Back to the future: What digital editors can learn from print editorial practice.tags: computers, database, digital humanities, editorial studies, history, queries, textual studies
A ersion of this essay was published in literary and Linguistic Computing
Digital Editing and Contemporary Textual Studies
The last decade or so has proven to be a heady time for editors of digital editions. With the maturation of the digital medium and its application to an ever increasing variety of cultural objects, digital scholars have been led to consider their theory and practice in fundamental terms (for a recent collection of essays, see Burnard, O’Keeffe, and Unsworth 2006). The questions they have asked have ranged from the nature of the editorial enterprise to issues of academic economics and politics; from problems of textual theory to questions of mise-en-page and navigation: What is an Edition? What kinds of objects can it contain? How should it be used? Must it be critical? Must it have a reading text? How should it be organised and displayed? Can intellectual responsibility be shared among editors and users? Can it be shared across generations of editors and users? While some of these questions clearly are related to earlier debates in print theory and practice, others involve aspects of the production of editions not relevant to or largely taken for granted by previous generations of print-based editors.
The answers that have developed to these questions at times have involved radical departures from earlier norms1. The flexibility inherent to the electronic medium, for example, has encouraged editors to produce editions that users can manipulate interactively, displaying or suppressing different types of readings, annotation, and editorial approaches, or even navigate in rudimentary three-dimensional virtual reality (e.g. Railton 1998-; Foys 2003; O’Donnell 2005a; Reed Klein 2001; Ó Cróinín nd). The relatively low production, storage, and publication costs associated with digital publication, similarly, have encouraged the development of the archive as the de facto standard of the genre: users of digital editions now expect to have access to all the evidence used by the editors in the construction of their texts (assuming, indeed, that editors actually have provided some kind of mediated text): full text transcriptions, high-quality facsimiles of all known witnesses, and tools for building alternate views of the underlying data (e.g. Kiernan 1999/2003; Robinson 1996). There have been experiments in editing non-textual objects (Foys 2003; Reed-Kline 2001), in producing image-based editions of textual objects (Kiernan 1999/2003), and in recreating digitally aspects of the sensual experience users might have had in consulting the original objects (British Library nd). There have been editions that radically decenter the reading text (e.g. Robinson 1996), and editions that force users to consult their material using an editorially imposed conceit (Reed-Kline 2001). Even elements carried over from traditional print practice have come in for experimentation and redesign: the representation of annotation, glossaries, or textual variation, for example, are rarely the same in any two electronic editions, even in editions published by the same press (see O’Donnell 2005b, § 5)2.
Much of the impetus behind this theoretical and practical experimentation has come from developments in the wider field of textual and editorial scholarship, particularly work of the book historians, new philologists, and social textual critics who came into prominence in the decade preceding the publication of the earliest modern digital editorial projects (e.g. McKenzie 1984/1999; McGann 1983/1992; Cerquiglini 1989; Nicols 1990; for a review see Greetham 1994, 339-343). Despite significant differences in emphasis and detail, these approaches are united by two main characteristics: a broad interest in the editorial representation of variance as a fundamental feature of textual production, transmission, and reception; and opposition to earlier, intentionalist, approaches that privileged the reconstruction of a hypothetical, usually single, authorial text over the many actual texts used and developed by historical authors, scribes, publishers, readers, and scholars. Working largely before the revolution in Humanities Computing brought on by the development of structural markup languages and popularity of the Internet, these scholars nevertheless often expressed themselves in technological terms, calling for changes in the way editions were printed and organised (see, for example, the call for a loose leaf edition of Chaucer in Pearsall 1985) or pointing to the then largely incipient promise of the new digital media for representing texts as multiforms (e.g. McGann 1994; Shillingsburg 1996).
Digital Editing and Print Editorial Tradition
A second, complementary, impetus for this experimentation has been the sense that the digital editorial practice is, or ought to be, fundamentally different from and even opposed to that of print. This view is found to a greater or lesser extent in both early speculative accounts of the coming revolution (e.g. McGann 1994; the essays collected in Finneran 1996 and Landow and Delaney 1993) and subsequent, more sober and experienced discussions of whether digital practice has lived up to its initial promise (e.g. Robinson 2004, 2005, 2006; Karlsson and Malm 2004). It is characterised both by a sense that many intellectual conventions found in print editions are at their root primarily technological in origin, and that the new digital media offer what is in effect a tabula rasa upon which digital editors can develop new and better editorial approaches and conventions to accommodate the problems raised by textual theorists of the 1980s and 1990s.
Of course in some cases, this sense that digital practice is different from print is justified. Organisational models such as the Intellectual Commons or Wiki have no easy equivalent in print publication (O’Donnell Forthcoming). Technological advances in our ability to produce, manipulate, and store images cheaply, likewise, have significantly changed what editors and users expect editions to tell them about the primary sources. The ability to present research interactively has opened up rhetorical possibilities for the representation of textual scholarship difficult or impossible to realise in the printed codex.
