst 4 (1937) 3-15.
Nongbri, Brent. “The Construction and Contents of the Beatty-Michigan Pauline Epistles Codex (P46).” Novum Testamentum 64 (2022) 388-407.
After reading these articles, I recognized that I was not going to be able to approach early Christian literature (or indeed, ancient Roman literature more generally) in quite the same way again. Gaining a greater appreciation of the the ubiquity of enslaved labor in the Roman world, and especially the role of shorthand specialists in the production of writing, casts all of early Christian literature in a different light.
Now Moss has synthesized her findings and produced what is probably the most important book in New Testament studies written in the last half century. God’s Ghostwriters: Enslaved Christians and the Making of the Bible manages to touch upon and reframe nearly every “classic” question of critical New Testament scholarship–issues of authorship and pseudepigraphy, sources and editing, transmission and textual variation, reading and reception, and even theology. And it’s written in a way that will be readable to those outside specialist circles (for scholars, a companion website provides amplified footnotes and a set of additional resources, and the articles mentioned above offer further documentation and nuance).
The topic of slavery in the New Testament is of course not new. There are many studies about what early Christian writers had to say on the topic of slavery and also about the metaphorical use of the language of slavery and freedom in early Christian writings. But in the last twenty years or so, scholarship has begun to focus more on actual enslaved people in the Roman world, including early Christians. Moss is meticulous in noting her debts to the work of predecessors and contemporary colleagues working on ancient slavery. Yet, she has pushed this conversation in interesting new directions by focusing especially on the role of literate enslaved laborers in the production, transmission, and consumption of literature in the Roman world.
Sarcophagus of Valerius Petronianus, side relief showing an elite figure reclining while an assistant holds writing tablets, Museo archeologico di Milano; image source: Wikimedia Commons
Part of what makes the book intriguing is the complication of the idea of authorship. When we talk about things like “Paul’s letters,” we’re actually talking about the letters of Paul and his co-authors. Several of the letters are clear about this in their opening lines: Paul and Sosthenes (1 Corinthians); Paul and Timothy (2 Corinthians, Philippians, and Philemon); Paul, Silvanus, and Timothy (1 Thessalonians). But Moss pushes further to highlight the different kinds of labor that went into composing writing in the Roman world. There are places in our sources where such labor flashes into view and then disappears just as quickly, such as Romans 16:22: “I, Tertius, the writer of this letter, greet you in the Lord.” We only see Tertius for a moment, but he has been there the whole time, “writing” the entire letter to the Romans. And acknowledging this fact feels different after reading Moss’s discussion of shorthand notation, a complicated set of signs memorized by literate workers, who were often enslaved, in order to quickly take dictation. The notary who took the dictation would later expand the shorthand copy to produce a copy in normal writing. It’s difficult to escape the conclusion that, much as we might like to imagine these letters as a direct conduit to “Paul’s beliefs,” what we are seeing here is more probably a situation of co-authorship. And it’s likely that Paul’s other letters, and most literary works in Roman antiquity, were written in a similar way. The praise of the shorthand writer attributed to the professor, poet, and statesman Ausonius evokes a partnership (albeit an unequal one): “No teaching ever gave you this gift, nor was ever any hand so quick at swift stenography: Nature endowed you so, and God gave you this gift to know beforehand what I would speak, and to intend the same that I intend” (Ephemeris 7).
Wax tablet with a student’s practice in shorthand symbols; British Library Add. MS 33270; image source: H.J.M. Milne, Greek Shorthand Manuals (Egypt Exploration Society, 1934)
Was Tertius enslaved? Moss demonstrates that people who performed secretarial work in Roman antiquity often were. And names like “Tertius” (literally, “the third one”) were common for slaves. But more elite Romans also had such “numerical” names. At the end of the day, we don’t know the status of Tertius with certainty; we simply don’t have enough data. Indeed, one of the themes that runs through the book is the problem of invisibility. Had Tertius not decided to add his ten words of greeting, we wouldn’t know he had ever existed. A similar situation applies with almost all literate enslaved labor in the ancient Roman world. We have to work with a small corpus of surviving evidence, because when these kinds of workers did their jobs well, they became invisible. Moss highlights the irony: “They performed their work so perfectly that they wrote themselves into nonexistence.”
