Though his claim that "poetry is the spontaneous overflow of powerful feelings" might tempt us to favour texts closest to the original sources of inspiration, Wordsworth is equally famous for the inveterate revision of his poems over the remaining half-century of his life. Recollection in tranquillity cannot help but overlay spontaneity with deliberation, but if perfection is the work of time then we may be justified in preferring the final authorized texts. Then there are the versions in between, where a poem finely polished may be marred by subsequent over-refinement. Which to choose has always been the editor's dilemma; establishing a text has always meant that we must privilege one version over others, and settle for a static representation of what might be better understood as a dynamic process. De Selincourt's landmark edition confirmed the authority of Wordsworth's final texts. "No poet ever paid more meticulous or prolonged attention to his text than Wordsworth," wrote De Selincourt° in justification of choosing the last lifetime edition of 1849-50 as his copy-text, though he was careful to indicate evidence of development by citing early manuscripts in an apparatus criticus at the bottom of the page. Unease with this portrait of the poet in old age has become more acute among his recent editors, who have attempted a snapshot of a Wordsworth closer to the moment of inspiration by printing the earliest completed version of the poems. A leader of this movement is Stephen Gill, whose Oxford Authors edition aims to restore the "original identity" of the poems.° But the idea of a poem's "identity", like that of the poet's, is destabilized when we begin to regard it as a work in progress. In his recent book, Revision and Romantic Authorship (Oxford University Press, 1996), Zachary Leader argues that the protean Romantic self cannot be captured in the earliest versions of poems. All writers revise their texts, and perhaps none more so than Wordsworth. In a review of this book, Frank Kermode writes that Leader “wants to know what notions of identity underlie the assumption that a poet in his twenties could be identical with the poet who, in his seventies, was still tinkering with his early writings, as if they were essential to the expression of the singleness of a life or a life-work, rather than leaving them alone as virtually the work of a different person, or at any rate of a person in no need of being assimilated to a later one.” ° Revision for Leader is a vital part of the creative process, and reveals a complexity in acts of composition that are at odds with comfortable notions of a unitary self. It is dubious not only to think that age is best which is the first, another instance of that bias Jerome McGann warned against, stemming from our "uncritical absorption in Romanticism's own self-representations".° Besides the merits of mature reflection, more than one hand might be involved in the making of changes (as is clearly evident in many Wordsworth manuscripts), further undermining the notion that the Romantic poet was a solitary genius who sings as artlessly as a bird. But the print medium is not well-equipped to show change over time except by the cumbersome apparatus of footnotes, cannot possibly for reasons of space present all versions as they evolved, and can only with difficulty make evident the extent to which composition and revision can be a social activity. Wryly, Kermode observes that these early, late, and intermediate versions "may one day be represented by hypertext, ....[although] these plural texts are not likely to be of much use to people who simply want to read Wordsworth and leave it to the experts to give them a text." True, an electronic Wordsworth may not be eagerly sought out by the general reader; I've met very few lovers of reading who prefer a screen-image over an affordable book you don't even have to plug in to use. Indeed, such a Wordsworth edition may not in fact be meant to be "read" at all, at least not in the sense in which we still use the term today.°Instead, a hypertext Wordsworth would be meant to be explored, studied closely in a fashion that the linear structure of print makes difficult, and used at a distance by scholars who don't enjoy the privilege of proximity to a major research library. Even more important, Wordsworth in hypertext may be the most effective way yet to represent Wordsworth in development. This medium reinscribes textual stability as a series of moments in a lengthy creative process, for, by adding motion to comparative views, hypertext enables us to represent change. His transformation from Romantic rebel to Victorian sage makes Wordsworth an ideal choice for exploring the potential of this fluid means of representation. Such a rendition of his text would be a key example of the way digital media are "contributing to a general reconsideration of traditional, unitary notions of identity", as Sherry Turkle, a student of the self in the computer age, has written.°Thus, when Bruce Graver and I announce here that Cambridge University Press has agreed to publish our electronic edition of Lyrical Ballads on CD-ROM, we do so conscious of the enormous value of previous print editions and the scholarship that has produced them, but also confident that an edition in this form will do what print never quite could to show Wordsworth as an evolving self. There have been several other pioneers on this particular frontier of cyberspace, but none so far have conceived their project as we have. Chadwyck-Healey's English Poetry Full-Text Database is perhaps the best-known of the CD-based products, but it by no means focuses on Wordsworth nor does it print more than one version of any poem, and that usually from an out-of-copyright late edition. Their student version, English Poetry Plus on CD- ROM, contains just 39 Wordsworth titles, and never acknowledges the source of any of these texts. Using software designed by Electronic Book Technologies of Providence RI, it does have a pleasing user interface, an effective search function, and links to brief biographical sketches. A more scholarly work is promised by David Miall of the University of Alberta; his Romanticism: CD-ROM forthcoming from Blackwell's is an electronic anthology offering a "generous selection" of poems by Wordsworth, but how these are to be chosen and what texts they are to be based on is not specified. A networked environment promises much wider dissemination than a single-user CD, so it is gratifying to see