But the sense that digital practice is fundamentally different from print has been also at times more reactionary than revolutionary. If digital theorists have been quick to recognise the ways in which some aspects of print editorial theory and practice have been influenced by the technological limitations of the printed page, they have been also at times too quick to see other, more intellectually significant aspects of print practice as technological quirks. Textual criticism in its modern form has a history that is now nearly 450 years old (see Greetham 1994, 313); seen more broadly as a desire to produce “better” texts (however “better” is defined at the moment in question), it has a history stretching back to the end of the sixth century BCE and is “the most ancient of scholarly activities in the West” (Greetham 1994, 297). The development of the critical edition over this period has been as much an intellectual as a technological process. While the limitations of the printed page have undoubtedly dictated the form of many features of the traditional critical edition, centuries of refinement—by trial-and-error as well as outright invention—also have produced conventions that transcend the specific medium for which they were developed. In such cases, digital editors may be able to improve upon these conventions by recognising the (often unexpressed) underlying theory and taking advantage of the superior flexibility and interactivity of the digital medium to improve their representation.
The Critical Text in a Digital Age
Perhaps no area of traditional print editorial practice has come in for more practical and theoretical criticism than the provision of synthetic, stereotypically eclectic, reading texts3. Of course this criticism is not solely the result of developments in the digital medium: suspicion of claims to definitiveness and privilege is, after all, perhaps the most characteristic feature of post-structuralist literary theory. It is the case, however, that digital editors have taken to avoiding the critical text with a gusto that far outstrips that of their print colleagues. It is still not unusual to find a print edition with some kind of critical text; the provision of similarly critical texts in digital editions is far less common. While most digital projects do provide some kind of top-level reading text, few make any strong claims about this text’s definitiveness. More commonly, as in the early ground breaking editions of the Canterbury Tales Project (CTP), the intention of the guide text is, at best, to provide readers with some way of organising the diversity without making any direct claim to authority (Robinson nd):
We began… work [on the CTP] with the intention of trying to recreate a better reading text of the Canterbury Tales. As the work progressed, our aims have changed. Rather than trying to create a better reading text, we now see our aim as helping readers to read these many texts. Thus from what we provide, readers can read the transcripts, examine the manuscripts behind the transcripts, see what different readings are available at any one word, and determine the significance of a particular reading occurring in a particular group of manuscripts. Perhaps this aim is less grand than making a definitive text; but it may also be more useful.
There are some exceptions to this general tendency—both in the form of digital editions that are focussed around the provision of editorially mediated critical texts (e.g. McGillivray 1997; O’Donnell 2005a) and projects, such as the Piers Plowman Electronic Archive (PPEA), that hope ultimately to derive such texts from material collected in their archives. But even here I think it is fair to say that the provision of a synthetic critical text is not what most digital editors consider to be the really interesting thing about their projects. What distinguishes the computer from the codex and makes digital editing such an exciting enterprise is precisely the ability the new medium gives us for collecting, cataloguing, and navigating massive amounts of raw information: transcriptions of every witness, collations of every textual difference, facsimiles of every page of every primary source. Even when the ultimate goal is the production of a critically mediated text, the ability to archive remains distracting4.
In some areas of study, this emphasis on collection over synthesis is perhaps not a bad thing. Texts like Piers Plowman and the Canterbury Tales have such complex textual histories that they rarely have been archived in any form useful to the average scholar; in such cases, indeed, the historical tendency—seen from our post-structuralist perspective—has been towards over-synthesis. In these cases, the most popular previous print editions were put together by editors with strong ideas about the nature of the textual history and/or authorial intentions of the works in question. Their textual histories, too, have tended to be too complex for easy presentation in print format (e.g. Manley and Rickert 1940). Readers with only a passing interest in these texts’ textual history have been encouraged implicitly or explicitly to leave the question in the hands of experts.
The area in which I work, Old English textual studies, has not suffered from this tendency in recent memory, however. Editions of Old English texts historically have tended to be under- rather than over-determined, even in print (Sisam 1993; Lapidge 1994, 1991). In most cases, this is excused by the paucity of surviving witnesses. Most Old English poems (about 97% of the known canon) survive in unique manuscripts (O’Donnell 1996a; Jabbour 1968; Sisam 1953). Even when there is more primary material, Anglo-Saxon editors work in a culture that resists attempts at textual synthesis or interpretation, preferring parallel-text or single-witness manuscript editions whenever feasible and limiting editorial interpretation to the expansion of abbreviations, word-division, and metrical layout, or, in student editions, the occasional normalisation of unusual linguistic and orthographic features (Sisam 1953). One result of this is that print practice in Anglo-Saxon studies over the last century or so has anticipated to a great extent many of the aspects that in other periods distinguish digital editions from their print predecessors.
Cædmon’s Hymn: A Case Study
The scholarly history of Cædmon’s Hymn, a text I have recently edited for the Society of Early English and Norse Electronic Texts series (O’Donnell 2005a), is a perfect example of how this tendency manifests itself in Old English studies. Cædmon’s Hymn is the most textually complicated poem of the Anglo-Saxon period, and, for a variety of historical, literary, and scholarly reasons, among the most important: it is probably the first recorded example of sustained poetry in any Germanic language; it is the only Old English poem for which any detailed account of its contemporary reception survives; and it is found in four recensions and twenty-one medieval manuscripts, a textual history which can be matched in numbers, but not complexity, by only one other vernacular Anglo-Saxon poem (the most recent discussion of these issues is O’Donnell 2005a).
