This is an electronic version of a chapter in Poetry, Place, and Gender: Studies in Medieval Culture in Honor of Helen Damico, edited by Catherine E. Karkov. Medieval Institute Publications (Kalamazoo: Western Michigan University, 2009), pp. 98-131. The black-and-white figures are replaced with color screenshots from the third edition of Electronic Beowulf.

The nathwylc Scribe and the nathwylc Text of Beowulf

Kevin Kiernan

In 1882 Julius Zupitza succinctly stated as a matter of fact, rather than as a hypothesis, that “all that is distinct in the FS. in fol. 179 has been freshened up by a later hand in the MS.”1 Considering how little it explains, it is surprising that this hypothesis has implicitly prevailed, if only in corrections of supposed freshening-up errors, in editions of Beowulf over the past century and a quarter. Zupitza never described the features of the hand that convinced him it belonged to a later scribe. He did not give any indication of how much later the later hand was than that of the principal scribe who copied the folios before and after fol. 179. Zupitza did not even hazard to guess why or how the original text disappeared in the first place; nor why some areas contain overlapping traces of letters and even lines of text; nor why the later hand was so selective in choosing what to freshen up and what to leave untouched, especially when all editors agree that many faint readings are still legible today. Finally, Zupitza did not describe in support of his hypothesis the unique condition of the surface of the vellum, which appears scoured overall, with napped stains over the most illegible passages, perhaps because it was not as obvious in his autotypes facsimile as it was in the manuscript itself. With the advent of digital technology, and superior digital images of the manuscript evidence, we are now able as a scholarly community, rather than as an individual scholar with special access to a rare manuscript, to reevaluate the evidence and to offer and assess other hypotheses to explain it.2

My own understanding of what seems to have happened to fol. 179 has evolved over the past twenty-five years as digital images, digital image processing, and digital tools for analysis changed the way I was able to evaluate the evidence. In an effort to arrive at a comprehensive explanation, I have had to abandon some of my earlier views about this folio and to refine others. There are, I believe, four major stages that account for fol. 179 as it now exists. In stage one the original text was completely removed from both sides of the folio by an overall scouring. As there is no sign of collateral damage to either fol. 178v verso or fol. 180r recto, the person who made this palimpsest must have worked with the gathering that begins with fol. 179 detached from the preceding gathering. In stage two an Anglo-Saxon scribe (perhaps the same person), working with the two other Beowulf scribes, replaced the original text with a new and different one that bears some convincing signs of a draft still in progress.3 I call this person the nathwylc scribe, for the repeated use of the rare word nathwylc (“I know not who” and “I know not what”) three times in the space of sixteen lines on this folio.4 The nathwylc scribe has a very similar ductus to scribe two, but uses some later letterforms more characteristic of scribe one, which presumably led Zupitza to see the script as a later hand than that of scribe two. The nathwylc scribe leaves other paleographical markers, too, such as the unique occurrence of an “oc” form of a in brade in the first line;5 an unusually fat bowl for the wynn in weard two lines later; the use throughout of a rounded, more three-sided a rather than square insular a; a more frequent use of caroline instead of insular s; and some unexpected spelling and usage variants, which further distinguish the new hand from that of the second scribe.

The other two stages affecting fol. 179 happened, appropriately enough, some indeterminate time “later.” In stage three, parts of the nathwylc scribe’s new text on fol. 179 failed to adhere to the scoured vellum, obliterating parts of words and letters and leaving behind fragments, which Zupitza in many cases described, and editors have for the most part accepted, as incorrectly freshened-up text. It is impossible to know when this fading occurred, but it was plausibly caused by the 1731 fire, by the water used to put it out, or by a combination of the two, probably abetted by whatever concoction was used to dissolve the ink of the original text. We know the fading happened before Thorkelin and his scribe copied the manuscript around 1787, because both transcripts show that the text was illegible in the same places by then.6 In stage four someone dabbed and rubbed chemical reagents over the most illegible readings to try to resuscitate the ink in these places. Under ultraviolet these areas appear as dark stains, making it almost impossible to identify any letterforms beneath them.7 Circumstantial evidence that is discussed below suggests that staff at the British Museum applied these reagents to aid in the preparation of Zupitza’s facsimile.

Whether or not they have expressed doubts, nearly all editors in their readings have tacitly accepted Zupitza’s freshening-up theory. Embracing the theory when it was still relatively new, Frederick Klaeber categorically stated that fol. 179 “has been freshened up by a later hand, but not always correctly,” and referred readers to Zupitza’s notes for details.8 It is not surprising to find Zupitza’s theory still holding forth in the revision of Klaeber’s edition, in view of R. D. Fulk’s strenuous endorsements in two recent articles.9 Appealing to an unspecified consensus in the first article, Fulk says that “most of those who have examined the folio believe that the text was at some point retouched by a scribe who traced such letter forms as he could make out (sometimes badly) and left the rest untouched.” Fulk believes that “This [i.e., partial, sometimes inaccurate, retouching theory] explains the stark contrast between dark and faint text as well as the surprising readings, which may be attributed to the later scribe’s misapprehension of the faint letter forms he traced.” He concludes that, “given how much [this analysis] explains about the damaged folio, alternative analyses must account for a great deal if they are to rival it” (“On Argumentation,” p. 15). In the second article, Fulk again summons a mystery consensus, when he claims that “most observers believe that the rewritten letters are intended merely to trace the original ones,” but giving as examples only A. E. A. Werner and Zupitza himself.10 Norman Davis and Kemp Malone, who both spent a great deal of time studying the manuscript in preparing their respective facsimiles, were not members of this seemingly bogus consensus. Davis says that “it is remarkable that the ultra-violet photographs in the present edition do little to support the readings which Zupitza thought he could see in the ‘first hand’ underneath the ‘freshening up’” (Zupitza, Beowulf [1959], p. vi). Malone leaves it up to the reader to decide “how much touching up (if any) there has been” and advises the reader to consult Davis “for further discussion of the matter” (Nowell Codex, p. 83).

In fact, editors who accept Zupitza’s hypothesis (with or without examining the manuscript) have shown no curiosity about what his theory fails to notice, much less analyze. If one believes that some nathwylc or other freshened up all of the text that survives, one must accept the corollary that the original text somehow disappeared. Why was the original text removed? How does one account for the unusual appearance of the overall surface of the vellum? What might have caused the stains over completely illegible sections? Why would a brilliant forger not completely restore at least the faded text that anyone can easily see today? Why would someone tracing what was still relatively clear consistently create letterforms for which there were no models on the folio? Why do faded parts of the text show features of this later hand? What happened to disfigure the replacement text? Why are the missing sections particularly discolored, leaving a gray residue on some letters bordering the gaps? Editors must try to answer such textual questions concerning the transmission of the text if they hope to establish an acceptable modern edition of the poem.

Tilman Westphalen first asked many of these questions.11 Addressing Zupitza’s assertion nearly a century after it was made, Westphalen suggested that someone looking for vellum deliberately erased all of fol. 179, recto and verso, before the second scribe fortuitously rescued his former work and did the best he could to restore it. After a meticulous paleographical comparison of all letterforms on fol. 179 with those of other folios, Westphalen identified the later hand as the second scribe, who tried to restore the erased text about twenty years later. The second scribe’s later hand, Westphalen argued, shows some modernizing moves toward insular caroline script as well as some signs of aging in the declining control of his pen strokes. In 1981 in Beowulf and the Beowulf Manuscript I accepted Westphalen’s identification of the hand and its relative dating, but formulated a new hypothesis. Marshaling previously unnoticed codicological evidence, I argued that the second scribe was participating in a revision that first brought together two poems about Beowulf, one recounting his youthful exploits and the other his troubled reign as an old man and his heroic death.12 It will be useful to review briefly the wealth of circumstantial codicological support for this revision, as editors have never taken it into account in their implicit theories of transmission of the text. Any reasoned consideration of whether or not fol. 179 is a palimpsest in the context of other textual changes will be incomplete outside of this wider context.

All of the codicological support for a revision converges around fol. 179. Beginning with this folio, the physical format of the manuscript changes in the number of sheets to a gathering (five instead of four), in the sheet arrangement (hair outside all sheets instead of alternating hair and flesh sides), and in the number of rulings of the sheets (from twenty to twenty-one lines). The distinct change in format raises the possibility that these last two gatherings about an elderly king Beowulf might have once existed as part of a different collection. The motivation for a revision might have been to bring the two different stories together. On the folios preceding the palimpsest, a missing fitt-number XXIIII indicates that a revision of this section, beginning in the first scribe’s stint, might have involved the removal of at least the beginning of an original fitt XXIIII.13 The evidence is somewhat obscured by a clumsy postmedieval alteration of number XXV to XXIIII, and by the erasure, presumably by the same person, of the final I in numbers XXVI–XXVIIII in an abortive effort to correct the faulty sequence. This section begins with a gathering (fols. 163–70) anomalously ruled for twenty-two lines, rather than the otherwise unvarying 20 twenty lines in the first scribe’s stint, suggesting that the first scribe, with more text than usual to copy, might have prepared a gathering with thirty-two more rulings than usual to hold this part of the revision.

The second scribe takes over copying in the next gathering, fols. 171–78, in an unplanned transition three lines into fol. 172v. One might expect professional scribes to conceal the change of hands by starting the transition at the top of a verso, not three lines down. This careless transition in hand might suggest that the scribes knew the untidy mix would later be concealed in a formal, final copy. In any case, there is compelling evidence that the second scribe took over copying this gathering after he had already completed the last two gatherings, which begin with the palimpsest. The order of copying is evident because, in the gathering he completed for the first scribe, the second scribe deftly and surreptitiously compressed twenty-one lines of text on folios ruled for only twenty lines on four consecutive facing pages (fols. 174v–176r). In other words, the second scribe knew exactly how much extra text he had to fit into the gathering preceding fol. 179. The only obvious reason why the scribe would have added these extra lines in defiance of the rulings was if the following gathering had already been copied.14 A theory of the text of Beowulf should not ignore these paleographical and codicological anomalies, which among other things effectively disprove the unexamined but frequently adduced assumption that two scribes were simply copying from a well-established exemplar that had gone through a long transmission.15 However one interprets it, the extremely unsettled state of the physical makeup of the manuscript surrounding and including the palimpsest does not support the notion that the scribes were copying a settled text under normal conditions.