Wordsworth already prominent on the World Wide Web. The best guide to Wordsworth on-line is to be found in Alan Liu's gloriously encyclopedic Voice of the Shuttle Web Page for Humanities Research , a resource to which we are all indebted. At least a couple of what can best be described as "fan pages" have been posted: Thomas C. Gannon's Wordsworth Page (which features a thoughtful quote for the day) and Richard Darsie's Selected Poetry of William Wordsworth, an eclectic gathering of a dozen poems. More systematic collections have been made available by academic institutions. Columbia University's Project Bartleby reproduces all the poems in the 1888 edition of Wordsworth's Complete Poetical Works, indexing them chronologically and by first line. The Representative Poetry Archive at the University of Toronto offers a selection of 44 poems from a wide range of textual sources, all documented. Individual scholars have also begun to experiment with posting electronic texts for study purposes. Michael Gamer's collection of the prose associated with Lyrical Ballads is a good case in point, while Richard Bear (a Ph.D. candidate in electronic book design at the University of Oregon) has attempted to recreate the text of the rare 1798 Bristol first edition, first issue of Lyrical Ballads, probably based on a copy of the 1926 Noel Douglas replica. Our project differs from these in several significant respects. First, the poems we will present have not been arbitrarily chosen, but are a reasonably coherent group stemming from a single collection first published jointly but which came to be dominated by Wordsworth's poems in subsequent editions. Second, we will offer not just one version of each poem but all the versions of poems by Wordsworth and Coleridge appearing in the four lifetime editions of Lyrical Ballads, together with later important versions of poems revised by Wordsworth in his lifetime, and these will be linked hypertextually. Third, our copy-texts will be taken from the original editions themselves as held in libraries around the world, though of course our procedures will be informed by the findings of previous scholars, especially the editors of the Cornell Wordsworth series. Fourth, our e-texts will be "marked-up" or tagged using SGML (Standard Generalized Markup Language) in conformity with the principles of the Text Encoding Initiative (TEI). Fifth, we plan to link our transcribed e-texts to scanned images of the original printed editions in order to give the reader some sense of the look of the poems upon the page. Finally, this scholarly hypertext edition will be issued on CD-ROM in the first instance, with the intention of proceeding to network distribution as soon as it becomes practical. Cambridge University Press's CD-ROM series currently consists of four offerings: Peter Robinson's edition of The Wife of Bath's Prologue, a World Shakespeare Bibliography 1990-1993, Samuel Johnson's Dictionary, and the Works of John Ruskin. The first of these (the one with which I am most familiar) assembles all 58 of the pre-1500 manuscripts and printed editions of the prologue, using Robinson's Collate program to compare the different versions and to generate an apparatus of variants. The 58 versions, together with the apparatus and digitized images of 1200 manuscript pages, are linked together using Electronic Book Technologies' DynaText software to create a hypertext "book". Typically, a DynaText book is

presented on the screen in two windows (click on the image for a clearer view), one a narrow column on the left containing a table of contents and the other a wider data-window on the right containing the text or texts compiled from top to bottom in one large file document. Rather than accessing the contents of that file document linearly, a simple mouse-click on one element of the table of contents takes the user directly to the designated section of the text. Each such section may be opened as an independent window, and the windows so generated can be scaled and moved about the screen to permit the comparison of texts. Although DynaText is powerful software that can handle a huge mass of complex material, I find myself as the member of the editorial team responsible for hypertext design chafing at some of its limitations. DynaText was initially conceived as a commercial product designed to provide rapid indexed access to large prose documents such as catalogues and manuals; its table of contents window gives effective access to works with section titles and sub-headings, but is a blunt instrument when it comes to the presentation of poetry, where the user quite rightly demands access line by line. Short of embedding every line of each poem into the table of contents, there seems to be no easy way of navigating through the variant texts. The Collate-generated apparatus helps if opened into an independent window, but then other windows have to be opened and manipulated on the screen in order to compare lines once interesting variants are found. But the cluster of multiple windows thus generated can be awkward to scale and arrange on the screen, nor is it easy to keep track of which window holds which portion of the immense file document. With a multiplicity of texts such as the Lyrical Ballads offers, it is all too easy for the reader to get lost in cyberspace. The navigational tool so far proposed, the inclusion of a "base text for collation" at the top of the long file document in the data-window, is as much a relic of print culture as the model of the single-document "book" divided into subsections. In addition to the multiple versions of each poem, the editor is obliged either to construct a further ideal version of the of the poem to serve as the "base text" or, worse (because it violates the principle that this new medium should not merely reproduce the book but transcend it), to choose one of the existing variant texts, say from the first edition or the last lifetime edition, as the "base text". I had thought to address this difficulty by including texts of all the poems from all four lifetime editions of Lyrical Ballads as potential base texts, and then asking the reader at the beginning of each session to choose which of the editions he or she wished to use as the base text against which all the others would be collated. But this was not interactivity so much as an offloading of editorial responsibility, and furthermore resulted in an unnecessarily vast proliferation of alternative texts within the DynaText file. Instead, to address these navigational problems, I have begun to experiment with functionalities associated with Internet delivery of documents over the World Wide Web. An array of windows is essential to the display of the Wordsworth project, for as Sherry Turkle observes "windows have become a powerful metaphor for thinking about the self as a multiple, distributed system."°What seems most attractive about HTML (Hypertext Markup Language) standards at the present stage of development is that they enable the use of frames, a system of layout in which windows are generated on the screen according to a pre-determined pattern. Though this pattern can be easily altered by the user, its grid provides an intelligible starting point for the display of multiple texts of the same poem on the screen at the same time. My prototype pages for the Lyrical Ballads Hypertext Project (click to see it in full interactivity) take "Simon Lee" as a test-case, and consist