The poem also has been well studied. Semi-diplomatic transcriptions of all known witnesses were published in the 1930s (Dobbie 1937)5. Facsimiles of the earliest manuscripts of the poem (dating from the mid-eighth century) have been available from various sources since the beginning of the twentieth century (e.g. Dobiache-Rojdestvensky 1928) and were supplemented in the early 1990s by a complete collection of high quality black and white photos of all witnesses in Fred C. Robinson and E.G. Stanley ‘s Old English Poems from Many Sources (1991). Articles and books on the poem’s transmission and textual history have appeared quite regularly for over a hundred years. The poem has been at the centre of most debates about the nature of textual transmission in Anglo-Saxon England since at least the 1950s. Taken together, the result of this activity has been the development of an editorial form and history that resembles contemporary digital practice in everything but its medium of production and dissemination. Indeed, in producing a lightly mediated, witness- and facsimile-based archive, constructed over a number of generations by independent groups of scholars, Cædmon’s Hymn textual criticism even anticipates several recent calls for the development of a new digital model for collective, multi-project and multi-generational editorial work (e.g. Ore 2004; Robinson 2005).
The print scholarly history of the poem anticipates contemporary digital practice in another way as well: until recently, Cædmon’s Hymn had never been the subject of a modern critical textual edition. The last century has seen the publication of a couple of student editions of the poem (e.g. Pope and Fulk 2001; Mitchell and Robinson 2001), and some specialised reconstructions of one of the more corrupt recensions (Cavill 2000, O’Donnell 1996b, Smith 1938/1978, Wuest 1906). But there have been no critical works in the last hundred years that have attempted to encapsulate and transmit in textual form what is actually known about the poem’s transmission and recensional history. The closest thing to a standard edition for most of this time has been a parallel text edition of the Hymn by Elliot Van Kirk Dobbie (1942). Unfortunately, in dividing this text into Northumbrian and West-Saxon dialectal recensions, Dobbie produced an edition that ignored his own previous and never renounced work demonstrating that such dialectal divisions were less important that other distinctions that cut across dialectal lines (Dobbie 1937)6.
The Edition as Repository of Expert Knowledge
The problem with this approach—to Cædmon’s Hymn or any other text—should be clear enough. On the one hand the poem’s textual history is, by Anglo-Saxon standards, quite complex and the subject of intense debate by professional textual scholars. On the other, the failure until recently to provide any kind of critical text representing the various positions in the debate has all but hidden the significance of this research—and its implications for work on other aspects of the Hymn_—from the general reader. Instead of being able to take advantage of the expert knowledge acquired by editors and textual scholars of the poem over the last hundred years, readers of _Cædmon’s Hymn instead have been forced either to go back to the raw materials and construct their own texts over and over again or rely on a standard edition that misrepresents its own editor’s considered views of the poem’s textual history.
This is not an efficient use of these readers’ time. As Kevin Kiernan has argued, the textual history of Cædmon’s Hymn is not a spectacle for casual observers (Kiernan 1990), and most people who come to study Cædmon’s Hymn are not interested in collating transcriptions, deciphering facsimiles, and weighing options for grouping the surviving witnesses. What they want is to study the poem’s sources and analogues, its composition and reception, its prosody, language, place in the canon, significance in the development of Anglo-Saxon Christianity, or usefulness as an index in discussions of the position of women in Anglo-Saxon society—that is, all the other things we do with texts when we are not studying their transmission. What these readers want—and certainly what I want when I consult an edition of a work I am studying for reasons other than its textual history—is a text that is accurate, readable, and hopefully based on clearly defined and well-explained criteria. They want, in other words, to be able to take advantage of the expert knowledge of those responsible for putting together the text they are consulting. If they don’t like what they see, or if the approach taken is not what they need for their research, then they may try to find an edition that is better suited to their particular needs. But they will not—except in extreme cases I suspect—actually want to duplicate the effort required to put together a top-quality edition.
The Efficiency of Print Editorial Tradition
The failure of the print editors of Cædmon’s Hymn over the last hundred years to provide a critical-editorial account of their actual knowledge of the poem is very much an exception that proves the rule. For in anticipating digital approaches to textual criticism and editorial practice, textual scholars of Cædmon’s Hymn have, ironically, done a much poorer job of supplying readers with information about their text than the majority of their print-based colleagues have of other texts in other periods.
This is because, as we shall see, the dissemination of expert knowledge is something that print-based editors are generally very good at. At a conceptual level, print approaches developed over the last several hundred years to the arrangement of editorial and bibliographic information in the critical edition form an almost textbook example for the parsimonious organisation of information about texts and witnesses. While there are technological and conventional limitations to the way this information can be used and presented in codex form, digital scholars would be hard pressed to come up with a theoretically more sophisticated or efficient organisation for the underlying data.
Normalisation and Relational Database Design
Demonstrating the efficiency of traditional print practice requires us to make a brief excursion into questions of relational database theory and design7. In designing a relational database, the goal is to generate a set of relationship schemas that allow us to store information without unnecessary redundancy but in a form that is easily retrievable (Silberschatz, Korth, and Sudarshan 2006, 263). The relational model organises information into two-dimensional tables, each row of which represents a relationship among associated bits of information. Complex data commonly requires the use of more than one set of relations or tables. The key thing is to avoid complex redundancies: in a well designed relational database, no piece of information that logically follows from any other should appear more than once8.