It seems no coincidence, then, that textual anomalies also converge on fol. 179. The first line copied by the nathwylc scribe (fol. 179r1) abruptly announces, in what sounds like a non sequitur, the fifty-year reign of Beowulf.16 This sudden and surprising revelation comes in the middle of what is arguably the most convoluted sentence in the poem. When we last saw him, on the preceding folio, a young Beowulf was recounting his recent successes in Heorot and doling out to Hygelac and Hygd the gifts that Hrothgar had just given him. Hygelac in turn gave to Beowulf Hrethel’s incomparable sword, a sizeable grant of land,17 a hall, and a bregostol, or princely seat (2190–99). If there once were a separate poem on Beowulf as a young man, here on fol. 178 verso would be an appropriate place to end his story.

Then follows the unruly sentence, introducing in midsentence a seventy- or eighty-year-old Beowulf, who has already ruled for fifty years:

Eft þæt geiode     ufaran dogrum
hildehlæmmum,     syððan Hygelac læg,
ond Hear[dr]ede     hildemeceas
under bordhreoðan     to bonan wurdon,
ða hyne gesohtan     on sigeþeode
hearde hildfrecan,     Heaðo-Scilfingas,
niða genægdan     nefan Hererices —:
syððan [fol. 179] Beowulfe     brade rice
on hand ge(hwearf);     he geheold tela
fiftig wintra     — wæs ða frod cyning,
eald eþelweard —,     oð ðæt an ongan
deorcum nihtum     draca rics[i]an . . .
               (Klaeber 2200–2211)
After that it happened in higher days,
in clashes of conflict, when Higelac lay dead,
and to Heardred harmful-swords
under shield-guards became killers,
when they sought him among his succeeding-folk,
hardened fighters, Heatho-Scilfings,
for his hostilities assailed Hereric’s nephew —:
afterwards [fol. 179] to Beowulf the broad realm
passed into hand; he held it well
for fifty winters — was then a wise king,
an old land-warden —, until a certain one began
on dark nights, a dragon, to reign . . .
               (my translation of Klaeber’s text)

Klaeber implicitly acknowledges the awkwardness of this sentence, spanning 178v14–179r7, in his note to the last word on fol. 178v, the first word in line 2207: “syððan is used, in a way, correlatively with syððan 2201.” The reader must discover the tenuous way these different parts of speech are supposed to correlate from Klaeber’s glossary: the first syððan is the conjunction, “since, from the time when, when, after, as soon as (s[ome]t[imes] shading into because),” while the second syððan is the adverb, “since, thereupon, afterwards.”

Following Klaeber’s syntax (but omitting his strong pause after “Hererices”), Mitchell and Robinson in their recent edition more explicitly describe the syntactic problem. According to their note to these lines, “2200–8a þæt [demonstrative] anticipates the passage of dependent speech which begins at syððan [conjunction] ‘after’ l. 2201b (correlative with syððan [adverb] ‘then’ l. 2207a) and ends at gehwearf l. 2208a. The conj. þæt would normally have preceded syððan l. 2207a in prose and poetry (OES §1978) but must be understood before syððan l. 2201b to give an idiomatic translation.”18 In a recent discussion of this syntactically problematic sentence while presenting his own theory for the palimpsest, Carl Berkhout has observed, “some translators of Beowulf have simply rendered syððan here as ‘that’ and moved on.” Berkhout rightly objects that “nowhere else in Beowulf or in the entire corpus of OE poetry and prose does the adverb or the conjunction syððan, or any of its variants, introduce as a substitute for þæt a clause dependent on an impersonal construction with (ge)gan(gan), gelimpan, (ge)weorðan, or other such verbs.” He concludes that, “even if þæt were to be understood before adverbial syððan in Beowulf 2207a, as Mitchell and Robinson suggest, we would still be left with a syntactic anomaly here.”19

The complementary arguments of these translators, editors, and grammarians convincingly show that the most direct way to solve the syntactic anomaly and justify the translations is to assume that the second scribe simply omitted a crossed thorn, the abbreviation for þæt (“that”), before syððan: “After that it happened . . . when Hygelac lay dead and . . . the Heatho-Scilfings . . . assailed Hereric’s nephew, [that] afterwards ([þæt] syððan) to Beowulf the broad realm passed into hand.” The scribal solecism is easy to understand, particularly if the nathwylc scribe was in the process of revising the following folio. Berkhout uses his analysis of this easily rectified syntactic problem to explain why the original text on fol. 179 was entirely erased. He posits that the second scribe “accidentally omitted a full clause after syððan, amounting to about two or three poetic lines, plus the word þæt or þætte before beowulfe” and that he or someone else in the scriptorium erased all of fol. 179, recto and verso, to make room for the missing lines in a more compactly rewritten fol. 179 (“Beowulf 2200–08: Mind the Gap,” p. 54). Following this line of argument, Berkhout must assume, as he says, that the second scribe or his associate “became occupied with or distracted by other tasks, leaving Beowulf for the next four and a half centuries in far more depleted textual condition than it had been before” (“Beowulf 2200–08: Mind the Gap,” p. 56). His theory of a pseudo-palimpsest (a completely erased leaf, ready for a new text) is designed to make room, as well, for Laurence Nowell, the sixteenth-century antiquary who owned what we now call the Nowell Codex (including the Beowulf manuscript as its fourth item) and who, according to Berkhout’s theory, freshened up as much of the erased text as he could see.

Berkhout first proposed that Nowell was “Scribe C” (the nathwylc scribe) in a paper he read in 1986 at the annual meeting of the Medieval Academy of America. The published abstract of this paper lays out the essential argument:

there are good reasons to suspect that the readable text on the defective fol. 179 (182) of the Beowulf manuscript, involving lines 2208[sic]–2252, is in the hand of the antiquary Laurence Nowell, who, in or about the year 1563, did his textual and aesthetic best to recover the intentionally erased text on this folio. The pattern of linguistic errors and scribal discrepancies, along with other physical indications, strongly suggests that the final scribe of this folio was not Scribe B or any other Anglo-Saxon. It is not reasonable to suppose that the folio might have been freshened up in the post-Conquest or Middle English period. It is very reasonable, however, to suspect Nowell. It is certain — and provable — that he had the opportunity, the ability, the inclination, and the motive to freshen this page while he was in the service of William Cecil. Although the evidence for this argument remains circumstantial, this evidence is considerable and so far passes all tests of arguments against it. We must seriously consider the possibility that one of the most crucial passages in our most important Old English literary texts comes to us only through a good but imperfect scholar at work in the age of Shakespeare.20

Although there is no evidence that he ever read Beowulf, we can agree that Nowell, who wrote his name on the first folio of the codex in 1563, had the opportunity, and even that he might have had the inclination and motive, to repair the palimpsest. The style in which Nowell wrote his name at the start of the St. Christopher fragment provides a quick peek at his own formal penmanship:

Figure 1

[Fig. 1: Nowell’s signature in Cotton Vitellius A. xv, fol. 91(93)r]

Anyone even casually familiar with his many transcripts of Old English and other collected materials in British Library manuscripts Add. 43703–43710 must seriously doubt that Nowell had the “aesthetic ability” or basic artistic skill to counterfeit the duct of an Anglo-Saxon scribe, even one with the relatively inelegant style of the second scribe or the nathwylc scribe. Berkhout agrees with me that these transcripts do not support his thesis. What, then, is the “considerable” evidence that “so far passes all tests of arguments against it”? In a subsequent talk, “In Search of Laurence Nowell,” for the International Society of Anglo-Saxonists (ISAS) in 1987, Berkhout first identified the single manuscript on which he based his theory. There he revealed that Nowell’s “display” script occurs in British Library MS Davis 30, Nowell’s facing translation of the Laws of King Alfred. Berkhout’s confident assertion notwithstanding, this evidence cannot, in fact, pass even gentle tests against it, as a cursory comparison of the two scripts reveals:

Figure 2

[Fig. 2: Nowell’s “stylized insular hand” v. the nathwylc scribe]

Berkhout himself later characterized the script in BL MS Davis 30 as a “stylized insular hand,” but has never published a detailed paleographical explanation of why he believes that the “mess,” as he calls it, on fol. 179 “is chiefly the result of Nowell’s effort to recover an original text crudely erased not long after it was written.”21 Fortunately, in one of several impassioned attacks on my late dating of Beowulf, Johan Gerritsen takes Berkhout’s evidence and argues his case in some detail in “Have with You [i.e., me] to Lexington! The Beowulf Manuscript and Beowulf.”22 Gerritsen in effect undermines Berkhout’s case in his fairly thorough description of the features of Nowell’s formal script in BL MS Davis M30:

It is plain that [Nowell] is accustomed to a more rapid script, for he repeatedly lapses into more cursive forms (such as his characteristic ligature long s+t), though invariably recollecting himself again. Ascenders are clubbed; descenders are not differentiated as long and short and have no off-strokes, but single and letter-final minims generally have rounded serifs. The cross-strokes of insular f are often circle segments; u is a one-trace script character, not two minims; a is round-headed, its second limb at a variable angle and occasionally straight; the bowl of g is always open and low. The most remarkable feature is his tall e, both by itself and in combinations: its bowl is mostly a high loop not connected to the stem, and the tongue is frequently absent; the small variety is a vertical bow with a (generally far too small) figure 2 at the top. The pen is cut almost straight and fairly fine, enabling him to write long s as a single trace instead of three. (p. 31)

