the referenced text…. Readers can browse through linked, cross-referenced, annotated texts in an orderly but nonsequential manner" [Nicole Yankelovich, Norman Meyrowitz, and Andries van Dam, "Reading and Writing the Electronic Book,"IEEE Computer 18 (1985)]. Writers on hypertext trace the notion to a 1945 article by Vannevar Bush in Atlantic Monthly that called for mechanically linked information retrieval machines in the midst of what was already becoming an explosion of information. In the 1960s Douglas C. Englebart and Nelson began to design computer systems that could implement some of these notions of linked texts, and today hypertext refers almost exclusively to electronic hypertext systems that rely on computing equipment and software.

Originally the idea of hypertext did not depend upon computers. In fact, the standard scholarly article in the humanities or physical sciences perfectly embodies the underlying notions of hypertext as nonsequentially read text. For example, in reading a critical article on James Joyce's Ulysses, one reads through the main text, encounters a number or symbol that indicates the presence of a foot-or endnote, and leaves the main text to read it. The note might contain a citation of passages in Ulysses that support the argument in question as well as information about the author's indebtedness to other scholars, disagreement with them, and so on. The note might also summon up information about sources, influences, and parallels in other literary texts. In each case, the reader can follow the link to another text indicated by the note, thus moving entirely outside the text of the article itself. Having completed reading the note or having decided that it does not warrant a careful reading at the moment, one returns to the main text and continues reading until one encounters another note, at which point one again leaves the main text. This combination of excursionary reading outside the main text and returning to it constitutes the basic experience of hypertext.

Suppose now that one could simply touch the page where the symbol of a note or reference appeared and instantly bring into view the material contained in a note or even the entire other text—here all of Ulysses—to which that note refers. Scholarly articles situate themselves within a field of such relations, but the print medium keeps most of these intertexts out of sight and relatively difficult to follow because the referenced (or linked) materials lie spatially distant from the corresponding note. In contrast, electronic hypertext makes individual references easy to follow and the entire field of textual interconnections easy to navigate. Changing the rapidity and ease with which one can orient oneself within such a context of information radically changes both the experience of reading and ultimately the nature of what is read. For example, if we had a hypertext system in which our putative Joyce article existed linked to all the materials it cited, it would appear to the reader as part of a much larger system whose totality might count more than the individual document. The article would now be woven much more tightly into its context than would its traditional print technology counterpart. The ease with which readers traverse such a system has additional consequences for as they move through this web of texts, they continually shift its center—and hence the focus or organizing principle of their investigation and experience. Hypertext, in other words, provides an infinitely recenterable system whose provisional point of focus depends upon the reader, who becomes truly a user of knowledge.

Hypertext offers enormous possibilities to the student and teacher of literature, all of which derive from its fundamental connectivity, a quality that greatly speeds up certain processes involved in skilled reading and critical thinking, while also making them far easier to carry out. The greater speed of making connections in hypertext permits and encourages sophisticated forms of analysis….

Hypertext has the capacity to emphasize intertextuality in a way that page-bound text in books cannot. Scholarly articles, as we have seen, offer an obvious example of explicit hypertextuality in nonelectronic form. Conversely, any work of literature—which for the sake of argument and economy I shall here confine in a most arbitrary way to mean canonical literature of the sort we read and teach in universities—offers an instance of implicit hypertext in nonelectronic form. Take Joyce's Ulysses, for example. If one looks, say, at the "Nausicaa" section in which Bloom watches Gerty McDowell on the beach, one notes that Joyce's text here "alludes" or "refers" (the terms we usually employ) to many other texts or phenomena that one can treat as texts, including the Nausicaa section of The Odyssey, the advertisements and articles in the women's magazines that suffuse and inform Gerty's thoughts, facts about contemporary Dublin and the Catholic Church, and other passages within the novel. Again, one can envisage a hypertext presentation of the novel that would link this section not only to the kinds of materials mentioned, but also to other works in Joyce's career, critical commentary, and textual variants. Hypertext here permits one to make explicit, though not necessarily intrusive, the linked materials that an educated reader perceives surrounding the novel.

[In "Is There an Intertext in This Text?: Literary and Interdisciplinary Approaches to Intertextuality," American Journal of Semiotics 3 (1985)] Thaïs Morgan suggests that intertextuality, "as a structural analysis of texts in relation to the larger system of signifying practices or uses of signs in culture," shifts attention from the triad constituted by author/work/tradition to another constituted by text/discourse/culture. In so doing, "intertextuality replaces the evolutionary model of literary history with a structural or synchronic model of literature as a sign system. The most salient effect of this strategic change is to free the literary text from psychological, sociological, and historical determinisms, opening it up to an apparently infinite play of relationships." Morgan describes well a major implication of hypertext (and hypermedia) intertextuality: one gains an opening up, a freedom to create and perceive interconnections occurs. Nonetheless, although hypertext intertextuality would seem to devalue any historic or other reductionism, it in no way prevents those interested in reading by means of author and tradition from doing so. Our experiments [with students at Brown University using Context32, a series of linked documents,] thus far suggest that hypertext does not necessarily turn one's attention away from such traditional critical approaches. What is perhaps most interesting about hypertext, though, lies not in whether it fulfills certain claims of structuralist and poststructuralist criticism, but that it provides a rich means of testing them.

Intermedia does not work in the manner of most first-generation computer-assisted instruction. Most such programs and materials, which follow the model of printed workbooks, take the user through a pre-arranged sequence of exercises and experiences. Such systems constrain the users by forcing them to follow a single sequence or relatively few possible sequences. Intermedia, in contrast, reflects the fundamental characteristic of hypertext and hypermedia systems: they are bodies of linked texts that have no primary axis of organization. In other words, Intermedia has no center. Although this absence of a center can create problems for the teacher and researcher, it also means that anyone who uses Intermedia makes his or her own interests the de facto organizing principle (or center) for the investigation at the moment. One experiences Intermedia as an infinitely decenterable and recenterable system.

Hypertext then is related to the ideas of Jacques Derrida and Louis Althusser, both of whom emphasize the need to shift vantage points by decentering discussion. As Derrida points out in "Structure, Sign, and Play in the Discourse of the Human Sciences," the process or procedure he calls decentering has played an essential role in intellectual change. For example, "Ethnology could have been born as a science only at the moment when a de-centering had come about: at the moment when European culture—and, in consequence, the history of metaphysics and of its concepts—had been dislocated, driven from its locus, and forced to stop considering itself as the culture of reference." Derrida makes no claim that an intellectual or ideological center is in any way "bad," for as he explains in response to a query from Serge Doubrovsky, "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" [Richard Macksey and Eugenio Donato, The Structuralist Controversy: The Languages of Criticism and the Sciences of Man, 1972].

Intermedia, like all hypertext systems, permits the individual user to choose his or her own center of investigation and experience. What this principle means in practice is that Context32 does not lock the student into any kind of organization or hierarchy. For the person who chooses to organize a session on the system by making use of author overview files, moving from [the Keats overview or OV] to Tennyson OV, the system would represent an oldfashioned, traditional, and author-centered survey course. On the other hand, nothing constrains the student to work in this manner, and any students who wish to investigate the validity of period generalizations could organize their sessions in terms of such periods by using the Victorian and Romantic OV files as starting or midpoints, while others could begin with ideological or critical notions, such as overviews of Feminism or the Victorian Novel. In practice, however, students employ Context32 as a text-centered system, since they tend to focus upon individual works with the result that even if they begin sessions by entering the system at an individual author overview file, they tend to spend most time with files devoted to individual works, moving between poem and poem (Swinburne's "Laus Veneris" and Keats's "La Belle Dame sans Merci") and between poem and informational texts ("Laus Veneris" and files on chivalry, medieval revival, courtly love, Wagner, and so on).

As the capacity of hypertext systems to be infinitely recenterable suggests, they have the corollary characteristics of being antihierarchical and democratic in several different ways. First, as the authors of "Reading and Writing the Electronic Book" point out, in such systems "Ideally, authors and readers should have the same set of integrated tools that allow them to browse through other material during the document preparation process and to add annotations and original links as they progress through an information web. In effect, the boundary between author and reader should largely disappear." One sign of the disappearance of boundaries between author and reader is the fact that the reader, not the author, largely determines how the reader moves through the system, for the reader can determine the order and principle of investigation. Intermedia has the potential, thus far only partially realized, to be a democratic or multicentered system in yet another way: as students who use the system contribute their comments and individual documents, the sharp division between author and reader that characterizes page-bound text begins to blur with several interesting implications: first, by contributing to the system, student users accept some responsibility for materials anyone can read; second, students thus establish a community of learning, demonstrating to themselves that a large part of any investigation rests on the work of others.