The process used to eliminate redundancies and dependencies is known as normalisation. When data has been organised so that it is free of all such inefficiencies, it is usually said to be in third normal form. How one goes about doing this can be best seen through an example. The following is an invoice from a hypothetical book store (adapted from Krishna 1992, 32):
|Name:||Jane J. Smith|
|Address:||323 Fifteenth Street S., Lethbridge, Alberta T1K 5X3.|
|0-670-03151-8||Pinker, Stephen||The Blank Slate: The Modern Denial of Human Nature||$35.00||1||$35.00|
|0-8122-3745-5||Burrus, Virginia||The Sex Lives of Saints: An Erotics of Ancient Hagiography||$25.00||2||$50.00|
|0-7136-0389-5||Dix, Dom Gregory||The Shape of the Liturgy||$55.00||1||$55.00|
Describing the information in this case in relational terms is a three step process. The first step involves identifying what is that is to be included in the data model by extracting database field names from the document’s structure. In the following, parentheses are used to indicate information that can occur more than once on a single invoice:
Invoice: invoice_number, customer_id, customer_name, customer_address, (ISBN, author, title, price, quantity, item_total), grand_total
The second step involves extracting fields that contain repeating information and placing them in a separate table. In this case, the repeating information involves bibliographical information about the actual books sold (
ISBN, author, title, price, quantity, item_total). The connection between this new table and the
invoice table made explicit through the addition of an
invoice_number key that allows each book to be associated with a specific invoice9:
Invoice: invoice_number, customer_id, customer_name, customer_address, grand_total
Invoice_Item: invoice_number, ISBN, author, title, price, quantity, item_total
The final step involves removing functional dependencies within these two tables. In this database, for example, information about a book’s
item_price are functionally dependent on its
ISBN: for each
ISBN, there is only one possible
customer_id is associated with only one
customer_address. These dependencies are eliminated by placing the dependent material in two new tables, Customer and Book, which are linked to rest of the data by the
ISBN keys respectively.
At this point the data is said to be in third normal form: we have four sets of relations, none of which can be broken down any further:
Invoice: invoice_number, customer_id, grand_total
Invoice_Item: invoice_number, ISBN, quantity, item_total
Customer: customer_id, customer_name, customer_address
Book: ISBN, author, title, price
Normalising Editorial Data
The normalisation process becomes interesting when one applies it to the type of information editors commonly collect about textual witnesses. The following, for example, is a simplified version of a sheet I used to record basic information about each manuscript witness to Cædmon’s Hymn:
|Shelf-Mark:||B1 Cambridge, Corpus Christi College 41|
|Scribe:||Second scribe of the main Old English text.|
|Location:||Copied as part of the main text of the Old English translation of the Historia ecclesiastica (p. 332 [f. 161v]. line 6)|
|Recension:||West-Saxon eorðan recension|
|Text:|| Nuweherigan sculon
heofonrices weard metodes mihte
&hismod ge þanc weorc wuldor godes
From the point of view of the database designer, this sheet has what are essentially fields for the manuscript sigil, date, scribe, location, and, of course, the text of the poem in the witness itself, something that can be seen, on analogy with our book store invoice, as itself a repeating set of (largely implicit) information: manuscript forms, normalised readings, grammatical and lexical information, metrical position, relationship to canonical referencing systems, and the like.
As with the invoice from our hypothetical bookstore, it is possible to place this data in normal form. The first step, once again, is to extract the relevant relations from the manuscript sheet and, in this case, the often unstated expert knowledge an editor typically brings to his or her task. This leads at the very least to the following set of relations10:
Manuscript: shelf_mark, date, scribe, location, (ms_instance, canonical_reading, dictionary_form, grammatical_information, translation)
Extracting the repeating information about individual readings, leaves us with two tables linked by the key
Manuscript: shelf_mark, date, scribe, location
bq(code). Text: shelf_mark, ms_instance, canonical_reading,
bq(code). dictionary_form, grammatical_information, translation
And placing the material in third normal form generates at least one more:
Manuscript: shelf_mark, date, scribe, location
Text: shelf_mark, ms_instance, canonical_reading
Glossary: canonical_reading, dictionary_form, grammatical_information, translation
At this point, we have organised our data in its most efficient format. With the exception of the
canonical_reading keys, no piece of information is repeated in more than one table, and all functional dependencies have been eliminated. Of course in real life, there would be many more tables, and even then it would be probably impossible—and certainly not cost effective—to treat all editorial knowledge about a given text as normalisable data.
What is significant about this arrangement, however, is the extent to which our final set of tables reflects the traditional arrangements of information in a stereotypical print edition: a section up front with bibliographic (and other) information about the text and associated witnesses; a section in the middle relating manuscript readings to editorially privileged forms; and a section at the end containing abstract lexical and grammatical information about words in the text. Moreover, although familiarity and the use of narrative can obscure this fact in practice, much of the information contained in these traditional sections of a print edition actually is in implicitly tabular form: in structural terms, a glossary are best understood as the functional equivalent of a highly structured list or table row, with information presented in a fixed order from entry to entry. Bibliographical discussions, too, often consist of what are in effect, highly structured lists that can easily be converted to tabular format: one cell for shelf-mark, another for related bibliography, provenance, contents, and the like11.
