Nuvola apps important.png Attention, Ce wiki a été transféré sur le domaine Démo.Istex.

Voir :


H2PTM (1989) Delany

Aller à : navigation, rechercher
Elements, Links and Structures in Hypertext
Elements, Links and Structures in Hypertext
Paul Delany
Department of English
Simon Fraser University
Burnaby, B.C. V5A 1S6- Canada (Grand Vancouver)
actes du colloque H2PTM 1989 Paris
publié dans H²PTM89 : Communication interactive, Paris, France, 1990
It is not in acted, as it is in written history: actual events are nowise so simply related to each other as parent and offspring are; every single event is the offspring not of one, but of all other events, prior or contemporaneous, and will in its turn combine with all others to give birth to new; it is an ever-living, ever-working Chaos of Being, wherein shape after shape bodies itself forth from innumerable elements. And this Chaos, boundelless as the habitation and duration of man, is what the historian will decipt, and scientifically gauge, we may say, by threading it with single lines of a few ells in length.
Carlyle,"On History")


The written text is the stable record of thought, and to achieve this stability the text had to be based on a physical medium: clay, papyrus or paper; tablet, scroll or book. But the text is more than just the shadow or trace of a thought already shaped; in a literate culture, the textual structures that have evolved over the centuries determine thought almost as powerfully as the primal structure that shapes all expression, language. So long as the text was married to a physical media, readers and writers took for granted three crucial attributes: that the text was linear, bounded, and fixed. We can define Hypertext as the use of the computer to transcend the linear, bounded and fixed qualities of the traditional written text. Instead of facing a stable object enclosing an entire text, and held between two hands, the hypertext reader sees only the image of a single block of text on the computer screen. Behind that image lies a variable textual structure that can be represented on the screen in different ways, according to the reader’s choice of links to follow.

One of the great advantages of hypertext is that it makes visible and explicit mental processes that have always been part of the total experience of reading. For the text as the reader imagined it-as opposed to the physical text objectified in the book-never had to be linear, bound-ed or fixed. A reader could jump to the last page to see how the story ended; could think of relev-ant passages in other works; could re-order texts by cutting and pasting. The stubborn materiality of the text constrained such operations: they required some physical task such as flipping pages, pulling another book from a shelf, or dismembering the original text beyond repair. Nonetheless, over the centuries, readers developed a repertoire of aids to textual management, these aids operated both within a single volume, and in the relations between volumes. Within the individual volume of the traditional book we may find such internal hypcrtextual functions as tables of contents, page-numbers, chapters, verses, rubrications, footnotes, and indexes. Some of these may be assigned by the original author, others by specialists in textual organization such as indexers or printers, or by later generations of scholars. External hypertextual functions have traditionally been post-authorial, supplied by librarians and bibliographers. Taken together, these functions constituted a proto-hypertext, in which we can find important models for hypertext design today-though the special powers of the computer allow us to look beyond the textual aids that evolved during the long history of writing and printing.


Hypertext gives us an almost unlimited power to manipulate texts; but it also upsets our old reading habits, and presents us with difficult conceptual problems. These problems can be summed up by the questions: "What is a unit of text?"; "What are the relevant links between units?"; and, "How do units and links combine to form structures?’’ My essay will suggest some very preliminary answers to these three questions.

Among traditional textual units the best-recognized and most functional ones are the letter, the word, the sentence, and the book. To think of them as commensurable units on a linear scale of magnitude is natural, but misleading. A letter is an irreducible unit of the written code; a word is a conceptual unit, a sentence a syntactical one, a book a unit whose identity is largely determined by its traditional status as a physical object. Nonetheless, they are units that can be handled by many kinds of textual aids developed over the whole period of literacy. But between the sentence and the book (in terms of magnitude) we find such units as footnotes, paragraphs, chapters and essays; these are less amenable to definition, because they are largely informal means of organizing thought For the same reason, however, they are likely to be important elements for building hypertext structures: they are the kind of mental "chunks” that we use to break a complex issue into components and make it intelligible. In addition, we are beginning to imagine new textual units, not yet codified or even named, that will be specific to the hypertext environment. They will come into being and pass away in the dynamic virtual text of the computer, products of such broad cognitive principles as identity, association and structure.

