The great age for the discovery of papyrus manuscripts in Egypt was of course the late nineteenth and early twentieth century. But a handful of papyrus manuscripts, more specifically Latin papyrus codices, also survived from antiquity in European libraries.
In his paleographical handbook published in 1893, Edward Maunde Thompson gave a list of these. As it turns out, images of most of them are available online:
Avitus, Sermons and letters (Paris, BnF Lat. 8913 and 8914; CLA 573)
Augustine, Sermons and letters (Paris, BnF Lat. 11641; Geneva Ms Lat. 16; St. Petersburg, Russian National Library Lat. F. I. 1; CLA **614)
Hilary, De trinitate (Vienna, Austrian National Library Cod. 2160*; Vatican Library, Barberini Lat. 9916 (one folio); Sankt Florian Austria Stiftsbibliothek III.15.B (one folio); CLA 1507)
One aspect of the EthiCodex project involves a closer look at the materials that make up ancient codices: papyrus, parchment, leather, threads, glues, and inks. A few weeks ago, the Papyrus Collection at the University of Oslo welcomed Ira Rabin and Grzegorz Nehring of the Bundesanstalt für Materialforschung und -prüfung (BAM). We had a nice look at some of what their team is doing in terms of developing a more systematic approach to the chemical analysis of ancient inks.
Ira Rabin and Grzegorz Nehring at the UiO Papyrus Collection
It was an illuminating couple of days, both figuratively and literally. Many thanks to Ira and Greg, and also to our hosts at the Papyrology Collection, Anastasia Maravela and Federico Aurora.
In my previous posts about making a model of Beatty-Michigan codex of the Pauline epistles (P46, TM 61855), I discussed the sizes of the bifolia and the curious fact that P46 seems to lack any evidence for the presence of protective parchment or leather stays in the center of the quire. Another issue related to the absence of evidence for stays is a similar lack of evidence for the presence of a spine lining strip or a cover. Just as the stays prevent the binding thread or tacket from ripping the papyrus leaves at the center of the quire, the spine strip protects the outermost bifolia from the binding agent, and the cover protects the whole codex.
Several of the Nag Hammadi codices have well preserved leather spine strips that are now kept together with the covers. Here is the cover of Nag Hammadi Codex IX (TM 107749). Notice the papyrus scraps that still adhere to the spine strip.
Nag Hammadi Codex IX, inside of leather cover, showing spine strip at center with papyrus fragments still adhering; image source: The Claremont Colleges Digital Library
A spine strip also appears to have been preserved with the Crosby-Schøyen Codex (TM 107771), The darker leather is just visible along the spine in this photograph of the codex before it was disassembled:
Crosby-Schøyen Codex before conservation, showing spine lining strip at left; image source: William H. Willis, “The New Collections of Papyri at the University of Mississippi,” in Leiv Amundsen and Vegard Skånland (eds.), Proceedings of the IX International Congress of Papyrology (Oslo: Norwegian Universities Press, 1961), 381-392, plate 2.
In the case of P46, the absence of evidence for a cover and spine strip is understandable, since the outermost bifolia of P46 did not survive. So, even though there is no surviving evidence for either a spine lining strip or a cover, it seems reasonable to expect that they existed. Based on that reasoning (plus the practical point that threads cut right through the papyrus without a spine strip), I added a leather spine strip to the model I made.
Model of P46 showing a hypothetical spine lining strip, snake weight at right to keep the quire closed
I decided not to put a cover on this model, but I am pretty firmly convinced that P46 and indeed most papyrus codices, probably had covers. I recall reading somewhere (I can’t come up with the exact reference) that perhaps the presence of covers on papyrus codices was exceptional. I find this a bit difficult to believe for a couple reasons.
First, papyrus codices, especially single quire codices, tend not to stay closed. There is a snake weight in this image above to hold the codex shut because what it “wants” to do is spring open. A cover with a flap and ties keeps the codex codex closed.
Second, papyrus leaves do not do well without some form of protection. This quire endured just a few trips between office and house in my backpack, and it already has begun to show signs of wear at the edges.
Papyrus books with leather covers, however, are surprisingly durable. The model I made of Nag Hammadi Codex VI can be shaken around by the cover and generally abused without showing any significant damage. I think this kind of cover must have been the norm for papyrus codices in antiquity.
In my first post about making a model of P46 (TM 61855), the papyrus codex of Paul’s letters in Greek that is split between the Chester Beatty Library in Dublin and the University of Michigan, I discussed the different sizes of the bifolia. Now I want to turn to the elements that usually protect the central folds of papyrus codices, the stays on the inside of the quire. In the case of P46, these items do not survive (if they ever existed at all).
