This article is a slightly expanded version of a presentation in John Unsworth's session, "Creating Digital Resources in the Humanities," for DRH97, the conference on digital resources in the humanities at St. Anne's College, Oxford University, Oxford, England, 14-17 September 1997. The article concerns the William Blake Archive, a Web-based project supported by the Getty Grant Program and developed under the auspices of the Institute for Advanced Technology in the Humanities at the University of Virginia. The Archive's co-sponsors include the Library of Congress, Sun Microsystems, and the Inso Corp. In addition to the Library of Congress, collections that have so far contributed their works by Blake — more than a thousand images — include the Huntington Library and Art Gallery in California, the Pierpont Morgan and New York Public Libraries in New York, the Yale Center for British Art, the Glasgow University Library, the Essick Collection, and the Houghton Library at Harvard.

My collaborators and I have just passed the two-year point of what began as a three-year project — not a bad time to take an honest look back at our original vision of collaboration and measure it against the gritty street-level realities we've encountered. At this point we have chalked up a string of accomplishments that we think we're justified in feeling fine about — we may be a tiny bit infected still with techno-optimism. But getting here has chastened us, put our communication skills if not our friendship to the test, lowered our sights, and given new meaning to and new appreciation for "cooperation" and "compromise." The narrative of those alterations and the useful lessons we may have learned so far are the subject of this article.

John Unsworth's directive for the session at which this paper was first presented went like this: "The topic is collaboration in the production of digital resources. You're doing it in a way, and on a scale, that very few people in the humanities have done: Whatever seems most notable about that experience, that's what you should talk about."

That could be flattering, but when I heard it what stood out in my mind was a two-line e-mail message I got from my most technology-averse collaborator, Bob, a couple of weeks ago: "Morris — I can't read e-mail any more. Please send hard copies. Bob." Bob soon lifted his e-mail ban, but that should undercut unintentional utopian overtones while I do a quick sketch of a many-handed undertaking now going into its third year.

So: You should know that behind the scenes at the William Blake Archive are three of us, Bob, Joe, and Morris, who collaborate from a distance — California, North Carolina, and New York. We collaborate with each other but we also collaborate with another group in Virginia, the staff of the Institute of Advanced Technology in the Humanities, and of course at that end they collaborate with each other. Most of our collaboration is electronic. The only paper letters we ever send, I think, are to museums and foundations. But when partial, indirect, virtual means fail us, as they inevitably do now and then, we hold the face-to-face option in reserve. We've found the face-to-face option so essential, in fact, that we've institutionalized it.

"We felt and I see now that we were on the frontiers of something, but we weren't sure what."

The Blake Archive project began with collaborations. The first spark was struck almost accidentally in a 1992 conference on "Textual Technologies" deep in the heart of Texas A & M University. In an idle minute between sessions Jerry McGann [Jerome J. McGann, professor of English at the University of Virginia] mentioned the new institute at his university, which had been founded on the pretty undeniable premise that humanists and programmers need special mediation to work successfully together. Jerry said he had wondered about an undertaking for Blake's illuminated books more or less parallel to his own Rossetti Archive,, but with images more important in the mix. My first thought was No, I wouldn't touch it. My second thought was collaborative — maybe I could just barely imagine it with some trusty partners.

To face the unknown with, I looked to Robert N. Essick, and Joseph Viscomi, respectively professors of English at the University of California, Riverside, and the University of North Carolina, Chapel Hill, who were already my collaborators at the time on a printed volume of the new Blake Trust series [Blake's Illuminated Books, 6 vols., Tate Gallery/William Blake Trust/Princeton Univ. Press, 1991-95]. After talking over Jerry's idea for a bit, we agreed we wouldn't even understand what we were proposing until someone filled us in on the technological possibilities, which led to the founding act of collaboration in the Blake Archive project.

