At the 1995 ACH/ALLC conference in Santa Barbara, I announced a new electronic editing project: an electronic Wordsworth edition, co-edited by Ronald Tetreault and myself. Within a month of the announcement, we were contacted by Cambridge University Press; within another year, we had contracted with them to produce an electronic edition of Lyrical Ballads, which we then saw as a pilot project for a full edition of Wordsworth's poetical works. In this paper I will discuss the ways in which we have redefined our project; I will then describe some of the problems we have had in assembling it, focusing especially on the difficulties we have had encoding authorial manuscripts in such a way that they can be conveniently and accurately represented in existing software.

I. From Hypertext to Archive

We originally conceived of our edition as a kind of hypertext, modeled after descriptions of similar projects undertaken by Jerome McGann, Stuart Curran, and others. Our aim was to bring together in electronic form not only the texts of the Lyrical Ballads themselves, but all of the relevant manuscripts, much of the source material, all of the contemporary reviews, all authorial comments, and whatever else seemed pertinent to this collection of poems. These materials would be organized with a variety of hypertext links which would, we thought, easily enable the reader to switch from one version of the poem to another, from poem to source, from poem to review, and so forth. But it quickly became evident that such an edition would not be feasible. There were the usual problems with permissions, of course, but the main problem was one of audience and design. Would it make sense, in the early days of the electronic edition, to try to assemble such a massive amount of materials? Would scholars, many of whom are unfamiliar with the electronic environment, be able to negotiate an edition so complex? And, finally, would we ever finish such an edition, so that Cambridge could establish a market for electronic products and continue to support other endeavors like ours? As we worked, our skepticism grew. As it happened, the bibliographical complexity of the Lyrical Ballads came to our rescue. Several different versions of the collection survive. It went through four editions in seven years, each of which is substantially different from the others; there were two distinct printings of the first edition, and first printing of the first edition survives in three distinct versions; the second edition also contains significant printed variants from one copy to the next, in part because of printer's errors and in part because the authors revised their poems while the volumes were in press. And there are also unauthorized printings of the collection, including an American edition and an odd set of London editions made from the discarded sheets of the original printings. Scholars, of course, have known about these differences for some time, but few have ever seen the wide variety of printed variants because the volumes which contain them are very rare and are scattered in libraries and private collections from New Zealand to Grasmere to Colorado. So we decided to attempt to bring together in one place all of these variant printings, supplying full texts of specific copies of Lyrical Ballads, supplemented by photographs of the printed pages in all of their various states. At the same time, we decided to eliminate many of the supporting documents that we had originally intended to include, believing that they would distract and perhaps even confuse readers. In short, we stopped thinking of our work as a hypertext, and began thinking of it as an archive, a virtual library containing digital representations of printed volumes that few had ever seen. This project seemed doable and useful, and would have the added benefit of not burying the Lyrical Ballads under the sheer weight of its associated texts.

II. Representing Revision

Because we decided to focus our edition on the printed versions of Lyrical Ballads, we minimized the need for reproducing poetical manuscripts. We thereby avoided the need to get permissions to reproduce copyrighted material, and we also avoided the difficulties inherent in encoding and representing in software extremely messy draft work. But we could not avoid manuscripts altogether. Our edition seeks to preserve the history of the printed artifact, Lyrical Ballads. Part of that history is contained in the printer's manuscripts, the most substantial of which survive in the Beinecke Library at Yale. These printer's manuscripts have been given added importance by the most recent editors of the collection, James Butler and Karen Green, whose edition appeared in 1992 in the Cornell Wordsworth series. Butler and Green chose the printer's manuscripts of Lyrical Ballads (1800) as their copy text; for us to ignore the printer's manuscripts would be to ignore the texts of the poems that most scholars are now reading. So I set to work transcribing and encoding both of the Yale sets of printer's manuscripts: those for 1800 and 1802. The main problem I had to confront was how to encode the manuscripts in such a way that their subtleties could be represented adequately on a computer screen. There was only one model: Peter Robinson's Wife of Bath's Prologue, which I studied very closely. But the differences between a 15th century manuscript, drawn up by a professional scribe long after the author's death, and a printer's manuscript, drawn up by the authors themselves, are enormous, and it soon became clear that Robinson's model would be of little use. Where Robinson had just a handful of revisions to deal with (few of them with any authority), I had thousands, all in the autograph of either the author or his copyist, and I had to find a variety of ways to encode these revisions, in order to distinguish them from each other. For example, Wordsworth's 1802 printer's manuscripts were prepared from proof sheets of the 1800 edition, and thus the "manuscript" looks more or less like a printed book which has been marked up by its owner. But there is little consistency in the ways revisions were entered (the sort of consistency that computers, which depend on global commands, clearly understand). Sometimes revisions were written in the margins, sometimes at the foot of the page, sometimes in between lines, and sometimes as overwrites; where the revisions became too extensive to fit on the page, Wordsworth inserted loose sheets of writing paper and entered them there. For the 1800 printer's manuscripts, the complexities are even greater. These manuscripts are a series of letters, prepared in the Lake District by Wordsworth and Coleridge and sent to their printer in Bristol, hundreds of miles away. The letters were sent over a period of several months; the first ones were sent before the authors had determined the final order of the poems, and before several of the poems were even written. The manuscripts were revised as they were prepared, and later letters often include revisions to poems sent weeks earlier. There are also several different autographs in the manuscripts: the base text was prepared by Coleridge or Dorothy Wordsworth, usually, and then revised either by William Wordsworth or Coleridge. Definitive punctuation was added in Bristol by the brilliant young chemist, Humphry Davy, and there are also marginal notations made at the printers, including one that caused the omission of 15 lines of one of the poems. Now all of this is clearly part of the publication process, and it helps to explain the state of the published volumes. And it can be encoded, fairly accurately although not without great pains, with the TEI DTD. But can it be represented clearly in a browser? Can textual scholars, used to the way manuscripts are transcribed in printed volumes, see something vaguely like what they are used to seeing? This is a problem that we yet to solve, and I would like to describe some of the difficulties at the conference, in hopes that someone there will be able to offer suggestions.