3 - The Construction of Codex I: Scribal Errors as Clues to Context

Summary

This chapter approaches previously unexplored perspectives on the codicology of Nag Hammadi Codex I, a unique early Christian manuscript. Several hitherto overlooked features regarding scribal style and codicological features are discussed, and it is argued that these can help us understand the background of the codex, both within the history of early Christian book culture and in terms of who produced it. The chapter demonstrates that Nag Hammadi Codex I, one of the first multi-quire codices from antiquity, was most likely produced as an ad hoc solution to sloppy scribal work. Thus, it offers grounds for rejecting the hypothesis that Codex I was constructed on commission by a professional bookseller as a commercial product, as some scholars claim. Finally, it is argued that Nag Hammadi Codex I should be viewed as one of the earliest examples of what was to become the standard technique of ancient book manufacturing.

On the basis that material aspects of texts offer insights into the identity of the people behind them, this chapter surveys the binding and copying style of Codex I, which stands out in the Nag Hammadi collection as the only multi-quire codex.¹ As argued below, much of the evidence to be drawn from its materiality points to its production within a monastic setting, the work of a monk inexperienced in copying texts or one who placed other principals before meticulous production of the codex. In this particular case, the scribe’s disordered work and the difficulties he encountered while engaged in it were partly solved by turning the manuscript from a single into a multi-quire codex.

The Material Features of Codex I

Codex I consists of five texts: The Prayer of the Apostle Paul, The Apocryphon of James, The Gospel of Truth, The Treatise on the Resurrection, and The Tripartite Tractate.² Evidence from the cartonnage of Codex VII, copied by the same scribal team, usually dates it to around the last half of the fourth century. Three letters have been found inside the cover of Codex VII, dated to the years 341, 346 and 348, leaving us with a terminus post quem of 348. The cover of Codex VII could, of course, have been made later than the cartonnage papyrus, but not before. As for the leather cover of Codex I, recent radiocarbon dating indicates that it likely predates Codex VII.³ Thus, a date in the middle of the fourth century seems plausible for Codex I, considering the available evidence.

Codex I was put together by using three quires (see Fig. 3.1).⁴ The first, by far the largest, consists of twenty-two sheets made from two rolls of papyrus; the second consists of eight sheets made from one roll and the third of six sheets, also made from a single roll. Thus, the first 86 pages together with the front flyleaf make up the first quire; pages 87–118 make up the second, and the third quire starts with page 119 and ends with 142. The last four pages of the codex are not preserved, which leaves open the possibility of a sixth text. As Stephen Emmel has pointed out, there are ink marks at the presumed end of The Tripartite Tractate indicating that something most likely followed it.⁵

Figure 3.1 The quire structure of Nag Hammadi Codex I, illustrated as described in Robinson, Facsimile: Codex I, vi–xxxi.

Adapted from an image previously published in Paul Linjamaa, ‘Nag Hammadi Codex I as a Protective Artifact and an Accidental Multi-Quire Codex’, in The Scriptural Universe of Late Antiquity, ed. Emmanuel Grypeou (Madrid and Salamanca: Editorial Sindéresis/Universidad Pontificia de Salamanca, 2021), 105–126.

As noted, Codex I is the only multi-quire codex in the Nag Hammadi collection, but it does not follow regular multi-quire patterns.⁶ It is unclear why the scribal team, or those who bound the codex, did not simply make one large quire, as is the case with the other Nag Hammadi codices, or one additional quire with fourteen sheets instead of two unusually small quires in the end. As has previously been suggested, these facts indicate that the construction of the codex was unplanned, and that the two last quires were impulsively prepared.⁷ As it happens, this is not the only irregularity attached to the production of Codex I. In the two rolls that made up the first quire, the kollemata on the left overlap those on the right. This is typical, since the roll from which the sheets were cut would most likely have been constructed so that a scribe, writing from left to right, could comfortably do the job without getting stuck on the joints.⁸ In the two smaller quires the kollemata on the right overlap those on the left, suggesting that either the rolls used were rolled up from left to right (contrary to custom) or, and perhaps more likely, that the sheets were accidentally turned the ‘wrong way’ after being cut; maybe a period of time had elapsed between cutting and commencing the work on the codex.⁹ Either way, this is another detail that makes the two last quires different from what one would expect.¹⁰ As it happens, other irregularities found in the codex production can be attached to one of the scribes in particular.