We have, therefore, many different actual or potential textual elements to deal with. In print technology they are usually stable—thanks to the linear, bounded & fixed nature of the book— and explicit, in the sense that an experienced reader can easily recognise elements by the conventional clues that we all know. But in hypertext the elements become stable and explicit only in the degree that the hypertext designer has chosen to provide visual indications of their role. At the same time, it is possible to mark a path through the jungle of textual units. The path I will follow here is based on a founding opposition between textual and hypertextual systems. Instead of the multitude of traditional textual elements—the letter, the paragraph, the book, and the rest—we can reduce hypertext to a meta-textual form with only two elements, blocks and links.

Block may even be a misleading term here, because it implies a stable object. We may do better to speak of a frame; that is, an enclosure that marks off a textual unit, and makes possible relations with other units, but has itself no determinate content. In practical terms, such a frame will often be implemented as a window on the computer screen. Any given frame may contain a variety of traditional textual units, from a single letter upwards; in hypermedia, it may also have a graphic or musical content; but what counts at the hypertextual level is only the boundary of the frame and its relations with other frames.

Let me develop this argument a bit further, using the terminology of my colleague Tony Wilden. Wilden observes that "Tout message qui 'cadre' un autre message est, dans un contexte synchronique, d'un type logique supérieur au message qu'il cadre."[1] The concept of the "logical type," deriving from the work of Russell and Whitehead, means simply that discrete statements are on a different and "lower" conceptual level than statements that refer to whole classes and categories. "Cette chambre est bleue" is a communication; "Cette phrase est en français" is both a communication and a meta-communication, saying something about a class of statements, and the membership of this particular statement in it.

Russell’s famous example was the two utterances: "All Cretans are liars; I am a Cretan." We can only resolve the paradox by recognising that the utterances are of different logical types, existing in different conceptual universes, each of which is on its own level and has its own rules. So, in hypertext, each frame can be thought of as a block of text, organised by the traditional rules of the language in which it is written. But at the same time, the frame participates, at the level of its higher logical type or meta-communication, in an autonomous hypertextual syntax—that is to say, a set of rules or conventions for making intelligible relations between frames.

Two problems deserve mention here. First, hypertextual syntax does not have the collective and conventional form of a natural language. Hypertext designers are free to create and to link elements as they see fit; they will try to make their structures intelligible to users, but such intelligibility cannot be guaranteed in the same way that a well-formed sentence in a natural language can be reliably understood by a competent hearer. My collaborator George Landow has proposed some rules of thumb for the rhetoric and stylistics of hypertext, but the development of cceptable rules is still in its infancy-though to push forward with this development is one of the important tasks we face in the 1990s.[2]

Second, although hypertextual structures arc of a higher logical type than the blocks of traditional text they control, they are also relatively schematic or "impoverished" representations of the organic whole comprised of text and hypertext combined. To quote Wilden again:

On peut dire de "Cette phrase est en français" qu’elle est une métacommunication d'un type logique supérieur au message auquel elle réfère. C'est la classe de tous les messages en français. Mais le message auquel elle réfère est d'un niveau d'organisation plus complace. Il semble que le rapport hiérarchique entre les niveaux d'organisation et les niveaux de type logique soit inversement proportionnel. (Un règle ou un code sont d’un type logique supérieur mais d’un niveau d'organisation inférieur aux messages qu’ara en tire). {180]

If one thinks of a platoon of soldiers, this organisation of individuals into a visible class has obvious functional uses; but the class of "the platoon" is necessarily in opposition to the complex human reality of each soldier when considered as a person. In hypertext, we should always be aware of the dialectic--of levels or logical types-between the individual blocks of text, whose structures derive from the traditional reading process, and the hypertext structures that frame the blocks, link the frames, and propose a controlling spatial metaphor for the whole.


