Workshop of Filthy Creation, Cyberspace Division

Jack Lynch

Hypertext is the nineties' critical buzzword of choice, but many of its promises remain only promises. Dozens of recent articles and even several books exploring the intersection between hypertext and literary theory have raised all the right questions, but few bother to descend to the practical decisions on which any actual hypertext project must rest. Landow, for all his insights, cannot pretend to have the experience that comes from manually assembling tens of thousands of files. This is, by contrast, a report from the front on some of the practical -- even mundane -- decisions behind one of the most elaborate literary hypertext projects yet attempted, the Pennsylvania Electronic Edition of Mary Shelley's Frankenstein, due in 1997 as a CD-ROM from the University of Pennsylvania Press.

Stuart Curran is the editor for the project; two graduate students in English, Sam Choi and I, have worked as assistants, along with a number of undergraduate students. We settled on Shelley's Frankenstein because it was included in the Penn Reading Program: all of 1993's incoming freshmen were required to read the novel. Many of the reasons that guided the novel's inclusion in the Reading Program made it a good choice for an electronic edition: its appeal to readers from high school students through the academy, for instance, a range not to be found in, say, Catcher in the Rye on the one hand or Eikonoklastes on the other. The idea of multiple readerships has been a guiding principle from the beginning.

More important, the message is exceptionally well suited suited to the medium: as Curran writes, Frankenstein "manifests a series of problems, all of them needing to be worked through by a serious student of the novel, yet at the same time being in the aggregate beyond the capacity of print technology."1 The novel's pop-culture representations make it well suited to multimedia. It is situated in an elaborate web of themes and contemporary contexts. Shelley stood at the center of one of the greatest literary circles in English history, and the relations between the members of a literary circle are an almost perfect analogue for the hypertextual links between fragments of text. Even the words themselves present problems that lend themselves to hypertext solutions: the text exists in two substantially different versions, the first edition of 1818 and the thoroughly revised third edition of 1831. All of this makes for just the sort of rich hypertext we hoped for: an intertextual web integrating the texts and its contexts, including science, literature, philosophy, feminist thought, naval exploration, mythology, geography. Frankenstein comments upon, even incorporates into itself, an entire library, from Aeschylus's Prometheus Bound through Percy Bysshe Shelley's Prometheus Unbound, with stops along the way at Milton's Paradise Lost, the works of Rousseau and Goethe, the political philosophy of Godwin, the feminist thought of Wollstonecraft, the Gothic, and so on, almost without end.

Having settled on a work, we turned our attention to making the edition as widely available as possible. The choice of reader software was our first consideration, for with limited resources we could hardly develop one from scratch. The SGML standard is well suited to the presentation of texts, but there is no widely available and inexpensive reader. Several commercial packages have been used to produce very professional products, but they seem better suited to the multimedia aspect of the project than to the textual and hypertextual. We settled finally on HTML, the language of the World Wide Web. Its status as an emerging standard caught our attention, since it promises to become a more prominent player in the hypertext and multimedia world. We plan to distribute our CD-ROM containing the HTML-encoded files along with PC- and Macintosh-compatible diskettes containing a freeware browser, but to allow the user to use any browser he or she finds convenient. When it at last becomes convenient to provide licensed access over the Internet, we look forward to making the project even more widely available.

HTML is not, however, a perfect choice. Although the language is always developing, and features whose absence we lamented two years ago are now part of the ad-hoc standard, HTML does not readily allow for split screens, which would simplify collations of textual variants. It was not designed with verse in mind, making the lengthy extracts of poetry difficult to manage. Some foreign characters are supported but by no means all -- and those that are included aren't easy -- making editing and proofreading foreign criticism painstaking; the want of a code for the British pound sign, for instance, reveals both the shortsightedness of the developers and the problems we had to work around.

With the book itself, the reader software, and the markup language settled, we began work in earnest by establishing a text. Traditional editorial practice would have us choose one edition or the other as a copytext and then relegate the substantive variants to a textual appendix. But while critical print editions make the variants are available to researchers, they do not allow them to reproduce the experience of reading the text that is available only in the apparatus. Hypertext has no such limitation. We resolved therefore to include both texts entire, giving the reader the choice of which edition to use. Each text is fully annotated, with a link from each passage of 1818 to the corresponding passage of 1831 and vice versa, giving the reader the choice of reading either text through, or of jumping back and forth between the texts at any time. To point up the differences between the two texts -- some of the revisions are themselves significant in showing how the novel changed -- we provided a link from either 1818 or 1831 to an intermediate screen showing a collation of the two versions. The system can be seen here in schematic form:

To make the novel manageable, we broke the text into segments shorter than a chapter but usually longer than a paragraph -- a unit of no more than around thirty lines, never splitting a paragraph in the middle. Nomenclature was a constant problem at this stage of the project: "page" misleadingly suggests the paper equivalent; "file" sounds too technical. We settled on "frame," which seemed neither too complicated nor too imprecise.2

Then came the commentary. Stuart Curran was primarily responsible for the glosses on the text. The Pennsylvania Electronic Edition, even without the extensive critical and contextual material, is by far the most extensively glossed edition of the novel, or perhaps any novel ever: the "first-level" footnotes alone already stretch to over 150,000 words, thrice the length of the novel itself, a mass of commentary that could not be practically accommodated in any paper edition. And these textual glosses are accompanied by supplementary factual material on science, philosophy, history. Several dozen short essays on the history of research on electricity, polar expeditions, spontaneous generation, and the like serve to situate the reader in the early nineteenth century, while two hundred capsule biographies introduce him or her to the most important figures surrounding the novel. By our best estimate the complete edition, were it reproduced in a standard paper volume (of course without the multimedia materials), would run to nearly 20,000 pages.

Frankenstein is exceptionally rich in pointers to things beyond what was once called "the text itself." Several contextual materials suggested themselves immediately. Shelley's use of Paradise Lost, from the title page through the Creature's closing speech, made it an obvious candidate; we include it entire. Whereas even extensively annotated editions might provide ten apropos lines and refer the reader to another volume, our edition takes the reader to those lines in the context of the complete text. We include some of the works of Shelley's parents, such as Wollstonecraft's Vindication of the Rights of Woman and Godwin's Political Justice. The allusions to Wordsworth, Coleridge, Byron, and Percy Shelley led to our including the complete text of all relevant materials (such as "Tintern Abbey," The Rime of the Ancient Mariner, Childe Harold's Pilgrimage, and Prometheus Unbound).

All these texts, though, in spite of being too long to include in a paper edition of the novel, are readily available -- with the possible exception of Godwin, they should appear in every high school and public library -- while our conception goes beyond the easily available. Mary Shelley's scientific knowledge, for instance, was drawn largely from the works of Erasmus Darwin and Humphry Davy, now available only in the rare book rooms of research libraries. Volney's Ruins of Empire, from which the Creature draws his knowledge of much of the world, has not appeared in English in nearly a century. Much of our excitement about the project grew out of providing not merely an edition of a single text but a library of materials unavailable outside major universities, a library containing the equivalent of dozens of rare volumes available for not much more than the cost of a trade paperback.

In addition to original annotations and contemporary contextual materials, we have included about one hundred fifty scholarly essays, both on the text itself and on such related phenomena as the contexts and pop-culture reinterpretations. Here the only pratical distinction from the other parts is the need to secure reprint permissions: they are otherwise seamlessly integrated into the textual web. Annotations on each frame refer readers to relevant paragraphs in every article that comments on the passage, while the critical essays refer both to the text and to one another. Readers are offered a plurality of critical voices in a more complete variorum than has ever been attempted.

There are other contextual materials of a different sort that we could not afford to exclude from the edition. As early as 1823 Shelley's novel had been adapted for the stage, inaugurating the long tradition of pop-culture representation. Early in the age of film the story moved from the stage to the screen, most famously in Universal's 1931 Frankenstein with Boris Karloff, but also in many dozens of reinterpretations, spinoffs, and parodies. By one count, there have been over four hundred film interpretations of Frankenstein between Thomas Edison's 1910 silent short through Kenneth Branagh's recent large-budget adaptation.3 The Pennsylvania Electronic Edition includes histories and commentaries on these dramatic and cinematic adaptations, with material devoted to the creation of a complete popular-culture mythology surrounding the novel.

And context is more than text. HTML can incorporate graphics, which encouraged us to search for early nineteenth-century pictorial materials: maps on which one can trace Frankenstein's journeys across Europe and to the Pole, contemporary illustrations of eighteenth-century scientific equipment, illuminated alchemical treatises, portraits of the people mentioned or alluded to in the novel or the commentary, and facsimiles of the title pages of nearly every work from which we draw. The reference to Petersburg in the first sentence of the novel leads to contemporary maps of Petersburg, engravings of views of the city, and so on. And Frankenstein's long and rich heritage in popular culture suggests another sort of non-textual material, the film clip, which the multimedia capabilities of the Web and Netscape make practical. Wherever reproduction rights can be secured, we hope to include video clips from Frankenstein movies from Edison through Branagh, all integrated into the vast contextual library.