In other words, as Fig. 2 illustrates, Nowell’s script in BL MS Davis M30 does not support Berkhout’s hypothesis. The nathwylc scribe does not use clubbed ascenders; this hand differentiates long descenders (e.g., þ, wynn, f, insular s) and short ones (i.e., r in proximity to these other letters), sometimes with off-strokes; single and letter-final minims do not normally have rounded serifs; the cross-stroke of insular f is never a circle segment; u is a two-stroke character of two minims; a is indeed rounded (one of the distinguishing features of the “later” hand), but unlike Nowell’s its top is sometimes a flat, oblique hair-stroke in æ (e.g., þær r7, slæpende r10, fær v2) and especially in the ea-ligature; the bowl of g is often closed; the tall e-loop is thick and bold, not thin and spindly, and low e (with or without its tongue) always has a pronounced loop, not a tiny, fastidious one. The pen is cut broad, which accounts for the bold strokes so unlike Nowell’s display hand. Gerritsen does not mention the double hyphens, semicolons, and marks of abbreviation that bear no resemblance to the nathwylc hand. He admits, however, that “it should be evident that a comparison of Nowell’s work as outlined above with what we see in his codex [on fol. 179] could not be expected to yield any really positive result.” Yet he concludes, on the basis of “the tall e” that the hand “could have been Nowell.”23

Fulk calls Gerritsen’s fallacious logic a “well-reasoned argument” that points “to a post-medieval freshening” (“Some Contested Readings,” p. 201). Neither Berkhout, nor Gerritsen, nor Fulk explains how Nowell or some other postmedieval forger, if freshening up what was still legible on fol. 179, consistently managed to “freshen up” letterforms that were uncharacteristic of the second scribe, such as the unique oc-a r1, the unusually fat wynn-bowl r3, or (throughout) the rounder, more triangular a, instead of the second scribe’s characteristic flat-topped “square insular” a.24 If one accepts the freshening-up theory, one is faced with the absurdity of a later form of a as the original reading in wintra r2. According to Zupitza, “wintru is owing to the later hand, the u standing in the place of an original a.” Although I once agreed that u was written over a (Beowulf and the Beowulf Manuscript, p. 234), the high-resolution images in Electronic Beowulf convinced me that the first side of the “u” is a minim stroke that does not curve into the bottom of an intentional u. The most likely explanation seems to be that the nathwylc scribe started to write the wrong letter (perhaps a dittographic r), but corrected the mistake by writing the left side of his rounded a over it, and completed his a normally.

A thorough examination of his most careful attempts at insular script convincingly shows that Nowell lacked the skill of a facsimilist to accomplish what would amount to a brilliant, successful forgery. As A.H. Smith observed in 1938, when the handwriting on fol. 179 was “viewed through a self-illuminating microscope which is powerful enough to show the pen strokes, there did not seem to be that hesitation and lack of coincidence usually associated with freshening up, forgery, and the like.”25 High-resolution digital images corroborate his view. The images of the Electronic Beowulf, digitized both in bright light and in response to ultraviolet, have the effect of providing stable views for the microscope readings Smith examined. Zupitza’s interpretation that everything legible on the folio was merely “freshened up by a later hand” has paradoxically hampered the analysis of the script on this defective leaf. It is true that the script shows features (the sometimes open g-bowl, the preference for tall caroline s, and especially the rounder a) that are later than the square insular script of the second scribe. However, these later features are not later than the insular caroline script of the first scribe. It would be more accurate to say, therefore, that all that is legible was “written by another hand exhibiting characteristics of both of the two main scribes.” The nathwylc scribe successfully camouflaged the convergence of features by matching the cut of his pen to the work of the second scribe on the surrounding folios. There is no paleographical reason not to conclude that the three scribes were contemporaries.

In arguing for Nowell, Berkhout expresses disbelief that “the upper text on 179 could possibly be the work of a contemporary Anglo-Saxon scribe with a practiced hand, an idiomatic knowledge of [Old English], and reasonable familiarity with Germanic alliterative versification” (“Beowulf 2200–08: Mind the Gap,” p. 56). In fact, the nathwylc scribe wrote in a practiced enough Anglo-Saxon hand to confound scholarship for over a century. As all writers do, the nathwylc scribe made a number of mistakes, but in most cases he corrected them. He normally corrected mistakes immediately, in the course of making them, by overwriting rather than erasing the error, no doubt because of the precarious, perhaps even still damp, condition of the newly scrubbed vellum.26 These extemporaneous corrections testify to the speed and confidence of his work, and are manifestly not the tentative signs of a retoucher. As for his supposed lack of familiarity with Old English verse, no professional metrist has yet assailed the alliteration or meter of these lines. The forty-two lines on this folio display an accomplished, well-distributed use of the major metrical types according to the various analyses of Sievers, Bliss, Hutcheson, Hoover, and Russom.27 Indeed, in the face of the metrical evidence of competence, I have rejected two of my own conjectural restorations in Electronic Beowulf, because they present metrical problems.

There is a dittographic error involving alliteration, where the scribe, repeating an immediately preceding r, wrote rende for wende (179v10), but otherwise the alliteration in all legible cases is perfectly correct for forty-two lines of verse.28 The mistake of rende for wende is complicated by Zupitza’s misreading of the faded en for ih, a trompe l’œil induced by the fading of the hair-strokes of the e-head in en, as Sedgefield, Malone, and I have all explained.29 The shaft of this supposed h is much too short for a real h, and lacks the onset serif, as a comparison with any of the nearly fifty h's on the folio will make clear (see Beowulf and the Beowulf Manuscript, p. 241). It should not be surprising that the hair-strokes of the e-head have faded, in view of the general fading of wine earlier in the line. To illustrate what happened to the original en, I have in Fig. 3 digitally copied only the hair-strokes of the e-head of lengest v9, immediately above “rihde” v10 (top), and pasted them in the space where the hair-strokes faded in rende (bottom):

Figure 3

[Fig. 3: Conjectural restoration of “rihde” to rende 179v10]

The only grounds for doubting the nathwylc scribe’s command of idiomatic Old English would be if one accepts nonsensical spellings, such as rihde, as mistakes in freshening up rather than as subsequent fading, or a lack of adhesion, of parts of letters that left behind fragments resembling other letters, such as “ih” for en. To believe the freshening-up hypothesis we must therefore posit an expert forger who also understood the fine details of Old English meter, but who could not recognize simple Old English words and phrases (such as fea worda cwæð) because some letters (in this case, a in fea and w in worda), though still legible today, were partly faded. No English speaker today would be prone to transcribe incorrectly the comparable modern English phrase, “he spoke few words,” because the two w’s were faint but visible.

Until Westphalen, no one (including Zupitza) had ever attempted to explain why the original text disappeared from the folio, recto and verso, and why someone had to freshen up everything that is now legible. Yet if someone were simply scavenging vellum, as Westphalen tentatively proposed, one must ask why this person did not cut out the leaf or, if more vellum was needed and the text was considered worthless, pull out the sheet or completely disassemble the quire to ease the task of erasing the text. If, on the other hand, a revision was underway, these questions have uncomplicated answers. To judge by the highly economical use of the preceding two gatherings to accommodate additional text, the scribe either had no available vellum or wished to keep the following gathering, beginning with fol. 179, intact. Scribes would not normally want to use a replacement leaf at the vulnerable start of a gathering. To be sure, the decision to erase the entire original text on fol. 179 while it was bound between fols. 178 and 180 would have made the task quite difficult. Whoever erased the text would have had to protect the rest of the gathering from whatever liquid solution was used to help dissolve the ink before scraping it off. There is no sign of any spillover onto what are now the facing leaves (fol. 178v and fol. 180r) to fol. 179. Scouring fol. 179 would not have been as difficult as it might seem, however, if the nathwylc scribe had not yet attached (or could still easily detach) the last two quires, which as we have seen the second scribe must have already copied. Considering the codicological proof that the last two gatherings were copied before the immediately preceding one, the scribe would have been able to scour fol. 179 without harming the folios on either side of it.

Everyone who has described the physical construction of the codex has found that fols. 179–88 formed the outside cover of the penultimate gathering of Beowulf. There is no doubt, however, that fol. 179 underwent very different treatment from its presumed conjugate. The relatively rough and discolored condition of the entire surface of the vellum on fol. 179, compared to fol. 188 and to all other leaves in its gathering, shows that fol. 179 was not meticulously smoothed with pumice after the text was removed. The scribe may have feared that the rescraped vellum would become too thin by a thorough repumicing. Or he may have known, because of the widespread revisions, that he or someone else would soon recopy the manuscript, thereby concealing all of the paleographical and codicological evidence of work still in progress. There are various ways to account for why the ink failed to adhere to the palimpsest, especially on the smoother hair side of the recto. For instance, some waxy or greasy residue from whatever concoction the nathwylc scribe used to remove the previous text may have dissolved from the heat of the 1731 fire. Its dissolution could have left behind the unique fading, indiscriminately including bits of letters as well as larger sections of text, on this specially treated leaf. We know the fading had occurred by the late eighteenth century, because Thorkelin and his hired scribe were unable to read the same areas of the folio that are obliterated today.

