A practical method of testing hypotheses

§ 14 Digital reconstruction requires close attention to minute details, and bears resemblance to the traditional method of reconstructing text which involved piecing together missing letters and words by placing pencil tracings of letters on paper mounts in areas where lacunae exist. R. Fulk demonstrates this technique in his examination of the Beowulf MS and successfully retrieves lost readings of the damaged MS, whilst also disproving other suggested readings of problematic areas within the MS (2005, 192-223). Likewise, J. C. Pope used photocopies and a cut and paste method to fill in lacunae, and in the decades and centuries before Fulk and Pope, many textual editors, palaeographers and the like, were utilising the pencil-tracing method of analysis (Pope 1978, 25-65). So, what is the advantage of doing similar work like this using Photoshop or another digital manipulation technique? As long as the digital work is done using high quality images, a similar process of pencil-stenciling to reconstruct folia can be utilised. My method utilised Photoshop's rubber-stamp tool, while monitoring the scale, position, and colour of each letter. This technique proved effective in positioning letters within a damaged area in order to test out various hypotheses relating to possible readings of an incomplete line of text. For instance, line 30-31 reads as follows: Wene ic ful swiþe ond witod . . . . / . . . . . to dæge, dryhten wille [I fully believe and know . . . / today, the Lord will . . .] (Translation provided by the author of this article. Old English text used from ASPR 1936). The line is missing some words and several suggestions have been proposed by previous editors. While J. Cramer (1897, 163) and Grein-Assmann (1898, 176) suggest talige / þætte us [I reckon / that to us], W. Mackie (1934, 174) proposes telle as opposed to talige in l. 30b, followed by þæt usic. Further still, F. Holthausen (1935, 10) queries whether witod[lice] (certainly) might be appropriate for the half-line l. 30b. Holthausen also goes so far as to rearrange words that are already present on the MS and suggests that the half-line 31a should read as þæt to dæge usic (1935, 10). Line 30b only contains one stressed syllable, so lice would offer another stressed syllable and give l. 30b two stressed syllables as opposed to the one syllable that it contains (due to the lacuna). The surviving portions of letters still evident on the outer edges of the MS make Holthausen’s reading for l. 30b unlikely. Certainly, alliteration and meter should be taken into account when suggesting readings, but being unaware of the surviving evidence available on the MS makes reconstructing lines a pointless exercise. Some of the conjectural readings made by previous editors were done without them ever viewing the MS or a facsimile and worse still, some of the suggested emendations daringly rearranged words to suit a hypothetical reading of a line. Irving Jr. states: As editors, we need not go so far as to return to the quaint if lovable practice of the old scholars who cheerfully wrote in some verses of their own to plug such gaps, but it would be well to indicate prominently that a word or a phrase or a verse is missing, or that alliteration is not happening, and to suggest a remedy where possible or supply an obvious word. Scrapping or ignoring the formal system of verse is a poor solution (1998, 17).

§ 15 The problem with making conjectures without seeing a MS results in some hypotheses taking precedence over contradictory MS evidence. Segments of letters are sometimes visible around the outer edges of a lacuna (See Figure 6); however, without seeing or knowing about such MS evidence can result in rather bold and erroneous presumptions as exemplified in Holthausen's hypotheses. It is worth noting that the bulk of Holthausen's philological work on the Exeter Book (relating to fol. 119v-121v) was done in the 1880s and 1890s, so this was well before the 1933 facsimile was created. There seems to be no indication that he was working with pictures and in several of his articles or notes relating to the suggested hypotheses he was responding to other hypotheses from scholars who for the most part had not viewed the Exeter Book or were working from their own notes. This is not to suggest that Holthausen was undertaking guesswork, as he may have been working from B. Thorpe's 1842 edition of the Exeter Book.

Figure 6: Small letter fragments from f. 120r.

Figure 7: Reconstructed area containing letter fragments on f. 120r.

§ 16 Digital reconstructions offer a solution to previous scholarly conjectures without outright dismissals of earlier hypotheses. Using software to digitally reconstruct incomplete lines can help negate away from authoritative restorations of lines in which scholarly hypotheses undermine or ignore existing MS evidence. See Figures 8-11 below (the outline of the lacuna is in blue).

Figure 8: Portion of original damaged f. 120r.

Figure 9: Reconstruction of J. Cramer and Grein-Assmann hypothesis.

Figure 10: Reconstruction of W. Mackie hypothesis.

