The Napier fragment

In 1887 A. S. Napier published a semi-diplomatic transcript of a fragment from an Old English manuscript, which he reported finding the year before as a flyleaf at the end of Bodleian Library, Oxford, MS Junius 86 (Napier 1887). Medieval and Renaissance binders sometimes used the leaves of older manuscripts, when they had lost their interest, at the front and back of bindings to reinforce them and protect their contents (Ker 1957, xli). Junius 86 is the second part of an eleventh-century collection of, for the most part, Old English homilies.

For a description of the composite Junius 85+Junius 86, see Ker 1957, art. 336.

Unfortunately, the fragment was removed and temporarily mislaid soon after its discovery in 1886, and no one other than Napier has ever been able to examine it (Sedgefield 1899, xvi.). For the inaugural issue of the it is de rigueur to use digital tools for analyzing virtual manuscripts. For this job I will be wielding several digital tools originally developed for the () in order to create and analyze a virtual facsimile of the lost manuscript leaf Napier describes and transcribes.

The Old English , also known as Alfred's or (perhaps more accurately) the Alfredian , is a translation of Boethius's , one of the foundational texts of Alfred the Great's (871-899) cultural and educational reform at the end of the ninth century. Before Napier's discovery, there were only two known surviving manuscripts in two distinct versions, an early to mid tenth-century prose-and-verse one in British Library, MS Cotton Otho A. vi,

N.R. Ker dates s. x med (Ker 1957), while Humfrey Wanley assigns it to the time of Alfred or slightly later (Wanley 1705/1970).

and a twelfth-century all-prose one in Bodleian Library, Oxford, MS Bodley 180. The first part of the tenth-century manuscript, including a prose-and-verse preface, was totally destroyed in the Cottonian Library fire in 1731. Francis Junius, the great seventeenth-century Anglo-Saxon scholar for whom the Junius manuscripts are named, had fortunately made a collation of his own transcript of Bodley 180 with the Cotton manuscript before the fire. Even more providentially, Junius had copied in full all of the verse sections into what is now Bodleian Library, Oxford, MS Junius 12. His transcripts of the destroyed meters are our only source for them today.

We know from Junius 12 alone that the opening folios of the Cotton manuscript contained two prefaces that declared, in both prose and verse, that King Alfred first translated Boethius literally into prose and then geworhte hi eft to leoðe, swa swa heo nu gedon is, reworked it for verse, just as it is done here, that is, in a prose-and-verse manuscript.

The verse preface emphasizes Alfred's interest in poetry: Ðus Ælfred us ealdspell reahte, / cyning Westsexna, cræft meldode, / leoðwyrhta list.... Ic sceal giet sprecan, / fon on fitte, folccuðne ræd, Thus Alfred told us an old story, King of West Saxons showed off his craft, skill of verse-making. ... I yet must speak, fashion in fitts folk-shared philosophy.... (Krapp 1932, 153, my trans.; see Kiernan 1998, 11).

The later twelfth-century manuscript does not include any verse, even though its prose preface incongruously claims, too, that Alfred reworked it for verse, just as it is done here. Scholars since the seventeenth century have interpreted the prose preface to mean that the entirely prose version in the twelfth century manuscript is actually "older" than the tenth-century version. They have seen it, in other words, as a copy of the original prose draft, before Alfred reworked the Boethian meters from a literal prose translation to a verse translation.

Griffiths 1994 repeatedly refers to the prose draft (e.g. 41); cf. Sisam 1953, 294-295. Malcolm Godden has told me that he assumes that the all-prose version was a completed text intended for circulation and use.

It seems highly unlikely that a draft, which we know from the two prefaces that Alfred revised and replaced, would none the less survive a transmission of more than two centuries, while retaining a nonsensical comment that the all-prose draft contained verse. A more likely possibility, it seems to me, is that the twelfth-century version is a later revision that replaced the Alfredian verse passages, after they had lost their appeal as poetry, with prose paraphrases of them.