Whatever its cause, subsequent dissolution of parts of the new text, rather than partial refreshening of the original text, best explains why bits of still-recognizable letters survive. It is improbable that a refreshener would touch up only nat and the tips of the h and the l in nathwylc r8, as Zupitza says (“only part of hwylc freshened up”). It strains belief that an Anglo-Saxon scribe (or a postmedieval restorer, for that matter) would freshen up only nat- but not the faint but legible -hwylc and -hwylces in line 8 and again at lines 16–17 on the recto. Because there is no attempt at all to retouch these letters, it is compelling to conclude that something destructive later happened to hwylc and hwylces. To put it another way, because nathwylc is a high-frequency word only on this folio, and is almost nonexistent elsewhere in the Old English corpus, the dark ink of nat- beside the faint ink of -hwylc(es) adds some circumstantial support for fading of a new text on a palimpsest, rather than partial refreshening of an old, original text that mysteriously disappeared. Even Fulk eventually seems to acknowledge that the folio is in effect a palimpsest, as the OED defines the term. The folio, he says, “seems, for some reason, to have been washed clean of its original text and rewritten, either by the second scribe or by another with considerable skill in imitating that scribe’s hand” (“Some Contested Readings,” pp. 208–9). I cannot argue with this description, which I have advocated for over a quarter of a century. Later in this article Fulk suggests that the reason for erasing the original text was to write a corrected one: “Perhaps then the scribe’s initial error was simply to begin writing on the second line of the verso rather than the first, and it was the bungled measures subsequently taken to correct this problem (and other errors on the leaf) that led ultimately to the folio’s present condition” (“Some Contested Readings,” p. 219). One hopes that a more detailed explanation will be forthcoming in the revised edition of Klaeber.

In the meantime, Fulk confidently dismisses as “certainly mistaken” the theory advanced not only by me, as he implies (“Some Contested Readings,” pp. 192–93), but by several other scholars (including Malone, A. H. Smith, and Walter Sedgefield) who have studied the manuscript at first hand, that fading and flaking produced a number of paleographical optical illusions. Fulk considers this commonly held theory extravagantly implausible, “since the missing text must then be assumed to have come off the vellum not randomly but in such a way as to leave only fully intact letters or portions of letters that happen to make sense” (“Some Contested Readings,” p. 209). Fulk does not mean that these “letters” make sense in the words in which they occur, but that they look like other real letters. In fact they look like malformations of real letters and make nonsense in their semantic context: there are no Old English words fæs, rihde, fec, or bealc.30 There are, however, the real words fær, rende (and wende), fea, and beale or bealo, which can account for them if bits of the letters r, en, a, and e or o, faded or flaked off to create plausible (if only they made sense) versions of s, ih, and (twice) c.31 These few paleographical optical illusions that resemble malformed insular letters should be considered in the context of the overall appearance of the folio, especially on the recto, where one finds the “mass of meaningless strokes that,” as Fulk says, “one would expect to result from the random loss of portions of letters” (“Some Contested Readings,” p. 193) as well as of words and passages. General fading and some flaking are surely the simplest explanations for faint, but easily read, and uncontested readings such as -iod of ðiod r12, the final n in bolgen r13; -inc fæt in sinc fæt v2, ær in ærgestreona v4 and in ðær v9, -ine in wine v10, the second l in eall v12, -eald in heald v16, s in hruse v17, y in hyt v18, and w in gehwylcne v20. It makes more sense that these readings faded than that an otherwise diligent refreshener failed to retouch them. In short, an improperly prepared surface explains the subsequent loss of these letters and bits of letters far better than the still undefended theory that a “later” hand failed to freshen them up.

While an improperly prepared writing surface explains the general fading of the revised text, there are additional reasons for our inability to restore with confidence the faded text in the many gaps on the recto and in the first two lines of the verso. The most destructive damage to the text was done not in Old English times, but in modern times, when someone recklessly applied chemical reagents to these areas.32 The ultraviolet images prepared for the Electronic Beowulf (see Figs. 4 and 5) show that someone swabbed and drizzled some liquid solution on the faded sections of the text on these lines. C. L. Wrenn first suggested in 1953 that “the discoloration on these pages seems to have been caused by the application of some chemical reagent in the attempt to improve their legibility.”33 In 1981 I mistakenly dismissed this suggestion on the grounds that “the Thorkelin transcripts show that the text was in as bad shape in 1787 as it is today, and no one would have been treating the MS with chemicals before Thorkelin’s time” (Beowulf and the Beowulf Manuscript, p. 222, note 46). While true, these observations only establish that the fading of the text on fol. 179 predates the Thorkelin transcripts, while the application of reagents postdates them. My implicit assumption, however, was that officials at what was then the British Museum would not have allowed anyone in modern times to use chemicals on its manuscripts. However, the digital ultraviolet images are decisive, revealing swab-like marks and even drips over some of the letters in lines 5 and 8 on the recto, as I have illustrated and discussed briefly in the textual notes to the Electronic Beowulf.

Figure 4

[Fig. 4: 179r5–9 bright light vs. ultraviolet, showing use of reagent]

The blanks in the Thorkelin transcripts reliably identify the areas where the text had faded and where reagent was later applied to try to restore it. While the facsimiles of Zupitza, Davis, and Malone all obscure the stains, the ultraviolet digital images in the Electronic Beowulf show them clearly. The image on the right in Fig. 4 shows beyond doubt that some liquid was liberally swabbed between and around hea and hord on line 5, as well as above stearne sti- on line 6, drizzled between beorh and stearne and across the ð of uncuð in line 7, and then into the faded sections in lines 8–10 and beyond, most obviously including the first line and a half of the verso.

When in the mid-1970s I first examined these areas with an ultraviolet lamp in the old British Library, I was surprised to see heavy stains showing up in the cloudy areas of the Malone and Davis facsimiles of the Beowulf manuscript, where I most hoped to recover the illegible text. Neither Malone nor Davis mentions the stains, which may mean that neither used ultraviolet themselves, but instead relied on the photographer. Davis refers only to the “ultra-violet photographs,” not to direct examination himself (Zupitza, Beowulf [1959], p. vi), and Malone says he used the same photographs.34 Whether or not they themselves used the ultraviolet lamp, however, the ultraviolet photographs made for their facsimiles fail to show ultraviolet fluorescent effects, in particular the dark discoloration in the gaps in lines 5, 8–10, and 18–21 on the recto, and lines 1–2 on the verso. At the old British Library in the 1970s one had to examine ultraviolet effects on manuscripts in a tiny closet with a huge World War II–era ultraviolet lamp humming ominously overhead. To protect the manuscript and the user, who would fluoresce along with the text, there were strict (and welcome) rules to confine these viewings, including any necessarily hasty notes, to seven minutes at a time. Today, with excellent digital images, which permanently preserve ultraviolet effects, it is possible for anyone to study at leisure these transitory effects brought out by ultraviolet rays, and to reach new conclusions, and reevaluate old ones, about the manuscript evidence. When I first encountered them, I interpreted these areas, which had been rubbed when wet, leaving a napped and grayish appearance in bright light, as evidence of a second stage of revision. I now realize that these stained areas are not revisions-in-progress, but rather modern efforts to resuscitate faded ink with the use of chemical reagents, as Wrenn surmised.

The evidence from ultraviolet fluorescence is equally decisive in the other areas of the folio that were swabbed with reagent where the writing was gone (Fig. 5). There would be no reason, after all, to use reagent in these areas unless the writing was in fact illegible, as it must have been at least by the time that Thorkelin and his scribe copied the manuscript.

Figure 5

[Fig. 5: UV showing use of reagent on 179r18–21 and 179v1–2]

What is surprising are the many records showing that readers at the British Museum in the second half of the nineteenth century (and perhaps later) were allowed to use chemicals on manuscripts. P. R. Harris reports that “permission was granted by the Trustees on a number of occasions to apply chemicals to manuscripts to attempt to make faded writing legible. When C. T. Martin of the Public Records Office wished to treat some Cotton manuscripts in this way in 1870, [the Keeper of Manuscripts, Edward Augustus] Bond was of the opinion that no damage would thus be caused to the documents.” Harris also notes that “Professor William Wright of Cambridge (formerly Assistant Keeper of Manuscripts) was allowed to apply chemicals to various Syriac manuscripts in 1871 and 1872, and in 1873 William Stubbs was permitted to use a similar process on Cotton Tiberius G. xv.”35

After 1873, it appears that special permission from the Trustees was no longer required for the use of reagents. Andrew Prescott, who has searched both catalogued and uncatalogued archives in the Department of Manuscripts, has not found any references to the use of reagents on any manuscripts, including Beowulf, after this time. Yet Bond, who believed their use was harmless to manuscripts, continued on as Keeper for another five years, and his successor, Edward Maunde Thompson, was still endorsing their use twenty years later in his Handbook of Greek and Latin Palæography.36 In fact, as late as 1912, Thompson repeats his endorsements in his discussion of palimpsests in An Introduction to Greek and Latin Palæography. In both works he says, “even if, to all appearance, the vellum was restored to its original condition of an unwritten surface, yet slight traces of the text might remain which chemical reagents . . . might again intensify and make legible.” He advises that, “of modern chemical reagents used in the restoration of such texts the most harmless is probably hydro-sulphuret of ammonia.”37 In view of Bond’s and Thompson’s advocacy of reagents as well as the clear ultraviolet evidence of their unrecorded use on fol. 179 of the Beowulf manuscript, it seems safe to assume, as Prescott suggests, that the requirement for the Trustees’ approval was dropped after 1873.

Prescott has advanced in personal correspondence the persuasive theory that the Keeper himself, Edward Maunde Thompson, must have authorized, and may have supervised or even undertaken, the use of reagents on the palimpsest, when the facsimile was prepared for Zupitza by Praetorius (Zupitza, Beowulf [1882], p. xviii). Prescott believes that it would be entirely consonant with the attitudes of a Victorian curator, who was confident that at least some reagents were harmless to manuscripts, to apply the chemicals, photograph the results for a permanent record, and not worry too much about possible degradation afterwards. In his acknowledgments Zupitza, who did not participate in the photographing of the manuscript, says that his “warmest thanks are due . . . especially to my friend Mr. E. Maunde Thompson, the Keeper of the Manuscripts in the British Museum.”38 The “autotypes” prepared by Praetorius for Zupitza offer some support for this theory. In a few places they seem to show a bit more ink traces than modern photographs, and in one case a careful comparison of the high-resolution digital image with Zupitza’s autotype has convinced me that Zupitza was right about a restored reading that nearly everyone, including me, has rejected as impossible.

