The Epistemology of the Electronic Text: Scholarly and Pedagogical Considerations Julia Flanders Brown University Julia_Flanders@brown.edu John Lavagnino Brown University John_Lavagnino@brown.edu Carol Barash Seton Hall University barashca@pirate.shu.edu 1997 ACH/ALLC 1997 editor the secretarial staff in the Department of French Studies at Queen's University Greg Lessard encoder Sara A. Schmidt edition image pedagogy In invoking "the epistemology of the electronic text" this session means to open up a set of questions which have to do with how we conceptualize electronic texts as vehicles for information, and how we imagine their particular kinds of authority. These issues have important implications for how we use electronic texts in teaching and research, since they lie at the heart of the cultural position of the electronic text, both within the culture of the academy and outside it. We would like to ask not simply what people want electronic texts to do, or what they want them to provide, but what drives these desires and how they affect the actual use of the data and function offered by electronic resources. The papers in the session approach these questions from several angles, but also speak to each other's concerns. John Lavagnino's paper on the place of images in the electronic scholarly edition inquires into the role of the image as guarantor of textual integrity, and the special emphasis that images receive in discussions of electronic editions. His argument that images loom disproportionately large in the imaginations of both creators and users of electronic texts engages with Carol Barash's paper on the pedagogical use of electronic texts, based on her work with the Women Writers Project's textbase and other electronic resources. She investigates both the actual use of images in teaching early women's writing, and the methodological context that supports and motivates that use, arguing that images enable different kinds of textual study and also have significant pedagogical effects on students' use of the electronic materials. Julia Flanders' paper addresses the question of the use of images in the context of attitudes towards the edition: both as a product of human judgement and taste, and as an accurate point of access to other, epistemologically prior documents. Her paper undertakes to explicate the function of the electronic edition in terms of the sociology of the academy, and its ascription of different kinds of authority to different kinds of textual and physical evidence: evidence which the electronic edition must offer in unfamiliar or defamiliarizing ways. Together, the papers will encourage a more self-conscious discussion of how we imagine electronic text resources, and their claims to sufficiency and authority. Trusting the Electronic Text Julia Flanders Introduction The trustworthiness or untrustworthiness of electronic texts and electronic editions is a topic of acute concern to those who see conventional libraries supplemented by new electronic resources, many of which offer unparalleled access to rare materials if they can only be relied upon for accuracy and scholarly integrity. The public media also finds accuracy and integrity of great interest as a way of approaching the issue of electronic texts generally. Caveats about lack of depth, shoddiness, and unscholarliness have a more than factual force: they express as well the intangible concerns about the effects of electronic materials on the cultural status of texts, or their intimation of some underlying paradigm shift with unforeseeable consequences. The trustworthiness of the electronic text thus becomes the focus for deeper questions about its cultural authority. As the electronic text, and particularly the electronic edition, receives increasing attention and use, expectations about what it should look like and do become more specific, ambitious, and realistic. We not only have diverse examples before us of successful electronic edition projects (the Canterbury Tales Project, the electronic edition of Samuel Johnson's Dictionary, the Rossetti Archive, the Blake Archive) but we also begin to have feedback from users--themselves increasingly knowledgeable--about what they expect and desire from the electronic edition; we have, in short, the growth of a set of common expectations which are even on the verge of becoming dogmatically entrenched. The electronic edition, it seems, should include some or all of the following: transcriptions of most or all of the important witnesses; an edited text which represents some sort of "best text" derived by the editor from the witnesses; digital images of illustrations, manuscripts, or even whole printed texts; annotations and textual notes; secondary criticism; facilities for collation of variants; and text encoding which identifies the salient components of structure and content upon which scholarly study is based. Evaluation Criteria The criteria by which we judge and value these various possible components of the electronic edition vary considerably. For transcriptions and other textual material, we apply at least the notion of transcriptional accuracy, which is familiar enough from the realm of the printed text but takes on additional complexity when the text is transcribed using something like SGML. Secondary sources and annotations are additionally judged for their factuality--the degree to which the statements they make are true--and for their scholarly merit, credibility, and relevance. Likewise, the edited version of the text will be judged by the respectability and credibility of the scholarship that produces it. Facilities for textual analysis, which in projects like the Canterbury Tales may form a significant part of the editedness of the edition, can be judged according