P.Oxy. 44.3197, a contract for dividing up the slaves of the estate of Tiberius Julius Theon. Among the slaves are Heron the γραμματεύς (scribe), Ammonas the νοτάριος (shorthand writer), and Demas the προχειροφόρος (amanuensis); image source: University of Oxford Sustainable Digital Scholarship
One tactic Moss uses to address this problem of invisibility is to present historical reconstructions that are avowedly imaginative. This decision will no doubt invite some criticism. Yet, by explicitly and repeatedly highlighting the imaginative elements in her own account, Moss invites fellow historians to consider the imaginative work that lies behind many of our own easy assumptions about Roman literary culture.
The question really is: What should our “default” assumptions about the composition of early Christian literature be? Having read Moss’s work, I don’t think it’s historically responsible to picture Paul or the writers behind the gospels as lone theological geniuses. The realities of ancient practices of composition, at least to the extent that we can reconstruct and imagine them on the basis of limited and fragmentary sources, suggest that a rethinking of ancient authorship is going to be necessary.
And it’s not just the composition of early Christian literature that involved this often invisible labor. The copying of texts, the movement of books from place to place, and the reading of books all frequently involved enslaved or formerly enslaved workers. If Paul and co-authors wrote letters together, those who delivered letters also had a role in creating and conveying the meaning of the text. Cicero comments in a letter to a friend, “Your freedman Cilix was not well known to me before, but ever since he delivered me your letter, so full as it was of affection and kindness, he has himself by his own words followed up in a wonderful way the courtesy with which you wrote. It was a delight to me to hear him holding forth…” (Ad. fam. 3.1). Moss’s chapters detail the ways in which reading the ancient evidence with enslaved laborers in view can inform and transform the way we understand these aspects of early Christianity.
Clever and insightful observations abound. Throughout the book, Moss places various early Christian texts in dialogue with literature that depicts different aspects of ancient enslaved life. To take just one of many examples, Moss pairs Paul’s language–or perhaps better, the language of Paul, Sosthenes, and the others involved in the writing of 1 Corinthians–about the body (“Christ is the head of every man…You all are the body of Christ, and individually its limbs”) with the fascinating but neglected treatise on household management by Bryson, who theorized slavery for elite Romans: “Slaves resemble the limbs of the body…”
Excerpts from Bryson, Oikonomikos logos, Bibliothèque nationale de France, Grec 1984, fol 165 verso; image source: Gallica
The comparison prompts reflections on how early followers of Jesus may have understood their relationship with the man they called their κύριος, dominus, master. It adds a dimension to the self-designation of some followers of Jesus as δοῦλοι Χριστοῦ, “slaves of Christ” (Romans 1:1, Philippians 1:1, Titus 1:1, James 1:1, 2 Peter 1:1, Jude 1). And for all the ink spilled on the meaning of πίστις/fides in early Christian writings, Moss’s emphasis on Roman elites’ use of the term to describe the proper attitude of the enslaved to the enslaver (“loyalty”) is striking and suggestive.
Overall, this book has an effect that is similar to that of E.P. Sanders’s Paul and Palestinian Judaism (1977). Even if you don’t agree with every interpretive move the author makes, the collective force of the book’s examples leaves you with what can only be described as a new perspective. In other words, if the field takes God’s Ghostwriters seriously (and it should), then there is no going back. And the way forward will involve greater effort to pay attention to the enslaved people who played important parts in producing, promulgating, and preserving the writing of the Roman world.
Metal tag, likely from the collar of an enslaved person; the inscription reads: tene me ne fugia(m) et revoca me ad dom(i)num Viventium in ar(e)a Callisti Inscription translation: Capture me, lest I escape, and return me to my master Viventius on the estate of Callistus; image source: The British Museum, 1975,0902.6
Variant Readings 