Where Zupitza says for the illegible section on 179r5 that “what is left of the two letters after hea- justifies us in reading them ðo,” almost all editors have read um and confidently restored heaum. And while they are less confident about the following word, and variously restore it as hæþe, heþe, hofe, and hope, the editors are united in rejecting Zupitza’s hlæwe. According to Zupitza, “very little of hlæwe [was] freshened up; the h indistinct, læwe pretty certain, but the w may be easily mistaken for þ in consequence of the h of [nat]hwylc on fol. 179v being visible through the parchment.” Klaeber’s comments on Zupitza’s readings are representative. He says that all the letters are “very indistinct,” and that “ðo seems too short and hlæwe too long for the space in the manuscript.” Agreeing with these observations, I once thought that Zupitza’s “h indistinct” must be an error for “h is distinct,” because like Klaeber I interpreted the lone surviving dark ascender as the top of an h.

As we have seen, digital imaging and image processing programs open new possibilities for editors intending to conserve manuscript readings (as Zupitza, Klaeber, and I all intend in this case) to analyze this kind of evidence more accurately. Digital imagery makes it easy to compare the extreme differences between an image of the palimpsest acquired in bright light and one acquired using ultraviolet. Textual scholars can now use programs like Adobe Photoshop to cut and paste scribal letterforms to test conjectural restorations through its Layers function. John Pope used photocopies in the late 1970s to undertake a similar restoration of a damaged passage in the Exeter Book by physically cutting and pasting letters from other parts of the photocopy of the manuscript.39 Following his lead, I developed under the Electronic Boethius project an OverLay tool to assist editors to make digital facsimile editions including conjectural restorations, in order to show the paleographical justification for conjectural restorations in print editions.40 OverLay also facilitates the comparison and descriptive tagging of multiple sets of digital images, one set captured using bright light and the other using any number of other techniques, such as fiber-optic backlighting and ultraviolet fluorescence. An editor can also use OverLay to help isolate and digitally remove manuscript offsets and shine-through.

This flexible tool lets editors overlay one image with another and then minutely examine the differences between digital images. One can use it, for example, to analyze in detail letterforms written by the second scribe or Nowell with the nathwylc scribe. With its help I was able to superimpose Zupitza’s 1882 autotype of fol. 179 on top of a high-resolution image of the same page illuminated with bright light and captured with a powerful digital camera. The process does not in any way affect the original image but instead generates an independent digital file combining the original image and an overlay incorporating conjectural restorations. To acquire an overlay file of these two images, one first selects two corresponding points on each image (in this case I chose the tips of the h-ascenders in hea and hord), and then clicks the OverLay icon. By means of a slider that slowly increases and decreases, as desired, the transparency of the superimposed image, one can closely observe the differences between Zupitza’s autotype and the modern image. A selection feature allows the user to isolate the section from hea through hord for special examination. When associated with a corresponding transcription, one can then use a tagging feature to encode the selection with descriptive markup (including a textual note with automatically recorded coordinates of the place in the image) for later searching or display.

With the aid of these tools I found that the 1882 autotype in this case shows clearer ink traces than the digital image does today.

Figure 6

[Fig. 6: Highlight of OverLay of Autotype for 179r5]

Using the “transparency” slider in the lower left of Fig. 6, I could minutely compare the two images letter-by-letter, trace-by-trace. As a result of this investigation I now realize that Zupitza meant that the clearly visible ascender belongs to the l, not the h, which itself is faintly visible (or “indistinct,” as Zupitza said) to the left of l in the autotype. The h-ascender begins slightly to the left of the r-descender of deorcum in the line above. Because it failed to adhere, the lower half of the h-shaft appears to be the last minim of an m. Though faint, its ascender is clearly visible in the autotype, as is the top and most of the right side of the h-bow. The rest of hlæwe is clearer in the autotype than it is today in the manuscript. These results indicate that the autotype preserves an example of how a reagent can improve a reading, and then destroy the evidence, the very reason we now deplore the widespread practice of using chemicals on manuscripts during the nineteenth century.

Although most editors have considered the reading heaum here as virtually certain, there were in retrospect always reasons to doubt it. In the first place, the usual form of heah in the dative singular is weak hean (used four times) not heaum; strong heaum occurs nowhere else in Beowulf. Moreover, there are ten other words on this folio that end in –um, and in every case the scribe uses a macron over u as the abbreviation for m (deorcum r4, nihtum r4, eldum r7, hæðnum r8, geweoldum r14, willum r15, -gum AB v5, mælum v8, wæteryðum v13, cræftum v14). It is difficult to make a convincing case that the one exception occurs in a badly damaged, illegible part of the folio, especially after close scrutiny of the tops of the supposed minims. Klaeber’s expressed view that there was not enough room in the manuscript for Zupitza’s readings is, moreover, demonstrably unfounded. By cutting and pasting the nathwylc scribe’s letterforms from elsewhere on the folio, I tested Zupitza’s conjectural restoration. Before the advent of image-processing programs, editors could only make conjectural restorations in printed editions without having to show whether or not the reading actually fit the space or otherwise satisfied the paleographical circumstances. Now one can see in a virtual restoration of the scribe’s own hand that there is in fact sufficient space for heaðohlæwe with all restored letters coinciding with the remaining ink traces (see Fig. 7):

Figure 7

[Fig. 7: Digital restoration of heaðohlæwe, 179r5]

The conjectural restoration of heaðohlæwe, “battle-barrow, battle-burial-mound,” seems far superior to all previous restorations (heaum hæþe or heþe, “high heath,” heaum hofe, “high dwelling or court,” heaum hope, “high plateau”). Poetic compounds with heaðo- are among the most common in the poem, very frequently occurring in hapax compounds, as in this case. The second element, hlæw, is also a common word in the second part of Beowulf. This collocation is particularly appropriate for the barrow where Beowulf holds his last battle. The verse type, se ðe on heaðohlæwe, is described in various ways by different metrists in comparable instances: Ac ic ðær heaðufyres 2523(2522)a and nemne him heaðobyrne 1554(1552)a are Sievers C, Bliss d2c, Hutcheson 3C7I, Hoover nAn, and Russom x/Ssx.41

The same image-processing software that helps provide paleographical support for the conjectural restoration of heaðohlæwe makes it possible for editors to attempt to justify by means of a restored facsimile any conjectural restorations that purport to represent readings supported by the surviving ink traces.42 In this case I restricted my choices to letters from this single folio, as they are the only ones written by the nathwylc scribe. The OverLay tool can further enhance this process, as well. By overlaying the restored folio (the facsimile edition) on the original unrestored one (the original facsimile), scholars, using the slidebar in the lower left corner of the tool, can minutely examine any restored reading compared to the ink traces. As we have seen, by this means editors of electronic editions are able to provide paleographical support for conjectural restorations. Because the restored readings are digital, editors and textual scholars can and should continue to propose new restorations or defend old ones.43

The OverLay tool is an analytical resource easily adaptable to other manuscript problems, both on the palimpsest and throughout the manuscript. For example, the tool is especially useful on almost every folio for resituating along the damaged edges the hundreds of covered letters and bits of letters, restored by fiber-optic backlighting, on the versos. In the Electronic Beowulf these restored readings are currently available in browser frames as illustrated textual notes, supplementing and correcting Zupitza’s 1882 observations on covered readings. Often only slight traces or fragments of letters survive beneath the paper frames, and while fiber-optic backlighting reveals what remains, it is not always easy for a reader to see exactly where the backlit image belongs on the covered edge. By correctly matching two points on the edge of the paper frame, as illustrated in Fig. 8, an editor can use the OverLay tool to restore covered readings wherever they belong:

Figure 8

[Fig. 8: OverLay of covered edge with backlit g, 179v19]

As illustrated in Figs. 9 and 10, the OverLay tool can also help address more difficult editorial problems. Offsets (caused by ink transferring from a facing page) and shine-through (caused by text showing through from the opposite side of a page) are common in medieval manuscripts, both creating obstructive reverse or mirror images. Between lines 8 and 13 of the palimpsest there are many continuous offset letters from the facing folio, 178verso, which make it almost impossible to isolate the remaining legitimate ink traces from the faded text on 179 recto.44 Because they coincide with the areas where reagent was applied, these offsets were plausibly caused when the manuscript was closed while the vellum was still damp with reagent. The digital medium makes it relatively easy to discover and display how an offset from the text on fol. 178v further obscures the damaged readings between lines 8 and 13 on fol. 179recto of the palimpsest.

Figure 9

[Fig. 9: Inverse image of fol. 178v9–11 with background removed]

The OverLay tool has access to its own mini-image processor, which allows the editor to reverse an image, intensify the writing, and render the background vellum transparent. Using these features, I first created an inverse image of fol. 178v; then highlighted the text on the folio and removed the lightened background; next, after locating points where offsets correspond to places on the inverse image of 178v, I overlaid this reversed text onto fol. 179r:

Figure 10

[Fig. 10: OverLay of 179r10–12 revealing offsets from 178v9–11]

The overlay shows that the inverse of the text on 178v10, [þu]sendo. bold 7brego stol him wæs, is offset on 179r10. The inverse dl- of bold immediately precedes slæpende, and much of the rest of the inverse line shows up as spurious ink traces within and beneath 179r10. After identifying the spurious offset ink traces, the editor can use another feature of the OverLay tool, as shown in Fig. 11, to remove the offset of inverted him wæs below syððan in order to focus on the legitimate ink traces from fol. 179:

Figure 11

[Fig. 11: Removing offsets from 179r10]

With the offsets removed, an editor can more easily evaluate the ink traces that remain on the palimpsest and offer more plausible conjectural restorations for the faded text.