Figure 11: Reconstruction of F. Holthausen hypothesis.

§ 17 Further to the aforementioned hypotheses, ASPR (1936, 357) argues that because ll. 94, 96, and 126 contain the complete form of the pronoun usic with the ic erased in each case, it is likely that the same is true for l. 31a. Additionally, the descending stroke at least three spaces before the word to in l. 31a is visible, thus reinforcing the assumption of usic as the original reading. ASPR further suggests that þæt us or þ¯te us would fill the rest of the gap, and this reading is supported by Muir (2000, 679). Four spaces preceding the word to there is a long descender visible at the bottom of the hole, thus, it is likely that the surviving piece belonged to the s in usic. Because reckon or account are fitting translated words in terms of context with the rest of the line, telle and talige are both suitable possibilities for l. 30b and would fit on the MS. I argue that either þ¯te usic or þæt usic in l.31a is suitable, and because a descender in usic is still present on the MS, the full word þætte would not allow enough room for usic on the MS. Great care was taken to ensure that my conjectural reconstructions were palaeographically feasible in terms of MS spacing, meaning, meter, and alliteration (these considerations are discussed in detail in the Methodology section above). The word telle seems suitable in terms of context, there is evidence in the surrounding area of the lacuna to support the reading and, the word fits metrically. I digitally amended areas containing damage only in instances where enough internal evidence and reasonable linguistic support permitted piecing-together of letters or words; however, in other instances where conjectures seemed too risky and without feasible support such lacunae were left untouched.

§ 18 The above examples illustrate a virtual hands-on technique for testing possible hypotheses of incomplete lines. Rather than taking letters arbitrarily from the text, I scrutinised the areas surrounding the various lacunae. In doing so, I identified letters that matched the line most accurately, while further aligning and scaling the letters to approximate measures that corresponded with preceding and subsequent letters of lines partially visible due to lacunae. Certainly, manual stenciling on a facsimile could be used in a similar way, but the ease, speed, and accuracy of resizing of letters and words using Photoshop or other suitable programmes makes this type of reconstruction a viable option to test and examine conjectural line-readings in a virtually tangible way. It is also worth mentioning that having the ability to magnify to the extent that Photoshop allows is useful when viewing the areas surrounding the damage. See Figures 2 and 5 for examples of magnification and a pixel by pixel grid (I am aware that the reconstructed area appears somewhat pixelated in Figure 2, but the file size used was small. The example here merely illustrates the reconstruction possibility given the letter fragments visible in the original. Normally, I suggest using a file with a high resolution so that the area will not appear pixelated as shown above).

Concrete visualisation

§ 19 Another advantage of reconstructing MSS is that it allows a reader to visualise (concretely) what a folio, like f. 120r of the Exeter Book, may have looked like without large, gaping lacunae, burns, and worm holes disseminated around the vellum. Of course this is a what if and for many critics this conjectural reading or MS manipulation might seem like a pointless exercise to create nothing more than eye-candy out of a damaged medieval MS. We must keep in mind that this practice should not be overused, but in some instances reconstruction can add a more polished look. Certainly, the damage, as evidenced in the Exeter Book, adds to the MS's character and speaks of its history, not just as a text, but as an object that carried practical functions throughout history. However, when Bishop Leofric (1016-1072 AD) left the Exeter Book to the care and trust of Exeter Cathedral and its community before his death in 1072, there was no way of knowing that in the MS's history over the past millennium it would suffer the damage and misuse that it did. In some respects, virtually repairing areas of damage to a MS in places where loss is evident can serve to restore the MS to a less damaged condition without actually causing further physical damage or corruption to the surviving vellum. For example, on folio 120r, an approximately 2x2 inch circular burn-mark is situated nearly one-third of the way down the folio's right margin. Reconstructing the hole was relatively straightforward, as most of the missing letters around the lacuna were still visible. See Figures 12 and 13 below.

Figure 12: Original damaged f.120r.

Figure 13: Reconstructed f.120r

§ 20 Exemplified in Figure 13 above, I attempted to match the appearance of the vellum behind specific missing letters as precisely as possible. K. Kiernan notes that manuscript pages of vellum rarely lie flat and in extreme cases are severely distorted (2000). Making the vellum appear to contain natural creases, bends, folds, etc. rather than a flat surface is not simple, because there are no technical programmes available to create or manage MS manipulation to that extent.