of two successive contents pages which lead to a galaxy of pages each composed of five frames: one narrow vertical column on the left plays the customary table of contents role, while a grid of four squares on the right allows the reader to compare four different versions of the poem on the screen at once. The text in each window may be scrolled through manually, but by the use of internal anchors a simple click on a live hyperlink in the left-hand window causes all four texts in the squares to scroll simultaneously to the same line (click on the image for a clearer, though static, view). In this way, a balance is struck between exploration and direction in the reader's examination of textual complexity. There is a second feature of this scheme meant to help readers find their bearings. The left-hand column contains no mere list of contents but what I call a "variant map" of the poem being studied. The variant map is a guide to revisions that were made at various stages in the poem's development. It is based on the poem to the extent that it reproduces the text of the poem wherever changes were not made, but whenever a change in any of the versions under consideration is encountered it substitutes a descriptive hyperlink for the variants themselves. The reader is thus alerted that an alteration has been made, and by clicking on the "hotspot" can summon up the parallel passages. Replacing the base or reading text with a variant map turns the annotation process inside out in a way that seems appropriate to the dynamism of this medium, for the hiatus in reading caused by running across a link cues the reader to click on the spot and look to the right (in the direction of the normal flow of reading) to learn the word or character elided and to compare texts. Rather than footnotes which distract attention from a definitive text, the variant map is an abstraction of the poem which does not privilege one version of the text over another and that piques the reader's curiosity by means of gaps in the text to pursue the significance of revisions made in successive versions. Without doubt, there are numerous disadvantages to this proposed scheme of display. Currently available bandwidth is simply too slow to deliver the amount and complexity of digitized texts and images we envision. Besides, not everyone is yet using a frames-capable browser, such as Netscape 2.0 or better (and some users are still restricted to the text-only Lynx system). Furthermore, the 14-inch monitors commonly in use today do not have enough screen real-estate to display more than four versions of the text at a time, and even this minimum array for our purposes appears somewhat cramped. While these drawbacks may soon disappear given the rapid advances in computer technology we are experiencing, a more lasting objection is raised when we consider that HTML is a much simplified subset of SGML and therefore inadequate to represent the complexity of the texts in question. However, in the case of the variant map element, it should be recognized that it can function quite irrespective of the markup scheme used. The way its links interact with the frames is another matter, though, since DynaText will not process frames the way a web-browser can. Help is on the way from Electronic Book Technologies, as I have been informed by company representatives that their DynaWeb software is capable of browsing DynaText books tagged in SGML, conveying them across a network, and in the process reproducing frames. Perhaps the solution to the apparent incompatibility of Netscape extensions and SGML structures is to project the DynaText Lyrical Ballads through a DynaWeb browser. In whatever ways this project may develop, its goal is not to supplant books nor to discredit the achievements of earlier editors, but to attempt to do things through this medium that print cannot, and in the process to discover what unique capabilities it has, if any. The fixity of print has been able to give us one Wordsworth or the other, but is not capable of fully representing his evolving consciousness. The poet was well aware that each of us passes through a succession of selves, not completely fragmentary and unrelated to one another but sufficiently discontinuous for him to wish for our "days to be/Bound each to each" by some means or other. Even if no more successful than natural piety, digital media can at least enable us to give attention to the various stages of composition and revision and to give each of them the respect they deserve. In the end, we may find that we have done no more than translate the contents of one medium into another, but that will not be without value, as textual critic James Thorpe reminds us: “All editions carry the taint of time. They are for the here and now, whether that is a decade, a generation, or a century. Ultimately they must all be replaced for most purposes.” ° To the extent that computers can help us to embody postmodern concepts like the decentered self and show them at work in specific instances, they are the tools of the time and will certainly have an impact on the study of literature. At the least, digital hypertext editions can spark a renewal of interest in editing and textual scholarship, analytical and descriptive bibliography, and even assist in the realignment of literary studies from the theory and practice of critical interpretation to the new cultural history's focus on the literary marketplace and the material production and representation of "literature".°What consequences they will have we cannot know until we try.