Editors of print editions have customarily made conjectural restorations while paying attention to sense and meter, but often without sufficient regard to the manuscript evidence. In a recent example, Mitchell and Robinson acknowledge in their note to these lines that “before syððan [i.e., where the vellum has crumbled away at the end of 179r9] Thorkelin A has crossed thorn for þæt, Thorkelin B þæt,” but they still remove “that” from their text on the grounds that it “would make no sense” (Beowulf, p.126). They give no justification for doubting that a crossed thorn still survived on the burnt edge of the manuscript, which Thorkelin typically expanded to þæt, but which his scribe would accurately transcribe as a crossed thorn. The reason the demonstrative þæt seems to make no sense for Mitchell and Robinson is that, following it, they have proposed a conjectural restoration that renders it meaningless (i.e., “þah] MS illegible”). Klaeber’s conjectural restoration, bemað, which according to Fulk “seems the best guess on paleographical grounds,” rather surprisingly does not fit in the space between syððan and the isolated þ, as the digital restoration in Fig. 12 illustrates.

Figure 12

[Fig. 12: Conjectural restoration of bemað 179r10]

Although bemað is only five letters long, the letters m and ð are both wider than all other letters. Taken together, they add in effect the width of an additional letter. In Fig. 12 I have digitally cut and pasted letters from other parts of the folio to test Klaeber’s restoration. As the reconstruction makes clear, bemað is too wide to fit the space, and the cross-stroke of ð interferes with the þ-ascender. In this hand the cross-stroke of ð always extends beyond the bowl (see the two extant examples in this line in Fig. 12), and the cross-stroke would interfere with the þ-ascender if the last letter were indeed an ð (cf. cuð þær 179r7, three lines above). Even an unusually stunted cross-stroke would not provide enough room for bemað. In an edition that purports to be based on the manuscript, paleographical evidence should of course carry weight.

My own restoration, bohte (“paid for”), in Electronic Beowulf [2.0] fills the space, does not conflict with the remaining traces of ink, and makes sense in the context. However, while it is metrical in an on-verse, the type does not occur with anacrusis in an off-verse, as Fulk has pointed out. Models of normal meter in Beowulf should carry weight when a manuscript reading is not certain. I have accordingly abandoned bohte as an acceptable conjectural restoration, and in the third edition of Electronic Beowulf offer instead beget, third singular preterite of begytan, in the sense of “to get, obtain, acquire; to take hold, seize.” In Fig. 13 I have taken screen shots from the new overlay facility in the third edition of Electronic Beowulf to defend my conjectural restorations of 179r10–13:

Figure 13

[Fig. 13: Conjectural restoration of 179r10–13]

           (He þæt) syððan [beget]
þ[eah] ð[e] [he] slæpende   (be)syre[d] [hæf]de
þeo[f]es cræfte;   þæt si(e) ðiod [onfand],
b[u]folc beorn[a],   þæt he gebolge[n] wæs.

“He [i.e., niða nathwylc, “I know not who of men”] afterwards got that [i.e., the stolen item], though he had swindled the sleeping one [the dragon] by a thief’s craft; so that his countrymen found out, a nearby-folk of fighters, that it was infuriated.”45 A digital restoration provides paleographical justification for a conjectural restoration in an edition. The restoration beget fills the space, does not conflict with the few remaining traces, makes sense in the context, and also provides unobjectionable meter across many systems (e.g., Sievers B, Bliss 3B*1b, Hutcheson 2B2b, Hoover nAn, Russom x/Sxxs).

In 1981 I argued that a “new edition of the text on the palimpsest is desperately needed, but must await further attempts, by modern technological methods, to decipher all of the extant ink traces.”46 My view then as now was that a “conservative edition,” one that intends to conserve the manuscript text, “will strive to retain all of the readings that are clear in the MS, and will try to make sense of the text by emending conjectural restorations (for which there is little or no MS support), Thorkelin readings,47 and then doubtful MS readings” (p. 243).48 Over the past twenty-five years I have had many opportunities to explore these paths, first with a digital camera creating grayscale images for medical image-processing, and then in 1993, with the start of the Electronic Beowulf project, with an all-purpose camera that the British Library purchased for the project to prepare high-resolution 24-bit color images.49 Each stage of exploration has led me to refine, and sometimes recant, aspects of earlier arguments. I believe it is not a scholarly virtue to be inflexible or dogmatic in the face of new, or for that matter, uncertain evidence.

As I have tried to epitomize in various ways in this essay, I am also convinced that image-based electronic editions of a text like Beowulf have some clear advantages over traditional print editions such as Klaeber’s. With a high-resolution facsimile, there is no reason not to have every textual note linked to its source in the manuscript for all to examine and assess. Print has fostered the idea of an established text, whereas digital editions are now free to present texts in a less-settled state, the way they exist in manuscripts. As I have explained under Edition in the Help facility of the Electronic Beowulf: “The new editorial emendations and conjectural restorations are not meant to replace, but rather to supplement or challenge previous editorial changes that have become ingrained in modern editions. Anglo-Saxon readers of this manuscript did not have access to any modern changes and perhaps devised different solutions each time they encountered problems that demanded emendations. The conjectural restorations are informed guesses that try to fill gaps in paleographically and linguistically plausible ways.” The great value of digital technology is that it encourages continuing research of textual issues and obviates the need, required of print editions, to establish a text. The manuscript record left by the three Beowulf scribes, in particular the nathwylc scribe of fol. 179, continually reminds us that there was no established text in Anglo-Saxon times. There is no definitive edition of Beowulf today, just as there never was one in Anglo-Saxon times. Electronic editions will help us understand the pleasures of a constantly moving text more than printed editions. The high-frequency occurrence on the palimpsest of the otherwise low-frequency word, nathwylc, can perhaps help focus attention on some important theoretical and practical issues surrounding the editing of Beowulf.50

Notes

1. In 1882, Julius Zupitza published Beowulf: Autotypes of the Unique Manuscript Cotton MS. Cotton Vitellius A.XV in the British Museum (EETS 77). A revised edition was published in 1959, with a new reproduction of the manuscript and an introductory note by Norman Davis (see the bibliography). The first edition will be cited as Zupitza, Beowulf (1882). The revised edition will be cited as Zupitza, Beowulf (1959). Quotation at Zupitza, Beowulf (1959), p. 102.

2. Helen Damico offers a new hypothesis on the historical context of the poem within the dating range of the script in “Beowulf’s Foreign Queen and the Politics of Eleventh-Century England.” In (Inter)texts: Studies in Early Insular Culture Presented to Paul E. Szarmach, edited by Virginia Blanton and Helene Scheck, pp. 209-40. Medieval & Renaissance Texts & Studies 334. Tempe, AZ: AC-MRS, 2008).

3. The result of these two stages is a palimpsest, in its primary sense of “writing-material written upon twice, the original writing having been erased or rubbed out to make place for the second” (OED). In addition to a few uncorrected mistakes, its draft status is most evident from the retention of the catch-word, sceapen, the unerased remnants of what appears to be an extensive dittographic passage, at the end of the recto and start of the verso.

4. When it first appears on the palimpsest, the word nathwylc refers to the thief (niða nathwylc, “I know not who of men,” 2218); the second time it apparently refers to the theft (þeof nathwylces, “thief of I know not what,” 2226); and the third time it refers to the person who originally hid the treasure (gumena nathwylc, “I know not who of men,” 2234). Nathwylc is first used as a substantive in the second scribe’s section at line 2056: Nu her þara banena byre nathwylces, “Here now a son of I know not who of those slayers.” The word is also used once as an adjective in the first scribe’s part of the manuscript: þæt he niðsele nathwylcum wæs (“that he was in I know not what hostile hall”) line 1515.

5. I previously resisted this interpretation of the letter, considering it instead a defective æ in Beowulf and the Beowulf Manuscript, (New Brunswick, NJ: Rutgers University Press, 1981; Rev. ed. Ann Arbor MI: University of Michigan Press, 1996), p. 33. Now that a conservator has trimmed away the opaque tape on the top of this letter, I agree with Malone's identification of a unique occurrence in the Beowulf manuscript of the so-called “oc-a” (Nowell Codex).

6. Kiernan, The Thorkelin Transcripts of Beowulf, Anglistica 25 (Copenhagen: Rosenkilde and Bagger, 1986), pp. 80–81, 113–14, 144. See also Kiernan, Electronic Beowulf, for images and collations of the Thorkelin transcripts.

7. See the note on fol. 179 in the Kiernan, Electronic Beowulf 2.0 Help section under Electronic Edition: “my latest opinion of this folio, a palimpsest, is that it exhibits the handwriting of a third scribe, contemporary with the two main scribes, and that the text on 179 reflects a late revision-in-progress of the poem. Much of this new text failed to adhere to the poorly prepared vellum. I now believe that C.L. Wrenn was right when he proposed that someone in modern times applied a reagent to these faded areas (the reagent fluoresces strongly under ultraviolet).”

8. F. Klaeber, Beowulf, and the Fight at Finnsburg, 3rd edition (Boston: D. C. Heath 1950), p. 82, textual note.

9. R. D. Fulk, “On Argumentation in Old English Philology, with Particular Reference to the Editing and Dating of Beowulf,” Anglo-Saxon England 32 (2003), pp. 15–16; and following examinations of the manuscript, “Some Contested Readings in the Beowulf Manuscript,” Review of English Studies 56 (2005), 192-223. The fourth edition of Klaeber’s Beowulf and the Fight at Finnsburg, edited by Fulk, Robert E. Bjork, and John D. Niles, bases its readings on the theory without critically examining it (p. xxix). This edition appeared in 2008, after I had written this essay.