Godden has persuasively argued that there was an all-prose version in circulation a century after the death of Alfred, because Ælfric apparently draws on it briefly in (Godden 1985, 296-298); Bolton 1972 and Griffiths 1994 give other examples, while Griffiths suggests that Ælfric's use may also show familiarity with Metre 31 in its verse form (43-44).

To judge by his calculations on how much was missing, Napier assumed his fragment came from an all-prose, early tenth-century, manuscript. His estimate of lines missing between the fragments is based on Bodley 180, and his restorations of lost text within the fragments come from the early nineteenth-century editions of the prose manuscript by J.S. Cardale and Samuel Fox (Cardale 1829; Fox 1835). In his 1899 edition Sedgefield endorses Napier's assumption in his manuscript stemma, which describes the fragment as prose only? (Sedgefield 1899, xix).

Reproduction of stemma in Sedgefield 1899, xvii

Likewise, George Philip Krapp, who edited the verse meters from Cotton separately in what has come to be known as , describes the Napier fragment as a parchment leaf [sic] of the first half of the tenth century, containing part of the Anglo-Saxon prose version... (Krapp 1932, xxxv). Ker implies that it came from an early prose manuscript, too, for he includes the fragment under prose binding fragments of great interest in his introduction (Ker 1957, lxi), and bases his own description on Napier's without comment or modification (art. 337). No one to my knowledge has ever questioned Napier's hypothesis.

Despite these uncritical endorsements, the missing text between the front and back of the fragment contains one of the Boethian meters. The first side of the fragment ends just before the start of the eighth metrical passage, while the second side begins immediately after the words, Þa se Wisdom þa þis leoð asungen hæfde, þa ongan he eft spellian (when Wisdom had sung this verse, he then began to speak again). The fact that the missing part is one of the meters leaves open the possibility that the fragment came from another prosimetrical manuscript, like the extant tenth-century one, rather than from an earlier all-prose manuscript of Alfred's original draft. Napier could not easily examine this possibility, because at the time both modern editions were based on Bodley 180, not Cotton Otho A. vi.

Even if he considered the possibility, Napier had no easy way to gauge the number of manuscript lines the verse would take up between the recto and verso of his fragment. The situation was not much improved with Sedgefield's edition. Sedgefield theoretically based his text on the Cotton manuscript (see his preface, vii), but in practice he gave extraordinary authority to Bodley 180. In fact, he went so far as to extract the Cotton meters from his Cotton text and tack them on at the end of his edition in a section called (151-204), preceding an appendix for the Napier (N) fragment (205-206). As Kenneth Sisam dryly remarks, The arrangement of C is not easy to visualize from Sedgefield's edition.... (Sisam 1953, 294, n. 2).

Edition Production Technology EPT

The digital tools I will use to try to solve the general Napier problem are part of a comprehensive Edition Production Technology EPT workbench, powered by the Eclipse programming environment (see Kiernan et al. 2004 for details). The EPT's StaTend tool can quickly help to visualize the arrangement of both the Cotton and Bodley versions as they relate to the Napier fragment. Designed for the project, the tool computes simple statistics for folios (number of lines, characters per lines, spaces per line) and helps an editor use Junius's transcript to reconstruct virtual folios from this data for lost or damaged parts of a manuscript. It has general application, however, and adapts especially well to the Napier problem, because the Cotton and Bodley manuscripts supply variant texts that come between the two sides of the fragment. The tool first computes the number of characters and spaces per line in the fragment, following Napier's line boundaries:

Napier's line boundaries with statistics

The results show that the 16 lines of the fragment, recto and verso, average about 75 characters per line, including spaces. Using this data, the tool can easily provide two simulated reconstructions of the full folio, verso, one for a prose meter and one for a verse meter. StaTend reveals that the lost part of the page, between the end of the first side and the beginning of the second side of the fragment, presumably also held an average of about 75 characters (including spaces) per line. A simple query interface allows the editor to determine how many lines the full leaf held if all lines averaged 75 characters per line:

