This paper intends to draw attention to an interesting case of several fragments from what seems to be the same roll or hand interchangeably. I will argue that the obvious criteria used to establish the similarities and dissimilarities of a hand are relative, and that the criteria commonly used to identify two fragments as parts of the same manuscript (or to set them apart) should be used more cautiously when it comes to the identification of the scribe.[1] I will not attempt to produce a list of these criteria, but rather I will expose some of the problems that I encountered in my effort to determine the connections between a number of papyri and their place within the general formal round style. The classic studies on hand identification have set the methodological grounds.[2] However, the point on which all scholars agree is that, in view of the lack of external evidence and the relatively limited number of specimens available for such studies, the conclusions are undermined by a heavy reliance on assumptions.

    This case study of the formal round hand will present a total of six fragments, with at least four different inventory numbers, which are parts of at least three different rolls, by at least two different scribes. All of the fragments preserve works by Plato; five of them contain the Statesman and one contains Meno. Three of the Statesman's fragments were published as P.Oxy. XXVII 2468. The rest are to be published in vol. 76.

    I wish to draw attention to four obvious criteria typically used for the identification of hands:

    1. The way in which the letters are drawn
    2. The "feeling" or the general impression of the hand regardless of the elements that contribute to that impression
    3. The presence or absence of some optional elements (i.e. breathings, accents, lectional marks, punctuation, spelling conventions etc.)
    4. The format of the book.

    The inconsistency in the way certain letters are formed by a single scribe, often within the same manuscript, is well known and it is indeed used cautiously as an identification criterion. The danger in utilizing the formation of letters does not lie in using similarities to prove that a text is by the same scribe, but rather in using differences as proof that they are by two different ones.

    The overall feeling of a hand, its general character, is a deeply personal experience and it not properly defined in a scholarly manner, yet it is often encountered in palaeographical discussions. It is flexible, and therefore unreliable, as to what features one should focus on. The overall feeling depends, among other things, on the size of the letters and the spacing between them, the regularity, formality, bilinearity, fluency, decoration, thickness of lines and shading, whether it is upright or slanting, whether strokes are straight or curvy, and when of curves whether they are properly round or oval. Many of these terms are themselves vague and relative. Each of these factors is used separately as criteria for hand identification. But some of them contribute so drastically to the general impression that they would by themselves suffice to alter the "feeling" of a handwriting. Depending on the hand and other circumstances, some features alter more drastically than others the feeling, e.g. size and spacing. On the other hand, a feature that seems to influence the feeling gravely in one style is negligible in another, e.g. decoration and shading.

    Identification criteria can be divided into deliberate elements and elements applied subconsciously. This distinction is relevant to the intention of a scribe to maintain consistency throughout the roll or codex and it can also assist in revealing a scribe's habits (by identified inconsistencies within one roll and later using them to identify other books by the same scribe).[3] Those criteria perceived as deliberate are useful in establishing provenance from the same roll due to a professional scribe's effort for consistency within a single work. Obviously, the same scribe may have used different conventions in different books and the absence of such elements in two manuscripts does not exclude a common scribe. At the same time, the presence of the same "subconscious" characteristics in two fragments does not necessarily mean that they come from the same book.

    The third and fourth criteria noted above are of the deliberate type. They can easily be applied on demand and it seems relatively simple for a scribe to keep up with the relevant guidelines without being betrayed by habit.[4]

    In identifying fragments from the same roll, uniformity in the non-deliberate elements is a prerequisite, but deliberate ones also play a decisive role, as consistency is expected throughout the volume. This only applies to professional books,[5] which is the case with all the papyri treated here. Discrepancies in uniformity do not necessarily point to different rolls, as the tiredness of the scribe could account for that.[6]