to a standard of computational accuracy: the capacity of the software and encoding to produce correct, usable results to queries and collations. Similarly, the encoding of the text may transform aspects of the content into a more readily processable form (for instance, by encoding punctuation or delimiters as attribute values rather than as content) which again would be judged by its ability to reconstruct the text accurately and usefully. Finally, images are judged by their quantity and quality of data (resolution and accuracy) but also by the air of tangible, attestable reality which they bring with them. Reality and Models In the last two cases given above--images and what we might call computational features--we see an encapsulation of the polarities of the electronic edition. In the case of the image, data is provided which is only processable by the human eye, and which is nearly impervious (at least at present) to any kind of computerized retrieval or analysis. Almost in direct proportion to its intractability, the image is seen as substantiating the real-world existence of the text which the electronic transcription reproduces: without the visual evidence of the image, as John Lavagnino has argued, the electronic edition seems perversely ungrounded and untrustworthy. In the case of computational features, the original text is rendered far more accessible, in analytic terms, by the added encoding and the tools provided; we might even say that its textuality has been enhanced to the degree that the original sprawling data has been ordered and preprocessed by the work of encoding. This processing or functional modelling of the text, though, has a troubled status in relation to the edition's perceived authority, where that authority is construed along the axes of value which apply to the more familiar aspects of the edition derived from the realm of print. For one thing, the introduction of computational work into the process of editing the text (as for instance in the case of the collational features provided in the Canterbury Tales Project) seems to supplant what turns out to be a very important--though hard to quantify--component of the trustworthiness of the edition as a whole, namely the role of the scholarly editor. For another, the replacement of recognizable content by a computational model of it draws the text away from the realm of the tangibly real that things like images or sounds work to substantiate, with their irreducibility and untranslatability. These poles define in some sense the current parameters of the electronic edition--from unprocessable data as a facsimile of reality on the one hand, to highly processed data as a functional model of reality on the other. They also ask us to think about trustworthiness in the context of other criteria such as usefulness. Conclusion Scholarly and pedagogical use of electronic editions relies on the criterion of trustworthiness not simply as a kind of paltry intellectual crutch, but at a deeper level as a crucial component of the architecture of academic study. To the extent that the electronic edition is being assimilated into the general work of the academy, it is being naturalized there in a form which corresponds most closely to the systems of value already operative in that sphere. This paper concludes that these systems of value, while not irrelevant, may lead scholars away from the more functional analytical aspects of the edition, towards those which--though they feel familiar--ultimately have very little to do with the electronic medium. Bibliography John Lavagnino Completeness and Adequacy in Text Encoding Richard J. Finneran The Literary Text in the Digital Age Ann Arbor University of Michigan Press 1996 Peter Robinson Is there a text in these variants? Richard J. Finneran The Literary Text in the Digital Age Ann Arbor University of Michigan 1996 Kathryn Sutherland Looking and Knowing: Textual Encounters of a Postponed Kind Warren Chernaik Marilyn Deegan Andrew Gibson Beyond the Book: Theory, Culture, and the Politics of Cyberspace Oxford Office for Humanities Communication Publications 1996 The Place of Images in the Electronic Edition John Lavagnino Many of us are currently working on creating scholarly editions in electronic form, and a few of us have even managed to complete projects of this sort and make them publicly available. At this early stage, with so little experience of such editions---particularly experience with their use by scholars---it would seem unlikely that we'd be in a good position to issue any general rules about what an electronic edition should be, apart from those rules that can be carried over from the long tradition of scholarly editing for print publication (such as concern for accuracy of transcription and completeness of documentation). Yet there is actually one principle that has gained widespread agreement---a principle which is peculiar for the nearly complete absence of argument that has been presented in its support. This is the principle that states that an electronic edition must always include facsimiles of the sources on which it is based. The fullest statement of this position that I know is from G. Thomas Tanselle: the advantages of hypertext as apparatus will not be fully exploited unless its capabilities for visual reproduction are used. Digitized images of the original manuscripts and printed pages should always be provided, along with the more manipulable electronic texts (that is, keyboarded transcriptions of manuscripts and rekeyboarded or optically converted texts of printed pages). Just as a scholarly edition in codex form is considered deficient if it does not provide a record of variant readings, a hypertext edition (or ``archive'') should be regarded as inadequate if it does