§ 21 According to Professor Muir, there were issues that arose when attempting flat-field correction on the Exeter Book. When the MS was rebound in the 1930s, it was bound too tightly which consequently dipped in the gutters and caused parallax problems. To counter this issue, the programmer of the digitised manuscript was able to bend the text in the gutter upwards to some degree, although the results did not completely rectify the issue (Muir 2006). I attempted to obtain dimension-depth with digital fine-tuning tools such as Sharpening, Contrast adjustment, the use of Layers, and Opacity adjustment. Use of Layers and Opacity adjustment is discussed in greater detail in the Methodology section but it is worth noting that sharpening of any area within the manuscript should be done only when absolutely necessary and in small increments as the risk of over-pixelating the selected area will result in loss of fine detail and increase the contrast of the image. Sharpening may help to highlight the edges of a selected letter or word, however it can neutralise the natural bleed of the ink into the vellum, and be counterproductive. The appearance of over-sharpening can create an unnatural appearance making the alteration and its edges stand out on the manuscript when compared to unaltered areas. A useful tip is to use Sharpening as a temporary measure to help exaggerate the shape or edges of a selected letter or word for better viewing, and then reverse the sharpening using the undo function when close analysis is complete.

§ 22 In creating a visual reading of a damaged folio through digital retouching it is important to communicate this interpretation to readers (communicating a reconstructed interpretation is exemplified in Figure 16). L. MacDonald and R. Jacobsen note that the creator of the image has the purpose of communicating it through a suitable channel of one or more observers. Every element of the chain from creator to receiver affects the quality of the image, and hence the effectiveness of the communication process (2006, 352). In the process of digitally altering a folio an editor takes on another role in the chain of communication from the first mode of transmission with a text's (oral) poet or scribe to the ears and eyes of his twenty-first-century listeners and readers. In other words, the digitally-enabled interaction and communication goes beyond the original damaged MS and its restored counterpart. It is important that editors or those conducting digital restoration understand the relationship between the original, the surrogate, and the reconstructed counterpart and convey the relationship between the texts to an audience. Thus, it is crucial to identify the reconstructions as interpretations of original texts and allow readers to analyse the reconstructions as such. Using the approach of explicitly stating where changes are made encourages the audience to actively scrutinise both the original and its reconstructed counterpart and requires an audience to draw meaning and reach conclusions about the texts both separately and together. A similar model of active participation is found in new media technologies in museums that allow audiences to digitally handle or view objects, texts, and monuments (see Russo and Watkins 2007, 149-163). This process facilitates what S. Hazan calls the hands-on, minds-on scenario, in which interactive participation enhances the learning experience (2007, 144). As long as digital reconstruction is clearly identified it can in many ways be a catalyst for piquing interest and encouraging further analysis of a primary text or artefact. In a similar way that editors using and manipulating digital images of primary textual material have an obligation, as Terras argues, to understand the nature of the resources they are basing readings, transcriptions, and translations on (2011, 57), this responsibility is furthered because of the necessity to communicate alterations and changes to an audience explicitly. Essentially, the value of evidence should still point to the original MS or artefact and any digital alterations should be used to encourage further consideration and scrutiny of the original MS and/or its digital surrogate.

Continued preservation of the original

§ 23 In general, the most obvious benefit of MS digitisation is the continued preservation of the original MS, as digitised facsimiles can be used more often for scholarly analysis and teaching. This preservation of the surviving MS is also true when working with reconstructions of damaged lines of text. This hands-off approach to working with medieval MSS allows textual editors and literary critics opportunities to analyse and work with the MSS without having to physically handle the surviving medieval MSS. There are a number of ways to describe being given the opportunity to hold a millennia-old (or more) MS in one's hand, the words memorable and delightfully magical come to mind from personal experience. Yet, opportunities to work with some of the surviving Anglo-Saxon MSS are becoming less available to scholars not only as the results of logistical, financial, and geographical reasons, but also because of MS availability and fragility. Still, if one is given time to analyse any given MS, there are often issues that remain unresolved within the time allotted to examine the MS. Perhaps unresolved issues might lead to further visits to libraries or cathedrals that house Anglo-Saxon MSS, but more often than not, visits are limited and costly, and time allowed to analyse a MS quite often is insufficient. Thus, digital analysis and digital reconstruction of high-resolution facsimiles offers ways to examine a reliable copy of the original MS without physically handling or mishandling the original MS.