An uncatalogued library, of course, is useless. With this explosion of text into hypertext, we felt some need to rein it back in, to allow the reader to find specific information without stumbling upon it by accident. Though the obvious way to reach the context is from the text -- a link from the text might lead to a note which leads to Wollstonecraft which leads to Rousseau which leads to a critical discussion of Rousseau's theories of education which leads back to the Creature's education in volume two -- we see the importance of a more conventional hierarchical organization. Hypertext allows the two to be overlaid, so in addition to the circuitous path just described, there is also a page called "Contexts: Education," which refers to Rousseau and Wollstonecraft and the relevant critical articles; an index nominum includes all the people mentioned in the text or the commentary; an index of themes points toward major discussions of the novel; a list of critical approaches allows the reader to look at the biographical discussions of the novel, or the feminist critiques, or the materialist approaches, the psychological readings, and so on. And a full concordance of not only every word but every phrase of up to eight words in the novel will take the place of a search engine, which cannot yet reliably run on various hardware and software platforms. It should be possible to reach any piece of text in the edition by selecting only a few logically organized links.

As the edition grew, the greatest practical difficulty was in this management of something so vast and shapeless -- keeping the elaborate network of links straight in our own minds. Even though we had ourselves created every gloss and link, it soon grew beyond our power to keep tabs on its intricacies. Here some rudimentary computer skills came in handy. We created a number of simple programs (Unix shell scripts and PERL programs): one program kept track of the geographical locations in the book and the commentary; another looked for personal names; yet another a list of characters. These programs ran automatically every night, and their reports would list every frame (whether text, context, or commentary) in which the words appeared. We therefore had at any time an up-to-date index to work from and to monitor our own progress. Another program, also run daily, would hunt for illegitimate links (such as those caused by typographical errors or forgotten leads), which we realized would be nearly impossible to track down through simple proofreading.

We have done our best to resist the technological euphoria that often accompanies reports similar projects. While my comments show my excitement over the edition, we know the computer is no panacea. The limitations and problems of the technology, and our own failure to overcome some of the problems we set for ourselves, deserve at least some brief consideration.

The greatest difficulty for the reader of this edition is too obvious to be noticed by the cyber-utopianists: the computer screen. Our enthusiasm should never lose touch with the practical concerns forced on readers by media. Even the best screens are too grainy and flickery for comfortable prolonged use. And a book that cannot be scribbled in or read on a train ride cannot hope to be a commercial success. This led us to the seemingly retrograde move of including a thoroughly traditional paperback edition of the 1818 text bundled with the computer disk. The paperback is not annotated, although marginalia refer readers to the frames in the electronic edition. The idea is that the student can carry about a reading copy and then do more sophisticated research on relevant passages.

We are still concerned that some information may be difficult to find, and that some important connections are left unmade. There are nearly fifty thousand hypertext links, each placed by hand, covering every relation we were able to anticipate. But as Frankenstein discovered when his Creature left his workshop of filthy creation, some projects achieve a degree of complexity that puts them out of their creator's control. The most important problem, then, is inherent in the medium itself. It may be no problem at all, though it often feels like one: there is no end to what can be included in a web. Despite the often impressive theorizing on hypertext's creation of a new set of reading practices, we do not yet have any empirical data on how readers will in fact approach such a massive text. Ten thousand files can be linked to one another nearly a hundred million ways, and the editor's job is to determine which of these almost limitless possibilities is likely to prove useful to readers -- to choose fifty or sixty thousand good ones and assume the other 99.95 million are less important. We were sometimes forced to set arbitrary limits and felt like tyrants, and at other times let connections spin almost out of control and felt like anarchists.

In spite of all the problems, it is this sense of charting previously unexplored waters that makes us hopeful that the Pennsylvania Electronic Edition will suggest some of what an electronic edition might do. The final touches will be applied in the next few months, but we welcome feedback, for we hope to make this an ongoing project. And though of course we hope the edition is widely used among its most obvious audience, students of Frankenstein, the practical experience of assembling an edition of this magnitude -- with everything we've done right, and everything we've done wrong -- may prove useful for others taking advantage of the new medium.


1. Stuart Curran, "Frankenstein Revealed: The Pennsylvania Electronic Edition," Penn Printout, 11:3 (1994).

2. Since we began, Netscape has introduced an extension to HTML called "frames" which have little to do with our frames, but it was too late to settle on another name.

3. Stephen Jones, The Illustrated Frankenstein Movie Guide (London: Titan Books, 1994).