In the traditional printed text, we instinctively link one sentence or paragraph to the next as we proceed linearly through the book. This sequential chain of reading fulfils a silent contract: authors promise that their works will be comprehensible when read in linear order, while readers approach each new text with that order in mind. There are exceptions, of course—for example, when we begin with the index and search the text for particular references. Nonetheless, linearity is the rule; and it follows that there is no point in using hypertext when sequential order is indeed adequate to the ideas that the text seeks to express. Many things in the world have a linear order that we need not disturb! Further, there is a sense in which hypertext also conserves linear order since, even when links transport us from one block to another, they do so in a chronological sequence. But three crucial differences remain between reading hypertext and reading a book. First, in hypertext the links between blocks are made visually explicit, rather than silently taken for granted. Second, hypertext supports a multiplicity of reading sequences, according to the reader's choice of where to enter the structure and what links to follow. Finally, we need not assume that hypertext users will read all the text contained in a structure; they are more like people who consult a work of reference than those who read a book from cover to cover. A link in hypertext establishes a relation between one frame (and its enclosed text) and another. How, in principle, are these relations to be justified? A structure with ten blocks of text will have a potential forty-eight direct links between blocks, and a virtually infinity. number of possible ways of traversing the resultant web. To avoid such a "bad infinity," otherwise known as being "lost in hyperspace, ' we must have a rule of selectivity in assigning links. Further, we need to recognise and make explicit that there can be different kinds of links. The possible connections between one block and another are limited only by our ability to associate ideas; all the more reason why we need—and still largely lack—a rhetoric for representing and classifying such associations.

I will mention four kinds of links:
  1. From the general to the particular.
  2. From the particular to the general.
  3. A link that proposes an analogy.
  4. A link that proposes a causal or influential relation between blocks.

Evidently we can think of more kinds than these. But let us for the moment consider whether the differences between these kinds of link should be made explicit through some iconic code. We could do so with the form of links—for example, a solid or a dotted line--or of frames. In the book medium, we might establish such relations by using bold type for headings, normal type for main text, and reduced type for footnotes. With hypertext we have both a wider repertoire of devices for specifying relations, and the problem that most of the relations remain hidden "behind" the active screen. It is therefore not possible to make much progress on the issue of links without considering from the start a higher level of organisation, that of the overall hypertext structure.


I will discuss three approaches to hypertext structures: through spatial form, through the use of a center, and through the idea of citation.

Spatial form. It is a curious feature of human perception that we feel more comfortable visualising in three dimensions than in two. This means that a complex or "organic" object means more to us than a schematic one. We also have a well-developed facility for intuiting what is not directly visible--the other side of a human figure, and so forth. It follows that spatial metaphors in hypertext can be complex—provided that they correspond in some way to our visual perception of the natural world-and that we can readily extrapolate from visual clues to grasp a more complete structure "behind” the active screen. Some of these intuitive patterns that hold promise for the hypertext designer are the network, the tree diagram, the nest of Chinese boxes, or the web.

We must always be concerned that users may not find a logical or comprehensive path through the hypertext corpus, and may end up in the position of a kitten with a ball of wool. One way of avoiding this is to adopt a fairly simple "hub and spoke" or "tree diagram" design. A sentence in Joyce's Ulysses, for example, might be linked to three or four notes on different points of explication, to a parallel passage elsewhere in the novel to a visual image, and to a few bars of a song.[3] The implicit structure here is a series of spokes radiating out from each passage in the novel that requires explication. If each passage thus becomes a node in a hypermedia structure, then the overall form would be a linear series of nodes, proceeding sequentially through the novel, with each node having its own one-directional links to a cluster of supplementary files. Moving upwards in complexity, the visual metaphor of the network would allow anything defined as a node to be linked bi-directionally to any other node-which, of course, corresponds much more closely to the actual mental activities involved in reading and teaching Ulysses, where consideration of a particular phrase typically sets off a chain of connections through a space made up of both the text of the novel and a large body of commentary. The web corresponds to a "flexible" network in three dimensions, allowing for the foregrounding of various patterns without dismantling the system of links. We can think of this as a "cat's cradle" facility, that would present information in different configurations — corresponding to different modes of interpretation — while preserving the integrity of the whole. Finally, the Chinese Box uses the metaphor of one block of text folding into another, or expanding out of it, as in Jay Bolter's "Storyspace" program.