Database Views and the Critical Text
This analogy between the traditional arrangement of editorial matter in print editions and normalised data in a relational database seems to break down, however, in one key location: the representation of the abstract text. For while it is possible to see the how the other sections of a print critical edition might be rendered in tabular form, the critical text itself—the place where editors present an actual reading as a result of their efforts—is not usually presented in anything resembling the non-hierarchical, tabular form a relational model would lead us to expect. In fact, the essential point of the editorial text—and indeed the reason it comes in for criticism from post-structuralists—is that it eliminates non-hierarchical choice. In constructing a reading text, print editors impose order on the mass of textual evidence by privileging individual readings at each collation point. All other forms—the material that would make up the Text table in a relational database—is either hidden from the reader or relegated, and even then usually only as a sample, to appearance in small type at the bottom of the page in the critical apparatus. Although it is the defining feature of the print critical edition, the critical text itself would appear to be the only part that is not directly part of the underlying, and extremely efficient, relational data model developed by print editors through the centuries.
But this does not invalidate my larger argument, because we build databases precisely in order to acquire this ability to select and organise data. If the critical text in a print edition is not actually a database table, it is a database view—that is to say a “window on the database through which data required for a particular user or application can be accessed” (Krishna 1992, 210). In computer database management systems, views are built by querying the underlying data and building new relations that contain one or more answers from the results. In print editorial practice, editors build critical texts by “querying” their knowledge of textual data at each collation point in a way that produces a single editorial reading. In this understanding, a typical student edition of a medieval or classical text might be understood as a database view built on the query “select the manuscript or normalised reading at each collation point that most closely matches paradigmatic forms in standard primers.” A modern-spelling edition of Shakespeare can be understood as the view resulting from a database query that instructs the processor to replace Renaissance spellings for the selected forms with their modern equivalents. And an edition like the Kane-Donaldson Piers Plowman can be understood as a view built on basis of a far more complex query derived from the editors’ research on metre, textual history, and scribal practice. Even editorial emendations are, in this sense, simply the result of a query that requests forms from an unstated “normalised/emended equivalent” column in the editors’ intellectual understanding of the underlying textual evidence: “select readings from the database according to criteria x; if the resulting form is problematic, substitute the form found in the normalised/emended_equivalent column.”12.
How Digital Editors can Improve on Print Practice
If this understanding of the critical text and its relationship to the data model underlying print critical practice is correct, then digital editors can almost certainly improve upon it. One obvious place to start might seem to lie in the formalising and automating the process by which print editors process and query the data upon which their editions are based. Such an approach, indeed, would have two main advantages: it would allow us to test others’ editorial approaches by modelling them programatically; and it would allows us to take advantage of the inherent flexibility of the digital medium by providing users with access to limitless critical texts of the same work. Where, for economic and technological reasons, print editions tend to offer readers only a single critical approach and text, digital editions could now offer readings a series of possible approaches and texts built according to various selection criteria. In this approach, users would read texts either by building their own textual queries, or by selecting pre-made queries that build views by dynamically modelling the decisions of others—a Kane-Donaldson view of Piers Plowman, perhaps, or a Gabler reading text view of Ulysses.
This is an area of research we should pursue, even though, in actual practice, we are still a long way from being able to build anything but the simplest of texts in this manner. Certain processes can, of course, be automated and even improved upon electronically—we can use computers to collate readings from different witnesses, derive manuscript stemma, automatically normalise punctuation and spelling, and even model scribal performance (see Ciula 2005; O’Donnell 2005c). And it is easy to see how it we might be able to build databases and queries so that we could model human editorial decisions in relatively simple cases—reproducing the flawed dialectal texts of Cædmon’s Hymn discussed above, perhaps, or building simple student editions of small poems.
Unfortunately, such conceptually simple tasks are still at the extreme outer limits of what it is currently possible, let alone economically reasonable, to do. Going beyond this and learning to automate higher-level critical decisions involving cultural, historical, or literary distinctions, is beyond the realm of current database design and artificial intelligence even for people working in fields vastly better funded than textual scholarship. Thus, while it would be a fairly trivial process to generate a reading text based on a single witness from an underlying relational database, building automatically a best text edition—that is to say, an edition in which a single witness is singled out automatically for reproduction on the basis of some higher-level criteria—is still beyond our current capabilities. Automating other distinctions of the type made every day by human editors—distinguishing between good and bad scribes, assessing difficilior vs. facilior readings, or weighing competing evidence of authorial authorisation—belong as yet to the realm of science fiction.13.
This doesn’t let us off the hook, however. For while we are still far away from being able to truly automate our digital textual editions, and we do need to find some way of incorporating expert knowledge into digital editions that are becoming ever more complex. The more evidence we cram into our digital editions, the harder it becomes for readers to make anything of them. No two witnesses to any text are equally reliable, authentic, or useful for all purposes at all times. In the absence of a system that can build custom editions in response to naïve queries—“build me a general interest text of Don Juan”, “eliminate unreliable scribes”, or even “build me a student edition“—digital editors still need to provide readers with explicit expert guidance as to how the at times conflicting data in their editions is to be assessed. In some cases, it is possible to use hierarchical and object-oriented data models to encode these human judgements so that they can be generated dynamically (see note 14 above). In other cases, digital editors, like their print predecessors, will simply have to build critical texts of their editions the old fashioned way, by hand, or run the risk or failing to pass on the expert knowledge they have built up over years of scholarly engagement with the primary sources.