In the summer of 1993, we met in New York City and drove together to the University of Virginia for a meeting with Jerry, who showed us some samples from his Rossetti work, and John Unsworth, the newly appointed director of the Institute, who hadn't even moved to Charlottesville yet. He and his then-staff showed us some things they might do to make a Blake project. They talked with assurance about the World Wide Web and Java. We had barely heard of the Web and had never actually seen it; we had never heard of Java. We felt and I see now that we were on the frontiers of something, but we weren't sure what, and we got the first glimmerings of an understanding that has grown: collaboration is about lots of things besides the division of labor, like reassurance in the face of technical panic.

From there we proceeded — with trepidation — to the next collaboration, on an application to The Getty Grant Program in 1994, which we composed and revised by e-mail, following the patterns that had evolved in editing the Blake Trust volume. Our vision was of a searchable archive of texts and images drawn from the world's greatest public and private collections — searchable right down to the small details of Blake's words and his pictures. It would be open and free to all who could access it on the Web. To The Getty we said one reason for giving us money was that useful lessons would be learned. We even said they could see us as "a model for future attempts to use the networks for collaborative editing and the presentation of visual and textual materials . . . [T]he Blake project will produce sophisticated tools and techniques for electronic image archiving and editing that will serve a wide range of art-historical and literary purposes." We'd be explorers on a new scholarly frontier; we'd build a sophisticated product, and we'd be adventurously introspective about the process.

We got the money. Then we spent the first of it to get together again in Charlottesville, that time to discuss what a Blake Archive might be and do and plan the first steps for getting from here to there. We were the first Institute project whose creators weren't part of the University of Virginia. All the computer knowledge was going to be hanging out there; all the Blake knowledge was going to be scattered all over America. So the collaboration was going to have to be intricate and the coordination tight. The recipe had these basic parts:

  • A project manager at the Institute.

    With a humanities background, technical knowledge, experience, appropriately anal personality, and stiff spine, the manager would keep an eye on the literary, art-historical, and digital pieces of the project at once. (For the first two years our project manager was Amy Sexton; Matt Kirschenbaum took her hard-to-fill place earlier this year.) The project manager would have a part-time student assistant who would also have a humanities background and some technical skills.

  • Programmers and others.

    At the Institute our project would have access to specialized technical expertise as needed from the staff there. Of course our needs would have to be coordinated with the needs of several other projects. That would be John's responsibility.

  • Project directors.

    The three of us needed to sort out our internal division of labor. We had our recent experience with the Blake Trust volumes to guide us. Joe would deal with the pictures — the digital scanning and color-correcting — and with the transcriptions of Blake's etched texts, which have to be edited and corrected, while Bob and I would divide the commentary, which would include detailed descriptions of every component of every picture in order to make the pictures as searchable as the texts. In consultation with the Institute, we would arrive at a design of the site and set our standards as a group, cross check each other's work, and divide the other jobs opportunistically.

And there always have been many other jobs — or call them opportunities. So it was that Bob negotiated permissions with the Huntington Library and Art Galleries because he's next door and has connections; I worked on the major British collections because I was in London last year; and Joe negotiated with the other American collections because he's dealt with them before. But in each instance we first decided together by e-mail what works we'd ask for and settled the terms of our formal agreements with the owners. At one time or another we have all done demonstrations and presentations at conferences, sometimes together and sometimes not, depending on who's where and who's interested. Who "tags" (puts information in SGML codes) the "copy" and "plate" headers that structure the information about the "objects" in the Archive? Bob usually does. Who revises the Help document and writes new grant applications? I usually do — and it goes on and on from there. Fortunately, we're a flexible and temperate lot — in this department of our lives anyway — with a fairly vast collective reservoir of experience in the art of collaboration to draw on.

"No one multi-purpose channel can work for everyone and everything."

