1. Introduction

This paper describes an intermediate point in the development of an electronic edition of a unique and complex early modern book. At this stage, we the authors have done extensive research on the book itself and have also begun thinking about appropriate ways of rendering information about the book electronically. This essay describes the book, a gospel harmony or ‘concordance’ made for King Charles I in the mid-1630s by the Ferrars of Little Gidding, England, particularly in light of its construction and intended functions. It goes on to explore the appropriateness of electronic forms for describing the book, ending by postulating possible tools for delivery.

2. The Little Gidding concordances

When we think of disassembling books, it seems we tend to think of a relatively recent activity, such as that of Otto Ege, whose dismantling and distributing of manuscripts is discussed in Peter Stoicheff’s paper in this collection. Going back to the later 18^th century, the popular practice known as grangerizing involved individuals collecting engravings from various sources and adding them to books as illustrations. But long before this, there is at least one notable instance of people cutting up not only illustrations, but also textual pages, and reassembling the pieces to make a new book. In the 1630s, in the countryside northwest of Cambridge, the extended Ferrar family, in an intentional retreat from the world, found a “new way of printing” admirably suited to the family’s needs. This wealthy merchant family, faced with severe financial difficulties (brought on in large part by the Crown's seizure of the Virginia Company in which they were heavily invested), devoted themselves to a monastically-shaped life.

Nicholas Ferrar, one of the family’s leaders, devised an activity that would keep his nieces occupied both physically and mentally: the construction of ‘harmonized’ gospel books, made by hand from printed resources. The family called these books ‘concordances,’ though they did not list words as concordances commonly do, but rather brought the four gospel accounts into concord. The books are thus generically ‘harmonies.’ These books were the site of physical and spiritual labour. In addition to keeping their makers from the sin of idleness, the books kept them thinking about the words of the Gospel. Also, their organization of the four gospels into a continuous 150 chapter account of Christ’s life provided a single narrative that could be memorized and recited at five daily hours of prayer, and thus repeated monthly, in its entirety. Finally, in their advanced form (the books grew in complexity as the community made more) the books provided a way to study the four gospels at once. The concordances were suitable, then, as a communal work project, as matter for public worship, and as a resource for personal biblical study. To these uses another was added from outside: the use of these books as gifts, most notably to King Charles I, at his request.

The Little Gidding concordances and their history have been described at length elsewhere, but we will give a short account of their physical and textual construction here in order to establish the challenges and opportunities involved in producing an electronic edition of one of these books.¹ The most striking fact of the concordances is that they are made almost entirely out of pieces of other publications, either books or collections of engravings. The present-day book they most resemble, if only in their method of construction, is the scrapbook. But they go far beyond anything in the scrapbook mode in their attempt to create a serious or ‘true’ book by such a method. They are meant to be books, produced locally and individually, but with the effect of print. Keeping this effect in mind, it is remarkable that these pages usually contain not large pieces taken from other books, but rather sometimes hundreds of small fragments of text and image, combined into something new. They include hundreds of pictures, including the occasional repetition of particular prints and the repetition of subjects in multiple sizes. Often, these prints have been combined in collage, along with text, creating a cornucopia of image and text.

3. The Little Gidding markup

Most chapters of the king’s concordance—the most complex of the fourteen extant Little Gidding concordances and the subject of our project—include three arrangements of the gospel text:

Figure 1: Chapter XXXV (c) British Library Board. All Rights Reserved (Shelfmark C23e4)