Less time consuming and convenient

§ 24 The old cut and paste and stenciling techniques to analyse lines are laborious and time-consuming. Working with high resolution digital images in Photoshop or with other image-altering software offers an improved, efficient and conveniently accessible method of analysis.

Reader confusion

§ 27 The risk that a reader might mistake a reconstructed MS for its original counterpart should not be taken lightly. A point that I identified in my own reconstructions was that, left unidentified, the reconstructed folia gave no indication of where digital manipulation had occurred. This posed a very serious issue, as any reader whether academic or general, might misidentify a reconstructed folio for the original. Some 80 years ago, cultural critic W. Benjamin brought attention to this problem in relation to reproductions of classical art and contended that reproductions and recreations of artifacts posed a threat to the primary objects of focus (refer to Benjamin [1936] 1968). A remedy to the problem of mistaking a reconstructed folio as the surviving text might involve highlighting the reconstructed area. Emphasising the area does not negate from the reconstruction; rather, it serves to identify the digital additions as well as distinguish the folio itself from its damaged counterpart. Examples in Figures 14-16 below exemplify the risk of not identifying areas in which reconstruction has occurred.

Figure 14: Untouched f. 120r of the Exeter Book.

Figure 15: Reconstructed f.120r.

Figure 16: Identified reconstruction of f. 120r.

§ 28 Without illuminating the area in which the digital reconstruction occurs (see Figure 15) readers might confuse Figure 15 for the surviving MS (Figure 14) or they might even miscalculate where the digitisation appears on the MS. Consequently, Figure 16 offers a more sensible solution, as the highlighted area clearly identifies the reconstructed text.

§ 29 Beyond the possibility of confusing the edited text for its original counterpart, we must be aware of creating further separation or remoteness between our audience and the original text. This issue is not just restricted to manuscript studies, since digital reconstruction and modelling is not new in the world of ancient and classical studies and archaeology (see Duranti and Schaefer 2012 and Schroeder et al. 2007. Refer also to suggested readings in subsection 2 Overuse of Digitisation below). It is somewhat ironic that endeavours to digitally reconstruct a document or in the case of archaeologists, create a model of an ancient site, building, or artefact with the aim of reconnecting an audience or readership with a place, time, and/or piece of history in the process creates additional distance between the original and its audience. This is not to suggest that reconstructions and/or models are without merit or purpose, but in creating another text we are not allowing our audience to draw closer to the time or place when the original text was recorded; rather we are offering another twenty-first interpretation of an ancient or medieval text or site. Benjamin noted that even the most perfect reproduction of a work of art is lacking in one element: its presence in time and space, its unique existence at the place where it happens to be ([1936] 1968, 218). More recent studies by Terras (2011, 55) and F. Cameron reassert Benjamin’s argument in the age of digital reproduction by highlighting the fact that reproductions can lead to the loss of auraic, iconic and ritualistic qualities (2007, 50) and by creating reconstructions based on digitised versions of original MSS, there is a further distance to the original text (See Turner 2012, 135-144). The very essence of a work of art's creation, whether it be a text, artefact, building, or musical composition cannot be fully captured in a digital reproduction. A scan neither takes into account nor can explicitly communicate the time and space taken to create the original object it represents. A digital reproduction offers a reflection of the object as it exists, but the context of the reproduction is separate from the original. Readers should always remain mindful of the complex relationship between a digital scan, a manipulated text and an original source. Reconstructions can never take the place of the original and altered texts should be used for comparative purposes to test conjectural readings or interpretations of the original source.

Overuse of digitisation

§ 30 While digital retouching software can be used to scrutinise a MS, there are instances in which digital reconstruction is not practical and serves little purpose to attempt to eliminate a lacuna. Other fields have grappled with issues concerning what to do with virtual reconstructions of archaeological artefacts and sites when little evidence of the original remains. For instance see A. Bentkowska-kafel, H. Denard and D. Baker's Paradata and Transparency in Virtual Heritage which addresses the cognitive and technological issues faced when using visualisation techniques and 3D modeling in archaeology and cultural heritage projects, and offers practical solutions and techniques for dealing with various challenges (2012) (see also Lamb 2007, 29-44. Other relevant readings include those listed in the previous section on Reader Confusion). J. A. Barceló argues: Finding solutions by inverse engineering is a particularly difficult task because of the nonuniqueness difficulties that arise. Nonuniqueness means, in effect, that the true solution cannot be selected from among a large set of possible solutions without further constraints being imposed. This undesirable behavior is due to noise in the measurements, and insufficient numbers of measurements. That is to say, we generally do not know the most relevant factors affecting the shape, size, texture, composition, and spatio-temporal location of material consequences of social action. Instead, we have sparse and noisy measurements of perceptual properties, and an incomplete knowledge of relational contexts. From this information, we need to resolve the causal distribution of nonperceptual properties to adequately interpret archaeological observables. We also must take into account the circumstances and contexts (social and natural) where actions were performed and the processes (both social and natural) that acted upon that place after the original cause, because they may have altered the original effects of primary actions (2007, 439).