not offer images of the original documents, both manuscript and printed. Important physical evidence will obviously still be unreproduced, but at least the range of paleographical and typographical evidence made available will be far greater than has been customary in editions of the past---even in ``facsimile'' editions, which have usually been limited to single documents. [Tanselle 1995, 591] It is easy to observe in this statement the problem that afflicts many discussions on this subject: what starts out as a potential advantage of the electronic edition somehow turns into an absolute requirement, without any argument to support this transformation. And yet it seems perfectly clear that all we are really talking about is a technical possibility whose exploitation is valuable for many sorts of materials, but is by no means essential to every scholarly edition. If reproductions of the sources are really so necessary, then there is no reason apparent here why the rule applies only to electronic editions: it is perfectly possible for print editions to contain such information as well, and if it's really necessary it's necessary for those editions too. (See also Litz (1996) for another instance of this position.) But although no reasons are given for this curious position, it is possible to infer some. One reason is no doubt the submerged idea that an electronic edition is somehow not as real as one on paper (McKenzie 1991 is one publication that actually advances a form of this argument directly rather than assuming it without discussion). A variant of this view, and one that does have some reasonable grounding, is the idea that transcriptions in some kinds of electronic editions are not terribly reliable because they were produced by scanning or offshore typists rather than scholars, and that we therefore need the images to check the transcriptions; Womersley (1996) seems to have this in the back of his mind, in reviewing such a production, though he doesn't fully articulate it. This is a point that only applies to editions created in this particular way, though, and not to electronic editions in general. A larger reason is the lingering strain of positivism that afflicts both humanities computing and textual editing: in both disciplines the idea still persists that it could be possible simply to establish the facts without any element of interpretation, and that indeed it would be best if we could eliminate all interpretation from our work. Yet any informed understanding of digital images involves an awareness of the large number of choices that go into decisions on just how to do it---on the resolution, color spectrum, and lighting, for example (see Robinson 1993 for a survey, and Tanselle 1989 for a general account of the pitfalls of image reproduction). And, for works produced in the era of print, it is in general impossible to collect and digitize the entire range of extant sources: since each copy of a book can be different, a truly definitive collection of data requires the imaging of every last copy. The images that go into an electronic edition are necessarily the product of scholarly selection from a wide range of materials and ways to reproduce them; such processes of selection are inescapable in editing, and they are among the ways in which an edition creates a new representation of the text, rather than simply transmitting information about it. A more reliable basis for thought on just what an electronic edition needs would focus on the particular nature of the texts in question, and on the representation of them that the edition seeks to create, rather than on an attempt to raise up general rules applicable to every edition. For some texts, the importance of reproductions of original sources has been well argued; their significance for the study of manuscripts has been made particularly clear in several cases. For other texts, an editor may well judge that different kinds of materials are more important than images of sources, such as annotations or the texts of related works and adaptations. There are choices involved in doing this, choices with which other scholars might disagree. But the avoidance of choice is not an alternative, because it's not possible. Works Cited A. Walton Litz Afterword Richard J. Finneran The Literary Text in the Digital Age Ann Arbor University of Michigan Press 1996 245-248 D. F. McKenzie Computers and the Humanities: a Personal Synthesis of Conference Issues May Katzen Scholarship and Technology in the Humanities: Proceedings of a Conference held at Elvetham Hall, Hampshire, U.K., 9th-12th May 1990 London Bowker 1991 157-169 Peter Robinson The Digitization of Primary Textual Sources Oxford Office for the Humanities Communications Publications 1993 Thomas G. Tanselle Critical Editions, Hypertexts, and Genetic Criticism The Romanic Review 86 3 581-593 May 1995 Thomas G. Tanselle Reproductions and Scholarship Studies in Bibliography 42 25-54 1989 David Womersley Delightful ways to cheat learning Times Higher Education Supplement vii 14 June 1996 Textual Studies, Cultural Studies: Text, Image and the Production of Knowledge in the Interdisciplinary Literature Classroom Carol Barash Introduction This paper is based on pilot courses using the Brown University Women Writers Project (WWP) textbase in basic literature courses at Seton Hall University. I will discuss the relationship between texts and images in the use of a full-text database to teach several aspects of literary study: close reading and analysis of the text's language and structure (textual studies), the relationship between the text and its historical contexts (cultural studies) and literary history, and writing about literature. Students are clearly attracted to the visual aspects of electronic