The first, called the “comparison,” gives the separate gospel accounts in parallel columns; the second, called the “composition,” gives a single account made of multiple gospels. The third arrangement, called the “collection,” is the most complicated and also the most central to the book’s purpose, in that it combines the other two arrangements, giving in a single column both a readable composition from the four gospels and everything left over as a basis for comparison. It does this through two textual categories marked by typeface. The first, in black letter, is named “context,” and the second, in roman letter, is named “supplement.” The context consists of a composition of the various gospels into a single story, while the supplement consists of all the left-over matter, primarily repeated material. This third arrangement, the “collection,” is the sole arrangement of most of the other Little Gidding gospel concordances, and is the primary focus of our electronic work. It, in particular, stands out as an elegantly encoded text, unobtrusively demonstrating its complicated structure through the aforementioned alternation of typefaces and by marginal alphanumeric markers indicating gospel source and chapter (e.g., A⁹ means Matthew chapter 9). Notably, this Little Gidding markup has been layered on top of the canonical markup of the biblical text, that is, the marking of the text into books, chapters, and verses. This is all to point out that the Little Gidding community began with one of the most structurally-encoded texts available and then restructured it, using the biblical text’s canonical markup as an ordering system, but also using typeface as markup. This latter use is striking, in that black letter and roman faces were culturally significant as codes of authority, but had also previously been used as structural marks: for example, in the use of roman faces in early black letter ‘King James’ Bibles to indicate additional material, such as chapter summaries. In their books, the Little Gidding community gave neither typeface precedence of authority (in some other concordances that the community made, the context is in roman, and the supplement in black letter), but rather used it innovatively, to encode a structure within the biblical text itself.

4. Beginning our markup

The Little Gidding concordances are sufficiently unique and also strangely reflective of their historical moment (and thus remarkably insightful) that there should be a widely-available edition of them, or at least of a representative concordance. The king’s gospel concordance, being the most highly developed instance of the form, would be the obvious choice because it includes all the features found in the other concordances. Since these books are unique objects, the only way to see them is in person (and only by special permission), and the only representations available are on microfilm, which does not capture texture, and so does not represent the cut-and-pasted quality of the books. So, we have begun to work on an electronic edition of the king’s book. We think that the electronic medium is the obvious modern choice for making a representation of the book widely available, but more importantly, that the Ferrars’ method shares with today’s electronic compositional culture some important and mutually-enlightening similarities. Most notable is the transportability of the text or image fragment, and its redeployment in new compositions. While subsequent centuries witnessed the addition of images to printed texts and the assembly of various cut-outs in scrapbook form, the Little Gidding concordances stand out as serious books made entirely of fragments from other books. They are the apotheosis of the common place book; the proof that the world’s most important text can be compiled out of commonly-available materials.

Our goal is to produce an electronic edition that will combine a full photographic facsimile with a deeply encoded transcription of the text and description of the images, and that will, as much as possible, use the first as the interface for the latter, so that the photograph becomes the way into the database functions of the edition. For instance, the book itself provides marginal links between fragments of text that were originally together but now have been placed on different pages:

Figure 2: Hypertext (c) British Library Board. All Rights Reserved (Shelfmark C23e4)

This figure shows the end of Chapter XIII, in column 40, which ends with Luke 3:18. The hand-written ‘link’ indicates that the Luke narrative continues with the words “But Herod . . .” in Chapter XXII, column 57. It would be true to the book’s nature to make this part of the facsimile a live link, not only to the next facsimile page, but also to the uninterrupted text, as continued in Chapter XXII. In some places, clicking on textual fragments should also produce images of the original pages from which they came. Likewise, the most natural way into the cornucopia of images in the book is through the photograph of the page: clicking on any given image could helpfully produce a number of database-driven results, such as a thumbnail list of other images from the same series; or images by the same artist, the same engraver, or the same printer; or, it could also produce a list of images of the same biblical scene. Ideally, clicking on a fragment of an image would also make available the entire original, so that one could reconstruct the process of combining images that took place at Little Gidding.

So far, we have observed—even at the note-taking stage—that the encoded electronic environment is well suited to describing the complex content of the king’s book. When Paul began studying the book, both on microfilm and in person, he used a word processing document with a table to record chapter titles and numbers, scriptural contents, images, manuscript additions, and so on. At the point, however, when we started using XML to describe the book, the description became much more precise. Our first move was to create a document that, as simply as possible, described the book’s textual compilation: harmony chapters and their constitutive gospel chapters and verses, in order, noting typeface:

Figure 3: Textual compilation

This we did using the canonical reference system of the Bible, so no biblical text itself was yet involved. Note that this XML does not use the Text Encoding Initiative (TEI) markup structure, but rather, it uses a document structure we designed to match closely the text being encoded. Our goal is to produce a fully TEI-compliant document, but this will be a transformation from our more "native" encoding. We next did another pass on the book, adding tags and descriptions for the images. The fact that no gospel text was yet present made the document relatively small and easy to negotiate. When Paul returned to the British Library, he source-checked the details already recorded and added many fine details that are not legible on the microfilm. For instance, the chapter headings frequently use red and gold ink for select letters and words. Working with our XML transcription in the archive allowed us to take notes within a publication environment, using the TEI highlight element with a render attribute to indicate which text is coloured and which colour:

Figure 4: Our markup

For the Little Gidding books, we created a DTD that would most accurately and efficiently describe the materials, without reference, at first, to other markup structures. We designed this DTD as we worked with the book, using the oXygen XML editor’s “learn structure” feature, which produces a DTD to fit the document. Throughout this initial encoding, we changed our XML document structure (via the DTD) as the source material dictated. At times while working with the book in the archive, Paul made several DTD changes a day. The XML editor served not so much to arrange previously gathered information within a previously established structure as an arena in which to record information as that information was gathered and simultaneously to experiment with encoding structures. In this way, sophisticated XML editors such as oXygen create an easy to use XML compositional environment. As a tool for gathering information on the Little Gidding book, our XML editor was vastly superior to the word-processor that we had used earlier, especially in that it could contain the entire original textual structure (and later the whole gospel text itself) without overwhelming the writing space. Our Word file used a table that captured the bare-bones structure of the king’s book; whenever we wanted to note anything about the book, we needed to first describe what part of the book we were noting. Using an XML editor, we were able instead to notate and develop a facsimile of the king’s book, a facsimile that began as a skeleton, giving the structure of the document without the actual gospel text, which we later fleshed out with the text itself. Starting with a structural skeleton allowed us to work backwards through the Little Gidding process, establishing first what they had established first, which was a plan of how the four gospels would fit together. As mentioned above, this skeleton also facilitated our documenting of the book’s images. Before we added text to the structure, we added notes on the images, an order that allowed us to specify the images’ textual locations without us having to navigate the text at the same time.

Working with the skeletal structure also gave us a sense of what might be possible and desirable in a finished edition. The expansion and contraction of detail strikes us as an important ordering principle for interface. The document’s markup will allow an interface mechanism to draw out and present key elements on their own. An obvious example is a table of contents made up of the chapter heading element; but other searches, in particular those drawing both upon the text and the images, have the potential to shed new light on the document. The principle here is not ultimately to limit the user’s reading, but to enable further exploration by presenting the many structuring lines that run through the book.

The canonical structure of the gospels and their free availability on the Web led to one other aspect of this project so far. Once we had encoded the basic structure of the king’s book, Stuart developed our own cut-and-paste technique as a mid-way step in the development of our transcription of the gospel text. He wrote a program (in Python) to pull a modern, public domain version of the King James text into the Little Gidding structure. Our XML reference thus becomes populated with the full text:

Figure 5: From reference to text

So far this automated, high-speed re-enactment of the Little Gidding concordance-making has allowed us to work with an approximation of the final text, but it has also allowed us to work without the text present at all, depending on what work we were doing. It also provides a base text that we are presently conforming to the editions used in the king’s book.

5. Moving toward common structures

Of course, the nature of markup is that it must bridge the particular and the general, that is, render the particular generally. While Peter Robinson notes that the difficulty of learning the TEI markup is an impediment to the growth of electronic scholarly editions (Robinson 2005: para 14), Jerome McGann’s more radical observation comes to the point: “All text is marked text.” Further, “digital markup schemes do not easily ? perhaps do not even naturally - map to the markup that pervades paper-based texts” (McGann 2004: 199). While McGann is looking toward a semiotically dynamic digital markup that is outside of the scope of this project, his comment foregrounds here the fact that the biblical text is not only full of semiotic markings, but has been thoroughly marked with a structure not unlike that of the TEI, but preceding it. Allan Renear’s comment that “The principal goal of the TEI” is to develop “an interchange language that would allow scholars to exchange information” also nicely captures the nature of biblical markup: a structure of textual divisions, known as chapters and verses, that allows scholars to exchange information, particularly the sort of information captured in concordancing and cross-referencing (Renear 2004: 235). So, the standard biblical markup and the TEI markup are strikingly similar: they both mark the text as an Ordered Hierarchy of Content Objects (OHCO) in which objects nest within objects. The XML Gothic Bible of the Wulfila Project done at Antwerp University enacts this similarity by using TEI division tags for both biblical books and chapters, and ‘ambiguous block’ tags for verses: “The text follows the biblical canonical reference system and is divided into books (div type="book"), chapters (div type="chapter") and verses (ab type="verse"). For the sake of simplicity, the minor fragments have been segmented ? somewhat artificially ? using the same three-level structure. The type-attribute specifies the nature of each segment (e.g. div type="leaf" for the Skeireins)” (online).