    In identifying fragments by the same scribe and not necessarily from the same roll, a level of consistency in the non-deliberate elements is again a prerequisite. Johnson's study suggests that a scribe carries on with the same format from one book to another and that format could be used for the identification of a scribe, but he admits that the number of papyri involved in this study is not sufficient. Although such a conclusion is useful in strengthening the identification of a scribe, I do not find it compelling.[7] There is no decisive evidence that deliberate elements such as punctuation, formatting and formality are characteristics of the scribe rather than the book.[8]

    The three papyri examined in this paper (P.Oxy. XXVII 2468 + LXXVI 5106, LXXVI 5105, and 5107) are all of similar format. Of these, 2468+5106 and 5107 are similar in format, but not by the same scribe. 5107 and 5105, which are different in format, were possibly written by the same scribe.

    The three fragments of 2468 are now in the British Museum (all the rest are in Oxford). Their initial inventory number is not recorded and there is no archaeological evidence as to their physical proximity to the other fragments during the excavation. Within 2468, fragments 1 and 3 were calculated to be some 11–13 columns away from fr. 2, which was thought to belong to the first column of the roll.[9] This was proved with the discovery of another fragment preserving more of the first column (5106), together with a large left margin indicating the beginning of the roll and parts of what was obviously column ii. The hand is very similar to that of 2468. What proves beyond any doubt that this is the same roll as 2468 is that fr. 2 of 2468 fits exactly into the loop formed by the two pieces of the newly found papyrus (these two pieces are connected to form a single fragment). The editor of 2468 did not deem it necessary to prove that the three fragments originate from the same roll, perhaps because he had at his disposal a common inventory number for all three of them, now lost.

    The last fragment from Plato's Statesman is 5107, which I believe is the only one of the Statesman papyri that can be possibly identified with Johnson's scribe A1. The editor of 2468 compared the hand with 844, the only well-known A1 hand, but did not assign them to the same scribe.

    5107 was initially thought to be the same manuscript as 5106 due to the similarity of the two hands and the fact that they both preserve the same work, with what seemed to be very similar formatting and similar quality of papyrus. The immediately obvious dissimilarities were thought to be due to the difference in the sharpness of the pen or the copyist's tiredness. There was a further problem in the number of lines per column, indicating that they are not part of the same roll. The different inventory numbers also suggest that they originate from a different roll, although one of them gave no hints as to the time and place of its finding.

    5105 preserves a portion from Plato's Meno and is very similar to 5107. In both papyri, all letters are drawn in the same way and the decoration is similar. The letters in 5105 are larger and the formatting seems different.[10] Although letter and line spacing is more liberal in 5105, the ratio of spacing to the size of letters is the same in the two papyri. Very little evidence is offered in terms of punctuation and lectional marks.

    The general impression of 5107 is immediately different from 5106+2468: the handwriting is more disciplined, and the letters could be inscribed in a square more regularly than in 5106+2468. The pen is obviously sharper in 5106+2468, a feature distinctive at first sight. The decoration is with blobs at the tips of the letters and not with serifs as in the other two papyri. The strokes in 5107 are straighter as opposed to some curviness in 5106+2468. This creates a feeling for 5107 obviously different from that of 2468 and 5106, always within the context of a formal round hand.

    The scribe's tiredness could be the explanation for the less obvious differences between what is thought to be col. i of 2468 (that is fr. 2 of 2468 plus 5106) and columns xii–xiv. Fragment 2 was not large enough a sample to compare, but with the discovery of 5106, the task has become easier. Comparing 5106 with 2468, the general impression of the hand is undoubtedly very similar, with the first column being more compressed both in the shape of letters and the space between them, perhaps due to the physical condition of the papyrus. Decorations are with serifs and finials in both fragments, the pen is equally sharp, the distance of the letters does not allow them into squares, and some curviness in the strokes is equally present. On the other hand, coll. xii–xiv are obviously less carefully executed. The writing is occasionally cursive, with some letters connecting to each other and of variable size. Decoration is more occasional, but of the same character.