It is here, however, that digital editors can improve theoretically and practically the most on traditional print practice. For if critical reading texts are, conceptually understood, the equivalent of query-derived database views, then there is no reason why readers of critical editions should not be able to entertain multiple views of the underlying data. Critical texts, in other words—as post-structuralist theory has told us all along—really are neither right nor wrong: they are simply views of a textual history constructed according to different, more or less explicit, selection criteria. In the print world, economic necessity and technological rigidity imposed constraints on the number of different views editors could reasonably present to their readers—and encouraged them in pre post-structuralist days to see the production of a single definitive critical text as the primary purpose of their editions. Digital editors, on the other hand, have the advantage of a medium that allows the inclusion much more easily of multiple critical views, a technology in which the relationship between views and data is widely known and accepted, and a theoretical climate that encourages an attention to variance. If we are still far from being at the stage in which we can produce critical views of our data using dynamic searches, we are able even now to hard-code such views into our editions in unobtrusive and user-friendly ways.14. By taking advantage of the superior flexibility inherent in our technology and the existence of a formal theory that now explains conceptually what print editors appear to have discovered by experience and tradition, we can improve upon print editorial practice by extending it to the point that it begins to subvert the very claims to definitiveness we now find so suspicious. By being more like our print predecessors, by ensuring that our expert knowledge is carefully and systematically encoded in our texts, we can, ironically, use the digital medium to offer our readers a greater flexibility in how they use our work.
And so in the end, the future of digital editing may lie more in our past than we commonly like to consider. While digital editorial theory has tended to define its project largely in reaction to previous print practice, this approach underestimates both the strength of the foundation we have been given to build upon and the true significance of our new medium. For the exciting thing about digital editing is not that it can do everything differently, but rather that it can do some very important things better. Over the course of the last half millennium, print editorial practice has evolved an extremely efficient intellectual model for the organisation of information about texts and witnesses—even as, in the last fifty years, we have become increasingly suspicious of the claims to definitiveness this organisation was often taken to imply. As digital editors, we can improve upon the work of our predecessors by first of all recognising and formalising the intellectual strength of the traditional editorial model and secondly reconciling it to post-structuralist interest in variation and change by implementing it far more fully and flexibly than print editors themselves could ever imagine. The question we need to answer, then, is not whether we can do things differently but how doing things differently can improve on current practice. But we won’t be able to answer this question until we recognise what current practice already does very very well.
Bart, Patricia R. 2006. Controlled experimental markup in a TEI-conformant setting. Digital Medievalist 2.1 <http://www.digitalmedievalist.org/article.cfm?RecID=10>.
British Library, nd. Turning the Pages. <http://www.bl.uk/onlinegallery/ttp/ttpbooks.html>.
Cavill, Paul. 2000. The manuscripts of Cædmon’s Hymn. Anglia 118: 499-530.
Cerquiglini, Bernard. 1989. Éloge de la variante: Histoire critique de la philologie. Paris: Éditions de Seuil.
Ciula, Arianna. 2005. Digital palaeography: Using the digital representation of medieval script to support palaeographic analysis. Digital Medievalist 1.1 <http://www.digitalmedievalist.org/article.cfm?RecID=2>
Dobbie, Elliott Van Kirk. 1937. The manuscripts of Cædmon’s Hymn and Bede’s Death Song with a critical text of the Epistola Cuthberti de obitu Bedæ. Columbia University Studies in English and Comparative Literature, 128. New York: Columbia University Press.
───, ed. 1942. The Anglo-Saxon minor poems. The Anglo-Saxon Poetic Records, a Collective Edition, 6. New York: Columbia University Press.
Dobiache-Rojdestvensky, O. 1928. Un manuscrit de Bède à Léningrad. Speculum 3: 314-21.
Finneran, Richard J., ed. 1996. The literary text in the digital age. Ann Arbor: University of Michigan Press.
Foys, Martin K., ed. 2003. The Bayeux Tapestry: Digital Edition. Leicester: SDE.
Greetham, D.C. 1994. Textual Scholarship. New York: Garland.
Jabbour, A. A. 1968. The memorial transmission of Old English poetry: a study of the extant parallel texts. Unpublished PhD dissertation, Duke University.
Karlsson, Lina and Linda Malm. 2004. Revolution or remediation? A study of electronic scholarly editions on the web. HumanIT 7: 1-46.
Kiernan, Kevin S. 1990 Reading Cædmon’s Hymn with someone else’s glosses. Representations 32: 157-74.
───, ed. 1999/2003. The electronic Beowulf. Second edition. London: British Library.
Krishna, S. 1992. Introduction to database and knowledge-base systems. Singapore: World Scientific.
Landow, George P. and Paul Delaney, eds. 1993. The digital word: text-based computing in the humanities. Cambridge, MA, MIT Press.
Lapidge, Michael. 1991. Textual criticism and the literature of Anglo-Saxon England. Bulletin of the John Rylands University Library. 73:17-45.
───. 1994. On the emendation of Old English texts. Pp. 53-67 in: D.G. Scragg and Paul Szarmach (ed.), The editing of Old English: Papers from the 1990 Manchester conference.
Manly, John M. and Edith Rickert. 1940. The text of the Canterbury tales. Chicago: University of Chicago Press.
McGann, Jerome J. 1983/1992. A critique of modern textual criticism. Charlottesville: University of Virginia Press.