With the Blake Trust volumes there hadn't been much we ever needed to know that one of us couldn't find out. Some of the printed format had been set in advance for the whole series, and we tailored the rest of an editorial apparatus to suit ourselves by choosing elements from familiar patterns. The Blake Archive was an altogether different kettle of fish: We had to assume that together, all ten of us (more or less) had the sum of knowledge to make what had to be made. But in one camp was little knowledge of Blake, and in the other camp was little knowledge of markup systems, servers, scanners, display technology, browsers, networks, and search engines. Those knowledge gaps aren't so easy to bridge, and they have the potential to create confusion and misunderstanding at every stage: How can you plan a trip by car if you've never driven and don't have a map?

The inescapable answer is talk, show and tell, and talk until a common understanding emerges. That can only happen gradually, by a slow process of dawning collective realization. The communication system that can support such sustained collaboration has to be circumstantial — with enough different channels to match the needs of the moment, including the particular issue, its particular resistance to solution, the particular degree of specialized knowledge required to grasp it (hence its distance from the knowledge of others in the group), right down to the personality profiles of the people involved. In that sort of communications situation it really matters that some people are laid back, others impatient, others suspicious; that some read closely and respond overconscientiously while others skip important information; that some tune in ten times a day while others tune in once every ten days. To get the best messages out of ten people who are not quite fused in one project, you have to set up accommodating and complementary channels such that one takes up the slack left by another. No one multi-purpose channel can work for everyone and everything.

So, in a few words, here's what we did to help integrate our ten-part division of labor, knowledge, and skill into a single collective effort across the gaps of time, space, and know-how:

  • We agreed that it would be useful for Joe to spend the first year as a fellow in residence at the Institute, laying groundwork.

  • We agreed that e-mail could be the main glue in our system. That's actually worked out remarkably well much of the time. From mundane record-keeping to revising drafts to conducting barely civil debates about the way the search engine works, the speed, flexibility, and low cost of e-mail have made it the medium of choice in most circumstances.

By exploiting the scalability of e-mail we gave the project its central intersection, blake-proj, our closed electronic mailing list to which everyone involved in the project, including Jack Myers, our Getty liaison, subscribes and contributes as called upon. Blake-proj was John Unsworth's brilliant idea, and after a short break-in period, it has proven over and over its utility to the point of indispensability. I think of it as the top level of our communications structure: almost all significant lines of thought, whether conducted by mail, e-mail, phone, fax, ftp, or otherwise, begin and end at blake-proj.

More futuristically, John originally suggested some devices that work in "real time" — group discussions in the Institute's MOO (we even spent some real time practicing) and maybe some video conferencing. Those seemed appealing in theory, but they turned out to be nonstarters, perhaps because the chat-room metaphor behind the MOO was unappealing to people who don't enjoy games, and the setup for video conferencing has been too elaborate to fool with. Anyhow, more down-home real-time modes — telephone and face-to-face conversation — have seemed sufficient so far.

Another nonstarter was a very clever idea for an online Web-based tagging system that used HTML fill-in forms to allow us to create SGML code for the picture descriptions that were the key to image searches. Unfortunately, the limitations of HTML made the prototype less convenient than elementary templates and macros created with our everyday word processing programs.

Any account of our collaboration does, after all, have to include the thing we collaborate over, the Archive itself. The first two years were devoted mainly to research and development. By the end of the first year there was a public Blake Archive site that was handsome, but thin compared to what we were aiming for. All the real action was behind the scenes at our private, password-protected, work-in-progress site. The WIP site has three main divisions: one for banking things one of us might need, such as image files; a second for testing things, including our two Java applications, the logic that our search engine applies to find texts or images, the SGML architecture, or the elaborate page format; and a third division for proofing and tweaking works just before publication, when they are moved to the public site.

"No sooner would John post a proud message proclaiming the latest programming victory over some stubborn problem than Bob, Joe, or I would post a counter-note saying that the thing wasn't doing what we wanted it to do at all."