The single-quire papyrus codex is a relatively straightforward construction. A stack of papyrus sheets folded in half and bound. The binding can either be directly through the central fold or “stabbed” through the codex from front to back.
For larger (that is to say, thicker) single-quire codices, it seems as though binding through the central fold would be the more practical option. From ancient examples that have survived in a relatively good state of preservation, like several of the Nag Hammadi codices, we have learned to expect leather or parchment stays at the center of the quire in order to protect the papyrus from being torn by the pressure of the binding agent, whether it is thread or a leather tacket.
In the image below of Nag Hammadi Codex VI (TM 107746), which was bound with two leather tackets through two leather stays, the lower stay is visible, still in place:
Nag Hammadi Codex VI, opened to the center of the quire, with lower leather stay in place (the upper stay is obscured by a feather; the hand belongs to Marianne Doresse; image source: The Claremont Colleges Digital Library
(Side note: Does anyone know if there are other images showing the Nag Hammadi books with the stays in place before they were disassembled?)
Stays have survived in several of the Nag Hammadi codices. When the papyrus leaves of the Nag Hammadi books were removed and mounted, the stays were stored together with the covers. The cover and stays of Codex VI are now in the Bibliotheca Alexandria:
In other surviving single-quire codices, the stays have survived only occasionally. But when such stays are absent, there is often evidence that they were once present in the form of stains or discoloration on the central bifolium in exactly the place(s) where we might expect stays. We see this in the Bodmer Menander codex (TM 61594):
Bodmer Menander Codex, center of quire; image adapted from separate images from the Bodmer Lab
A similar discoloration (lightening) can be seen in the central bifolium of the Crosby-Schøyen codex (Schøyen MS 193, TM 107771). With the Berlin Proverbs codex (Staatsbibliothek zu Berlin, Ms. or. oct. 987, TM 107968), which has been rebound in a modern binding, we have a glass frame that preserves a parchment strip that was almost certainly the stay from the center of the quire:
Remains of the parchment stay, unidentified piece of leather, and binding thread from Staatsbibliothek zu Berlin, Ms. or. oct. 987; image source: Paola Buzi, The Manuscripts of the Staatsbibliothek zu Berlin preussischer Kuluturbesitz, Part 4 (Stuttgart: Franz Steiner, 2014), plate 5a.
This all makes sense. When papyrus is folded, it weakens. When it is punctured for binding, some reinforcement would seem necessary around the points of friction between the papyrus and the binding agent. I have learned from unpleasant experience that binding threads or leather tackets will rip right through a whole stack of papyrus sheets if they are unprotected. So, it’s quite curious to me that there does not seem to be evidence for any protective stays at the center of the single quire that makes up P46.
If we digitally reunite the now separated halves of the bifolium at the center of the quire, we see no evidence of stays (the color mismatch between the two folia here seems to be due to different conditions during photography):
There is evidence of two pairs of binding holes through the central fold of the codex, but there is no evidence (at least as far as I am aware) for the presence of stays. I’m genuinely puzzled as to how the binding would have worked without stays. For my model, I used linen thread and included a pair of parchment stays just to make the codex stable and usable:
Model of P46, open to the center of the quire
If I were to do it again, I would probably use leather stays, as the parchment stays have already begun to tear slightly under the pressure of the binding threads.
In the next post, I’ll talk about some other aspects of the model.
One of the most useful exercises to do when trying to understand the construction of an early codex is to build a model. With early codices, it is always the case that we don’t have all the information that we need to accomplish this with total accuracy, but the process of making a model helps bring to light some of the unanswered questions. I’ve mentioned on the Variant Readings blog some of the models I’ve made in the last few years–for instance, Nag Hammadi Codex III (TM 107743) and Codex VI (TM 107746).
In connection with an article I recently wrote, I made a model of the papyrus codex containing Paul’s epistles in Greek split between the Chester Beatty Library and the University of Michigan (a.k.a. P46, TM 61855). It is another single-quire codex. It’s the largest (that is to say, the thickest) codex that I have tried to make. Nag Hammadi Codex VI had 20 bifolia, Codex III had 40 bifolia, and P46 is estimated to have had 52 bifolia. Only 43 bifolia survive, but the pages of the codex were numbered, so the reconstruction of the size of the quire is reasonable (though not certain, as I discuss in the aforementioned article).