───. 1994. Rationale of the hypertext. <http://www/iath.virginia.edu/public/jjm2f/rationale.htm>
McGillivray, Murray, ed. 1997. Geoffrey Chaucer’s Book of the Duchess: A hypertext edition. Calgary: University of Calgary Press.
McKenzie, D.F. 1984/1999. Bibliography and the sociology of texts. Cambridge: Cambridge University Press.
Mitchell, Bruce and Fred C. Robinson, eds. 2001. A guide to Old English. 6th ed. Oxford: Blackwell.
Nicols, Stephen G. Jr., ed. 1990. Speculum 65.
Ó Cróinín, Dáibhí. nd. The Foundations of Irish Culture AD 600-850. Website. <http://www.foundationsirishculture.ie/>.
O’Donnell, Daniel Paul. 1996a. Manuscript Variation in Multiple-Recension Old English Poetic Texts: The Technical Problem and Poetical Art. Unpubl. PhD Dissertation. Yale University.
───. 1996b. A Northumbrian version of “Cædmon’s Hymn” (eordu recension) in Brussels, Bibliothèque Royale MS 8245-57 ff. 62r2-v1: Identification, edition and filiation. Beda venerabilis: Historian, monk and Northumbrian, eds. L. A. J. R. Houwen and A. A. MacDonald. Mediaevalia Groningana, 19. 139-65. Groningen: Egbert Forsten.
───. 2005a. Cædmon’s Hymn: A multimedia study, edition, and archive. SEENET A.8. Cambridge: D.S. Brewer.
───. 2005b. O Captain! My Captain! Using Technology to Guide Readers Through an Electronic Edition. Heroic Age 8. <http://www.heroicage.org/issues/8/em.html>
───. 2005c. The ghost in the machine: Revisiting an old model for the dynamic generation of digital editions. HumanIT 8 (2005): 51-71.
───. Forthcoming. If I were “You”: How Academics Can Stop Worrying and Learn to Love “the Encyclopedia that Anyone Can Edit.” Heroic Age 10.
Ore, Espen S. 2004. Monkey Business—or What is an Edition? Literary and Linguistic Computing 19: 35-44.
Pearsall, Derek. 1985. Editing medieval texts. Pp. 92-106 in Textual criticism and literary interpretation. Ed. Jerome J. McGann. Chicago: U Chicago.
Pope, John C. and R. D. Fulk, eds. 2001. Eight Old English poems. 3rd ed. New York: W. W. Norton.
Railton, Stephen, ed. 1998-. Uncle Tom’s Cabin and American Culture. Charlottesville: University of Virginia. Institute for Advanced Technology in the Humanities. <http://www.iath.virginia.edu/utc/>.
Reed Kline, Naomi, ed. 2001. A Wheel of Memory: The Hereford Mappamundi. Ann Arbor: University of Michigan Press
Robinson, Fred C. and E. G. Stanley, eds. 1991. Old English verse texts from many sources: a comprehensive collection. Early English Manuscripts in Facsimile, 23. Copenhagen: Rosenkilde & Bagger.
Robinson, Peter. nd. New Methods of Editing, Exploring, and Reading the Canterbury Tales. <http://www.cta.dmu.ac.uk/projects/ctp/desc2.html>.
───, ed. 1996. The Wife of Bath’s Prologue on CD-ROM. Cambridge, Cambridge University Press.
───. 2004. Where are we with electronic scholarly editions, and where to we want to be? Jahrbuch für Computerphilologie Online at <http://computerphilologie.uni-muenchen.de/ejournal.html>. Also available in print: Jahrbuch für Computerphilologie. 123-143.
───. 2005. Current issues in making digital editions of medieval texts—or, do electronic scholarly editions have a future? Digital Medievalist 1.1 <http://www.digitalmedievalist.org/article.cfm?RecID=6>
───. 2006. The Canterbury Tales and other medieval texts. In Burnard, O’Brian O’Keefe, and Unsworth. New York: Modern Language Association of America.
Shillingsburg, Peter L. 1996 Electronic editions. Scholarly editing in the computer age: Theory and practice. Third edition.
Silberschatz, Avi, Hank Korth, and S. Sudarshan. 2006. Database system concepts. New York: McGraw-Hill.
Sisam, Kenneth. 1953. Studies in the history of Old English literature. Oxford: Clarendon Press.
Smith, A.H., ed. 1938/1978. Three Northumbrian poems: Cædmon’s Hymn, Bede’s Death Song, and the Leiden Riddle. With a bibliography compiled by M. J. Swanton. Revised ed. Exeter Medieval English Texts. Exeter: University of Exeter Press.
Wuest, Paul. 1906. Zwei neue Handschriften von Cædmons Hymnus. ZfdA 48: 205-26.
1 In a report covering most extant, web-based scholarly editions published in or before 2002, Lina Karlsson and Linda Malm suggest that most digital editors up to that point had made relatively little use of the medium’s distinguishing features: “The conclusion of the study is that web editions seem to reproduce features of the printed media and do not fulfil the potential of the Web to any larger extent” (2004 abstract).
2 As this list suggests, my primary experience with actual practice is with digital editions of medieval texts. Recent theoretical and practical discussions, however, suggest that little difference is to be found in electronic texts covering other periods.