Although students, particularly beginning students, do not have sufficient knowledge of either primary materials or their contexts to create adequate treatments of more complex issues, they often produce excellent brief discussions of relatively limited, specific topics, such as aspects of technique in specific texts or ways in which one text relates to others. Although students in the survey contributed relatively little original material to Context32, those in English 61, the Victorian poetry seminar, contributed much more. Members of this and subsequent classes have created a wide range of materials including brief comparative essays, annotated bibliographies, analyses of particular passages, annotations to maps or other graphical materials illuminating individual works, and their own versions of concept maps and literary relation files….

Intermedia's emphasis upon connections and relations encourages students to integrate materials from a single course with everything else they know.

There is another form of democratization or absence of hierarchy: in hypertext systems, links inside and outside a text—intertextual and intratextual connections between points of text (including images)—become equivalent, thus bringing texts closer together and blurring the boundaries among them. Consider what happens to the distinction between intra- and intertextual links in Milton in a hypertext system like Intermedia. Examples of hypertext intratextual links could be created by Milton's various descriptions of himself as prophet or inspired poet in Paradise Lost or by linking his citations to Gen. 3:15 within the poem. Intertextual links, in contrast, are exemplified by links between a particular passage in Paradise Lost that mentions prophecy and his other writings in prose or poetry that make similar points as well as those between this passage in the poem and biblical texts, scriptural commentaries throughout the ages, comparable or contrasting poetic statements by others, and scholarly comment by students of literature. Similarly, Miltonic citations of the biblical text about the heel of man crushing the serpent's head and being in turn bruised by the serpent link to the biblical passage and its traditional interpretations as well as to other literary allusions and scholarly comment upon all these subjects. Hypertext linking simply allows one to speed up the usual process of making connections while providing a means of graphing such transactions—if one can apply the word simply to such a radically transformative procedure. The speed with which one can move between passages and points in sets of texts promises to change the way we read and write, just as high-speed number-crunching computing changed various scientific fields by making possible investigations that before had required too much time or risk. We do not know all the ways hypertext will affect reading (and production) of texts, but one effect is already clear: the distinction between intratextuality and intertextuality will become harder to maintain than it is with book technology, and this crucial change comes from the fact that electronic linking permits the reader to move with equal facility between points inside a text and those outside of it. Once one can move with equal facility between the opening section of Paradise Lost and a passage in book 12 thousands of lines "away," or between that opening section and a particular anterior French text or modern scholarly comment, then, in an important sense, the discreteness of texts, which print culture creates, has radically changed and possibly disappeared.

These observations about hypertext suggest that computers bring us much closer to a culture whose qualities have more in common with those of preliterate humans than even Walter J. Ong has been willing to admit. In Orality and Literacy he argues that computers have brought us into what he terms an age of "secondary orality" that "has striking resemblances to the old [oral, preliterate culture] in its participatory mystique, its fostering of a communal sense, its concentration on the present moment, and even its use of formulas." Nonetheless, although Ong finds interesting parallels between a computer culture and a purely oral one, he still insists, "The sequential processing and spatializing of the word, initiated by writing and raised to a new order of intensity by print, is further intensified by the computer, which maximizes commitment of the word to space and to (electronic) local motion and optimizes analytic sequentiality by making it virtually instantaneous." In fact, hypertext systems, which insert every text into a web of relations, produce a very different effect, for they allow nonsequential reading and thinking.

Such nonsequential reading weakens the boundaries of the text, either correcting the artificial isolation of the text from its contexts or violating one of the chief qualities of the book, depending on one's point of view. Another possible result of hypertext may also be disconcerting. As Ong points out, books, unlike their authors, cannot really be challenged: "The author might be challenged if only he or she could be reached, but the author cannot be reached in any book. There is no way to refute a text. After absolutely total and devastating refutation, it says exactly the same thing as before. This is one reason why 'the book says' is popularly tantamount to 'it is true.' It is also one reason why books have been burnt. A text stating what the whole world knows is false will state falsehood forever, so long as the text exists." The question arises, however, if hypertext situates texts in a field of other texts, can any individual work that has been addressed by another still speak so forcefully? One can imagine hypertext presentations of books (or the equivalent) in which the reader can call up all the reviews and comments on that book, which would then inevitably exist as part of a complex dialogue rather than as the embodiment of one voice or thought that speaks unceasingly and authoritatively.

As one might expect at this relatively early stage in the history of hypertext systems, those involved in developing them have devoted most attention to the simple fact of linking and to the effects upon discourse of such electronically linked text. Now we need to develop a rhetoric and stylistics of hypertext. To begin with, we must recognize that although hypertext redefines some of the basic characteristics of page-bound, printed discourse, such as the rigidly hierarchical distinction in scholarly works between a main text and its annotation, it still depends upon many of the same organizing principles that make page-bound discourse coherent and even pleasurable to read.

Designers of hypertext and hypermedia materials confront two related problems, the first of which is how to indicate the destination of links, and the second, how to welcome the user on arrival at that destination. Drawing upon the analogy of travel, one can say that the first problem concerns exit or departure information and the second arrival or entrance information. In both cases the designer must decide what users need to know at each end of a hypertext link to make use of what they find there. The general issue here is one of interpretation—namely, how much interpretation in the form of encoding or markup must the designer-author attach to the points at which one both leaves and enters a text?

John Slatin (essay date 1991)

SOURCE: "Reading Hypertext: Order and Coherence in a New Medium," in Hypermedia and Literary Studies, edited by Paul Delany and George P. Landow, The MIT Press, 1991, pp. 153-69.

[An American educator and critic, Slatin has written several essays on hypertext. In the following essay, he remarks on the continuities and differences between hypertext and conventional text.]

The basic point I have to make is almost embarrassingly simple: hypertext is very different from more traditional forms of text. The differences are a function of technology, and are so various, at once so minute and so vast, as to make hypertext a new medium for thought and expression—the first verbal medium, after programming languages, to emerge from the computer revolution. (The computer has spawned new media in the visual arts and music as well.) As a new medium, hypertext is also very different from both word processing and desktop publishing, two other computer technologies which have had an enormous impact on the production of texts. Both word processing and desktop publishing have as their goal the production of conventional printed documents, whereas hypertext exists and can exist only online, only in the computer. A new medium involves both a new practice and a new rhetoric, a new body of theory. I hope this paper will serve as a step in that direction.

The first requirement for a rhetoric of hypertext is that it must take the computer actively into account as a medium for composition and thought—not just as a presentational device and not simply as an extension of the typewriter. This means different things, depending on the level of abstraction at which one stands. At some levels, for example, one has to deal not only with knowledge as constructed by humans for humans, but also with data structures, with the representation—or, more accurately, the reconstruction—of knowledge (if the term still applies) in hardware and software. It means, also, that the rhetoric itself must be abstract, like Wallace Stevens's Supreme Fiction, in order to permit movement up and down the ladder of abstraction and to permit the articulation of principles that will enable practice. By the same token, the rhetoric of hypertext will have to be capable of change: for it is tied to a still immature (perhaps perpetually immature) technology which is itself changing at an exponential rate.

On the assumption, then, that description is the first step toward theory, I will contrast hypertext with more traditional text. I will focus on the assumptions each makes about what readers do and the ways in which those assumptions about reading affect the author's understanding of composition. For the purposes of this discussion, taking the computer into account means that we have to find ways of talking about "documents" that have multiple points of entry, multiple exit points, and multiple pathways between points of entry and exit points. Moreover, we must find ways to talk about the still more exciting kinds of activity fostered by this proliferation of possibilities: I mean interactive reading and its more or less inevitable concomitant, interactive writing, or co-authorship.

Widespread literacy is a comparatively recent phenomenon—that is to say that in Western societies such as those of Europe, and North America, general literacy is at best a couple of hundred years old. But Western culture was a print culture long before the coming of general literacy, and the text environment we are all familiar with is the product of fully mature, highly stable manuscript and print technologies which have been in place for many centuries. Our principles and strategies for effective written communication are therefore based on long-established assumptions about readers and reading. It will be helpful to consider these assumptions briefly before going on to discuss the different assumptions embedded in the design of hypertext and hypermedia systems.

The assumption that reading is a sequential and continuous process is the foundation on which everything else rests. The reader is expected to begin at a clearly marked point whose appropriateness has been determined by the author—usually with considerable effort: one of the hardest moments in any writing project is to figure out where to start—and to proceed from that beginning to an ending which is just as clearly marked and which has also been determined by the author in accordance with his or her understanding of the subject matter and the reader. The reader's progress from the beginning to the end of the text follows a route which has been carefully laid out for the sole purpose of ensuring that the reader does indeed get from the beginning to the end in the way writer wants him or her to get there.

All but the most naive and inexperienced writers recognize that all but the most naive and inexperienced readers inevitably and rightly make inferences about what's going to happen next, on the basis of what they have already read—not only in the current text, but in other texts resembling it. The reader's perception of the predictability of a given text is an important factor in his or her qualitative evaluation of the text.