Napier fragment in an all-prose manuscript

Basing his estimates on an all-prose , Napier estimated that the original leaf held approximately 38 lines, which both Sedgefield and Ker accepted. The StaTend figures show, however, that the original leaf must have had at least 42 lines. Moreover, if its formatting was like Bodley 180, fols. 19v-20r, the leaf would have had even more lines, because of the two sets of large capitals that announce the new chapters, XV and XVI, preceding the fragment.

The Cotton verse meter is somewhat longer than the prose version. Following the same procedure, the StaTend tool reveals that if the fragment came from a prose-and-verse manuscript, the folio would have taken up at least 49 lines. If its formatting was like the Cotton manuscript, fols. 20r-21v, it too would be somewhat longer to accommodate the large capitals at the beginning of the verse and prose.

Napier fragment in a prose-and-verse manuscript

These reconstructed virtual folios illustrate that there is no evidence to support Napier's and Sedgefield's assumption that the fragment came from an all-prose manuscript. There is simply no way of knowing whether the Boethian meter that came before the fragment in the original folio was prose or verse. Both versions would fit in a small folio volume. It seems that the only argument in favor of the meter being in prose is the circular one that it would then support the hypothesis that the twelfth-century prose manuscript is more authoritative than the prosimetrical one coming so close to the time of King Alfred himself.

Wanley believed that the manuscript was written in Alfred's lifetime or shortly after his death (Wanley 1705/1970, 217).

The one shred of evidence that the Napier fragment came from a prosimetrical version is that it shares one reading, gearod (past participle of gearwian, to make ready, prepare, equip), with the Cotton manuscript, where Bodley has gegyrewod (past participle of gierwan / gyrian, with the same meaning).

Sisam notes that Sedgefield reads gegyrewod with B, where C and the still earlier Napier fragment have the correct gearod (Sisam 1953, 294 n. 2).

While preparing his edition, Sedgefield naturally sought to study the manuscript itself of the fragment. After all, its paleographical features, as described by Napier, had made the missing fragment the earliest surviving witness in his stemma. It was presumably a shocking disappointment to learn that Napier had already lost it. Some years ago, Sedgefield reports in his introduction, the Napier fragment was taken out and bound separately, but it has since been temporarily mislaid, so that the present editor has not been able to see it (Sedgefield 1899, xvi). No one has found it after more than a century of searches. Ker, an expert sleuth, implies that he made a thorough search himself. Clearly he had no hope for its recovery. It was mislaid before the publication of Sedgefield's in 1899, he says, and is now not to be found (411).

Somewhat lamely thanking Napier in his Preface for supplying me with some valuable notes on several points (Sedgefield 1899, ix), Sedgefield was obliged to represent Napier's important discovery at second-hand. Sedgefield gives a word-for-word translation of Napier's introduction:

Professor Napier says—the following fragment of the Alfredian translation of Boethius in a hand of the first half of the tenth century forms the last leaf of MS. Bodl. 86.[

Here Napier correctly has hs. Junius 86 der Bodleiana (Napier 1887, 52).

] This leaf, which evidently has been used previously in the binding, was placed in its present position by the binder, and originally belonged to a small folio Boethius manuscript. The fragment formed the lower half of a leaf, and judging by the part missing between the two sides each page must have contained about thirty-eight lines. The writing is in parts very indistinct, as the letters are frequently blurred; the parchment is also perforated here and there, so that some letters are quite gone.

Sedgefield for some reason omits Napier's comment that he supplied in brackets text he could not clearly read from Bodley 180, and that he marked line boundaries with upright strokes.

Sedgefield also omits Napier's inclusion of restored readings from Cardale 1829 and Fox 1835 in his appendix.