§ 31 Inverse engineering can work to facilitate digital conjectural readings of damaged or missing lines, but in identifying a possible solution or accurate reading, further constraints are imposed. The problem with filling in a lacuna with suggested letters and words is compounded because the altered space or hypothetical work is replaced with perfect text. In the case of fols. 119v-121v the scribe (or someone after the scribe) made a number or corrections to the text, so there is evidence of both scribal errors and corrections. However, in creating a digital reconstruction, there is no basis to determine if the scribe would have made errors or corrections in the damaged lines, thus a true solution amongst a set of possible solutions cannot be attained. Essentially, in gaining the appearance of text to fill in an existing lacuna, there is still something lost in transmission from the original manuscript and its reconstructed counterpart. Although letters are reused from the scribe’s own hand, we are still overlooking the originality of each letter and stroke first recorded and by reusing letters we are undertaking our own regulation of lettering in the manuscript. Still, this is not to suggest that any apparent overlooking is due to ignorance; rather, we limit the amount of noise by working with material within the MS itself. It remains an important feature to work with material that has survived from within the MS and any letters or words created from out width should be carefully scrutinised and supported with codicological and palaeographical evidence from the MS in question. It bears repeating that this type of work is always going to render another interpretation or interpretations of a text unless there is another copy of the original source from which to make comparisons.

§ 32 There are a variety of reconstruction choices for existing lacunae that could be used in a digital edition or in analyses of missing content within a hole, but assisting one segment of the scholarly community might prove futile for another. For instance, philologists, metricists, and literary critics might benefit from viewing reconstructions of various scholarly readings of lines of text within the same lacunae. Creating several reconstructions of a folio (or the area on the folio) that contain a lacuna or lacunae will offer support for or against viability of hypotheses. However, this type of work will most likely prove unbeneficial to art historians or palaeographers examining the content as it exists. Another issue with filling in lacunae when no evidence exists regarding wording is to what extent one fills in the gap. Should reconstructions involve digitally adding in parchment around the edging of the lacuna as far in as evidence will allow? Should digitally reconstructed parchment fill the entire hole but leave the reconstructed area without text? Or is it more appropriate to leave the entire lacuna untouched? Digital reconstruction or restoration is very much a test in restraint because of the various options available to recreate a space on a MS that tempt one to overuse applications in reconstruction projects. It is important to state explicitly the intent of the reconstructions, as the finished product will reflect the primary purpose of the reconstructions and will affect how the images are read and interpreted by an audience.

§ 33 Overuse of Photoshop can make a medieval MS appear not unlike the equivalent of an airbrushed celebrity. It is crucial not to play a guessing game and in instances when a hypothesis does not work, then the most obvious conclusion would be to leave the area on the MS untouched. The main objective of digital reconstruction is not to give a medieval MS a twenty-first-century makeover. Likewise, in archaeology, Barceló contends that archaeologists use 3D models and digital reconstructions to test hypotheses. He argues for a different approach where archaeologists do not 'paint' the past, but are involved in social science inverse engineering (2007, 454), to understand the history of our society. With relation to manuscript studies and digital reconstructions, this approach is important, as we can use inverse engineering to understand the creation and transmission of a text in its historical context.

Ultimately, when conjectures are too far-fetched and devoid of solid evidence within the MS to justify digital emendation, the most sensible conclusion is to admit uncertainty and leave the section unaltered. In my experience, I excluded digitisation in instances where lacunae were without enough verifiable support. In other cases, I simply made particular lacunae smaller by cautiously amending the areas that contained enough evidence to justify, what seemed to me to be, reasonable emendations.