Hypertext designers face the challenge of representing, in the form of nodes and links, the actual complexity of building interpretations through reading, association and instantiation. In the mind of the critic, a link is not just a line between two blocks: it is a particular kind of connection, related to a whole family of kindred modes of explanation (such as, for example, Catholicism, Irish nationalism or Freudianism). Although computer-based visualization is now an important means of interpreting complex data in the sciences, its value in literary criticism has yet to be established.[4]The ability of hypertext to re-organize the traditional form of the book is surely useful on the level of information-handling, but more questionable when used to present interpretations in spatial or diagrammatic form. Progress in this area would require critics to become more self-conscious about spatial metaphors they currently use—for, say, the structure of narrative, or for levels of the mind in Freudian criticism—and also to take advantage of new opportunities to build models of interpretation on the computer screen. At the same time, any attempt to formalize mental activity will have its own pitfalls. A "designed" structure for intellectual work may provide only a somewhat shallow and barren environment, like a modem suburb or shopping mall; whereas the print media has developed organically with features added according to need in different eras, like an ancient city. Such an opposition resembles the one between levels of organisation and levels of logical type that we explored in the section on "Elements."

Centering, Hypertext can be thought of as a de-centered system, in the sense that it has no single, fixed point of focus or origin; and this de-centering can be cited in support of claims that hypertext has a special affinity with post-modern concepts of free-floating textuality. However, the center is not so easily banished. In hypertext, the dominance of the active screen over the virtual structure that lies behind it makes for a reading experience that is more centered, or focused, than the reading of a traditional book. Even Derrida, in a well-known exchange with Serge Doubrovsky, has observed that: " I didn't say that there was no center, that we could get along without a center. I believe that the center is a function, not a being—a reality, but a function. And this function is absolutely indispensable. "[5] It is more useful, then, to think of hypertext as having a mobile center-i.e. the currently active screen-rather than being radically de-centered.

The interesting issues of practical hypertext design concern whether we are to be content with the mobile center, or whether it should be supplemented with more traditional kinds of stable or hierarchic centers. One kind of hypertext imagines the reader as a browser, who moves through a web of blocks and links, making his or her own interests the de facto organizing principle (or center) for the investigation at the moment. Such a system will be freely and infinitely re-centerable; its points of focus—always provisional-will depend upon the choices made by a truly active reader. Another hypertext, however, puts a classical linear text, with its order and fixity, at the center of the structure. The designer then links various supplementary texts to this center, including critical commentary, textual variants, and chronologically anterior and later texts. In this case, the original text, which retains its old form, becomes an unchanging axis from which radiate linked texts that surround and modify it

A hypermedia stack that I designed with Jody Gilbert, on Fielding's Joseph Andrews, relies heavily on fixed or hierarchical centers.[6] The entry-screen to the stack remains as a privileged center, in two ways. First, this screen is a fixed point from which all subsequent screens converge. Second, the screen itself establishes a visual metaphor, that of the hub and spoke, that is repeated on succeeding screens and guides the reader's responses to their contents:

H2PTM (1989) Delany.png

If visualised in three dimensions, the stack is a series of tree structures, repeating a motif of nodes, links, and levels. At the top level, the master-screen is a control and distribution node leading to six sub-nodes at the next level down, each of which in turn controls a varying number of third-level functions. The stack therefore works as a hierarchy of trees and sub-trees; to make a horizontal link at the second of third levels, one must go up a level to the control node of the tree, then back down to the destination function. Given the small Macintosh screen and the inability of HyperCard to display more than one card at a time, potential horizontal destinations are not visible from the active card. With a bigger screen and, hopefully, further development of HyperCard, we would want to have the next-level-up control node always displayed in an ancillary window (perhaps even with active buttons included).

H2PTM (1989) Delany2.png

We can compare this with the British Inter-City rail system, a hub-and-spoke layout, in which all trains either start or finish in London (except that our structure has three dimensions rather than two). There are, for example, no direct trains from Oxford to Cambridge; instead, one must go from Oxford to London and from London to Cambridge. The aims of the system are speed and economy; trains that go along the spokes are expresses with large passenger capacity, whereas a decentralized network would have to have many thousands of minor trains. But there is also a useful conceptual simplification for the train user, since all possible choices of route are reduced to two orientations: towards London, away from London. In HyperCard, there is no speed advantage in making a horizontal link by way of an intermediate node; rather, the advantages of our layout are that its logical structure is easily grasped, and that it can be easily expanded by adding levels and nodes either at the current top or bottom. If we wanted to add two more Fielding stacks—for example, on Tom Jones and Jonathan Wild— it would be simple to add a new top-level node controlling access to all three stacks.