Prediction operates on a number of different levels and is determined by different things at different levels of abstraction. The predictability of a given text is a function of the relationships among phenomena at microscopic, macroscopic, and metatextual levels of abstraction. At the microscopic levels, the reader's ability to predict the course of the text from moment to moment is a function of such factors as paragraphing, sentence length, complexity of phrasing, vocabulary, and so on—that is, the factors that are evaluated in producing a so-called "readability index." Indeed, one might descend even further down the ladder of abstraction and argue that prediction takes place at the graphemic or phonemic levels as well. At the macroscopic levels, the reader is aware of such things as general subject matter, topics and subtopics and the structural devices organizing the text as a whole—sections, chapters, and subchapters, and so forth. At this level the reader is also at least subliminally aware of such things as how much more material there is to read. On what we might call the metatextual level, the reader makes inferences about the text as a whole, based on his or her understanding of the larger context to which s/he regards the text as belonging. These inferences are often implicit, or else they may take the form of mental or marginal annotations; in any case, they are outside the text and separate from it. But, as I shall explain later on, they are integral to hypertext.

The end product of the writing process is a text written or printed on paper, and then, often, sewn or glued or otherwise bound between covers. This is obvious, I know, and yet it needs to be said. Every writer has had the awful experience of opening a book or article hot off the press, only to stare in horror at a glaring error in a crucial passage that somehow escaped the most agonizing scrutiny. The fixity of the printed text as an object in physical space makes the text as an object in mental space seem equally stable and fixed. Or at least that's how we tend to want it. As Richard Lanham has said, "It was establishing the original text that the Renaissance scholars thought their main task, and generations of textual editors since have renewed their labors. The aim of all this was to fix the text forever" ["Convergent Pressures: Social, Technological, Theoretical," Conference on the Future of Doctoral Studies in English, Wayzata, Minnesota, April 1987]. The continuing controversy over Hans-Walter Gabler's edition of Joyce's Ulysses makes abundantly clear just how intense the desire "to fix the text forever" can be. Gabler's reconception of the editing process may have occurred "independently of his decision to use the computer," but the controversy over his "synoptic" version of Ulysses offers a clear illustration of the way computers will revolutionize our understanding of text. Every word in Gabler's synoptic text was written by Joyce himself—and yet the final "reading text" is a text no one ever wrote—it had never existed prior to its publication. (In this sense, Gabler's Ulysses resembles Benoit Mandelbrot's fractals, in which recursive mathematical formulae are graphically plotted to produce visual structures that, while in some cases resembling phenomena in the day-to-day world, have no counterpart in that world.) Gabler's is a simulated Ulysses, like the Don Quixote that would be produced by Borges' Pierre Menard if Menard were real and if he had a computer. This is what becomes of the work of art in an age of electronic reproduction. Text is always mutable, always subject to inadvertent error and deliberate change, and it has to be coerced into standing still—that's why publishers charge you money if you make too many changes in a text after it's been typeset.

For all these reasons—because a text looks like a permanent thing, because readers expect to begin at the beginning and end at the end and to know which is which (that's why students so often begin the last paragraph with In conclusion), because readers expect to get from beginning to end via a clearly-marked route—sequence is of paramount concern to a writer. Much of his or her effort goes into figuring out the correct sequence for the material that's going to be presented. The writer's job in this context is to contrive a sequence that will not only determine the reader's experience and understanding of the material but will also seem to the reader to have been the only possible sequence for that material; you want it to seem to have been somehow inevitable.

Of course this inevitability has a good deal to do with the issue of predictability I raised earlier. Readers have to be able to predict what will come next, at least up to a point, or they start to feel lost, which makes them start to feel nervous, which makes them want to put down what they're reading and go watch a football game or something—at which point the writer has failed miserably. But the flip side is just as bad, and the end result is going to be the same.

Writing that's too predictable is governed by a presupposition succinctly expressed by a certain hotel chain's ad campaigns. The presupposition that there should be "no surprises" may be fine for hotels. But it becomes a fundamental conceptual error where writing is concerned.

The informational value of a given document is not simply a function of the quantity of data it presents or the facts it contains. At one level of abstraction, what we call information may indeed consist in numbers, dates, and other data, other facts. But as Gregory Bateson says [in his Mind and Nature: A Necessary Unity, 1980], "All receipt of information is necessarily receipt of the news of difference." At a somewhat higher level of abstraction, therefore, none of these data can be considered information until they have been contextualized, arranged in such a way that both the significant differences and the significant relationships among them may become apparent to the intended reader. In Christopher Dede's terms, this is when information becomes knowledge. In other words, as literary artists and their readers have always known, there can be no information without surprise.

Rhetoric typically has little to say about the physical processes by which a text is brought into being. Or I could put it even more strongly and say that rhetoric has traditionally been indifferent to the technology of communication. One reason for this indifference is that the technology is so mature that it's simply taken for granted, so that it is essentially invisible as technology. There was a point in history, of course, when writing itself was a radically innovative technology and was regarded as such, as Eric A. Havelock, Father Walter J. Ong, and Richard Lanham have shown us. The computerization of writing has similarly made the technology itself highly visible, especially in the cases of desktop publishing and hypertext/hypermedia. By contrast with traditional text, hypertext and hypermedia depend upon an emergent technology which is still immature and still subject to radical transformation; indeed, all indications are that accelerating change is an inherent characteristic of this technology. It may never stabilize. Thus rhetoric for hypertext cannot afford to disregard the technological substrate upon which composition and reading depend.

There are many continuities between conventional text and hypertext. Anyone involved in creating a hyperdocument will still have to worry about the problems I've outlined so far. But hypertext is a very different kind of beast than a conventional text, and creating a hyperdocument poses some very different problems as well. The remainder of this paper will concentrate on those differences and their implications.

First of all, the hyperdocument may well contain material from several different media such as text, graphics, video, and sound. While this is an important factor, I don't think it's decisive. After all, printed books often contain text, line drawings, tables of data, reproductions of visual images, and so forth—though of course they cannot manage full motion video or sound. Besides containing different types of materials than those to be found in printed text, the hyperdocument is likely to contain considerably more material than a printed book. Again, this is not a decisive difference in itself: encyclopedias also contain an enormous quantity of material. The quantity of material in a hyperdocument does pose problems, and it does make for complexity. But the greatest difference between text and hypertext is not in the relative quantity of material each form handles: it's in the technology that handles the material.

What makes all the difference in the world is the fact that hypertext exists and can exist only in an online environment. This is crucial, not just because it substitutes monitors, keyboards, and mice for the customary physical apparatus associated with text—paper, books, pencils, and so forth. The fact that hypertext exists only in the online environment is crucial because, as Douglas Hofstadter says [in his Metamagical Themas: Questing for the Essence of Mind and Pattern, 1986], "It is the organization of memory that defines what concepts are." Hypertext uses machine memory in a way that has no analogue in the traditional text environment, where composition relies on the organization of human memory. It is the organization of memory in the computer and in the mind that defines hypertext and makes it fundamentally different from conventional text.

In such an environment, the problem is not simply to develop effective strategies for implementing well known and long established principles of effective communication. On the contrary, one of the chief functions of rhetoric in the hypertext environment is to discover the principles of effective communication and then develop ways of implementing those principles through the available technology.

The rapidly evolving technological environment makes hypertext possible by permitting the embodiment of a very different set of assumptions about readers and reading—and about thinking. These assumptions in turn form the basis for decisions made in the process of creating a hyperdocument.

Reading, in hypertext, is understood as a discontinuous or non-linear process which, like thinking, is associative in nature, as opposed to the sequential process envisioned by conventional text. Associative thinking is more difficult to follow than linear thinking. Linear thinking specifies the steps it has taken; associative thinking is discontinuous—a series of jumps like the movement of electrons or the movements of the mind in creating metaphor. This discontinuity is not fortuitous; rather, as Stewart Brand points out [in his The Media Lab: Inventing the Future at MIT, 1987], it is a basic aspect of the digital encoding of information. Brand offers the illuminating contrast between the surface of a traditional phonograph album, with its continuous grooves, and the surface of a compact disc, with its distinct, discontinuous pits.

Reading in this sense has little to do with traditional notions of beginning at the beginning and going through to the end. Instead, the reader begins at a point of his or her own choosing—a point chosen from a potentially very large number of possible starting points. The reader proceeds from there by following a series of links connecting documents to one another, exiting not at a point defined by the author as "The End" but rather when s/he has had enough. Accordingly, the most common metaphors in discussions of hypertext equate reading with the navigation or traversal of large, open (and usually poorly-charted) spaces. As Jeff Conklin has pointed out [in his A Survey of Hypertext, 1986], because the hyperdocument contains so much material, and because relations between the components of the hyperdocument are not always spelled out, there is a significant danger that the reader will get lost or become badly disoriented.

The difficulty is compounded because hypertext systems tend to envision three different types of readers: the reader as browser, as user, or as co-author. The relationship between these three classes can be fuzzy and therefore difficult to manage. One function a rhetoric for hypertext will have to serve will be to provide ways of negotiating it.