    If we are to look at the formation of the letters as an important criterion for the identification of a hand, then perhaps we should consider the possibility that the five fragments come from at least three different rolls. There is little doubt that 2468 fr.2 comes together with 5106, but for the rest of 2468, further examination is needed.

    Phi is diamond shaped while in 5106 it is round. Upsilon in 2468 coll. xii–xiv is formed in one movement with a loop or serif at the bottom, whereas in 5106 it is in two movements, with the right diagonal continuing downwards to form the stem. Sigma in 2468 has a peculiar upper right part with a decorative serif. In 5106 the serif is vertical to the main stroke and at the bottom right. Pi is also different, made in three strokes in 2468, with the upper part extending beyond the two verticals. Kappa is again in three movements with the upper diagonal touching the lower one instead of the back. Mu and omega have their upper parts decorated with serifs continuing from the vertical strokes.

    Comparing both with 5107, the latter aligns with 5106 in phi, sigma and omega, whereas its kappa and mu are more similar to those of 2468. The v-shaped upsilon looks more like that of 2468. Theta is different than that of 2468, but we do not have one from 5106 to compare. Pi is of both types.

    The text transmitted by all fragments is unremarkable, not pointing particularly to any direction of the manuscript tradition and being of no use in establishing connections between them. In terms of lectional signs and punctuation, 2468 has high and low stops, two dots to indicate the change of speaker, filler marks, and two examples of the diple. 5106 does not contain low stops, only middle ones. Moreover, it does not have the diple, instead it has the diple obelismene. In 5107, on the other hand, there are only high points. The absence of punctuation in such short portions of text, however, is not a reliable factor.

    The shading due to the position of the pen is one of the non-deliberate characteristics when it is not applied for decorative purposes, which is the case with all our fragments. Formatting is very similar in all three. The difference in width between 5106 and 2468 is insignificant. This is not unusual in the progression of a roll, however. The size of the letters is comparable in all three manuscripts.

    Thus, 2468 and 5106 have a very similar "feeling" against 5107, but 5106 and 5107 have more letter shapes in common with each other than any of them do with 2468. If we accept that 5106 and all three fragments of 2468 are similar enough to be by the same scribe, does it necessarily follow that they come from the same manuscript as well? Accordingly, although 5106 and 5107 are not parts of the same manuscript,[11] and probably not even by the same scribe, could their similarities indicate any relationship between them? Could it be that one is the apograph of the other, or originate from the same scriptorium, or that they share an exemplar, or that 5107 was modelled after 5106? Unfortunately, inventory numbers and archaeological records are of no help in this case. In addition, in most parts, the fragments are short, or the quality of the text is of no particular interest.

    The most obvious case of deliberate scribal practice is formatting and decoration (always within a general style). In our example, decoration carries the weight of proof by itself in determining that 2468 (together with 5106) and 5107 come from different rolls.

    2468 and 5106 are not as similar as they initially seem, although they are proven to be part of the same roll. At the same time, both are quite similar to 5107, which almost certainly comes from a different roll. In other words, changes in handwriting in the course of copying a single roll can be more easily noticeable than differences between two different hands. It is the type of differences or similarities, and the combination of characteristics that identify or tell apart two specimens.[12] Lastly, 5105 is probably not part of the same roll as either 2468+5106 or 5107.[13]

    If decoration is decisive evidence for two fragments coming from the same roll, it is also responsible for setting apart the hands. If we remove the decoration from 5107, then the general feeling is not that different any more.[14] The illusion of the thick lines created by the blobs disappears and the letters look smaller, slimmer, and perhaps even less square. I believe that if we could do the same for the other fragments – I found it to be very difficult due to finials being parts of the actual strokes – the result would be even closer. Of course, several differences remain. However, the possibility that 5106 and 5107 are by the same scribe, no longer seems that remote. This does not apply to the possibility of dealing with the same roll. The different decoration, a deliberate element, would not make sense in a professional book. The punctuation, also a deliberate element, seems different as well.