However, when you get down to what might be called the "information delivery level," as opposed to the "information organisation level," the hub-and-spoke template will probably no longer be appropriate. To deliver primarily graphic information, for example, a map may provide the best center, with appropriate buttons:

H2PTM (1989) Delany3.png

Since this is a hypermedia stack, at another "delivery point" we have the complete text of the novel, integrated with programs for searching and analysing the text At this level, a block of the novel's text belongs in the center:

H2PTM (1989) Delany4.png

The stack thus combines a strong and consistent structure at the level of hypermedia organisation (compare the discussion of "logical types" earlier), with a variety of modes of presentation at screens delivering information. This design has proved pleasing and effective to undergraduate students using the stack. I recognise, of course, the potential disadvantages of imposing such a strongly centralised structure; not everyone was happy with the feudal administrative pyramid and no, everyone in France, I am told, is happy with the centrality of Paris, and its dependent sub-centers! Nonetheless, one can still browse through the Joseph Andrews stack, arriving eventually at any other node from the point of departure-though to do so, one must observe the hierarchical principle because there are no "horizontal links (as if, in Britain, there were no local trains).

Citations. Now, some comments on a quite different approach to hypertext structure, which we may call statistical rather than spatial. A prototype for this approach can be seen in the Arts and Humanities Citation Index (AHCI), now available on-line as well as in print form. For any given creative or scholariy text, the AHCI lists where it has been cited in scholarly texts during a particular period The index thus provides a schematic record of scholarly intertextuality; but it is only a schema, because, like an on-line libraiy catalogue, the AHCI gives access to locations but not the underlying full-texts. Furthermore, AHCI has no thematic or subject ordering, so that its main use is in tracing links between authors (though, within the author category, there is a break-down to the level of individual texts).

A recent study of the Arts and Humanities Citation Index found the five most frequently cited works to be, for the period 1976-1983, in order: R. Kuhn, The Structure of Scientific Revolutions; Joyce, Ulysses; N. Frye, Anatomy of Criticism; L. Wittgenstein, Philosophical Investigations; N. Chomsky, Aspects of the Theory of Syntax.[7] The authors or works that are most frequently cited are, in a statistical sense at least, the ones that are most culturally central. There is, of course, a circularity here: are works important because they are cited, or are they cited because they are important? But in the relativist world of modem thought, we take such mutual determinations more and more for granted. For our present purpose, we can make a straightforward transformation of this frequency list into a hypertext structure: these five texts, considered as nodes, are the ones with the most links to other nodes. Using the AHCI, we might also produce a large-scale graphic representation of scholarly intertextuality, a kind of genetic code of our profession. We could thus visualise how a collective intellectual system generates its own centers and priorities through the cumulative choices of its adherents—evolving intelligibly, but without a single founding intentionality, like Saussure's model of language development

Another such model can be seen in the English literature hypertext developed by George Landow and others at Brown University, Context32. Starting from the premise that any distinctive culture is made up of a complex and "overdetermined" structure of relations, the unit encourages its student users to explore the multiple links between elements—political, religious, literary, etc.—of Victorian society, and thus come to appreciate the interrelatedness of the whole. The implicit underlying model is that the most representative Victorian texts- say, In Memoriam -are those that have the greatest number of relevant links to other nodes. These links operate, also in two directions. Outwards, there is much in In Memoriam that needs to be explained by citing major tenets of Victorian culture. Inwards, many central intellectual issues of the time, such as skepticism or the nature of nature, can be effectively instantiated from passages in the poem.

Again, we might say that this kind of centrality for crucial blocks of text provides a relatively fixed and objective hypertext structure, to be played off against the moving center of individual readers' choices of a path through the web.

Hypermedia and Macro-Structures

Expository prose, with its linear and propositional structures, has been too much identified with the privileged form of reason itself. Hypertext provides a better model for the mind's ability to re-order the elements of experience by changing the links of association or determination between them. But hypertext, like the traditional text from which it derives, is still a radical reduction — to a schematic visual code - of what was originally a complex physical and intellectual experience, engaging all the five senses. Consciousness itself is a continuous linking and restructuring of images selected from past, present and future; from the real & the imaginary; from the internal and external realms of experience. Hypermedia takes us even closer to this complex inter-relatedness of impressions; it extends hypertext by re-integrating our visual and auditory faculties into textual experience, linking graphic images, sound and video to verbal signs.