The browser is someone who wanders rather aimlessly (but not carelessly) through an area, picking things up and putting them down as curiosity or momentary interest dictates. In this respect the browser is someone who reads for pleasure, with this important difference: there is no expectation that the browser will go through all of the available material; often the expectation is just the reverse. It is difficult to predict the browser's pathway through the material—and in fact it is less important to predict the pathway the browser will take than it is to provide a backtracking mechanism, what Mark Bernstein calls [in "The Bookmark and the Compass: Orientation Tools for Hypertext Users," in ACM SIGOIS Bulletin (October 1988)] a Hansel-&-Gretel trail of breadcrumbs to allow the browser to re-trace his or her steps at will. (Of course this same mechanism is essential for readers in the two remaining categories as well.)

By contrast with the browser, the user is a reader with a clear—and often clearly limited—purpose. He or she enters the hyperdocument in search, usually, of specific information and leaves it again after locating that information. The user's path is relatively predictable, provided those who have created the hyperdocument have a sufficient understanding of the task domain. In these respects, then, the user resembles a typical student doing the assigned reading for a course. But there is also an important difference between the user and the student, which is most clearly recognized from the vantage-point of the author rather than the reader. The author(s) of a hypertext documentation system (for, e.g., a software product like Microsoft QuickBASIC 4.5) will have met their goal when the user finds the information s/he needs and returns to the work in progress. The instructor designing a set of hypertext course materials may well not be satisfied with such an outcome, however: the instructor aims at a dynamic process, in which the student moves among three different states: from a user the student becomes a browser (and may then become a user once again). Ultimately, the student becomes fully involved as co-author. Thus what looks like a hierarchy of readers collapses.

One of the most important differences between conventional text and hypertext is that most hypertext systems, though not all, allow readers to interact with the system to such an extent that some readers may become actively involved in the creation of an evolving hyperdocument. Co-authorship may take a number of different forms—from relatively simple, brief annotations of or comments on existing material, to the creation of new links connecting material not previously linked, to the modification of existing material or the creation of new materials, or both. Both literary theorists (e.g., Wolfgang Iser, Paul Ricoeur, Stanley Fish) and cognitive scientists like Jerome Bruner have talked for years about the reader's involvement in the construction of textual meaning. But hypertext's capacity for literally interactive reading and co-authorship represents a radical departure from traditional relationships between readers and texts. The implications of this departure from traditional relationships between readers and texts are enormous, both for the creative arts and for education: as many theorists now agree, understanding comes about when the mind acts upon the material. Marshall McLuhan's distinction between hot and cool media is relevant here—a cool medium being one that invites active participation, a hot one being one before which one sits passively. McLuhan was thinking, of course, about the difference between print and television, but one might argue that hypertext combines the heat and visual excitement of film, video, and television with text's cool invitation to participate.

"Writing," in the hypertext environment, becomes the more comprehensive activity called "authoring." Authoring may involve not only the composition of text, but also screen layout and other things that fall under the general rubric of interface design; it may also involve a certain amount of programming (as in Apple's HyperCard, where complex navigational and other processes are scripted by the stack's author). Perhaps most importantly, authoring involves the creation and management of links between nodes.

Ted Nelson, who coined the term "hypertext," defines hypertext [in his Literary Machines, 1987] as "non-sequential writing." This means writing in which the logical connections between elements are primarily associative rather than syllogistic, as in conventional text. One implication of this is that the hyperdocument "grows" by a process of accretion, whereas the conventional document tends to have been winnowed out of a larger mass of material. That is, in preparing to write a conventional document, you almost inevitably assemble more material than you can possibly use; the closer you come to final copy, the more you find yourself excluding material that "doesn't fit" the subject as you've finally defined it. Hypertext, by contrast, is an inclusive medium. Thanks to the capability of creating nodes and links, material not linearly related to the point being discussed at the moment but still associated with that point may be placed in a node of its own and linked to other nodes as appropriate; the material need not be thrown away. In much the same way, no individual point of view need be excluded.

This inclusiveness makes it unlikely that any one individual will see all the elements making up the system. It also means that the hyperdocument is in fact a collection of possible documents, any one of which may be actualized by readers pursuing or creating links between elements of the system.

The end product of the authoring process, the hyperdocument, is not a closed system, like a book; it is rather an open and dynamic system. The hyperdocument is an online system or network whose constituents are of two basic types: nodes and links. Nodes may consist of documents, images, or other materials electronically connected—linked—to one or more other documents or images. Very likely, the different nodes will represent the work of quite a few individuals, who may have been working at different times and in different locations. Indeed, one impetus for the development of hypertext systems has been in the need to address exactly this issue among members of development teams. The development of protocols and procedures for co-authorship thus becomes an important issue. So does the development of procedures for moving through the system.

The reader's progress through a conventional text is governed by the arrangement of the material; the burden of prediction falls more heavily upon the reader than on the writer. This situation becomes considerably more complicated in hypertext. Given a system of discrete and interconnected nodes, the reader/user must decide which links to follow; in order to make that decision intelligently, s/he must be able to make reliable predictions about the consequences of particular choices. But the freedom of movement and action available to the reader—a freedom including the possibility of co-authorship—means that the hypertext author has to make predictions as well: for the author, the difficulty at any given moment is to provide freedom of movement and interaction, while at the same time remaining able to predict where the reader/user will go next. The most effective solution here, I think, will be to treat each node as if it were certain to be the reader's next destination. This is time-consuming in the short run, but in the long run probably saves time by creating a more readily usable system.

This brings us to the issue of linkage, the mechanism that creates the hyperdocument and allows the reader to move through it. Douglas Hofstadter has suggested that the perception of relatedness is a defining characteristic—perhaps the defining characteristic—of intelligent behavior. Hypertext embodies this idea, for everything in hypertext depends upon linkage, upon connectivity between and among the various elements in the system. Linkage, in hypertext, plays a role corresponding to that of sequence in conventional text. A hypertext link is the electronic representation of a perceived relationship between two pieces of material, which become nodes once the relationship has been instantiated electronically. That is, the link simulates the connections in the mind of the author or reader; and it is precisely because the electronic link is only a simulation that problems will arise.

The interdependency of links and nodes is such that it is impossible to talk about one without talking about the other. Thus the question of how to define a node (block is an alternative term in the literature on hypertext) leads to additional questions about linkage. These lead in turn to questions about structure and coherence, and so back again to the issue of prediction.

A node is any object which is linked to another object. It may be as large as an entire book or as small as a single character (theoretically, it could be as small as a single pixel [picture element] on a display), though such extremes seem hardly practical; it may consist of a document or a block of text within a larger document; it may be a drawing, a digitized photograph, a (digitized) detail of a painting, a sound recording, a motion picture, or a scene—even a frame—from a motion picture.

There is no set answer to the question, how big should a node be? just as there is no set answer to the question, how long is a paragraph? Like the paragraph, the hypertext node is a way of structuring attention, and its boundaries, like those of the paragraph, are somewhat arbitrary. A node may contain a single paragraph; it may contain many; it may contain something else entirely. Jeff Conklin offers criteria for guidance in determining node size. To what extent, he asks, is the information in question so tightly bound together that (a) you always want to view it together; (b) you never want to take it apart; and (c) you rarely even want to reference parts of it outside of the context of the rest. In other words, a node is an integrated and self-sufficient unit; its size will be a function of the complexity of the integration. This in turn is contingent upon the author's perception of the nature of the material it contains and the relation of that material to other things in the hyperdocument.

The individual node, then, behaves in certain respects like a more conventional text. But the node is not just a self-contained unit. A node cannot, by definition, be entirely free of links—a node is a knot, is always embedded in a system—and that connectedness in turn gives the node its definition. "A node is something through which other things pass, and which is created by their passage" [John M. Slatin, "Hypertext and the Teaching of Writing," in Edward Barrett's Text, Context, and Hypertext, 1988].

A text becomes a node when it is electronically placed in relation to other materials (documentary or otherwise), which may (or may not) already contain links to other elements within the system. The difficulty here, of course, is that what are to me self-evident associations may not be even faintly apparent to you, and vice versa. This imposes an obligation on the author(s) of a hyperdocument which has no exact parallel for the writer of conventional text: the nodes must seem complete in themselves, yet at the same time their relations to other nodes must be intelligible. The problem of relationality here is analogous to the problems of intertextuality confronting readers of, say 20th-century poetry. This problem becomes increasingly challenging as the hyperdocument expands. Links exist for many different reasons—that is, to represent many different kinds of relationships between objects; the more links there are between the current node and other elements of the hyperdocument, then, the greater the necessity of identifying the attached material clearly—especially when the reader is allowed considerable freedom in choosing among the available links.

These identifiers carry an enormous burden. Indeed they are often asked to do the kind of explanatory work that ordinarily takes several sentences or paragraphs. Not surprisingly, there are several different methods of identifying link or node types. Some hypertext systems, such as MCC's gIBIS, use "typed nodes," while others—Xerox PARC's NoteCards, for example—employ "link types." These systems allow the co-author creating a new node to choose from a list of predefined relationship categories, whose names then become part of the node or link.