    A similar situation creates the illusion that 5107 and 5105 are not as similar as they actually are. Size of letters and letter spacing in 5105 is more generous than in 5107. The general impression is considerably different between 5107 and 5105, but this changes dramatically if we electronically balance the size difference between the two. It then becomes conceivable that 5107 and 5105 are by the same scribe.

    Having established that we have three different books (5106+2468, 5107 and 5105) and three (perhaps only two) scribes, the difficulty lies in placing them within the history of the formal round style.[15]

    The editor of 2468 dates the papyrus in the second century, but Prof. Cavallo considers it an earlier, not fully developed type of the "Roman Uncial" and places it in the first century.[16] This chronologically linear placement of the different types of the formal round and the distinction between early, transitional and mature hands should not be dismissed, but I feel it is not soundly supported. 2468 is Cavallo's example of an early, not fully developed "Roman Uncial" hand.[17] The argument for dating it in a period earlier than other fully developed hands (perhaps like our 5105) is its lack of canonization, the closed cup epsilon which is not completely round, the eta with the high middle stroke and the alpha with the vertical or oblique middle line. However, this lack of canonization rather seems to be a failure to comply with the standards of the most elegant examples and the most numerous representatives of the formal round hand that have come down to us. Many second century papyri feature the closed cup epsilon and eta (e.g. P.Oxy. III 454). In 2468 eta is found with both upper and middle height middle stroke.

    If the papyri studied here had to be placed linearly, the order would probably be 2468+5106, then 5107 and then 5105. With the characteristics mentioned in the previous paragraph scattered among the three manuscripts, they present an ascending level of formality and regularity, rather than of canonization. What could be perceived as progression in time, could also be the evolution of a single scribe's hand throughout his career or different degrees of formality in his writing. It was common in medieval manuscripts that a certain known scribe would produce two significantly different books (a well-known example of the scholar Triclinius or the scribe Nicholas the Stoudite, who had his formal and his informal hand). We also know of scribes from Oxyrhynchus (not necessarily professional ones) who were trying their hand in different styles and types of the formal round script.[18]

    This is where patronage could also come into the picture. If a copy was paid for, for which there is good evidence, then it is conceivable that the person ordering it had a say on the form, effect, style and format, possibly even providing the exemplar himself. Prof. Parsons vividly describes an imaginary transaction of this kind.[19]

    The chronological arrangement of these features of the formal round style is a possibility authorised by the experience of Prof. Cavallo and backed by some editors of relevant papyri. However, the very limited number of objectively dated specimens offers no solid evidence as to the internal relationships of the formal round style. One should also keep in mind that the features of a style against which we compare a hand are modern conventions, not necessarily corresponding to the effect that a scribe was aiming for. Thus, what we perceive as "fully developed" style is not necessarily a reliable starting point for classifying early and late examples.

    As in many palaeographical studies, the most important remark to be made is that ancient scribes and their profession are notoriously poorly documented, and a great many of our conclusions are therefore based on assumptions. Turner notes: "[the object of the palaeographical analysis of a text] is not so much to discover whether the indications they offer about the writers of literary manuscripts, their place in society, level of education, literary taste and manner of work support the picture already elicited from the documents and the archaeological remains, it is to frame a body of critical principles according to which texts may be assessed when there is no other evidence available than the palaeographical data they themselves provide. The task is full of pitfalls."[20]

    Plate I: P.Oxy. LXXVI 5106
    Plate I
    P.Oxy. LXXVI 5106
    Plate II: P.Oxy. LXXVI 5107
    Plate II
    P.Oxy. LXXVI 5107
    Plate III: P.Oxy. LXXVI 5107 (decorations removed)
    Plate III
    P.Oxy. LXXVI 5107 (decorations removed)
    Plate IV: P.Oxy. LXXVI 5105
    Plate IV
    P.Oxy. LXXVI 5105