Some hypermedia programs, notably Apple's Hypercard, can also include launching or controlling programs that permit not just following links between units, but also the active manipulation of text, sound, graphics or video at the point of arrival. In such cases, we would no longer speak of traversing or reading a hypertext structure; rather, each node in the web can become an active window where different processes are invoked. When the "control level" of programs is used to access remote resources and integrate information acquired there, hypermedia units become truly unbounded, and merge conceptually with the "trawlers" or "personal information managers" described below.

A comprehensive hypermedia environment for literary study would be based on scholars workstations with appropriate communications and peripherals. The textual foundation (or "textbase") of this environment would be a large corpus of literary texts in machine-readable form. Access to the corpus would be controlled by a customized hypermedia program that would both operate on the corpus directly, and control a variety of other resources. [8] These resources could include text analysis programs,[9] hypermedia units such as Context32 and Perseus, HUMANIST (an international e-mail forum with over 400 participating scholars), the institution s on-line library catalogue, and the Modem Language Association bibliography on CD. A literature department, working collaboratively, could steadily enrich the information base of the environment, and develop complex links between its elements..

The French textual archive, Frantext, already has in machine-readable form about 200 million words of text, eighty percent of them literary. By the middle of the decade, it is reasonable to assume that most of the major English literary texts will be on-line; the main obstacle to building such an archive will be the copyright laws, which need radical adaptation to the era of electronic storage and reproduction.[10] For more than two thousand years, literary scholars have in the three-fold work of gathering selected texts from the archive, re-ordering and supplementing them, then re-inserting these new textual entities into the archive through some form of publication. All of these three stages have now been revolutionized by the computer: the first by the development of machine-readable text archives, the second by the word-processor, the third by desk-top publishing, network file transfer, and electronic mail. Still in its early stages, however, is the application of hypertext to integrate these scholarly activities, by providing a controlling metaphor for text-handling. Theorists have spoken of "The New Alexandria,” an etectronic archive holding a dominant proportion of the world's texts. The role of hypertext in the New Alexandria would be to provide a flexible and efficient means of defining nodes and links— that is, to impose structures on what would otherwise be a shapeless mass of "words, words, words." It goes without saying that the implementation of such a hypertext facility would be a highly difficult intellectual task. Nonetheless, the first stage could be based on established methods of textual mapping, ones that work with such relatively objective markers as books, articles, authors, subject descriptors and page numbers. Conceptually, the problem here would be simply one of re-classification: to take the navigational markers that help users find their way around a traditional library, and adapt them to a textbase stored electronically rather than as marks on paper. In the second stage, the aim would be to support nodes and links as defined by the individual user, so (hat the scholar could do electronically what she now does manually in, for example, assembling the relevant texts to be cited in a scholarly article.

Il is already clear that there are no major technical or financial obstacles to the establishment of very large electronic bases of text, sound, graphics and video in the near future. To make this call usable, however, will require many different kinds of programs-so different that the umbrella term of "hypermedia" for such programs may become dangerously vague. To speak only of texts: some programs will be straightforward adaptations of traditional library classification systems, with similar advantages (uniformity and objectivity) and disadvantages (lack of responsiveness to individual interests). More advanced programs will try to provide some of the flexibility and intuitiveness used by an expert critic when combing through texts and developing a thesis.

We can expect a "toolbox" of hypertext programs to be developed to support this kind of critical work. [11] Some tools will be "trawlers” that go into the textbase in search of relevant lexia. In the first generation these will be conventional search engines, working algorithmically to precisely defined specifications.[12] Later, artificial intelligence faculties will be added: these will make the search process reciprocal, by identifying a user's established interests and constantly searching the textbase for matching blocks of text. In turn, the user will be able to "train” her trawler by giving feedback on whether hits are relevant or not. Other tools will combine blocks by rules of association to produce, not interpretation itself, but the kind of clustering of evidence around a thesis that forms the elements from which interpretations are made. Others will make it possible to present alternative structurings of literary works, perhaps with a visualization facility to bring out patterns in complex sets of thematic or syntactic data. In all this work, the first challenge will be to formalize, in whatever degree is feasible, the kinds of mental operations that a literary critic performs when analyzing a text. Once these operations have been modelled in a hypertext program, however crudely, it should be possible to sharpen the tool by the kind of iterative movements-from textbase to structure and back—already described. We emphasize the term "tool" because a large body of modern literary theory warns us against the presumption that interpretations might be generated mechanically from a textbase by any computer program whatsoever. [13]