10. Fulk, “Some Contested Readings,” p. 209. Werner, from the British Museum Research Laboratory, was drafted by Norman Davis to “judge of the mere physical appearance of the vellum.” He does not presume to verify that a later hand freshened up everything that is distinct on the folio. In a brief statement he says only that “it would appear that attempts have been made in the past to revive the ink in certain areas which appeared to have been damaged. There would also appear to be definite signs of subsequent intensification of certain letters, which now appear darker and have a characteristic sheen; the most obvious example is the initial N on line 14” (Zupitza, Beowulf [1959], pp. vi–vii).

11. Tilman Westphalen, Beowulf 3150–55: Textkritik- und Editionsgeschichte (Munich: Wilhelm Fink, 1967) pp. 58–69.

12. Kiernan, “The Palimpsest and the New Text of Folio 179,” Beowulf and the Beowulf Manuscript, 219-243. Fulk falsely asserts, “The ‘revised edition’ . . . is actually a reprint with two new prefaces and a reprint of a 1983 article” (“On Argumentation,” p. 9, note 19). My text includes revisions (see, for example, note 47 below) and my preface, “Re-Visions” (i.e., revisions), even briefly discusses Fulk’s theories (pp. xxiv–xxvii). The appendix, “The State of the Beowulf Manuscript, 1882–1983” from Anglo-Saxon England 13 (1983), adds important textual evidence, not available when the book first appeared, on hundreds of covered readings along the damaged edges of the restoration binding. “Until the advent of digital technology,” I explain in “Re-Visions,” “there was no practical way to reproduce the hundreds of backlit readings. With the aid of a digital camera and digital image-processing, however, I am now preparing digital facsimiles of all of these covered readings as part of the Electronic Beowulf project” (p. xxviii). The project won an international award for innovation in information technology for developing this method. In fact, Fulk uses the method I developed for disclosing covered readings with fiber-optic backlighting in his article on readings he contests. However, rather than providing visual evidence for his readings for other scholars to examine, as I do in Electronic Beowulf, Fulk merely asserts that he saw something else when he looked at the manuscript.

13. Both scribes continue to copy the following fitt-numbers as if XXIIII still existed. For a full discussion, see Kiernan, Beowulf and the Beowulf Manuscript, pp. 264–70. For a different accounting of the numbering, see Patrick Conner, “The Section Numbers in the Beowulf Manuscript.” ANQ 24.3-4 (1985): 33-38. His argument also suggests that the two scribes were following different exemplars.

14. If the scribe were simply ending a stint, there would be no reason not to stop in the middle of another gathering, as the first scribe did on fol. 172v3. There is an extensive discussion in Beowulf and the Beowulf Manuscript, pp. 133–50, and an extracted argument in my “The Eleventh-Century Origin of Beowulf and the Beowulf Manuscript,” The Dating of Beowulf, ed. Colin Chase. Toronto Old English Series 6 (1981; Toronto: University of Toronto Press, 1997), 9-21.

15. For an unduly influential example see Michael Lapidge, “The Archetype of Beowulf.” Anglo-Saxon England 29 (2000), 5-41. In a circular argument, Lapidge uses modern conjectural emendations as his only evidence of scribal error, rather than the many scores of examples of scribal corrections of undoubted mistakes. Moreover, by citing only edition line numbers (e.g. “MS 3154”), Lapidge fails to locate for his readers the supposed errors in the manuscript by folio, folio line, and edition number (e.g., fol. 198v3:3154). In fact, Lapidge confuses matters further through a confused understanding of the multiple foliations of the Beowulf manuscript. He first describes the “manuscript of Beowulf,” by which he must mean the Nowell Codex, as comprising “fols. 94–209” (p. 7), but in a footnote on the same page he refers to “the last page of Beowulf” (not Judith) as fol. 209. Beowulf of course ends on fol. 201v in this foliation, or on 198v of the foliation written on the manuscript. Lapidge later inadvertently uses this manuscript foliation, when he says, “I cite no example from the most illegible leaves, such as 179r (= lines 2207–39) and 198v (lines 3150–82)” (p. 9), although he subsequently uses examples from these leaves, which he again cites by the manuscript foliation (p. 21, note 56). For a comprehensive discussion of scribal corrections, see my “Proofreading of the Scribes,” in Beowulf and the Beowulf Manuscript (pp. 191–218), which Lapidge does not cite.

16. The spelling Beowulf, instead of Biowulf, is in itself unexpected in this part of the manuscript. The second scribe uses the io spelling fourteen of sixteen times in his stint, and the two exceptions, at 173r14 and 185v13, first had io spellings, too, but either he or, more likely, the nathwylc scribe, later emended them to eo. For illustrations of these two scribal corrections, see “Re-Visions” in the second edition of Beowulf and the Beowulf Manuscript, pp. xxvi–xxvii.

17. Higelac here gives Beowulf seven-thousand unspecified units of land (seofon þusendo), but he later gives Wulf and Iofor one hundred thousand of land (hund þusenda landes, 2994–95) and at the same time gives Iofor his only daughter to wed (2997–98). From these figures it is hard to justify the widespread assumption that Beowulf somehow ended up with more property without the daughter and the dowry.

18. OES §1978 in Mitchell’s Old English Syntax II (Oxford: Clarendon, 1985) does not discuss these lines. The unusually elaborate section title, “Part Two: Beowulf the King (lines 2200–3182): The Fight with the Dragon (lines 2200–2751): [The accession of Beowulf. The plundering of the dragon’s hoard],” highlights the importance of this passage in their edition (Bruce Mitchell and Fred C. Robinson, eds., Beowulf: An Edition (Oxford: Blackwell, 1998) p. 125. The manuscript fitt-numbers, however, present a different structure, as this turning point actually occurs in the middle of fitt XXXI in the manuscript.

19. Carl T. Berkhout, “Beowulf 2200–08: Mind the Gap,” ANQ 15.2 (2002): pp. 52–53.

20. Berkhout, “Anglo-Saxon Studies in the Age of Shakespeare,” Old English Newsletter (1986), A-28. It is not clear why Berkhout rules out a post-Conquest Anglo-Saxon scribe for his refreshener. The Anglo-Saxons did not abruptly stop speaking and writing Old English the day (or year) that the Normans conquered them.

21. Carl T. Berkhout, “Laurence Nowell (1530-ca. 1570),” Medieval Scholarship: Biographical Studies on the Formation of a Discipline, Vol. 2, ed. Helen Damico (New York: Garland, 1998) pp. 12–13.

22. Johan Gerritsen, “Have with You to Lexington! The Beowulf Manuscript and Beowulf,” In In Other Words: Transcultural Studies in Philology, Translation, and Lexicology presented to Hans Heinrich Meier on the Occasion of His Sixty-fifth Birthday, ed. J. Lachlan Mackenzie and Richard Todd (Dordrecht: Foris, 1989) 15-34. According to Berkhout, his 1987 remarks at ISAS “are more or less embedded in Gerritsen’s 1989 article” (“Beowulf 2200–08: Mind the Gap,” p. 57, note 1).

23. Gerritsen, “Have with You to Lexington!” p. 31, his italics.

24. There are accurate illustrations in Westphalen (Beowulf 3150–55 pp. 66–67). The one Michael Lapidge provides of “the square-shaped box-like letter a . . . the shibboleth by which Anglo-Saxon Square minuscule is identified” (“Archetype of Beowulf,” pp. 12–13) is a letterform used by neither the second scribe nor the nathwylc scribe of Beowulf. When they use a flat-topped a, both scribes begin with the left side and bottom of the square a, but draw the top hair-stroke second (not third) and continue its apex downward for the right side and off-stroke to complete the letter. Because of the oblique hair-stroke, the result for this a is often more triangular than square.

25. A. H. Smith, “The Photography of Manuscripts,” London Mediæval Studies 1 (1938), p. 200.

26. In addition to wintra r2, see stearne r6, innan r7 (where light erasing was also done), he (?) r12, long v11, wunode v12, “mæstan” [i.e. mostan] v17, and feorh v19.

27. The results are somewhat biased in that Sievers and Bliss simply accept the traditional editorial restorations, which of course reinforce standard metrical solutions. Hutcheson, Hoover, and Russom use the same restorations in their metrical data, but openly mark them as uncertain readings.

28. Fulk erroneously says that I explain rihde “as wende with portions of w and e faded” (“On Argumentation,” pp. 15–16); in fact I explicitly say, “there is no trace of an original wynn under the r” (Beowulf and the Beowulf Manuscript, p. 241).

29. Davis inclines to this view, too: “the form of letters usually read ih suggests that they may have been en”. He then cites Sedgefield: “MS. rende; in the first e the fine upstroke is visible, but the middle horizontal bar is missing; certainly not rihde, as Zu. says” (Zupitza, Beowulf [1959], p. vi). Cf. Malone, Nowell Codex, p. 86, and Kiernan, Beowulf and the Beowulf Manuscript, pp. 240-241.

30. Because both innan and innon are attested in Beowulf, I do not count innan v14, where the digital images seem to confirm that the off-stroke of a is simply faded. Nor do I count what editors have read as þana v20, which I believe is þame, another trompe l'oeil caused by fading of the top of the last minim of m and the e-head and e-tongue. The second “a” is not a true letterform. The issue is whether or not there is a better way to fix the textual problem than the traditional emendation, þara, properly þa[ra]. This cascading “solution” depends not only on emending n to r, but also on interpreting the following strokes as a unique form of a; and finally on the additional interpolation of the word lif later in the same half-line, an extreme emendation-cluster based on an uncertain reading. If one instead conjecturally restores þa [me]ðe þis ofgeaf (“then weary left this behind”), there is alliteration on leoda minra. Some might prefer to regularize the alliteration and the syntax by emending leoda minra to minra leoda (cf. eowra leoda 633b, þinra leoda 1675b, þara leoda 2036b; minra nowhere else comes second in similar constructions in Beowulf).