      1. See P.J. Parsons, "The Scripts and their Date," in E. Tov, The Greek Minor Prophets Scroll from Nanal Hever (8 Hev XII gr) (Oxford 1990) 19–26. return to text

      2. First set out by E.G. Turner, "Scribes and Scholars of Oxyrhynchus," in Akten des VIII. Internationalen Kongresses für Papyrologie, Wien 1955 (Vienna 1956) 141–146, reprinted in A. Bowman et al. (eds.), Oxyrhynchus: A City and its Texts (Oxford 2007) 256–261; W.A. Johnson, Bookrolls and Scribes in Oxyrhynchus (Toronto 2004).return to text

      3. P.J. Parsons, "Copyists of Oxyrhynchus," in Bowman et al. (eds.), op.cit. (above, n. 2) 265.return to text

      4. This seems more difficult for the way the letters are drawn. The scribe is careful at the beginning, but gradually loses concentration and starts to revert occasionally to the way he usually draws the letter, or perhaps to his practice in the last manuscript that he copied. return to text

      5. As defined in Parsons, op.cit. (above, n. 3) 263.return to text

      6. For instance, there is evidence of format changes in the middle of a roll: Johnson, op.cit. (above, n. 2) 37–39.return to text

      7. Ibid., 57. It is also a circular argument, since similar formatting is used as a criterion for scribe identification. Parsons numerous questions-hypotheses about scribal practices and scriptoria are appropriate in this case; Parsons, op.cit. (above, n. 3) 265–266. return to text

      8. Dr. Colomo brought to my attention the fact that Johnson's scribes show different practice in punctuation and accentuation from one roll to another (e.g. A5, A28 and B4). return to text

      9. Of fr. 2 in P.Oxy. 2468, pl. IV.return to text

      10. Only a small part of 5105 survives, which shows that the columns are wider than in 5107 by one cm. We cannot compare their column height, because that of 5107 cannot be determined. return to text

      11. See introduction to P.Oxy. LXXVI 5107.return to text

      12. An approach commonly found in modern forensic science, see R.A. Huber and A.M. Headrick, Handwriting Identification: Facts and Fundamentals (1999) 46: "There are a number of more commonly encountered elements of writing that may be described as the designs, inventions, and developments of the writer that, when considered in combination as a group, give to a writing a uniqueness. In this sense, it is the composition of the combination that is responsible for the individuality the hand writing acquires."return to text

      13. See introduction to P.Oxy. LXXVI 5107.return to text

      14. With the right amount of image enhancement it is possible to get a rough idea of how the hand would look without the decoration (Plate III). In this plate, blobs and serifs were erased in order to demonstrate how decoration affects the overall feeling of a manuscript. Of course, such a task cannot be performed with absolute accuracy; in this case I opted for leaving remains of the decoration rather than erasing part of the body of the letters.return to text

      15. See introduction to P.Oxy. LXXVI 5107.return to text

      16. G. Cavallo, "Osservazioni paleografiche sul canone e la cronologia della cosidetta 'onciale romana'," in G. Cavallo, Il calamo e il papiro. La scrittura greca dall' età ellenistica ai primi secoli di Bisanzio. Pap.Flor. XXXVI (Florence 2005) 154 (originally published in ANSP II, 36 [1967] 209–220). return to text

      17. G. Menci, "Scritture greche librarie con apici ornamentali (III a.C. – II d.C.)," S&C 3 (1979) 23–53. Menci's demonstration of a date-related use of decorations is not particularly helpful for the dating of our papyri, because she accepts Cavallo's dating for 2468 and uses it as evidence towards the chronological distribution of decorations.return to text

      18. See examples in Parsons, op.cit. (above, n. 3) 269–270.return to text

      19. Ibid., 266. return to text

      20. E.G. Turner, Greek Papyri: An Introduction (Oxford 19802) 88.return to text