The codex was copied by two scribes. Scribe A copied all the texts except The Treatise on the Resurrection, which is in the hand of a second scribe, the same who copied the first half of Codex XI. Contrary to Scribe B – whose hand is legible and regular – Scribe A has copied with significant variation in quality, as well as word and line count per page. As just one example, we can compare page 41 (The Gospel of Truth) with page 94 (The Tripartite Tractate), both copied by Scribe A. Page 41 has 35 lines, each with between 14 and 21 letters, just over 600 letters in total. Page 94 has 40 lines and each line has between 20 and 26 letters, a total of over 900 letters. This means that Scribe A’s word count fluctuates with a difference of up to 30 per cent (see Fig. 3.2 for another example).¹¹ Could there be an explanation for this, other than just viewing it as the work of a careless or novice scribe? A closer look at where in the codex we find the cluttered pages with a relatively high word count and where we find most of the pages with a low word count and an airy scribal style reveals a possible pattern. Many of the pages with a relatively low word count are at the very end of the codex, while the cluttered pages, often with a high word count, are mostly found in the second quire.¹² Especially cluttered (and with many mistakes¹³), compared with the other parts of the codex, are pages 113–118, the last pages of the second quire. From the second half of the third quire until the end of the codex (pages 130–137), the word count drops considerably to around 650 per page, from an average of around 800 up to 900 in the second quire.¹⁴ Furthermore, there is a noticeable difference between the length of the lines in the first quire compared to the second. In the first quire, the lines average between 8 and 10.5 cm, while the lines in the second quire average between 10.5 and 12 cm, a difference of over 18 per cent (see Fig. 3.2).¹⁵ In most parts of the third quire, the long lines have been kept, but in the last eight pages (those that are intact enough for us to see a whole line) there is a considerable drop in line length (to about 9.5–10.5 cm), as well as word count, resulting in larger letters and the appearance of an airier style.¹⁶

Figure 3.2 There is a word count difference of approximately 30% and line length difference of about 15% between the right page and the left page. Left page 118 in Quire II; right page 134 in Quire III. Notice the airy style on page 134, the end of Quire III.

Photo by Jean Doresse. Images courtesy of the Institute for Antiquity and Christianity Records, Special Collections, Claremont Colleges Library, Claremont, California.

These observations give us further clues concerning the production of this somewhat awkward multi-quire codex. The Tripartite Tractate is divided by decorative markings into three separate parts, which raises some interesting points. James Robinson has suggested that only Part I (51–104) was originally meant to be included in the codex and that the second quire had to be added to finish this part.¹⁷ The scribe then decided to include Part II (104–108) of The Tripartite Tractate in order not to waste papyrus and then continued with Part III (108–138), which required yet a third quire.¹⁸ If this were the case, one would have expected to find the pages with cramped style, high word count and long lines in the first quire, or at least in the second half of it. Instead, these are found in the second quire, particularly towards the end. Thus, I would argue that Parts II and III of The Tripartite Tractate were most likely meant to be included from the beginning. Why else would the scribe have felt pressured for space after the second quire was added, which would have provided ample room for finishing Part I if that were all that was intended? If only Part I was the original plan, the scribe could simply have ended the second quire in an airy and relaxed style, which he used, instead, in the third quire.¹⁹

The airy scribal hand towards the end of the third quire indicates that The Tripartite Tractate was meant to be the last text of Codex I (Fig. 3.2). Yet it is still possible, even likely, that some ad hoc writing was inscribed on the last four pages, similar to the seemingly improvised inclusion of The Prayer of the Apostle Paul which was placed in the front by adding a flyleaf (Fig. 3.1). Furthermore, as already noted, the ink marks following The Tripartite Tractate suggest a sixth text and, considering the drastic drop in word count in the third quire, it would indicate that the scribe miscalculated the number of leaves needed to finish the section. This left whole pages empty, allowing a sixth text to be inscribed, presumably one not considered the most urgent. There are a number of reasons why it would have been difficult to calculate the exact number of pages needed to copy The Tripartite Tractate. If the Vorlage of The Tripartite Tractate, a very long text, did not have the same dimensions as Codex I – if it were contained in a different medium such as a scroll or a smaller or larger codex, for example – it would have made the calculations harder.²⁰

