Bryn Mawr Classical Review 2005.01.04
William A. Johnson, Bookrolls and Scribes in Oxyrhynchus. Toronto: University of Toronto Press, 2004. Pp. 371; pls. 18. ISBN 0-820-3734-8. $85.00.
Reviewed by S. A. Stephens, Stanford University (email@example.com)
Word count: 2013 words
This is a lavishly produced book on the subject of ancient book production. It launches the author's ambitious study of the ancient Greek book in roll form, which he calls "voluminology" to distinguish it from study of the book in codex form or codicology. In Bookrolls and Scribes we get the evidentiary base upon which he plans to build in future work (p. 4). The bulk of the study consists of the data, presented in charts and tables, in combination with Johnson's discussion of selection criteria, controls, analysis, and inferences. Its main interest, as he acknowledges, will be to specialists and to those wishing for an accurate database on which to test their own hypotheses about ancient book production. This book is not particularly easy to read, but for the persistent reader there is much to be learned about the earliest forms of Greek books to which we have access.
The book consists of a preface, introduction, and two chapters (pp. 3-230) + three appendices (pp. 231-339), glossary, bibliography, list of papyri, index, and plates. The preface on terminology, conventions, and sigla is essential for the reader to understand the charts, tables, and appended arguments, and the illustration of the anatomy of the book roll facing this preface provides visual reinforcement of the terms, though oddly no. 9 (the kollêma or papyrus sheet) was not labeled.
The Introduction sets out the rationale for the book, namely, that previous discussions of book rolls have been impressionistic or based on very small data sets. The author sets out to remedy this by presenting evidence from 317 papyrus rolls of known literature found in Oxyrhynchus and published in the volumes of the Oxyrhynchus Papyri (listed in Appendix 1A). He also creates a control group of 95 papyri1 from non-Oxyrhynchite sites ranging in date from the Ptolemaic (40) and Roman periods (55) (listed in Appendix 1B). 317 + 95 represent the largest sample of literary rolls from which Johnson could extract the required information by his own measurements and/or extrapolation: height of column, height of roll, width of column, width of intercolumnium, and leading (the vertical distance from the base of one line to the base of the line immediately below). By limiting the study to known literature Johnson has been able to reconstruct an entire roll from the surviving fragments, thus providing valuable information about average lengths of book rolls. To contextualize his main data sample: there are 1366 rolls of literary papyri currently listed in the Leuven database with an Oxyrhynchite provenance. Johnson's sample of 317 represents about 22% of the whole. 184 texts are of prose, 133 of poetry. His non-Oxyrhynchite control set, in contrast, consists of 29 prose texts and 66 of poetry.
Chapter one falls into two main parts: initially Johnson considers the evidence of multiple rolls written by the same scribe. He selects 21 scribes to whom at least 60 rolls have been assigned (for a full list see table 2.1). It should be stressed that these assignments are based on scholars' assessments of hands, not scribal self-identification, and are therefore to some degree subjective. Still Johnson is rather conservative in assigning rolls to a single scribe and his observations are likely to be correct. His surprising conclusion from these data is that scribes who copied texts in more than one prose genre, e.g., history and oratory, tended to use the same column widths across genres and that column width was remarkably uniform throughout the roll (that is, little to no variation in column width from beginning to end of the roll). Johnson infers that scribes used some sort of tool (a notched stick perhaps, p. 34) to mark column width (or more probably column width + intercolumnium). The tool was scribe or scribal group specific (whether we want to think of this as a scriptorium or as a group of associated scribes). The second half of the chapter considers how scribes copied. Using the evidence from scribal error when available, Johnson concludes that although he can point to some examples of line-by-line copying, in the main scribes did not copy in this way or even column by column, but in copying from an exemplar used their own preferred layouts. He also concludes, reserving fuller treatment for future work, that basic punctuation was part of the exemplar, though there is considerable evidence for readers' intervention in texts as well.
The second chapter takes up the formal characteristics of the roll now focusing on the visual impression made by the layout and the differences between prose and poetry. Here Johnson discusses the column tilt (usually to the right) that is found in rolls (so-called Maas' law), and concludes that it was a deliberate stylistic feature, not the result of copying posture or inattention. Johnson points to two kinds of evidence: first, in a select number of papyri a series of dots are to be found occasionally in margins. These he calls alignment or ruling dots and infers that their purpose was to "guarantee an even leading among lines," but also to "guarantee a left margin along a particular (sloped) line" (p. 98, and see pls. 9, 11, and 12). But these alignment dots are found infrequently, and in the Arden Hyperides (plates 16 and 17), they seem to perform a different function. In that text the dots seem to block out the writing area within a whole papyrus sheet. However, the best evidence that Maas' law results from design is its sheer persistence. For the measurable columns of the data set (192 rolls) 134 have observable right tilts and 34 have an upright or a slight right tilt (p. 91). Johnson concludes that this represents a different aesthetic from the codex, one that "creates a fitting impression of vigorous forward movement along the extent of the bookroll" (p. 93).