Over the next few posts, I’ll describe some the issues I encountered in making the model. The feature item I’ll discuss is the size of the bifolia.
When a thick single quire is folded in half, the innermost sheets protrude. Ancient bookmakers solved this problem by making the outermost bifolia of the quire broader than those in the center of the quire. In the case of the better preserved Nag Hammadi codices, this change in size can be clearly documented with leaf by leaf measurements. P46 is a bit more damaged, and so we must do some estimation to determine the original dimensions of its bifolia. The breadth of the central bifolium of P46 is reasonably well preserved, with just a bit of wear on the edges. The original breadth was likely about 26.8 cm. The outermost bifolia of the codex have not survived, and the outermost bifolia of the codex that do survive are quite damaged. In his thorough PhD dissertation on P46, Edgar Battad Ebojo has estimated that the outermost bifolium of the quire likely measured about 32 cm (based on the measurements of the largest surviving bifolia). So, between the innermost and outermost bifolium, there was likely a difference of just about 5 cm, as illustrated below.
Outermost (top) and central (bottom) bifolia for a model of P46
With a total of 52 bifolia, each bifolium should be about 1 mm less wide than the previous one from outside to inside. So I decreased the breadth of each sheet by 1 mm as I cut them. The resulting stack looked like this:
Stack of bifolia with decreasing breadth to be used in a model of P46
Even with this adjustment, however, I found that the central bifolia still protruded a bit when the quire was folded.
Model of P46, view of the top of the quire showing protruding central bifolia
I had similar results with my model of Nag Hammadi Codex III:
Model of Nag Hammadi Codex III, view of the top of the quire showing protruding central bifolia
I suspect that these large single-quire codices in antiquity probably did have this uneven fore-edge. I am aware of only one good photograph of an ancient intact single-quire codex that clearly shows the fore-edge; this is the Berlin Akhmimic Proverbs codex (TM 107968). The codex is of course damaged, but it does appear that the central bifolia protrude quite noticeably (and this in spite of the fact that the outermost bifolium is about 4 cm broader than the central bifolium). (Side note: I’m unsure of the ultimate source of this photo; I found it in Sharpe’s chapter, but I don’t know where he got it. I would love to have an image developed from the original negatives, as there are many details that cannot be made out in this scan.):
Image source: John Lawrence Sharpe, “The Earliest Bindings with Wooden Board Covers: The Coptic Contribution to Binding Construction.” Pages 2.455-2.478 in Carlo Federici and Paola F. Munafò (eds.), International Conference on Conservation and Restoration of Archival and Library Materials; 2 vols. (Rome: Istituto centrale per la patologia del libro, 1999)
It is sometimes suggested that the fore-edge of single-quire codices could have been trimmed after binding to achieve a flat endge (e.g. in James Robinson’s writings on the Nag Hammadi codices). This is in theory possible, but I wonder how, in practical terms, such trimming would have been carried out in antiquity (I’m genuinely curious; suggestions are welcome!). It seems like it would be a tricky procedure with ancient cutting tools. For now, my assumption is that the bifolia of single-quire codices were cut to size before folding.
In the next post, I’ll discuss the binding of P46 and the use of stays in single-quire codices.
As I work through the Leuven Database of Ancient Book looking at early codices, I continue to see cases of ambiguous format. I discussed one of these, P. Rainer Unterricht (MPER N.S. 15) 7+8+10, in an earlier post.
Here are a couple additional examples of unclear cases. BKT 10.6 (P.Berol. inv. 21313) is a small fragment of papyrus with lines of what looks like a commentary on Thucydides written along the fibers. On the back side are what appear to be the ends of a few lines.
The website of the Berlin collection describes this piece as a roll. The Leuven Database describes the papyrus unambiguously as a codex. This case is not clearcut. The first editor of the papyrus, Luciano Canfora, observed that the format of the fragment was actually ambiguous: “It could be either a fragment of an opisthograph roll or a fragment of a papyrus codex (un frammento di rotolo opistografo ovvero del frammento di un codice papiraceo).” The most recent editors, Peter Parsons and Natascia Pellé, doubted the possibility that this fragment came from a codex (“Alquanto improbabile appare l’ipotesi che si tratti di un codice papiraceo”). They noted that, although the traces that can be seen of the hand on the back of the papyrus do resemble the script on the front of the papyrus, the written columns on either face of the fragment do not seem to be aligned. That is to say, the writing on the vertical fibers appears to be the end of a column of writing, while the writing along the fibers does not seem to be the beginning of a line, which is what we would expect in dealing with a typical codex. In any event, this is an ambiguous case. Perhaps “Codex(?)” would be the most accurate description.