Within this structure, the biblical verses in their normal order can be identified only with a single “ambiguous block” tag, since those tags nest within the divisions that are biblical chapter and book. Complicating matters for our project, though, the Little Gidding books use the biblical markup, but they also rearrange the biblical text in such a manner as to place together small sequences of verses and even single verses from multiple gospels. This mixing of materials outside of their canonical nesting structure makes it important for every verse to be marked with its full canonical identification. The Ferrars did exactly this in their concordances using a marginal alphanumeric system, with letters A through D standing for the four gospels, together with numbers to indicate the chapters. Each verse retains the original print markup designating its number. As we sought to imitate this extra layer of original markup—that is, the rearrangement of the canonical system into a new set of 150 chapters—our point of difficulty with the TEI was one of awkwardness. For each verse to carry full markup is onerous, at least at the working stages of the project. The TEI markup, for instance, for the beginning of the extant Gothic Bible (which is missing its first few chapters) follows.

<div type="book" id="B1" n="1">

<head>Matthaeus</head>

— <div type="chapter" n="5">

— <ab type="verse" n="15">

After which follows Matthew 5:15. Given the cut-and-pasted nature of the Little Gidding text, it seems that each verse would need this many tags to be TEI conformant. Our own markup uses an “hchapter” tag for each chapter of the Little Gidding concordance, which in turn is populated with “excerpt” tags (see figure 4: Our markup). (Presumably these two containers can be transformed into TEI divisions with relative ease.) Each “excerpt” is populated by “verse” tags, each of which contains that verse’s canonical reference in an id attribute. Notably, the “excerpt” tag also has an attribute giving the full canonical reference of that excerpt’s range of verses, and both tags indicate, as an attribute, whether that text is marked in the original as “context” or “supplement.”

Given our claim that the Bible’s canonical reference system treats the text as nested, or an OHCO, it is important to note that this system can handle ambiguity. For example, the verses divisions of the Bible frequently occur within sentences, presenting the text in two overlapping markup structures: one for grammar and one for referencing. The Little Gidding arrangement also makes the term “chapter” ambiguous, in that there are the canonical chapters of each gospel and also the 150 chapters or “heads” of the Little Gidding gospel book. Thus, the Bible comes to us as an ambiguously marked text, and the Little Gidding versions of it add to the ambiguity. Given that the TEI is in part a tool of disambiguation, the would-be encoder is left in a peculiar position. On the one hand, the Bible seems to defy the TEI, but on the other, trying to use the TEI throws light on the strangely marked nature of the biblical text. We would like, in the end, not so much to disambiguate this document, as to present it so that its ambiguities can be seen as part of the creative tension that enables the document as a multi-dimensional reading machine.