Sedgefield adds that the words in Professor Napier's transcript, which accurately represents the manuscript, are much run together, and no capitals are used, even though there are in fact nine capitals (A, N, N, W, S, F, S, I, F) in the transcript. These comments about the accuracy of the transcript and the retraction of the suspiciously freqent use of capitals are perhaps among the valuable notes Napier personally supplied Sedgefield, in lieu of the fragment. In any case, Sedgefield correctly observes that contractions and accents are relatively frequent, and concludes with the crucially important observation that the vowels a and o, following h, m, and n, are in some cases formed by a looped prolongation of the last stroke of these consonants below the line, a characteristic of the age of the fragment (Sedgefield 1899, xv-xvi). On the basis of this dating criterion, Sedgefield accepted the lost Napier fragment as the earliest witness in his account of the transmission of the text.

Napier's printed representation of the fragment in 1887 attempts in several ways to disclose the technical means of production of a medieval manuscript. First, his transcript uses the abbreviations (the crossed thorn for þæt, that; the macron over various vowels to represent abbreviated spellings) and copies the scribal word divisions of the manuscript, combining for example three words in one in the sequence hwonanhisien (i.e., hwonan hi sien, from where they might be) in line 6r. Second, the transcript indicates line boundaries by upright bars. Third, it conveys the damaged state of the manuscript by the lacunas, filled with readings from the twelfth-century prose manuscript via Cardale and Fox. Fourth, it records a scribal correction by superscript (swa in line 8v). And fifth, it presents paleographical information and a rough dating criterion by showing the existence of subscript a and o by means of italics. Two of the tools in the EPT workbench, the xTagger and the xMarkup, can exploit Napier's explicit markup to provide richer access to the information.

The EPT's xMarkup tool allows a researcher to encode a text for a wide range of purposes under the broad headings (each with its own set of elements and attributes) Start Transcript, Codicology, Paleography, Condition, Restoration, Collation, and Edition. One of the first stages of encoding an image-based electronic edition is to tag folios and folio lines to structure the transcript and facilitate searches and other forms of interchange.

xMarkup templates for tagging folios and folio lines

Even without the leaf, Napier's markup makes it possible and desirable to provide encoding for the fragment, recto and verso; the folio lines; scribal punctuation; capitals; accented letters; scribal corrections; implicit natural word boundaries; lacunas caused by damage; and the abbreviations (including ampersand, crossed thorn, macrons for suspensions, and subscripts).

The EPT's xTagger tool is able to manage all of the potentially conflicting markup and display it under similar broad hierarchical headings. The researcher can show or hide any aspect of the markup during and after encoding, as desired.

xTagger Show Markup for folio lines

The xMarkup template for abbreviations under Paleography helps the editor encode the subscripts, which like all abbreviations are used to save space.

xMarkup template for encoding subscripts

As with the markup for folios and folio lines, the xTagger can show or hide the XML encoding that enables searches, transformations for presentation, and other computer generated operations, in this case for the subscript letters that so clearly dated the fragment in the absence of the actual manuscript.

xTagger Show Markup for subscripts

It may well have seemed at the end of the nineteenth century that the subscript letters Napier and Sedgefield used to date the fragment were a relatively common feature of early tenth-century manuscripts. However, we now know that the use of subscripts in surviving Anglo-Saxon manuscripts was extremely rare. With a comprehensive knowledge of over 400 surviving manuscripts containing Old English, Ker in his exhaustive has identified only two other Old English manuscripts that use these subscripts. In his discussion of ligatures in the introduction, Ker says in the context of rare ligatures appearing as relics of earlier practices that h, m, n are combined occasionally with following a, i, o, the a, i, o being subscript, in 39 [Corpus Christi College, Cambridge, MS 173, ] hand 2, 133 [British Library MS Additional 47967, ] and, apparently, 337 [the Napier fragment] (Ker 1957, xxxiii).