HyperTIES, developed by Ben Shneiderman at the University of Maryland and distributed by Cognetics Corporation, offers a variation on this approach, by encouraging the author to compose a brief (two-line) description of the linked node. In HyperTIES, where reading is defined primarily as browsing and where browsing is completely separate from authoring, activating a link is a two-step process. The first step, clicking on a highlighted word or phrase, brings up the description of the attached node; the second step either brings up the attached node or returns the browser to the current screen.

The developers of Intermedia at Brown University have chosen a third alternative. Links belong to "webs" rather than to documents; webs are displayed onscreen as visual maps. An Intermedia document is first displayed as if it were freestanding; then, when the user opens a web, the links belonging to that web are displayed. This approach allows an individual node to be placed within multiple frames of reference, or "webs," while avoiding both the screen clutter and the mental clutter that can accrue so easily when multiple links radiate to and from a node. Which links the user sees will depend upon which web s/he has elected to open.

As the Intermedia approach suggests, one way to address the question of how many links a node should have is by turning it into the question of how many links should be displayed at any moment. Research on memory suggests that we can hang on to between five and seven "chunks" of information at a time, and that creating links between these chunks is a way to increase the effective size of the chunk. However, the number of chunks that can be retained decreases in inverse proportion to their size.

Probably no single method of identifying nodes and their relationships to other nodes is adequate to all needs; some combination will be needed. Choosing from a list of predefined relationships has certain advantages, since there is a strong likelihood that a newly created node will fall within an existing classification. However, the list must offer real choices without becoming so big as to make choice impossible. And co-authors will still want to be able to define and use new categories. Co-authors may also wish to give a fuller description of the attached node than the list approach permits; something like the HyperTIES strategy becomes appropriate here. And when a given node has multiple links to and from other nodes, it may be advisable to use an Intermedia-style mapping strategy.

The approach you choose to the problem of identifying links and nodes will depend on several factors: your understanding of the ways in which the material is related: your sense of who your readers are (are they primarily browsers? users? co-authors?). Your sense of what you want those readers to do is especially important.

You don't have to worry about interactive readers when you write a conventional text—the only thing you want the reader to do is to go on to the next sentence. But in hypertext, where there are a number of possible "next sentences" or nodes for the reader to go on to, you do have to make some decisions about what ought to happen next. That is, do you care whether the reader (a) opens a specific node or sequence of nodes; (b) chooses more or less randomly from the available links; (c) creates a new node, linking it not only to the current node but also to such other nodes as the reader—now a co-author—deems appropriate?

If you want the reader to open a specific node or sequence of nodes, you can either try to influence the reader's course of action, for example by highlighting a "preferred pathway" through the material, or you can simply pre-empt the reader's choice by automating the sequence or hiding links you don't want the reader to pursue. (Though if you take that route you give up many of the advantages of hypertext, it seems to me.) Or if you don't have a preference about the sequence the reader follows, you may opt not to give directions, leaving the choice of which links to activate—or whether to activate any link at all—entirely up to the reader. If you want to encourage response—that is if you want the reader actually to get involved as a co-author—you should say so somewhere on the screen and make it as easy as possible for the reader to change roles. In HyperCard, for instance, you can script a button to open up a text field with a date/time stamp and the name of the new co-author; you can even let the co-author enter keywords to make later searches faster.

I've already said that the author of a hyperdocument has a hard time trying to predict where the reader will go from any given point. The reader who activates a link often has a hard time, too, because it can be so difficult to predict what the result will be. The more cryptic the link or node identifiers are, the harder it is for the reader to predict the results of activating a particular link. The harder it is to make such predictions, the greater the likelihood that the reader will simply opt out of the process in frustration. And even if the reader does go ahead there is no guarantee that s/he will know the place when s/he gets there.

The reader has to make several different kinds of predictions. First, s/he has to predict the kind of material s/he will encounter upon activating a link. It would be quite distressing, for example, to activate a link in the expectation of moving to a narrative explanation of some issue, if in fact the associated node contains only raw data in tabular form. Second, s/he makes some predictions about the content of the node: at the most general level, the reader makes some kind of assumption about the closeness of the relationship between the current node and the material linked to it.

The questions of link and node labeling which are obviously of such central importance here impinge on the issue of predictability at what I have been calling the macroscopic level. At the microscopic level, predictability revolves around such things as screen-design: typography, visual effects, layout of information on the screen, and so forth, all have an impact on the reader's ability to organize the material in his or her own mind, and thus on his or her ability to operate effectively within the hypertext environment.

Typographical conventions are such that the reader of a conventional text has a pretty good idea what kind of object will appear next. The period at the end of a sentence leads one to expect a capital letter and the beginning of a new object belonging to the same class—a new sentence; the blank space at the end of a paragraph signals the beginning of a new paragraph; and so on. The signals in hypertext systems aren't nearly so clear.

Because the technology isn't mature enough yet to support a single set of conventions, each hypertext system has to develop its own conventions. For that reason, it is probably necessary to incorporate procedural discussions about these conventions into the hyperdocument itself; thus one element of the hyperdocument will be an ongoing critique of its own procedures. Participants might consider, for example, whether to assign specific fonts to individual co-authors so that their contributions can be readily identified; text styles might be assigned to specific node or link types; a question might always be italicized, for instance, while an explanation might be underlined; and so on. Color can be used for similar purposes in systems where color is available. So can visual effects such as wipes, dissolves, and zooms, which are available in HyperCard. These and other special effects can easily become distracting or even annoying, but if particular devices are consistently and intelligently used in association with particular node types, they can also function as more or less subliminal aids to prediction, helping the reader to perceive the hyperdocument as coherent.

We regard a conventional text as coherent to the extent that all the material it contains strikes us being (a) related in an appropriately direct way to the subject and to the author's thesis; and (b) arranged in the appropriate sequence. The perception of coherence in hypertext seems to me much more problematic, however, though I don't have time to do more than suggest what might be involved. Nor do I know enough to do more than that.

I think of hypertext coherence as appearing at the metatextual level—that is, at the level where the reader perceives what Gregory Bateson calls "the pattern which connects." The "pattern which connects" is the organizing notion around which all the disparate elements of the hyperdocument revolve. (An author's feel for "the pattern which connects" plays a significant part in decisions about node size and linkage as well.) This can be a relatively straightforward thing—a given hyperdocument might contain all the materials generated during a particular design project, for instance. This metatextual level is perhaps best represented by a visual map of some kind, whose nodes would open up to map subordinate patterns. This map ought to be readily accessible from any point in the hyperdocument, which suggests that it might be "iconized" and placed at a consistent screen location. This sounds simple enough, perhaps; but it becomes problematic again when we remember that we're dealing with a fluid system and multiple participants, and we start to ask whose understanding such maps represent. Maybe there needs to be a facility to allow any user to create such a map, whether for private consideration only or for public use might be up to the reader/co-author.

Conceptually, hypertext has a place, I think, in any environment where it's necessary or desirable to bring together large, complex, highly diversified bodies of information in such a way as to emphasize their interconnectedness—especially if physical space is at a premium, as of course it is on board a space station or an orbital device, or in a control room—or, for that matter, in a classroom.

Perhaps the greatest value of hypertext is in its ability to link enormous quantities of material that, in a conventional text environment, would be kept separate, perhaps even in different buildings, so that things which someone perceives as being related do in fact become related. Hypertext is weakest when it comes to spelling out what these relationships entail; it is important to say this, because the techniques for explanation are quite highly developed within traditional rhetoric, and it would be a mistake to abandon them as outmoded.

Hypertext places different demands on both readers and authors than those facing readers and authors of conventional text. The principal reason for this, in my view, is that hypertext is truly a new medium. Employing the full resources of technology to represent and correlate information, hypertext grants both readers and authors an unprecedented degree of freedom to arrange materials as they deem best, and it permits interaction between readers and authors to an unprecedented degree. In so transforming the methods of organization which have served traditional text for millennia, hypertext requires authors and system designers to find new methods of indicating relationships, representing and constructing knowledge, and achieving coherence.

Jay David Bolter (essay date 1991)

SOURCE: "Interactive Fiction," in Writing Space: The Computer, Hypertext, and the History of Writing, Lawrence Erlbaum Associates, 1991, pp. 121-46.

[In the following excerpt, in which he discusses Michael Joyce's interactive text "Afternoon" (1987) at length, Bolter places interactive fiction within the context of Modernism and suggests possibilities that the genre offers to authors and readers.]

Bibliographic databases and technical documents have long been regarded as legitimate texts for the computer: novels, short stories, and poems have not. It is true that many novelists now use word processors to prepare manuscripts for publication as printed books. But very few have attempted to write fiction to be read in the electronic space—that is, nonlinear fiction, which invites the reader to conduct a dialogue with the text. Yet fiction belongs just as naturally in this new technology as all the more pragmatic forms of writing. Fiction, at least modern fiction, is by nature open to experiment, and being open and open-ended is precisely the quality that the computer lends to all writing. The flexibility of electronic text makes for a new form of imaginative writing that has already been named "interactive fiction."