culture. Do the visual images lead students to a better understanding of cultural differences? Can pictures and sound, because they seem comfortable and knowable, enable students to learn what is different about other cultures and other historical periods, and learn to read and understand the cultural referents in a literary text? Those questions lead one to a complex matrix of assumptions about knowledge, reading, and culture, which this paper will develop in relation to teaching literary and cultural studies via the electronic medium. Textual Studies By "textual studies" I mean work with the language and structure of a text. The electronic medium greatly enhances close reading of the literary text. From easy access to "virtual reference room" sources (dictionary, bible, myth, time lines) to textual analysis (word counts, collocates, structural analysis), electronic texts allow students to work closely with the language and structure of a literary text, to analyze local structures and meanings, and to study how linguistic patterns play out in a longer work. SGML enables students to work structurally and formally, and particularly to study the relationship between language and structure in a literary work. Students are encouraged to test out their own ideas about the text, and to see whether and how those ideas are manifest in the text's language. It would seem that one does not need images to perform strictly "textual" studies in this sense. Writing about Literature It would also seem that one does not need anything more than words to teach students how to write about the language of a literary text: take the data they have from close work with the text's language and structure and teach them how to translate that data into an argument about the text. One can even teach them to perform textual analysis on their own work (which words and phrases appear repeatedly? what are they close to? what are they code for? do key words appear in the key sections of the text?), and to think about their own writing as linguistic and rhetorical structure. One might imagine that student's writing is enhanced by close work with the language of a literary text, particularly in the absence of all the hypertext capabilities they like so much, but tend to use like a video game. Cultural Studies, Literary History, and Electronic Texts In order to understand why one needs the images--even to teach students to perform linguistic and textual analysis--we need to think about the electronic text in relationship to other historical and cultural variants of the literary text. I take cultural studies and literary history to be necessarily overlapping domains. By "cultural studies" I mean the relationship between the text and its own history, and also its relationship with other literary and cultural texts. Here one begins with the materiality of the text: what it looked like, which texts it was responding to, how it was produced (by the author, the publisher, the audiences who first read it), what other texts (either words or images) appeared close to it in its original form, how it changed over time (how the words were arranged or packaged differently, what other texts appeared near it in new contexts). The cultural history of a literary text quickly leads the scholar or teacher to the text as a material object, and its relationship to other texts as material objects (a bible that fits inside a lady's purse is very different from a folio produced for reference in a library; one gets a much clearer sense of what Katherine Phillips's poems are trying to do when one sees the early 17th-century emblem books she was imitating; the allusions in Jane Barker's manuscript poems come from anti-Catholic broadsides of the 1680s; etc.). Some of this information could be conveyed (and in some cases, conveyed better) by verbal description, and some things will never be knowable without access to the original text. However, images of material included with the published versions of a literary text--title page, front matter, engravings--allow the student to feel that much closer to the original "text," the book that people held in their hands, read to one another, and experienced--in some sense--as a physical object. The feeling of comfort and proximity associated with printed books, and with the way knowledge is conveyed through books, extends to the pedagogical situation. Conclusions We are clearly in a transitional moment, and many of our students learn better in the electronic medium than they do by traditional means. Images (visual and aural) are crucial to teaching what is actually different about literature from different historical moments. I do not mean to be offering a simple paean to hypertext, or to multimedia, or to electronic dazzle of any kind; rather, I wish to suggest that we and our students already think about--and inevitably study--literary texts as cultural, material, and visual artifacts. We study (and teach and produce) layers of textual meaning that work back and forth between text as reproducible verbal structure and text as multiple, ambiguous and conflicting instances. While much crucial cultural information about a text resides outside it--and is thus the responsibility of the teacher or scholar, but not the textbase editor--those aspects of the text that were originally visual ought to be included as part of a textbase that is intended as a pedagogical resource. Rather than recreate formalism in the electronic medium, I want to argue that we should think of this new textual domain as part of cultural and interdisciplinary history, and therefore a place in which the words of a text always exist in relationship to other verbal and visual texts, which ideally one ought to be able to have access to for teaching and scholarship.