Digital facsimiles are not foolproof

§ 34 Despite the availability to view superior quality scans of texts the reliability of digital scans of MSS can still be put to the test. High resolution scans of MSS are still no substitute for vellum. However, the argument has been established and reiterated by academic researchers in a variety of disciplines from musicology to classical studies, digital humanists, historians, librarians, and archivists alike, that scans allow for convenient and dependable opportunities of research. S. Lee argues that digitisation liberates the document (albeit a facsimile) from the constraints of traditional access methods (2001b) and L. Capell further emphasises that digitization is a valuable access tool that allows archives, libraries, and other cultural heritage repositories to make items in their collections available to a worldwide, Internet-connected audience from anyplace at any time (2010, 235) (see also Lee 2001a). Offering access of manuscript images to a wider audience is demonstrated by initiatives involving institutions such as The Hill Museum & Manuscript Library (HMML) that partners with libraries around the globe to digitise and catalogue manuscript collections with the aim of preserving manuscripts, especially those at risk in unstable countries, while also making such collections accessible to researchers. According to W. W. Torborg, T. M. Van, and C. Stewart, there are countries in which research with primary sources is a risk for both researcher and text and because sources like medieval manuscripts have intrinsic value, the cheapest way to protect them from thieves is to keep knowledge of them secret (2012, 852). Digital scans in these instances offer preservation and opportunities for research without jeopardising the safety of the primary material or the researchers. Digitisation initiatives can also further research initiatives as exemplified in the Electronic Beowulf project in which K. Kiernan's use of high-end digital images and fiber-optic lighting revealed letters underneath the protective paper frame added to the manuscript in the mid-nineteenth century (2013). Evidently, digitisation plays an important role in preservation and research schemes; however, there are still instances when viewing an original, damaged MS reveals codicological, linguistic, and/or palaeographical ambiguities although such obscurities go undetected when viewing scans.

§ 35 Because various lacunae in the Exeter Book have been covered over with glue to preserve the folia from further damage, the appearance of some pieces of letters are altered or eliminated in the outer edges of lacunae. On folio 121v a large lacuna precedes the syllable –lum and earlier editors attempted to identify any letters visible by looking at facsimiles or made educated guesses as to what the missing letters could be. ASPR identifies the letter e three spaces before –lum (1936, 222), and cautiously claims that fragments of the letters preceding the syllable –lum make it reasonably certain that the word to which –lum belonged was englum [and that] Cramer's restoration . . . is probably correct (1936, 359). Still, no editors could claim with complete certainty that the syllable was conclusively eng.

Figure 17: A portion of a lacuna on f. 121v of the Exeter Book.

§ 36 The high-res scans reveal some evidence of letters (see Figure 17), but not enough evidence to make a definitive claim; however, upon close examination of the MS in the Exeter Cathedral Library (the reverse side of the leaf being f. 121r), I discovered that not only was the e evident, but the top of the n and much of its two shafts were clearly visible as well (for further details see Rambaran-Olm 2007). It is not that the facsimiles are poor quality, but the glue used to prevent the lacunae from deteriorating does not show through on the scans in a way that would make the letters of some of the surviving text on the outer edges of the holes beneath the glue distinguishable. Similar issues with glue used on papyri have plagued the study of papyrology; however, experts have utilised infrared imaging to view areas underneath glue (Bertani and Consolandi 2006, 234-237). D. Bertani and L. Consolandi explain that papyri are mainly written with a carbon-based ink which strongly absorbs infrared, making it easy to capture reflectograms, but the papyri itself has become brittle and is very delicate (2006, 234-237). The vellum within the Exeter Book is still reasonably robust, and perhaps a new digitisation using multispectral (ultraviolet to infrared) imaging could be justified, as it might bring to light some of the hidden letters underneath the glue, or shed light on the indistinguishable darkened letters within the burned edges of existing lacunae. This is not to suggest that the MS undergo extensive handling because its state is not as fragile as papyri. Having UV imaging done to the MS might shed new light on the codex, its history, and transmission while encouraging new research on the MS as a whole and the individual texts within it.

§ 37 In my experience, the actual hands-on approach including examination of the edges of the lacuna in the Exeter Book with a flashlight revealed some show-through (the edges of the lacuna that reveal letters or letter fragments when examined from the reverse side of the folio are indicated in the circled area on Figure 17), so I was able to conclude decisively that the three letters eng- are visible on the MS and that the complete word on f. 121v is in fact englum. Despite being a rather small find, it is still a reminder that the MS has much left to discover and that facsimiles do not always illuminate all the surviving evidence of a damaged MS.