We are just emerging from the nickelodeon era of interactive fiction, and the computer's equivalent of the nickelodeon is the adventure game (which costs rather more than a nickel). In an adventure game the player has a mythical world to explore—a dungeon or an enchanted forest or valley. The computer describes the scene, and the player issues simple commands such as "go ahead," "enter the room," "pick up the dagger," "get gold," and the like. The goal is to amass treasure and dispatch monsters, although sometimes the game is more sophisticated, casting the player in the role of a detective who must solve a murder or other mystery. Even the simplest of these games is a kind of fiction. The computer presents the player with a text, and the player's job is to understand and respond to that text. Depending upon the response, the computer presents more text and awaits a further response. The player, then, is a reader, but an unusually powerful reader, for his or her decisions determine what text will next appear. Admittedly the text of most of the current games is simple-minded, but the method of presentation is not. While a printed novel presents its episodes in one order, the electronic writing space removes that restriction for fiction as it does for the essay. Instead of a single string of paragraphs, the author lays out a textual space within which his or her fiction operates. The reader joins in actively constructing the text by selecting a particular order of episodes at the time of reading. Within each episode, the reader is still compelled to read what the author has written. But the movement between episodes is determined by the responses of the reader, his or her interactions with or intrusions into the text, and the reader's experience of the fiction depends upon these interactions.

In its simplest form, interactive fiction requires only [two elements]: episodes (topics) and decision points (links) between episodes. The episodes may be paragraphs of prose or poetry, they may include bit-mapped graphics as well, and they may be of any length. Their length will establish the rhythm of the story—how long the reader remains a conventional reader before he or she is called on to participate in the selection of the next episode. At the end of each episode, the author inserts a set of links to other episodes together with a procedure for choosing which link to follow. Each link may require a different response from the reader or a different condition in the computer system. The reader may answer a question posed in the text; there will be one link for each possible reader response. The computer can also keep track of the previous episodes readers have visited, so that they may be barred from visiting one episode before they visit another. Many other tests are possible, but even with the simple matching technique and the tracking of previously visited episodes, the author can create a fictional space of great flexibility. Readers may be allowed to examine a story in chronological order, in reverse chronology, or in a complicated sequence of flashbacks and returns. They may follow one character through the story, and then return to follow another. A reader might play the role of the detective trying to solve a murder, a role familiar from the computerized adventure games. A reader might be asked to influence events in a novel by choosing episodes that promise to bring two characters together or to punish an evil character for his or her deeds: each choice would define a new course for the story. Such multiple plots, however, are only one possibility for interactive fiction. The electronic writing space can accommodate many other literary strategies. It could offer the reader several different perspectives on a fixed set of events. In this case the reader would not be able to affect the course of the story, but the reader could switch back and forth among narrators, each with his or her own point of view. An electronic text could also establish relationships among episodes that are not narrative at all: a poet could define multiple reading orders for an anthology of his or her poems—according to theme, image, time of the year, or other criteria under the poet's or the reader's control.

It is important to realize that electronic fiction in this sense is not automatic fiction. The computer does not create the verbal text: it presents that text to the reader according to the author's preconditions. The locus of creativity remains with the author and the reader, although the balance between the two has shifted. Nor is electronic fiction necessarily random. The author may put any number of restrictions on the reading order. The extent of the reader's choices and therefore the reader's freedom in examining the literary space depends upon the links that the author creates between episodes. The reader may have to choose from among a few alternatives or may range widely through the work. Each author can relinquish as much or as little control as he or she chooses; the author has a new literary dimension with which to work.

What the author and reader can do with this literary dimension is shown in "Afternoon" by Michael Joyce (1987), one of the first examples of this new genre of interactive fiction. "Afternoon" combines the literary sophistication of a printed work with the immediacy of a computerized adventure game. "Afternoon" is a fiction and a game at the same time, and yet its visual structure is very simple. The reader confronts a window on the computer screen: episodes of "Afternoon," containing from one to a few hundred words, will appear successively in the window. At the bottom of the screen is a small bar, where the reader types replies in order to move to the next episode; the reader may also initiate movement by selecting a word from the current episode in the window. All the text of the episodes was written by Michael Joyce, but the particular order in which the episodes are visited is determined at the time of reading.

"Afternoon" begins:

I try to recall winter. 'As if it were yesterday?' she says, but I do not signify one way or another.

By five the sun sets and the afternoon melt freezes again across the black top into crystal octopi and palms of ice—rivers and continents beset by fear, and we walk out to the car, the snow moaning beneath our boots and the oaks exploding in series along the fenceline on the horizon, the shrapnel settling like relics, the echoing thundering off far ice.

This was the essence of wood, these fragments say. And this darkness is air. 'Poetry' she says, without emotion, one way or another.

Do you want to hear about it?

If the reader types yes, then a link will flash another episode on the screen, whose first sentences are:

She had been a client of Wert's wife for some time. Nothing serious, nothing awful, merely general unhappiness and the need of a woman so strong to have friends….

If the reader types no, another episode will appear beginning:

I understand how you feel. Nothing is more empty than heat. Seen so starkly the world holds wonder only in the expanses of clover where the bees work….

"Afternoon" does not accept no as the reader's last word: instead, it moves the reader along a different path. But the reader has other choices as well: he or she can select a particular word directly from the text (such as "poetry," "winter," or "yesterday" in the first episode). Certain words in each window will "yield" and branch to another episode. If the reader defaults by hitting the Return key, then yet another link will be followed. Many different responses will cause the text to move, but until the reader responds with some action of the keyboard or the mouse, the text of an episode remains on the screen, conventional prose (occasionally poetry) to be read in the conventional way.

One voice in a later episode in "Afternoon" describes the reading experience this way:

In my mind the story, as it has formed, takes on margins. Each margin will yield to the impatient, or wary, reader. You can answer yes at the beginning and page through on a wave of Returns, or page through directly—again using Returns—without that first interaction.

These are not versions, but the story itself in long lines. Otherwise, however, the center is all—Thoreau or Brer Rabbit, each preferred the bramble. I've discovered more there too, and the real interaction, if that is possible, is in pursuit of texture….

This is the great difference between "Afternoon" and a fiction written on and for paper. There is no single story of which each reading is a version, because each reading determines the story as it goes. We could say that there is no story at all; there are only readings. Or if we say that the story of "Afternoon" is the sum of all its readings, then we must understand the story as a structure that can embrace contradictory outcomes. Each reading is a different turning within a universe of paths set up by the author. Reading "Afternoon" several times is like exploring a vast house or castle. The reader proceeds often down the same corridors and through familiar rooms. But often too the reader comes upon a new hallway not previously explored or finds a previously locked door suddenly giving way to the touch. Gradually, the reader pushes back the margins of this electronics space—as in a computer game in which the descent down a stairway reveals a whole new level of the dungeon. "Afternoon" is constructed so as to remind the reader of the origins of electronic fiction in the computerized adventure games.

"Afternoon" is metaphorically related to computer video games as well. It is like a video game in which the player pilots a spaceship around planetary obstacles that each exert a gravitational force. Readers of "Afternoon" move along paths with their own inertia, while at the same time experiencing the attraction of various parts of the fiction as they move by. It is easy to fall into an orbit around one set of episodes, one focus of the story. Readers find themselves returning to the same episodes again and again, trying to break free by giving different answers and therefore choosing different paths. When it succeeds, this strategy may then push the reader into another orbit.

The planets in "Afternoon"'s solar system are characters and key events. The characters include a poet named Peter who is working as a technical writer for a software company; his employer Wert, Wert's wife Lolly, and another employee in the company with the unlikely classical name of Nausicaa. Peter's is the narrative voice through most of the episodes, but the reader may follow paths in which each of the two women also narrate. Various paths concern the history of these characters, and the reader can find himself or herself parked in an orbit around one of them. The events in "Afternoon" may be significant in the traditional narrative sense: one is an automobile accident which may have killed Peter's former wife and his son. But one of the most gravitationally powerful events is commonplace: Peter has lunch with his employer. What makes the event important is that it is a structural crossroad: the intersection of many narrative paths. Peter's lunch may be the occasion for him to think about Wert's wife, or about his own affair with Nausicaa, or about the crazy computer project on which he is engaged. The significance of the lunch episode depends upon where the reader has been before and where the reader goes next. In some readings "Afternoon" is a picaresque novel in which the characters seem to stay put while the reader wanders.