3 Synthetic here is not quite synonymous with eclectic as used to describe the approach of the Gregg-Bower’s school of textual criticism. Traditionally, an eclectic text is a single, hypothetical, textual reconstruction (usually of the presumed Authorial text) based on assumption of divided authority. In this approach, a copy text is used to supply accidental details of spelling and punctuation and (usually) to serve as a default source for substantive readings that affect the meaning of the abstract artistic work. Readings from this copy text are then corrected by emendation or, preferably, from forms found in other historical witnesses. In this essay, synthetic is used to refer to a critical text that attempts to summarise in textual form an editorial position about an abstract work’s development at some point in its textual history. All eclectic texts are therefore synthetic, but not all synthetic texts are eclectic: a best text (single witness) edition is also synthetic if, as the name implies, an editorial claim is being made about the particular reliability, historical importance, or interest of the text as represented in the chosen witness. A diplomatic transcription, however, is not synthetic: the focus there is on reporting the details of a given witness as accurately as possible. For a primer on basic concepts in textual editing, excluding the concept of the synthetic text as discussed here, see Greetham 1994.
4 It is indeed significant that the PPEA —the most ambitious digital critical edition of a medieval text that I am aware of—is at this stage in its development publishing primarily as an archive: the development of critical texts of the A-, B-, and C-text traditions has been deferred until after the publication of individual edition/facsimiles of the known witnesses (Bart 2006).
5 Transcriptions, editions, facsimiles, and studies mentioned in this paragraph in many cases have been superseded by subsequent work; readers interested in the current state of Cædmon’s Hymn should begin with the bibliography in O’Donnell 2005a.
6 While there is reason to doubt the details of Dobbie’s recensional division, his fundamental conclusion that dialect did not play a crucial role in the poem’s textual development remains undisputed. For recent (competing) discussions of the Hymn’s transmission, see O’Donnell 2005a and Cavill 2000.
7 There are other types of databases, some of which are at times more suited to representation of information encoded in structural markup languages such as XML, and to the type of manipulation common in textual critical studies (see below, note 14). None of these other models, however, express information as parsimoniously as does the relational model (see Silberschatz, Korth, and Sudarshan 2006, 362-365).
8 This is a rough rather than a formal definition. Formally, a well-designed relational database normally should be in either third normal form or Boyce-Codd normal form (BCNF). A relation is said to be in third normal form when a) the domains of all attributes are atomic, and b) all non-key attributes are fully dependent on the key attributes (see Krishna 1992, 37). A relation is said to be in BCNF if whenever a non-trivial functional dependency → A holds in R, X is a superkey for R (Krishna 1992, 38). Other normal forms exist for special kinds of dependencies (Silbertschatz, Korth, Sudarshan 2006, 293-298).
9 In actual fact, the model for a real bookstore invoice would be more complex, since the example here does not take into account the possibility that there might be more than one copy of any ISBN in stock. A real bookstore would need additional tables to allow it to keep track of inventory.
10 In actual practice, the model would be far more complex and include multiple levels of repeating information (words within lines and relationships to canonical reference systems, for example). This example also assumes that the word is the basic unit of collation; while this works well for most Old English poetry, it may not for other types of literature.
11 Of course, critical editions typically contain far more than bibliographic, textual, and lexical/grammatical information. This too can be modelled relationally, however, although it would be quixotic to attempt to account for the infinite range of possible material one might include in a critical edition in this essay. Thus cultural information about a given text or witnesses is functionally dependent on the specific text or witness in question. Interestingly, the more complex the argumentation becomes, the less complex the underlying data model appears to be: a biographical essay on a text’s author, for example, might take up but a single cell in one of our hypothetical tables.
12 The critical apparatus in most print and many digital editions is itself also usually a view of an implicit textual database, rather than the database itself. Although it usually is presented in quasi-tabular form, it rarely contains a complete accounting for every form in the text’s witness base.
13 This is not to say that it is impossible to use data modelling to account for these distinctions—simply that we are far from being able to derive them arbitrarily from two dimensional relational databases, however complex. Other data models, such as hierarchical or object-oriented databases can be used to build such distinctions into the data itself, though this by definition involves the application of expert knowledge. In O’Donnell 2005a, for example, the textual apparatus is encoded as a hierarchical database. This allows readers to in effect query the database, searching for relations pre-defined as significant, substantive, or orthographic by the editor. See O’Donnell 2005a, §§ ii.7, ii.19, 7.2-9.
14 In the case of my edition of Cædmon’s Hymn, this takes the form of multiple critical texts and apparatus: several reconstructions of the poem’s archetypal form, and various critical views of the poem’s five main recensions and collations. The criteria used to construct these views is indicated explicitly in the title of each page and explained in detail in the editorial introductions. The individual editions were extracted from an SGML encoded text using stylesheets—in essence hard-wired database queries reflecting higher-level editorial decisions—but presented to the reader as a series of progressively abstract views. In keeping with the developing standard for digital textual editions, the edition also allows users direct access to the underlying transcriptions and facsimiles upon which it is based. The result is an edition that attempts to combine the best of the digital and print worlds: the archiving function common to most electronic editions (and traditionally the focus of Cædmon’s Hymn textual research in print), with the emphasis on the presentation of expert knowledge characteristic of traditional print editorial practice.
Posted: Friday February 9, 2007. 18:36.
Last modified: Wednesday May 23, 2012. 19:54.
Commenting is closed for this article.