In other instances, the Leuven Database simply seems to be mistaken. For example, the Leuven Database unambiguously describes P.Oxy. 78.5162 as a codex, but it is almost certainly not. The front of the papyrus preserves a Greek-Latin glossary.
The back side of the fragment does contain writing, but as the editors note, this writing is “upside down in relation to the text of the glossary” and consists of “two columns of Greek medical prose.” This papyrus is much more likely to be a reused roll and not a codex (there are some examples of codices in which this “upside down” phenomenon occurs–I discussed one here a few years ago–but it is very, very rare).
These may seem like somewhat trivial matters, but when we’re trying to get a sense of what the dataset looks like for surviving fragments of early codices, these details matter.
Sources cited:
Luciano Canfora, “P.Berol. 21313, MP3 1552.04: Un frammento inedito di commento a Tucidide,” Quaderni di storia 74 (2011), 97-98.
Peter Parsons and Natascia Pellé, “6. Commentario a Tucidide III?” in Fabian Reiter (ed.), Literarische Texte der Berliner Papyrussammlung (Berliner Klassiker Texte 10; Berlin: de Gruyter, 2012), 31-33.
R.-L. Chang, W.B. Henry, P. J. Parsons and A. Benaissa (eds.), The Oxyrhynchus Papyri, vol. LXXVIII, Nos. 5127–5182 (London: Egypt Exploration Society, 2012.
A long-awaited book has finally been published, and my copy has just arrived: Theodore C. Petersen’s CopticBookbindings in the Pierpont Morgan Library, edited by Francisco H. Trujillo (Ann Arbor: The Legacy Press, 2021).
Theodore C. Petersen (1883-1966) studied the leather covers and binding structures of the many Coptic codices in the Morgan collection in the first half of the 20th century. He never published the results of his work, but he produced a typewritten manuscript with a number of drawings. Many of us managed to get copies of copies of copies of snippets of this manuscript here and there over the years, but the full manuscript was not widely available to the public until now.
Francisco Trujillo has provided an introduction and edited Petersen’s text. The catalogue is accompanied by Petersen’s line drawings and excellent new color photos.
While this is a catalog, it should be noted that the volume is about more than just the Morgan collection. There is a detailed discussion of early bookbinding in general and Petersen’s analysis of early bindings outside the Morgan collection, such as that of the Akhmimic Proverbs codex in Berlin:
There are several more sample pages shown at the site of Oak Knoll Press, where the book is being sold. I’m very much looking forward to reading this volume. Thanks to Francisco Trujillo, the team at the Morgan Library & Museum, and the people at Legacy Press for producing such an important (and great looking) volume.
And finally, for those wanting a closer look at the Coptic bindings and some of Petersen’s drawings, visit the collection of digital images at the Morgan.
A number of challenges face anyone who undertakes a study of the early development of the codex. Two such challenges are “built in” to the corpus of surviving codices and codex fragments: Very few samples have precise and secure dates, and some samples are of ambiguous format (that is, it’s not clear they are actually codices). In this post, I want to look at a piece that raises both these issues. Consider the set of papyrus fragments known as P. Rainer Unterricht (MPER N.S. 15) 7+8+10 (TM 63194), G 26011 b, G 26011 c+d, and G 26011 e.
What we have here are a group of rectangular fragments now framed as three separate items. The fragments contain Greek alphabetic exercises copied against the fibers. On the side with horizontal fibers, fragments 26011 b and 26011 c+d have only faint traces of writing (they were perhaps washed clean in antiquity). Fragment e is inscribed on the horizontal fibers with a math problem. Below are images of the three frames, first showing the horizontal fibers of each piece (top) and then showing the vertical fibers (bottom):
The issue of dating is fairly straightforward: There is no firm basis for a date. The most recent editors relied on the analysis of handwriting, and given that the text is copied in a less than competent hand, assigning a precise date with confidence is very difficult, as the editors readily acknowledge (“Eine noch ganz und gar ungeübte Handschrift datieren zu sollen, stellt eine mehr als problematische Aufgabe dar”). The editors describe the hand as first century. This is possible, but it is certainly not necessary.
More interesting is the question of the format of these pieces. The online catalog of the Österreichische Nationalbibliothek classifies the fragments as a codex. The Leuven Database of Ancient Books classifies the pieces straightforwardly as follows: “Book form: codex; columns per page: 1.” But both parts of that description are open to question. If we assume that what we are seeing here are leaves of a codex, then the alphabetic exercises are written in multiple columns to the page (sometimes even with vertical lines drawn between columns):
But in fact we are almost certainly not looking at leaves of a codex.