Approaching the Owners and Creators of the Codex

Palaeographic investigations have shown scribal overlap between Codices I, VII and XI (and there are other groupings as well, based on palaeographical similarity).²¹ The sequence of the texts in Codex I also seems to have been of some importance. The pages of the fourth text, The Treatise on the Resurrection, are unpaginated and the bottom half of the last page is empty, indicating that Scribe A left these pages empty after copying the preceding texts, The Apocryphon of James and The Gospel of Truth, and before proceeding to copy the last text of the codex, The Tripartite Tractate.²² There must have been a reason not to copy The Tripartite Tractate directly after The Gospel of Truth and a reason why The Treatise on the Resurrection should follow The Gospel of Truth and not The Tripartite Tractate. Some attempts have been made to read Codex I as a collection with a particular purpose, and most have viewed the placement and topic of The Tripartite Tractate as designed to give the preceding texts contextualisation, placing the ‘message’ of Codex I in a bigger picture.²³ However, no single view has received wide scholarly acceptance.²⁴ If we can get closer to answering the question of how the codex was produced, how it was read and by whom, it would give much-needed contextualisation for further investigation into the meaning and purpose of Codex I as a whole.

As mentioned above, it has been suggested that the order of the texts in Codex I must have been important, since the scribe left eight pages blank after The Gospel of Truth instead of just copying The Tripartite Tractate directly after it and leaving The Treatise on the Resurrection for the end. Considering the uncertainty that seems to have surrounded the construction of Codex I as a whole, it might have been thought safer to leave eight pages empty in the first quire instead of copying the very long The Tripartite Tractate and risking running out of space. It has been assumed that the order was important, but it could just as well have been a question of priority, that the copying of The Treatise on the Resurrection took precedence over copying the whole of The Tripartite Tractate. Leaving eight pages empty, the scribe made certain that The Treatise on the Resurrection would fit. It is also possible that Codex I was copied on several different occasions, which could explain the fluctuation in style, word count and size, as well as the multiple quires. Some have suggested that the owners of the Vorlagen of the different texts in Codex I might have been travellers who passed by only occasionally,²⁵ or perhaps Codex I was the result of a number of separate visits by the scribes to the place where the different Vorlagen were kept. After Scribe A finished copying texts 1–3 and 5, he could have given/sent the codex to Scribe B who had access to The Treatise on the Resurrection. There are multiple possible scenarios, and we will probably never get to the bottom of the precise circumstances behind the production of Codex I.

What conclusions can be drawn from the above findings, in regard to the context in which the Nag Hammadi codices were actually produced? The indications of carelessness by Scribe A, the fluctuating word count and page lines and the sometimes erratic style, do not seem to support the hypothesis that this was work done by a professional scribal team working on commission, as suggested by some,²⁶ because if the codex had been a commercial product, its cost would have depended on the quality of the material and the quality of the writing. Roger Bagnall categorises print quality along a range from calligraphy quality (the best) to documentary quality (the poorest).²⁷ The cost of a commercially produced book was dependent on the number of lines the scribe needed to copy, with a sum being agreed upon per copied line. Considering the line-length fluctuations of up to 18 per cent throughout the work done by Scribe A, who also produced a very mixed quality of writing, if Codex I had been a commercial product it would have been very difficult to calculate the price for such an asymmetrical work.²⁸ The observation that it was produced carelessly or with other intent than manufacturing a commercial product with the purpose of bringing in as much profit as possible fits better with the hypothesis that it was produced by the people who planned to use it.²⁹ As described in the introductory chapter, the latest suggestion about the texts’ background – advanced by Lundhaug and Jenott – is that they were part of a monastic book exchange network. As Lundhaug and Jenott have demonstrated, some of the codices were copied at the request of monks who wished to read texts to which they did not have access in their own library. It was not unusual for texts to be copied and sent to friends at their request, or for texts to be lent out to be copied by those who borrowed them;³⁰ however, Codex I was most probably not a book produced at the request of outsiders. Let us now turn finally to the reasons why the context of a book exchange network is not reflected in Codex I and why the likeliest scenario is that of a monastic setting.