Napier would have known that the early hands of the Parker chronicle were dated in or near the time of Alfred, so it is surprising that he did not cite them to explain the dating significance of the subscripts. The Parker manuscript was used as a touchstone for dating scripts, because the chronicle entries roughly coincide with changes in handwriting between the late ninth to the mid tenth century. Ker explains the development of the first two scribal stints by describing the first scribe's work on the annals from the year 1 to 891 as an upright hand of s. ix/x, intermediate in character between the pointed minuscule of s. ix and the larger squarer script of s. x (Ker 1957, art. 58). Sedgefield was thus on safe ground to date the Napier fragment, no doubt with Napier's encouragement, to the early tenth century.

What Napier could not have known is that expert paleographers after the nineteenth closely tie together the handwriting of the Parker chronicle and the Tollemache Orosius and assign both manuscripts to the same scriptorium. For the Tollemache Orosius, Ker says that the square Anglo-Saxon minuscule is throughout probably in one hand contemporary with and from the same scriptorium as the hand (or hands) of the annals for 892-924 in the Parker chronicle (Ker 1957, art. 165). In his description of the Parker manuscript for the same section (fols. 16v-25v) Ker says that the hand varies a good deal in appearance like the closely similar and possibly identical hand of Orosius (58). Ker's linking of these two manuscripts through a common scribe is corroborated and further developed by Malcolm Parkes. In Parkes argues that, by contrast with other Anglo-Saxon manuscripts of this period, this particular group of manuscripts [including Parker and Orosius] have a common characteristic: the palaeographical features of the different manuscripts in the group reflect the various stages of a particular pattern of evolution[,] and conformity to this pattern forms the basis for attributing them all to a single scriptorium (Parkes 1976, 158).

Parkes explains that the way in which the incidence of certain forms—in particular half-uncial a and s, cursive underslung l, cursive ligatures, and subscript letters—gradually diminishes, even within the handwriting of an individual scribe, suggests that the process of standardization was achieved slowly and entailed the elimination of variants (Parkes 1976, 158). Parkes specifically uses the subscripts of the Parker manuscript as an illustration. In the thirty-one pages of [scribe 1's] stint in the first booklet there are only ... three [instances] of subscript letters (159).

While it does not affect Parkes's essential point, I have noted seven examples in the first scribe's stint: miercna 12b32, his 12b34, suna 13a31, seaxna 13a32, mierce 13b27, monna 15a14, and sona 15b3.

He further observes that in the part copied by scribe 2, the same person who copied the , there are only three instances of subscript letters, each of which occurs at the end of a line to save space (159). All three are examples of mo (i.e. m with subscript o). In the Parker manuscript the scribes' use of subscripts averages one about every 165 lines. While neither Parkes nor I have made a tally of the use of subscripts in , the usage is not heavy. There is a single instance at the end of line 16 of the frontispiece to Janet Bately's edition of . In his facsimile of the Tollemache Orosius, Alistair Campbell mentions the practice of suspending a, o, and i to the final stroke of m and n and gives some widely dispersed examples. He also mentions the very rare subscript e. In stark contrast to the diminishing use of subscripts in the Parker and Orosius manuscripts, the scribe of the fragment Napier found used the widest variety of subscripts (na, mo, ma, ha) 16 times in 32 lines.

In addition to the use of rare subscripts, the Napier fragment has something else in common with the Parker and Orosius manuscripts. Along with Alfred's , the Anglo-Saxon chronicle and Orosius's were core texts of Alfred's reform. The evidence would seem to suggest that the Napier fragment came from the same Winchester scriptorium. Following Parkes's argument, it might even be argued that the extremely high incidence of the use of subscripts, with no signs of abating, suggests that the Napier fragment is the earliest manuscript of the three.

In any case, there is good reason to use the Parker and Orosius manuscripts as models to simulate the script of the Napier fragment. The EPT's DucType tool, designed to examine the paleographical features of scribal letterforms, can be pressed into service to assemble a complete set of letters from these contemporary manuscripts to help us visualize the Napier fragment.