The development of literary hypertext will take place in an emerging global information environment of the archive, the computer, and telecommunications. Unlike the hub-and-spoke pattern of time-sharing mainframes with terminals, the new environment will be one in which information access and processing will be not just distributed, but also de-centered: any individual workstation will be able to achieve virtual presence at any other node on the global network. The computer network itself can be seen as a powerful metaphor, corresponding to a kind of animate hypermedia structure. The elements of the network include electronic mail, file transfer, computer conferencing, remote access to textbases and manipulation of them. These facilities enable each individual scholar to be linked to a constantly shifting structure of information sources, and of relevant other users. Instead of having to rely on a few dictionaries of quotations that index memorable phrases, scholars will be able to search the textbase for any sequence of words relevant to their current projects. As in Lévi-Strauss's metaphor of bricolage, units of knowledge will be extracted from one hypertext structure, with the aid of the computer, and re-assembled into another. Though much remains to be done, technological development is working rapidly, and on many fronts, to make the entire cultural archive of text and images accessible, and to allow users of the archive to associate freely and effectively to communicate their results.


  1. Wilden, Système et Structure: Essais Sur la Communication et l'échange (Montréal : Boréal Express, 1983), 180.
  2. Landow, "Relationally Encoded Links and the Rhetoric of Hypertext,” in Proceedings of HyperTEXT 87 Workshop (Chapel Hill, NC, 13-15 November 1987), 331-343.
  3. P.Delany & J. Gilbert arc considering design issues for a hypermedia unit on Ulysses, tentatively called "Penelope's Web.”
  4. Some of the stylistic analysis of Jane Austen in John F. Burrows, Computation into Criticism: A Study of Jane Austen s Novels and an Experiment in Method (Oxford: Clarendon Press, 1987), is presented in graphic form. See also the diagrams of Freudian response theory in Norman Holland, The Dynamics of Literary Response (New York: Oxford University Press, 1968).
  5. "Structure, Sign and Play in the Discourse of the Human Sciences," in R. Macksey & E. Donato, eds., The Structuralist Controversy: the Languages of Criticism and the Sciences of Man (Baltimore. Johns Hopkins University Press, 1972), p. 271.
  6. P. Delany, J. Gilbert, "A Hypermedia Stack for Fielding's Joseph Andrews: Issues of Design and Content," in P. Delany, G.P. Landow, eds., Hypermedia and Literary Studies (MIT Press, forthcoming).
  7. E.C. Garfield, "A different Sort of Great-Books List: the 50 20th Century Works Most Cited in the AHCI, 1976-83, Current Contents (16), 20 April 1987,3-7.
  8. Current work on this concept, by P. Delany & J. Gilbert, uses HyperCard in a networked Macintosh laboratory.
  9. For a survey of available programs see S. Stigleman, "Text Management Software," Public Access Computer Systems Review, 1, 1 (1990).
  10. What is requiered is some kind of monitoring or sampling system that will automatically give credit to originators for the use of their texts. An example ofsuch a crude but workable system is Canada's "Public Lending Right": this searches the holdings of ten major librairies with compurized catalogues, and makesan annual payment (currently$40) for each "hit" of a book by a registred author. The maximum annual payment is$4,000a year. funds to support the scheme are provided by the federal government, about$5 million in 1989-90. Norway collects about$7 million (U.S) a ayear in fees for xerox copying;the money isthen distributed to authors through writers unions. Each bon-fiction writer currently receives about $2,000 a year from this sheme.
  11. For some useful hints see Mark Zimmerman, "Notes on Free Text Information Retrieval" (unpublished; available on the Gutnberg listserver).
  12. For example. Information Lens, a program developed at MIT for filtering incoming e-mail according to the interests of the recipient.
  13. See, for example, Stanley Fish, “What is Stylistics and Why Are They Saying Such Terrible Things about it?” in S. Chatman, ed Approaches to Poetics (New York: Columbia University Press, 1972); Hubert L. Dreyfus, What Compute Can't Do: A Critique of Artificial Intelligence (New York: Harper & Row, 1972).