The most recent editors assign the pieces three distinct numbers, but they note that all three items seem to be of the same material (“aus dem gleichen Material”), and they describe the group as a roll (“Rolle”). In a 2008 study, however, Jean Lenaerts argued that the papyri were part of a “carnet” (notebook):
“Je n’hésite pas à appeler cette entité un «carnet» scolaire, qui était vraisemblablement celui d’un apprenant adolescent ou adulte. Probablement taillés dans un rouleau, les trois coupons se succédaient dans l’ordre 7. 8 et 10, et leur fonnat quadrangulaire laisse à penser que le «carnet» était constitué de feuillets volants.”
It may be that Lanaerts’ description of the fragments as a notebook made of loose sheets has led others to describe this item as a codex, but I very much doubt these papyri ever formed part of a bound or unbound codex. Although the pieces as they are framed are indeed basically square, the squares are of quite different widths. But what makes the codex hypothesis especially improbable is the fact that the scripts on “fragment e” are upside down relative to one another:
That the writing on one face of the papyrus is upside down relative to the writing on the other side implies that we are most likely dealing here with a reused roll or simply loose sheets. It is therefore most unlikely that these fragments represent a first-century codex. It is possible (but hardly necessary) that they were produced in the first century. It is extremely unlikely that they ever were part of a codex.
P. Rainer Unterricht (MPER N.S. 15) 7+8+10 should not be considered part of the corpus of early codices.
Welcome to The Early History of the Codex: A New Methodology and Ethics for Manuscript Studies (EthiCodex). This blog will be a space for updates on the project, which officially begins in August 2021 and runs through July of 2026.
This first post is intended to introduce the project and provide some background. Early Greek and Latin codices (defined for the purposes of this project as those produced between roughly the second and the seventh centuries CE) are usually thought of as individually important because they preserve some of our most ancient copies of Christian and classical texts. But as a group, these books also have the potential to help us understand the development and spread of the technology of the codex. The last effort at a comprehensive overview of early codices was Eric G. Turner’s groundbreaking book, The Typology of the Early Codex, published in 1977.
Since the appearance of Turner’s Typology, many new codex fragments have come to light, and the fields of papyrology and book history have developed in a number of ways.
First, there is a new level of appreciation for the book as physical object. In addition to the main features that interested Turner (page size, writing surface, and quire structure), other material aspects of ancient books have come to be seen as worthy of note: the sewing of the binding, cover construction, and “paratextual” features, such as titles, chapter divisions, decorations, and marginal notes, to name a few. These kinds of features were not always high on the list of priorities for editors in the early twentieth century, who were mainly (and understandably) interested in publishing texts. The result is that it can be difficult to study these features systematically across the corpus of early codices.
Second, there is a growing recognition that the most common method of assigning dates to early codices (analysis of handwriting, or palaeography) is highly subjective and not as reliable as previous generations believed. Thus, up to the present, most conclusions about the development of the codex have been based on insecure foundations.
Third, recent years have also witnessed an increased awareness of ethical problems with the study of ancient manuscripts. That is to say, the study of ancient books is the study of cultural heritage property, much of it acquired under dubious circumstances.
What is needed is an ethically responsible and efficient way to handle a growing body of complicated evidence. We need to more systematically describe and catalog the physical features of early codices and codex fragments and build a larger corpus of securely dated samples. At the same time, we need to recognize that more advanced studies of items that are unprovenanced or that have been obtained through questionable means are problematic. To that end, The Early History of the Codex plans the following activities thanks to the generous support of the Research Council of Norway:
Conduct provenance research into the ownership histories of early Greek and Latin codices.
Produce detailed physical and codicological descriptions of the make-up of the earliest Greek and Latin codices.
Design an open-access database making codicological data and provenance information for an estimated 2500 early Greek and Latin codices easily searchable and freely available online.
Make a systematic canvassing of museum and library collections containing ethically acquired early papyrus and parchment books to determine willingness to have AMS radiocarbon analysis carried out on their early codices and then fund this analysis.
The ultimate goal is a new history of the development of the codex. But we also hope that the data we produce will be useful to scholars in many academic fields, from ancient history and classics to art crime and cultural heritage.
Thanks for reading, and keep an eye on this page for updates on the project.
The Early History of the Codex 