31. Fulk espouses meaningless bealc instead of damaged bealo or beale, a weakened spelling of bealo (cf. bealewa, bealwa, where one would expect bealuwa, bealowa). Here -beale is the second element in the compound feorhbeale in the formula feorhbeale frecne. According to Kaluza’s “law,” which Fulk attempts to revive as a criterion of early date, this weak syllable is “resolved” (i.e., not pronounced) in certain cases in the meter. Roberta Frank has recently challenged Fulk’s dating argument, pointing out that an Anglo-Dane in late Old English times would not pronounce these light final vowels, as well (“A Scandal in Toronto,”: The Dating of ‘Beowulf’ a Quarter Century On,” Speculum 82.4 (2007), pp. 858–59. I also mention North Germanic, or Anglo-Danish, spoken dialects in eleventh-century England as alternative explanations for early dating arguments in “Re-Visions” (pp. xxv–xxvi).

32. It appears from the ultraviolet images that two different kinds of reagent might have been used on fol. 179r recto. Compare the ultraviolet effects on hwylces r17 with those in the following lines in Kiernan, Electronic Beowulf.

33. C.L. Wrenn, “Introduction: 1. The Manuscript,” in Beowulf, with the Finnesburg Fragment, 3rd ed. rev. W.F. Bolton (New York: St. Martin’s Press, 1973) p. 12. Fulk suggests without citing new evidence that A. E. A. Werner, not Wrenn, was the first to note that “a chemical of some sort has been applied to certain portions of the text” (“Some Contested Readings,” p. 209). But Werner's brief report to Davis, which does not mention chemicals, was included in a 1957 letter, four years after Wrenn’s edition (Zupitza, Beowulf [1959], pp. vi–vii).

34. Malone explains in his “Note to the Photographs” that the official photographer of the British Museum photographed the Beowulf manuscript in 1957 for the second edition of Zupitza’s facsimile, and that he used the same negatives for his facsimile. Thus he and Davis list fol. 179 (182 in their numbering) among those photographed by ultraviolet. See Malone, Nowell Codex, p. 120; and Davis, “Note to the Second Edition,” in Zupitza, Beowulf [1959], p. v.

35. P. R. Harris, A History of the British Museum Library, 1753-1973 (London: The British Library, 1998) p. 276.

36. Edward Maunde Thompson, Handbook of Greek and Latin Palæography (London: Kegan Paul, Trench, Trübner & Co., 1893) p. 76.

37. Edward Maunde Thompson, An Introduction to Greek and Latin Palaeography (Oxford University Press 1912; reprint, New York: B. Franklin, 1965) p. 65. It is possible that Raman microscopy (“a non-destructive, non-invasive and extremely specific technique,” according to David Jacobs, senior conservation officer in the British Library) can safely and definitively identify the chemical reagent used on the Beowulf palimpsest. See Jacobs’s contribution, “An Introduction to The British Library and University College London Raman Microscopy Project” in Michelle P. Brown, The Lindisfarne Gospels,: Society, Spirituality and the Scribe, The British Library Studies in Medieval Culture (London: The British Library, 2003) pp. 430–33.

38. There may well have been additional pressure to try reagents around this time, as Eugen Kölbing in 1876 and Alfred Holder in 1882 both published results from collations with the manuscript, which Zupitza duly lists on the same page as “books referred to” in his notes (Zupitza, Beowulf [1959], p. xx).

39. Pope, “Paleography and Poetry: some solved and unsolved problems of the Exeter Book,” Medieval Scribes, Manuscripts and Libraries: Essays Presented to N.R. Ker, ed. M.B. Parkes and Andrew W. Watson (London: Scolar Press, 1978) 25-65. See my discussion of this article in “Old Manuscripts / New Technologies,” Anglo-Saxon Manuscripts: Basic Readings, ed. Mary P. Richards. Basic Readings in Anglo-Saxon England 2 (New York: Garland Press, 1994; New York: Routledge, 2001) pp. 44–48.

40. “The Electronic Boethius: Alfred the Great’s Old English translation of Boethius’s Consolation of Philosophy” was funded by the National Endowment for the Humanities and the Andrew W. Mellon Foundation, with the cooperation of the British Library and the Bodleian Library, Oxford. When the editing project was overtaken by “The Alfredian Boethius Project: Anglo-Saxon adaptations of the De Consolatione Philosophiae,” directed by Malcolm Godden and Susan Irvine, I redirected the resources of the Electronic Boethius project to develop the editing software. In recent years Ionut Emil Iacob and I, with the support of the Andrew W. Mellon Foundation, have extended this software as the generic, open source, image-based XML suite of tools called EPPT (Edition Production & Production Technology), which is now developed as IBX (Image-based XML). For a discussion of the early stages, see Kiernan, “Remodeling Alfred’s Boethius with the tol ond andweorc of Edition Production Technology (EPT),” Making Sense: Constructing Meaning in Early English, edited by Antonette diPaolo Healey and Kevin Kiernan. Publications of the Dictionary of Old English 7 (Toronto: University of Toronto Press, 2007), 72-115.

41. Despite the bewildering array of codes for describing metrical types, metrical aficionadios are fond of saying that they essentially agree with one another’s metrical analyses. It would be a fascinating and salutary advance in metrical studies of Old English poetry in general, and of Beowulf in particular, if a member of the guild of metrists would organize comprehensive metrical statistics for the leading metrical theories, with each type (including “hypermeter”) linked to its specific instance. [See now Meter tooltips and Meter Statistics in Electronic Beowulf 3.0.]

42. The best software for this purpose is, in my experience, Adobe Photoshop, which provides a range of sophisticated image-processing tools for creating conjecturally restored facsimiles that an editor can then use as an OverLay image with the unrestored manuscript image. While generously acknowledging many of the advantages of Electronic Beowulf, J. R. Hall complains that one cannot alter the lighting of the color images in the main frames nor even magnify the ultraviolet images in the side frames (“Three Studies on the Manuscript Text of Beowulf,”: Lines 47b, 747b, and 2232b,” Beatus Vir: Studies in Early English and Norse Manuscripts in Memory of Phillip Pulsiano (Tempe, AZ, 2006), p. 457). In fact, one can use Photoshop to magnify, or change the lighting or the contrast, or otherwise process any of the images in Electronic Beowulf.

43. After spending some time with Electronic Beowulf at the British Library, Frederick Biggs remained agnostic over whether a þegn or a þeof stole from the dragon’s hoard in “Beowulf and Some Fictions of the Geatish Succession,” Anglo-Saxon England 32 (2003), p. 61, note 29. Both words fit the space, but the vestiges of ink do not serve both readings equally well. Digital tools like Photoshop and EPPT’s OverLay can help adjudicate such disputes by providing evidence for everyone to examine. [See now the Conjectural Restorations facility in Electronic Beowulf 3.0.]

44. In Electronic Beowulf 1.0 and 2.0, I used Photoshop’s Layers to disclose the offset image from fol. 178v verso. OverLay is now specifically adapted for editors using EPPT for this and many other purposes. [The results of many overlays, including comprehensive conjectural restorations of fol. 179 using the nathwylc scribe's letterforms, are now available in Electronic Beowulf 3.0.]

45. Kiernan, Electronic Beowulf, lines 2220b–23; Klaeber, Beowulf, 2217b–20. In my edition of these lines Thorkelin readings are in round brackets in roman, while my virtual conjectural restorations are in square brackets with italics (crossed thorn abbreviation is rendered by þæt).

46. Kiernan, Beowulf and the Beowulf Manuscript, p. 243. Fulk surprisingly comes to a similar conclusion after discussing the last line on this folio: “It is my belief,” he confesses, “that improved technologies will make it possible in the future to retrieve some readings in the MS that are currently in part or in whole indecipherable, and particularly on this leaf” (“Some Contested Readings,” p. 223).

47. There are still discoveries to be made from the Thorkelin transcripts. A recent example in the Electronic Beowulf is the new reading eorðsele on fol. 179v3–4, which editors have always restored as eorðhuse (see for example Mitchell and Robinson, Beowulf, “2232 eorðhuse] eorð...”), following Zupitza. Thorkelin A does not give a reading, but Thorkelin B has “se..”, not “..se”. Thus B supports the restoration of eorðsele, not eorðhuse, which is used in prose, but nowhere else in Old English poetry; eorðsele, on the other hand, is used twice again in Beowulf. After requesting and reviewing my evidence, the Dictionary of Old English duly added this occurrence to its list. Fulk has recently adopted eorðsele without attributing it to the Electronic Beowulf (“Some Contested Readings,” p. 220). In his history of this recalcitrant reading, Hall observes that I corrected the Thorkelin B reading in my text in the 1996 revised (pace Fulk) edition of Beowulf and the Beowulf Manuscript (“Three Studies on the Manuscript Text of Beowulf,” p. 466).

48. A conservative editor will not emend the manuscript solely on the basis of meter, but will reasonably use standard meter as a guide for conjectural restorations where manuscript readings are not certain.

49. For a chronology of my use of a digital camera with the Beowulf manuscript from 1982 to the Electronic Beowulf project, see “Digital Image-processing and the Beowulf Manuscript.” Literary and Linguistic Computing 6.1 (April 1991): 20-27.

50. I am indebted to Andrew Prescott, in particular for his contributions to the section on the reagent and his preparation of high-resolution images of Zupitza’s autotype of fol. 179r.