Inexperience or Carelessness in Copying Codex I

The many untoward features of the codicology of Codex I suggest that it was produced by a scribe who did not place the production of a legible text to the fore. Inexperience or simple carelessness would explain the fluctuation of style and quality, fluctuating word count and line length, as well as the awkward codex construction. Yet we know that the setting in which the Nag Hammadi codices were generated had the experience to produce legible and well-structured texts, as evidenced by most of the other codices. The ample examples of homoioteleuton³¹ throughout the Nag Hammadi codices point to the fact that the scribes did not necessarily read the text, at least in a cognisant way, while copying; the main purpose was most likely to produce as close to a flawless product as possible. The scribe did not need to be cognisant of the content while copying, as Paul Saenger has argued. Scriptura continua was not a copying style that demanded a high level of comprehension by the scribe,³² as text devoid of punctuation and spaces between words is easier to copy, although more demanding to read. Transposition and other common scribal errors in texts copied after the textual revolution introduced spaces, and punctuation does not appear as often in texts composed in scriptura continua.³³ Meanwhile, both reading and copying scriptura continua was aided by introducing more spaces between each letter, rather than between words. Medieval manuscripts contain much less space between letters.³⁴ So, it is likely that most of the Nag Hammadi texts were chiefly copied by scribes who put their effort into manufacturing a text as close to flawless as possible.

Codex I is an exception. Scribe A did not place prime importance on the legibility of the texts, at least as far as The Tripartite Tractate is concerned, as indicated by the crowded pages in Quires II and III.³⁵ As it turns out, there are other indications in addition to the many errors in Codex I which indicate that it was not produced by a person focused on the copying task. As will be discussed in detail in the following chapter, there are scribal markings in the left margin of The Tripartite Tractate which highlight passages of interest. As these markings are parallel to the left margin, it appears that they can only have been made by the scribe himself while copying. A private person wealthy enough to produce a text of this magnitude would undoubtably have had a slave or professional scribe make the copy, expounding on the text only after its production. A monk, however, who was deemed spiritually mature enough, could have taken it upon himself to produce the text, even allowing himself the liberty to read, ponder and make notes in the text while copying. The Tripartite Tractate obviously piqued the interest of the scribe enough to distract him from meticulous copying to reflect on the text and insert notes. No other texts produced by the scribal team indicate that they were read and contemplated as they were copied, as Codex I seems to have been. In a monastic setting the production was left to those who wished to read and use the texts being copied, or their immediate superiors. Copying a book which contained potentially compromising material – like the extracanonical texts that Codex I could be viewed as containing – was laden with dangers to the spiritual integrity of the copyist.³⁶ However, as Palladius’ Lausiac History tells us, the ‘perfect’ monks who had reached sufficient spiritual maturity to keep them on the right path were permitted to read, write and copy whatever they saw fit.³⁷

Thus, it seems that Codex I was produced by a scribe without much scribal experience or one who did not see his main task as being that of producing a highly legible text; alternatively, he was carried away by what he read in The Tripartite Tractate. At the same time, the scribe obviously possessed enough theological and philosophical knowledge to be able to read and understand the text being copied to the point of making notes in it, and held a position within the scribal community which would allow him to read and comment upon potentially compromising material. These possibilities, therefore, suggest that this is a monk who was considered advanced in spiritual pursuits but without much experience in codex production who took it upon himself to copy a text of particular interest, rather than delegating the task to more conscientious scribes in the team, perhaps to protect their spiritual integrity.

Conclusion

The codicological facts surrounding the production of Codex I do not support the view that it was a commercial product, although it is hard to evaluate such an erratic work. If it were not a commercial product, nor one commissioned by a wealthy individual who kept scribes in his employment (which the notes in the margin in The Tripartite Tractate speak against), a monastic setting is the most convincing scenario in which texts with such intricate content might have been copied and commented upon at the same time. A monk without much scribal practice but with theological knowledge, who, rather than labouring mechanically, interacted with what was being copied, would explain the many unusual features of the final product. Meanwhile, it also safeguarded more experienced scribes without sufficient theological and philosophical training against being led astray.

All the Nag Hammadi codices except Codex I, produced by a scribe with little proficiency, followed the practice of codex production where the required number of sheets were measured and placed in a stack and folded in the middle. Thus, it is tempting to imagine that the advantages of using a multi-quire codex came to the knowledge of a group of experienced monastic scribes by way of sheer coincidence, by way of an inexperienced or neglectful monk (possibly due to distraction), who nevertheless managed to strike codicological gold. If it truly was distraction causing the inconsistencies, and not sheer inexperience or incompetence, what spiritual insights and theological topics could have been spellbinding enough to have caused the codicological neglect which we witness in Codex I? This is the topic of the next chapter.