I promised earlier to come back to the place of face-to-face in our scheme. It would be less true to say that we gave it a place than that it earned its place. The actual, like the virtual, is not good for everything; surrogates like e-mail are in many cases preferable: more efficient, more forgiving, less fraught. But we have found that, as a court of last resort, there is no virtual stand-in for meeting in person. Hence the annual institution we have come to call Blake Camp, which at first was a planning session but has become a problem-solving, assessing, and planning session that happens every summer in Charlottesville. The Blake Boyz, as John calls us, fly in for these annual summer get-togethers — even Bob, who hates travel the most. This summer's Blake Camp took up three solid days in June. The centerpiece was a roundtable gathering in the central office at the Institute. All the usual suspects were there, either participating or on call in their offices nearby. The Blake Archive was projected on the screen in front of us from 9 till 5 as we worked our way through a tricky and tedious agenda.

This summer's session came at the climax of a long series of tests that had led to some testy exchanges on blake-proj and marathon telephone conversations that hadn't clarified much. The goal was to publish, no later than this summer, the first fully enriched illuminated book, adorned with all our clever bells and whistles. When that had happened, we said, then and only then could we move from our research-and-development stage to a new and presumably much smoother and easier production phase where the main work would go into publishing all the works that had been piling up in the background while development and testing had taken over the foreground.

All the most important pieces had been in place for months. We knew pretty much what a page would look like, what you could do with the texts and images, and what tools you needed to do it with — all now diagrammed and labeled in the clever help-document graphics devised by our current project manager, Matt Kirschenbaum. We had settled on one short illuminated book, Blake's eight-plate Book of Thel, copy F, from the Library of Congress's extensive Lessing J. Rosenwald collection, as our guinea pig. Thel had been scanned, color corrected, and transcribed. You could enlarge the images, read the text right on the plates or in more readable transcriptions, search for texts, search for images, read descriptions of every image or part thereof, among other things.

But no sooner would John post a proud message to blake-proj proclaiming the latest programming victory over some stubborn problem than Bob, Joe, or I would post a counter-note saying that the thing wasn't doing what we wanted it to do at all. The three of us were approaching terminal frustration with search formulas that couldn't produce the right hits on the right things and unstable bumper-car Java applications that looked cool as hell but cluttered the screen with annotations that didn't match the picture-details being annotated. I'm sure the programmers were totally exasperated with what must have seemed to them an interminable list of demands from us contradicted by double messages. They were understandably pleased with the miles they had brought us; we were frustrated with the miles that hadn't been crossed. The atmosphere was charged.

It took Blake Camp to begin turning what was starting to feel like defeat — the three of us were talking among ourselves about how to lower our sights and cut our losses — into what felt like a chance at victory. But as we talked through the problems one by one with the Archive right in front of us all, with everyone who was relevant to a solution there to contribute, we discovered that many of the smallest problems could be dealt with on the spot, with John pounding the keyboard to change the SGML or the search logic or the format of a button as we exposed the source of one misunderstanding after another. Several larger problems would take longer. But along the way we regained our optimism and our confidence, I think largely because we realized that we had managed to come back to a common understanding of what we were after and what could be achieved.

When I look back I can see that the trick was in knowing when that common understanding had really emerged. In key instances it turned out that what we all thought was a common understanding wasn't intact. The shell of the understanding was full of lethal little leaks. Along the path of collaboration subtle cracks had opened in our e-mail communications, and in them misapprehensions had sprouted. This is hard to describe. I felt it in audiovisual terms as a combination of noise in the system and invisible gaps. And the cure was all too familiar: lengthy face-to-face hashing-it-out that puts us over and outside the more routine information channels that we had usually been inside. The medium of those channels had blurred the message just enough to seduce senders and receivers into an illusion of understanding.

After two of the three days of Blake Camp the air began to clear. The third day felt almost restful. On August 4 we issued this announcement:

The editors of the William Blake Archive — Morris Eaves, Robert Essick, and Joseph Viscomi — are pleased to announce that The Book of Thel, copy F, is now online in searchable form. This may seem a modest achievement, given that this is one of Blake's shorter works, and that its eight plates (with enlargements and accompanying transcriptions) have already been available on the site since November of 1996. In fact, however, this copy of The Book of Thel is a prototype for all future works to be added to the Archive (both illuminated books and other materials), and its appearance reflects the architecture and objectives of the Archive as they have taken shape over many months of development, testing, and refinement.