Colour management

§ 38 Another problem with the quality of scans involves the colour reproduction of the facsimiles (refer also to Hunt 2004). In general, Terras explains that colour is a thorny issue, with many issues such as illumination, and the differences between the way the human eye and computer systems record colour (2011, 53). P. Green explores colour management and its techniques and argues for the importance of colour fidelity. In broad terms he identifies five fidelity requirements which include: recognizable reproductions, aesthetically pleasing reproductions, accurate reproductions of original objects under reference viewing conditions, accurate reproduction of original under the viewing conditions of the original, and photogrammetric reproduction for the purposes not limited to visual interpretation (Green 2006, 304). In accordance with Green's classification as it relates to fols. 119v-121v, I worked with accurate reproductions under both viewing and reference viewing conditions and my digital reconstructions adhered to the rule of maintaining consistent colouring as represented in Muir's digital scans. This system of colour management is complex and in some ways flawed, as I cannot claim that my colour management is entirely accurate, nor can it be verified that the colour of the original scans are entirely accurate, but because the aim of the original scans were to produce replicas of the original as closely as possible, my objective was to maintain that aim for fidelity of the scans within the digital reconstructions. Techniques like saturation, relative colorimetric and perceptual tools are useful in viewing letter fragments and areas around existing lacunae, but as a rule colour manipulation should be kept to a minimum unless the goal of the reconstruction is to achieve an aesthetically more pleasing production (these techniques are explored in detail in Green 2006, 316-317).

§ 39 Reproductions of MSS can have various functions and requirements depending on whether the scan is to appear as a close replica of the original MS or otherwise; thus, the colour quality in a scan might affect an art historian’s work more than perhaps a linguist. Still, if the colour profiles are not calibrated precisely, then the quality of the scans (despite being high-resolution) will not provide accurate representations to use in any case. Each subsequent reproduction, as a result, will be varied to some degree; thus, making each electronic file more and more remote from the original MS. Making certain that the computer screen being viewed is calibrated to match the colour profile of the input and/or the outputs devices will ensure a closer likeness to the surviving MS.

Manipulation of the MS and its audience

§ 40 Much opposition to digital restoration stems from the argument that the altered piece might potentially mislead a casual reader into believing the reconstructed text is the original, and as mentioned above, such sentiment is justified. An additional concern is that a MS itself has come down to us in its form for a reason, so any alteration, even digital restoration, indicates some sort of disrespect for the MS’s history. In fact, applying the word manipulation carries a connotation that the MS is being exploited, misused, and forged, thus, lending itself to the idea that a MS is being mishandled. Problematic wording is compounded when one considers the use of the word restoration. Initially, I referred to the digital reparations I had made to ff. 119v-121v as restored, but thanks to some constructive criticism I soon recognised that identifying the images as restored exposed a type of academic dishonesty, albeit unintentional. Questions to consider include: how can one restore a MS when only a single copy survives? To what extent is a MS being restored? Not only might we consider that carefully controlled digital reparations and reconstructions are instructive and can aid in clarifying conjectures while also elucidating problematic lines, but limited reparations can further illuminate elements of what a damaged MS may have contained. Perhaps this type of comparative digitisation can bring balance to the idea that involves admiring a text as it survives without neglecting to consider the value and appearance that a MS maintained in the early phases of its compilation. Most critical concern must stem from the speculative nature of comparative restoration work and the dangers of over-theorising or, in this case, over-digitising a MS. It is vital that digital reconstruction proposes to address palaeographical, linguistic, and codicological concerns when reconstructing lines; thus, not only highlighting areas where reparations occur, but also emphasising that the digital work is hypothetical.

Wrong conjectural readings that look correct

§ 41 As a result of piecing together photocopied versions of the Exeter Book's The Husband’s Message, J. Pope realised that he had produced visually convincing results that may in fact be partly or even completely wrong (qtd in Kiernan 1994, 47). If a reconstruction is so convincing that it runs the risk of being mistaken as an unflawed original, then on the one hand, it is validation that an editor has some serious skills; yet, on the other hand, the danger is that the conjectural addition might be mistaken as the original, and worse still, the misread line will go unquestioned. Again, it is important to stress the necessity of identifying conjectural readings, so that scholars can assess the speculative additions without having to worry about mistaking an hypothesis for its original counterpart.