DucType for paleographical analysis

The EPT's RamSome, so named because its output requires some RAM and resembles an Old English ransom note, is an imaging facility of the StaTend tool, which earlier provided basic statistics about the Napier fragment (see figure 2, above). RamSome was designed for the to use letterforms from Cotton Otho A. vi to recreate virtual folios for the ones destroyed in the Cotton fire. Adapted for the Napier fragment, it draws on the set of letters collected from the Parker and Orosius manuscripts and on-the-fly transforms the transcript into an image, including the subscripts and other special characters, such as ampersand, crossed-thorn, and letters with macrons. The set of letterforms can include variant letters, and the RamSome interface is interactive, allowing an editor to pick and choose from the set, to select a Parker letter, for example, instead of an Orosius example, or to substitute one of the three forms of s for another one.

Detail from RamSome reconstruction of Napier fragment (cf. Figure 11).

[ View entire reconstruction ]

While he provides enough explicit markup to arrive at this virtual reconstruction of the lost leaf, Napier fails to convey several important facts about the actual fragment. We must infer for ourselves the technical means of production of the original manuscript, its dimensions, for example, the nature of the rulings and the size of the writing grids, the disposition of the scribe's handwriting, and so on. Napier does tell us that the manuscript that held the fragment as an endleaf was Bodleian Library, Oxford, MS Junius 86, a surprisingly small book considering the size of the text of the fragment. Junius 86 measures only 155 × 100 mm, or about 6 × 4 inches. To fit the space, the fragment must have been inserted sideways when it became a flyleaf for Junius 86. Even so, sixteen ruled lines of text had to fit in the 100 mm space, leaving little over 6 mm per line, from ruling to ruling, and making the written space of the fragment, without any margins at all, extremely cramped.

RamSome reconstruction of entire Napier fragment (Figure 10), rotated and sized to fit space available in Junius 86. Actual size (155 × 100 mm).

Napier tells us that the script is sometimes indistinct where the letters are blurred and that some letters are gone because of holes in the vellum. But he inexplicably fails to mention that the script is tiny—difficult to read in such proportions even in the clear and clean hand of the Orosius scribe. It is difficult to accept that the scriptorium that produced the spacious and highly legible pages of the Parker and Orosius manuscripts also produced such a tightly-packed and difficult to read manuscript of Alfred's in the early tenth century.

In view of the improbably small script, there appear to be several other circumstantial reasons to doubt the authenticity of the Napier fragment. It is odd, for example, that the scribe, who was otherwise so economical with vellum, did not use the right margin more efficiently. It stretches credulity that the Napier scribe would use 16 rare subscripts, relics of the past, in 32 lines, an average of one every other line. Although other space-saving features of the script that Napier does not describe (such as other cursive ligatures and underslung l) may have further compressed some of the lines, there do not appear to be letterforms in the overcrowded lines that would account for the great discrepancies in the length of the lines. From another angle, it stretches belief that Junius, who went to great lengths to collate manuscripts in Junius 12, was not aware of an Old English binding fragment, in plain view as an endleaf, in one of his own medieval manuscripts, Junius 86. This book formerly belonged to his nephew, Isaac Voss, who would have surely discouraged his binder from using Old English texts for end leaves, because he was an avid collector of Anglo-Saxon pieces. Junius was well acquainted with Voss's library, "for I have mett among that store my kinsman hath with diverse Francike, Anglo-Saxonike, and Gothic Antiquities, no where else to be found" (van Romburgh 2004, 876). It seems equally surprising that Humfrey Wanley failed to notice an endleaf containing Old English. If it leads to the recovery of such an important leaf of Old English, raising these legitimate doubts will be worthwhile. In the meantime, scholars are free to accept or reject the authenticity of the Napier fragment, but they should no longer assume that it proves the existence of an all-prose manuscript from the early tenth century.