Sometimes there is only one path leading from a given episode; sometimes there are several. The author's control of the narrative is inversely proportional to the number of paths and the kinds of responses expected of the reader. A single path gives the author the same degree of control as a printed book—in fact, even more control, because the reader cannot flip pages or turn to the back of the book; the reader must simply hit return and go on. An episode with many paths offers the reader the opportunity to head in any of several directions, although the reader may only be aware of this freedom after he or she has returned to the episode many times. And even after visiting all the episodes, the reader has still not exhausted the writing space. The significance of the episodes changes depending upon the order of reading. At one point several characters are invited to tell about themselves. If they do this early in the reading, then their subsequent words and deeds will be measured against this history. Nausicaa, for example, who seems a benign presence, turns out to have been a drug addict and prostitute. If the reader comes upon the self-revelatory episodes late in the reading, then these episodes must be read as explanation and justification of what went before. In either case, we find ourselves invoking familiar literary structures in the effort to make sense of this electronic fiction. The difference is that both (contradictory) structures coexist in the same electronic space, and, at least after several readings, the reader becomes aware of this double presence, so that he or she must reflect on the difference between the two possible readings.

The presence or absence of paths changes the feel of a particular part of the fiction. Wherever the choices narrow to a single path, "Afternoon" becomes a conventional story: it imitates the narrow space of the printed book before broadening again into its "natural" electronic space. The capacity to imitate the printed book is one important way in which "Afternoon" makes its comment on the nature of reading. Each of its paths seems to be a mixture of the accidental and the inexorable. Paths can have narrative force, as when Peter tries to find out whether his wife and son have been in an automobile accident. Following this trajectory makes the story quite linear. Peter fights his way through various telephone calls, in an effort to get information, at the same time reticent to make the most direct calls, that is, to take the most direct path to the information he anticipates. As readers we must do the same: we must be careful about our answers at the end of each episode in order to stay with this narrative strand. A wrong answer will shunt us to another path from which it may be difficult to return to Peter's frantic calls. This is a perfect example of the way in which electronic fiction can take into itself the experience of reading. The need of the reader to struggle with the story mirrors the struggle that the character goes through. "Afternoon" becomes an allegory of the act of reading. The struggle for meaning is enacted by the characters in "Afternoon," which is by now a conventional theme of 20th-century literature. What is new is that the allegory is played out by the reader as he or she reads. "Afternoon" becomes the reader's story in this remarkable way. Readers experience the story as they read: their actions in calling forth the story, their desire to make the story happen and to make sense of what happens, are inevitably reflected in the story itself.

"Afternoon" is about the problem of its own reading. This will certainly be the problem that confronts us with all interactive fiction, at least in the early days of this new form. As an example of topographic writing, "Afternoon" asks us to consider how we are to appreciate the multiplicity of such writing. The writing space of "Afternoon" can be represented by a diagram with squares for the episodes and arrows for the links between episodes. The whole diagram is vast, since "Afternoon" has over 500 episodes and over 900 connections….

The reader never sees this diagrammed structure: the reader's experience of "Afternoon" is one-dimensional, as he or she follows paths from one episode to another. Instead, the reader must gain an intuition of the spatial structure as he or she proceeds in time. This task is rather like that of a mathematician who attempts to envision a four-dimensional object by looking at several projections in three dimensions. Each projection is a snapshot, and the snapshots must be synthesized to win a sense of the whole. For the reader of "Afternoon," each reading is one projection of the geometry of the whole; the whole is the sum of all the possible ways in which "Afternoon" may be read.

The geometry of electronic fiction need not be defined solely in terms of the plot. In "Afternoon," the important events seldom vary on different readings. Instead, it is the characters' reactions and interactions that vary. The electronic writer can exploit other organizing principles of modern printed fiction, such as the stream of consciousness of one character or the points of view of several characters. Linked rings of episodes would be particularly effective for presenting multiple points of view. "Afternoon" has what we might call a subspace in which Lolly and Nausicaa tell their stories, each in several episodes and each ending more or less where she began. There is also the possibility of narrating the same events from different points of view, a technique familiar from Faulkner's The Sound and the Fury and Durrell's Alexandria Quartet. In the electronic medium the reader can be given more degrees of freedom than is possible in print; he or she may be allowed to flip back and forth among episodes, comparing one narrator's version against the others'.

The author might choose to build a narrative hierarchy by presenting the same events, for example, from three points of view, where the first point of view is in some sense superior to the second and the second to the third. The author might think of these as the divine, the heroic, and the satiric perspectives. From the satiric perspective, events are a confusion, as we normally experience them, making only a provisional sense that dissolves again into chaos. Imagine the dramatic story of Oedipus, told from the perspective of the shepherd who failed to obey the order to kill baby Oedipus and instead gave him away. After many years, this well-meaning shepherd is brought before Oedipus the king, threatened, and made to reveal what he did with the baby. He is then released and left to ponder the horror of Oedipus' crime and his own revelation. From the heroic perspective, events take on greater clarity and urgency, as Oedipus himself would tell the story. The third, divine perspective is omniscient and also detached, for the heroic sense of engagement is lost. It is the story of Oedipus as narrated by Apollo….

Each of the levels contains six episodes from one perspective, corresponding to the scenes of the play. The reader begins on the lower (the satiric) level, and the task is to break into a higher realm of understanding…. Thus, as the reader is visiting one episode on the satiric plane, he or she may succeed in jumping to the heroic plane and instantly see the same event in a new light. The divine level should clearly be harder to reach than the heroic, and yet its presentation of events might be so cold and crystalline that the reader may wonder whether attaining this level was in fact worth the effort. This geometry omits the choral odes which divide the scenes of Sophocles' play. We could include these and have them serve as foci where the change of point of view might occur. The odes, like operatic arias, often take place in a time that is outside the conventional time of the dramatic action, pulling the reader or listener away from the plot to reflect upon the significance of past and possible future action. They provide a perfect moment for a shift in point of view.

In this Sophoclean structure, each level is in the same narrative mode: each tells a story in the first person. We can also imagine a structure in which each level contains a different mode of discourse—prose, drama, narrative poetry, lyric poetry, and so on. There is no need to limit the modes to fiction: the same experience could also be treated in historical prose and in scientific prose with mathematics. Each level would then be presenting a different aspect of written reality. The fictive level describes a man seated at a mahogany desk writing a letter; the poetic describes the scratch of the quill pen upon paper and the sound of the sand used to blot the ink; the historical discusses literacy in Victorian England; and the scientific explains how ink disperses into the pores of the paper. In the electronic organization the author can refract reality into a series of such perspectives without destroying the rhythm or comprehensibility of the text. Readers do not have to contend with all facets of the event at once; instead, the order in which they examine the various facets determines their experience of the text.

The preceding geometries are suggestions; many others are possible. They simply show how we might envision the spatial expression of a multiplicity of temporal experiences provided by any one electronic text. Electronic literature is not static, nor is it "timeless": it exists in "real time," as computer specialists call it, the time that the user passes in front of his or her terminal. The fiction realizes itself on the computer screen and then rolls or blinks out of sight and is gone. The reader might produce the same sequence of screens on the next day, but for any large structure of episodes, exact repetition is unlikely. As readers develop an aesthetic sense for this new medium, they may no longer care about perfect repetition or long preservation.

On the other hand, repetition and return take on new significance in interactive fiction, because in its own way the computer is capable of endless repetition. If the writer is clever, he or she can arrange it so that the reader seldom duplicates a previous fiction, but can always have something like it again. We may judge the success of a work by its ability to adapt to new readings and yet preserve its essence. It may be with readings as with dreams: important ones may not return, but they are never entirely lost either. Conversely we may judge works by their ability to defy categories—to have no essence and so to be different for each reading. In no case can the author in the electronic medium claim to have erected a monument more lasting than bronze; the author must instead delight in writing, like Archimedes, in sand (or silicon).

Playfulness is a defining quality of this new medium. Electronic literature will remain a game, just as all computer programming is a game. We have seen that such literature grows out of the computer games that are popular today.

In video games, the kind depicting spacecraft and deadly robots, the player competes against the programmer, who has defined goals and put obstacles in the player's path. The emphasis is on the player, not the programmer, who is an anonymous employee of some software firm. Even in the simplest form of electronic fiction described above, the author/programmer is as important as the reader/player, and the relationship between author and reader may take a variety of forms: they may work cooperatively or competitively. No matter how competitive, the experience of reading in the electronic medium remains a game, rather than a combat, in the sense that it has no finality. The reader may win one day and lose the next. The computer erases the program and offers the reader a fresh start—all wounds healed. Anyone who has written a program knows that the computer has a genius for getting completely and hilariously off track. Fortunately, it is possible to drop everything and start over. This quality will carry over into electronic literature. Every reading of "Afternoon" can be a new afternoon, or the reader can choose to pick up the fiction where he or she left it in the last session. The impermanence of electronic literature cuts both ways: as there is no lasting success, there is also no failure that needs to last. By contrast, there is a solemnity at the center of printed literature—even comedy, romance, and satire—because of the immutability of the printed page.