Our particular XML description of this particular book is satisfying to us in how it begins to get the book ‘right’, but it cannot yet share in the great common project of the TEI. Would we have been better off using the TEI from the start? We do not think so, for what we have now is not the simple problem of reconciling our eccentric markup structure with a standardized one, but rather, the more basic problem of trying to describe a highly particular book with a standardized structure. We are not very far along the transformation to TEI, but we feel some confidence that our own DTD, which names things as the Ferrars named them, will help us stay true to the original document, for our larger aim is to name things without either abusing the original document or abusing the tag, the Scylla and Charybdis of textual markup. We have just begun to explore the use of TEI extensions to accommodate the textual peculiarities of the Little Gidding books. We think that we can address the engravings used in the books with the Visual Resources Association Core, which will allow us to name the multiple roles and persons involved in the production of each item (i.e., the artist, the engraver, and the printer). The interesting way that our project bumps against the limitations of existing encoding standards, though, is that we are trying to encode cut-and-paste, something no one has previously needed to do. For instance, the VRA Core offers all one could want for naming an individual work or a collection of works, but seems not to lend itself readily to describing the collages in the Little Gidding book, themselves made up of many parts of identifiable prints. Likewise, the TEI has not been conceived to describe a collage of texts. Thus, the most challenging and inspiring quality of this project is to imitate the Little Gidding community in using common structures to construct a particular meta-structure.

One question about our markup and its transition into TEI concerns XML Transformation Language, which enables large-scale and highly detailed transforming of documents, so that, for instance, one can make an XML document into an HTML document and display it on the Web. Our project involves two large-scale transformations: the transformation of a skeletal structure into a full text and the transformation of our Little Gidding document structure into TEI. An important question that Stuart keeps posing is whether we should see these transformations not as one-time transitions, but rather as a relatively permanent characteristic of the project: we may in fact choose to keep our text and its structure as separate documents, conserving each as working documents and then producing the product-text out of them; likewise, we may choose to preserve our own document structure and continue to transform it to TEI, so that the project’s particularity is not lost, both by the initial change to TEI and, looking down the line, in a compounded way, by future changes to TEI.

6. Mapping the textual world of the concordances

The king’s concordance is one of fifteen partial or complete extant Little Gidding concordances. Each one bears witness to the workshop and community in which it was made, and more broadly to the textual and imagistic world of early modern devotion. An edition of the king’s concordance should allow its readers to explore the Little Gidding workshop and the way it captured its historical moment. This was a group of people deeply engaged in the scriptural text, consciously developing habits of relating text to image to life. Particularly interesting are the occasions when the community used the same image in different textual contexts and image combinations. These multiple uses of images link disparate concordances. For instance, the community employed a print showing a story from the book of Joshua—the return of the spies with a huge bunch of grapes—in both their Whole Law of God, a harmonization of the Pentateuch in the St John’s College, Oxford collection, and the king’s gospel concordance.

As we imagine ways that our edition might represent the meta-book quality of the concordances, we are drawn to the idea of the map, and particularly the zooming and scrolling actions popularized by Google Earth. We usually assume that the book’s page should occupy most of the screen, but what if the reader could zoom out so that the page, or more accurately, the opening (the two pages visible whenever one opens a book), appeared as a smaller rectangle within the screen, revealing a map of relationships around it. The opening would remain in the center, but now it would be seen within a horizontal line of thumbnail images of the preceding and proceeding openings. Above and below this line would be thumbnail images of two kinds of books: the first would be all of the material sources used to make that page, including the pages of the various source Bible editions and prints in their whole state. The second would be openings of other Little Gidding concordances that use the same prints. In this way, our edition would set the stage for a larger project, the digitization of the entire group of extant Little Gidding concordances. This interface would map out the constellations of the Little Gidding textual world, a world marked by a rich sense of unending textual and imagistic combinations.

The Little Gidding Concordance project has the potential to make a contribution to our knowledge of early modern textuality and our thinking about textual structures in digital media. Strikingly, though, its most important aspect may be how it allows us to discover the relationship between these two areas. Our thinking about text today takes place in a very long textual tradition, and the change in media only makes our situatedness in this tradition all the more obvious. Some of the textual inventions most apparent in the Little Gidding books (chapter and verse divisions and the operations of cross-referencing) have long been part of the biblical tradition, but were developed more particularly between the 12^th and 16^th centuries. The idea of the Bible as a “book of starres” and the desire to see all of the text’s “constellations,” in the words of a friend of the Ferrars, George Herbert, profoundly informs what we do with text today and the tools we use to do it. The Ferrars of Little Gidding thought that text was incalculably important, and it is for this reason, not in spite of it, that they were textual and technological experimenters. In Little Gidding, then, we have a mind-stretching example not only of what strange things can be done with existing technologies, but what might motivate such strangeness.