Unlike its previous version, and unlike the other illuminated books currently available in the Archive, this copy of Thel has been tagged using SGML (Standard Generalized Markup Language). SGML tagging offers the Archive's users the opportunity to perform sophisticated searches, either on the text of the plates, or, more remarkably, on the content of their illustrations. Search results are retrieved and presented using DynaWeb, a product of the Inso Corporation. The text and image searching enabled by DynaWeb and the underlying SGML tagging is a powerful demonstration of the potential of electronic resources in the humanities.

However, there's more. Users with Java-capable browsers can now make use of Inote, Java-based software developed at the Institute for Advanced Technology in the Humanities, to assist them in their study of the Archive's visual materials. Users may employ Inote to examine editorial annotations of a given image independently of a search, or else, following a successful image search, Inote may be used to open the image, zoomed to the specific area containing the object of the search, together with the relevant editorial commentary. This is IATH's first public implementation of Inote, and its release marks a major advance for image-based electronic editing.

Users with Java-capable browsers can also take advantage of a second, equally innovative Java program developed at IATH, the ImageSizer. This is a feature that allows one to view Blake's plates and images at their true size, reproducing the object's actual physical dimensions on the screen, regardless of the resolution of a particular monitor; indeed, users can calibrate this feature to consistently display the Archive's images at whatever proportions they may wish.

Finally, the Archive's selective bibliography of criticism, reference materials, and standard editions, with about 500 entries, is now available. We hope to have the bibliography searchable by the end of the summer. We also hope to have David V. Erdman's Complete Poetry and Prose of William Blake online and searchable by that same time.

In the coming weeks and months we will turn our attention to placing other illuminated books online in searchable form, following on the model of The Book of Thel, copy F. We will begin with the other copies of Thel (copies H and O), as well as copies of Visions of the Daughters of Albion (copies C and J) now publicly available only in HTML — thus lacking any of the capabilities described above. We will then move on to other books: All Religions are One (copy A), There is No Natural Religion (copies C and L), The Marriage of Heaven and Hell (copy D), America (copy E), Europe (copies B and E), The Song of Los (copies A and B), The Book of Urizen (copy G), Songs of Innocence and of Experience (copy Z), The Book of Ahania (copy A), and The Book of Los (copy A).

The Blake Archive is located at:

Please forward this announcement as appropriate.

Matthew Kirschenbaum, Project Manager

The William Blake Archive

Institute for Advanced Technology in the Humanities

University of Virginia, Charlottesville

Since then our communications have returned to the usual channels, dominated as always by blake-proj. By the end of this year, 1997-98, the third and final year of the Getty grant period, we aim to have at least one copy of every illuminated book — including the 100 plates of Jerusalem — and multiple copies of most books online. After that, a new phase begins, with what funding we don't yet know. We shall begin to turn from the canon of illuminated books out toward the full range of Blake's work as a painter, printmaker, and poet. When do we expect to finish? I don't expect that subject to show up on next year's Blake Camp agenda, though it might keep one or another of us awake occasionally.

Morris Eaves is a professor of English at the University of Rochester and co-editor, with Robert N. Essick and Joseph Viscomi, of the William Blake Archive. He has been co-editor (with Morton Paley, University of California, Berkeley) of Blake/An Illustrated Quarterly since 1970. His editorial collaboration with his Blake Archive colleagues Robert Essick and Joseph Viscomi began with a printed volume in the new Blake Trust series of Blake's illuminated books published by Princeton University Press. As a student of the arts in their connections with commerce and technology, Eaves has published widely on Blake's theories of art in their commercial and technological contexts, most recently in The Counter-Arts Conspiracy: Art and Industry in the Age of Blake (Cornell 1992). As a Guggenheim fellow for 1997 he is writing a speculative book on editorial theory that investigates the challenges of editing across artistic forms and media.