The second part of the chapter considers the preferred column widths and heights and lengths for prose genres, using a sample of 178 rolls of prose divided into history (76), oratory (60), and philosophy (42), and of 130 for poetry divided into epic (77), dramatic trimeters (37) and other (16). Johnson argues that contrary to received views, philosophical texts, not those of the orators, tend to favour narrower columns than other prose genres. He also observes that rolls of philosophy (at least at Oxyrhynchus) had an idiosyncratic punctuation and other common features that may have indicated a distinctive tradition or perhaps control over copying (one shop, p. 155). He demonstrates (as one would expect) that prose and poetry prefer somewhat different formats, with poetic texts on average having rather shorter columns. Ptolemaic papyri of both prose and poetry tend to be shorter and broader than Roman, and what he calls deluxe books tend to have shorter columns with generous margins. Rolls display considerable variation in length (Table 3.7). Johnson regards 3-15 meters a normative range, and thinks that there may have been a correlation between content and roll length, with short rolls for books of novels like Achilles Tatius and longer, heftier rolls for histories. But the evidence for novel formats is quite limited (3 papyri). The most interesting conclusion, however, is that book rolls of literature in general are remarkably consistent in their formatting. Rolls of commentaries and Aristotle's Constitution of Athens provide some control. These latter show very wide columns and an uneven range of column widths (see pl. 14).
In his discussion of literary rolls, Johnson sorts his sample into three types on the basis of writing: (1) formal, semi-formal, pretentious; (2) informal and unexceptional; and (3) substandard or cursive (p. 102). These are important for his aggregated data on roll heights, column width preferences, etc. Yet at no point does he provide plates to illustrate his types, and he gives only a few references to paleographic handbooks. He does indicate the style he assigns to each roll in his sample in Tables 3.1-3.7, but the assignments are difficult to verify. The tables do not indicate where a plate is to be found, so that in order to check his categories a reader is reduced to randomly checking against P. Oxy. or other cited publications or using the on-line databases. Does it make any difference to his analysis? It might, because assigning a writing style can be quite subjective, and Johnson defines his categories differently from Turner's in the standard discussion (p. 102).
The chapter ends with an attempt to integrate the specific data of this study into work of other scholars on books and book production. The author argues for scribal shops copying to order, from masters either owned by the shop or more often provided by the customer or a public library. He further argues that the data supports the professional training of scribes in general, whether they found themselves employed on a private estate or working in a public scribal shop. Professional training then guaranteed certain uniformity of practice.
The remainder of the book consists of appendices:
Appendix 1A (pp. 231-50) contains the list of all papyri used in the sample organized by P. Oxy. number, author, work, assigned date, and whether written on recto or verso of roll. Appendix 1B is the non-Oxyrhynchus sample, now organized by Merten-Pack numbers. Appendix 1C breaks the samples out into authors and genres. Appendix 2 (pp. 251-336) is a list of corrections to the papyri used in the data sets. Corrections range from punctuation omitted or misidentified; iota adscript omitted or misidentified; spaces in conjunction with punctuation, occasional corrections of readings, more accurate restorations and reconstructions. The purpose of the index is to provide an accurate record of all the texts used in the study. It is of marginal use for a general reader interested in book production, though valuable for someone editing or using specific texts within the data set. One very useful result of this appendix is its more accurate record of punctuation. Appendix 3 considers seven rolls that could be reconstructed in more than one way. Johnson uses the conclusions from his earlier discussions on average line lengths and roll heights to suggest which possibility is the correct one.
The plates are a curious combination of very helpful and perplexing. There are excellent plates of a Hyperides papyrus showing column tilt, kollesis, synkollesis, and ruling dots that seem to mark out the writing area of the papyrus sheet. Another plate shows the considerable variation in column widths found in Aristotle's Constitution of Athens. A few other plates show ruling dots that apparently mark leading or margin alignment (plates 9, 11, 12), though not in sufficient numbers to demonstrate the author's argument conclusively. Other plates are simply confusing. Plates 1 and 2 are of papyri supposedly copied by scribe A7, yet printed back to back so that comparison of hands is difficult. Plates 3, 4, and 5 are three different manuscripts by scribe A5 (Herodotus, Plato's Phaedrus and a commentary on the Odyssey). Herodotus and Plato are supposed to have the same column widths (p. 20), yet they do not appear to do so on the plates. Also, there is no indication in the list of plates of what each plate is supposed to illustrate or where it is discussed.
In principle this book is a very important advance in our understanding of book production in antiquity. Johnson's approach is refreshing and his insights a stimulus to thinking more concretely about the technology of writing and the interrelated practice of readers. Its success, however, must rest in large part on the accuracy of its data samples, many of which are recorded to tenths of centimeters. These are impossible for the reader to verify without access to Oxyrhynchus papyri. I would like to trust the author's data, but the book has a disquieting range of easily correctible errors in copying numbers.2 I have been inclined to discount these on the theory that they are random and insufficient to alter the substance of the arguments. If I were depending on these data for my own work, however, in most cases I would be inclined to re-measure or at least spot check.
1. Given as 96 in text, e.g., p. 9, 102 n. 24, but 95 in Appendix 1B and 1C. 95 seems to be correct.
2. E.g., p. 102, n. 27: Turner's examples should be 14, 42, and 49 (not 46); p. 153, Chart 3.8a, 2: 178 examples should be 29 examples, and (b) oratory, 11 examples shows only 10 examples; p. 160, 94 Beard 1981 (containing the response to Harris 1989) should be 1991.