Interactive fiction is both innovative and traditional. It is certainly new to automate the presentation of text, so that the reader's decisions are automatically registered and cause other words to appear. However, in disrupting the stability of the text, interactive fiction belongs in a tradition of experimental literature (if we may use this oxymoron) that has marked the 20th century—the era of modernism, futurism, Dada, surrealism, letterism, the nouveau roman, concrete poetry, and other movements of greater or lesser influence. The experiments of Dada, for example, were aimed at breaking down all structures of established art and literature, and in that breakdown the Dadaists worked in the same spirit as writers now work in the electronic medium. Jean Arp wrote that in his poems: "I tore apart sentences, words, syllables. I tried to break down the language into atoms, in order to approach the creative." Tristan Tzara proposed a poetics of destruction, when he gave this recipe for creating a Dada poem: "To make a dadaist poem. Take a newspaper. Take a pair of scissors. Choose an article as long as you are planning to make your poem. Cut out the article. Then cut out each of the words that make up this article and put them in a bag. Shake it gently. Then take out the scraps one after the other in the order in which they left the bag. Copy conscientiously. The poem will be like you…."

Dada was an early and influential example of the modern will to experiment. The modern attack was often aimed at the conventions of the realistic novel that told its story with a clear and cogent rhythm of events. In the course of their attack, modern authors found themselves straining at the limitations of the printed page. Because the linear-hierarchical presentation of the printed book was so well suited to the conventions of plot and characters of the realistic novel, to attack the form of the novel was also to attack the technology of print. The French led the way with the nouveau roman and Philip Sollers and the Tel Quel group. From France and elsewhere, we have had programmed novels and aleatory novels. All these were instances of subversion: they worked from within, attempting to undercut the conventions of printed literature while themselves remaining printed books.

Most of the important literature of the 20th century has been accused of subversion. The avant-garde movements like Dada were never so radical as they claimed to be; they were instead extending or perhaps caricaturing the mainstream. Joyce, Virginia Woolf, Pound, T.S. Eliot, and others all participated in the breakdown of traditions of narrative prose and poetry; breaking with such traditions was the definition of being modern. Pound and Eliot set about to replace the narrative element in poetry with fragmented anecdotes or mythical paradigms. Joyce and Woolf devised new ways of structuring their works based on stream of consciousness or on multiple layers of topical and mythical organization. All of these writers were trying to set up new relationships between the moment-by-moment experience of reading a text and our perception of the organizing and controlling structures of the text. The surprising fact is that topographic writing in the space provided by the computer is a natural extension of their work. Topographic writing redefines the tradition of modernism for a new medium. To put it another way, modern authors have already been writing topographically, but they have been using the medium of print, which is not well suited to that mode of writing.

The experimentation with and fragmentation of the form of the novel is certainly older than modernism: it dates back at least to the 18th century. The whole tradition of experimentation needs now to be reconsidered in the light of the electronic medium, since each previous experiment in print suggests ways in which writing may now break free of the influence of print. In another sense all novels embody a struggle between the linear flow of the narrative and the associative trains of thought touched off by the narrative, and the electronic medium provides a new perspective on that struggle. The history of the novel itself will need to be rewritten, so that we understand works by authors from Laurence Sterne to Jorge Luis Borges not only as explorations of the limits of the printed page but also as models for electronic writing. It is as if these authors had been waiting for the computer to free them from print. And in fact, many of their works could be transferred to the electronic writing space and playfully reconstructed there. From this perspective, interactive fiction is the inevitable next step, like the take-off of an airplane that has been gaining speed on the runway. As with the speeding airplane, the moment of take-off for interactive fiction is a matter of definition….

There are so many experimental novels and novelists today that any choice among them is arbitrary. The will to experiment even extends to children's fiction: books in the series entitled "Choose Your Own Adventure" or "Find Your Fate" offer young readers something like the computerized adventure game. Each book contains a garden of forking paths, from among which the reader must choose. At the bottom of page 18, one may read: "If you choose to risk the curse, turn to page 29. If you press on down the tunnel, turn to page 42." Thus, the book sacrifices the linear order of pages to allow for multiple reading. There are many of these programmed novels, some for adult readers. The novel Rayuela by the Latin American writer Cortázar offers two possible reading orders: one a linear narrative and the other a satirical comment on that narrative. Other writers who have experimented with fragmented narrative include Raymond Queneau, Italo Calvino, and Michel Butor.

One experiment seems to have achieved an anonymous notoriety: everyone seems to have heard of this work, although very few know its title or author. Marc Saporta's Composition No. 1 was published in France in 1962, and the English translation, a work consisting of about 150 un-numbered pages, appeared in the following year. This fiction consists of loose sheets of paper, each containing one or a few paragraphs printed on one side. The sheets (at least in the English version) are somewhat larger than octavo, larger than a deck of playing cards, to which the author compares them. In the introduction, which is printed on the box and also appears as one of the pages, the author explains:

The reader is requested to shuffle these pages like a deck of cards, to cut, if he likes, with his left hand, as at a fortuneteller's. The order the pages then assume will orient X's fate. For time and the order of events control a man's life more than the nature of such events…. A life is composed of many elements. But the number of possible compositions is infinite.

Composition No. 1 consists of passages in the story of X, a character we learn about only through his reflection in the events and other characters of the narrative. X, an unsavory figure, is married to Marianne, has a mistress named Dagmar, and also rapes a girl named Helga. He apparently steals to support the compulsions of his neurotic wife, and he possibly dies in an automobile accident. By way of redemption, he seems to have played some role in the French resistance, although this strand in the story is the most obscure.

It is Saporta's tour de force that the reader is able to figure out this much from the experience of reading the pages one at a time. Each page is necessarily only a vignette. Even for a contemporary reader, accustomed to fragmentation and suspension, the natural impulse is to turn the page in order to find out what happens next. But the next page bears no immediate relation to the previous action. This fiction is not static: on most pages there is action or at least dialogue that promises to advance our knowledge of the characters. We get the impression that we are reading pages torn from a conventional novel. We find ourselves searching for their "proper" order—that is, for an order that makes chronological and causal sense. We become literary detectives, since the pieces in this puzzle are topical elements in a written text. Connecting these topics reminds us of papyrology or other scholarly detective work with the fragmentary remains of ancient writings. It is always the papyrologist's hope that newly found fragments will turn out to be part of a known or, better still, unknown, but identifiable ancient work—that both the text and its context will be revealed. We feel that hope as we read Saporta. Indeed, to call the work "fragmented" is to assume that it was originally whole, that the fragments belong in one order.

As 20th-century readers, we are prepared for the focus to shift from one page of a fiction to another. But we want to find a reason for such shifts. Each time we shuffle the pages, some intriguing juxtapositions happen by chance. It is a characteristic of our literary imagination that we can so often provide an interpretation for these juxtapositions. Whatever order falls out is a lesson in the nature of reading. We become active readers, fashioning texts as we shuffle pages. Saporta's trick makes us uniquely aware of the effect of narrative order upon the reading experience, as we see in practice what we are always told in theory. There is not one possible order for the episodes but a broad class of acceptable orders, each producing its own literary effect. Some orders are better than a simple chronological story, and many are worse. The number of reading orders for Saporta's composition is not infinite, as the author claims. If there are 150 pages in the set, then the number of possible presentations is 150!—a number with 263 digits, but still finite. Composition No. 1 has necessary limits: as Sharon Spencer points out [in her Space, Time, and Structure in the Modern Novel, 1971], in every possible rearrangement of its pages we will still have a portrait of X, the silent, second character in each fragmentary scene.

Composition No. 1 is an exercise in choice within a large but still limited fictional universe…. Saporta's universe is one of French existential romanticism, fascinated by the interplay of chance and fate. Moreover, Saporta seems to be working in a different direction from the one we have been considering for topographic writing in the computer. The author of Composition No. 1 apparently disavows responsibility for the structure of his work. In seeming to deny form, he displaces the formal responsibility onto his reader. In fact, Composition No. 1 has form; otherwise, it could not remain recognizable as a work of fiction. There are loose formal relations that exist among the pages when read in any order—relations established, for example, by the comparison of two female characters. Moreover, each paragraph or page is itself a conventional formal structure. Saporta's style is characterized by precise images and rhetorical effects: on each page, not a single word seems out of place. The precision on the page is in marked contrast to the disorder that prevails between pages and requires a different strategy of interpretation. On the page we may assume that the author is in complete control, and we may examine every mark of punctuation for nuances of meaning. Moving between two pages, we can only ask ourselves more general questions: how, for example, did the author arrange the beginning and ending of each page to make them fit together smoothly in any order?

Saporta's experiment in chance fiction seems to be an inevitable step in the exhaustion of printed literature. When all the other methods of fragmenting the novel have been tried, what remains but to tear the pages out of the book one by one and hand them to the reader? From the ideal of perfect structural control, Saporta brings us to the abdication of control. But we can also see in Saporta's experiment, and in others like it, not only the end of printed fiction, but also a bridge to the literature of the electronic medium.

Composition No. 1 brings us back to "Afternoon," for Saporta's work is an "Afternoon" under the limitations that print imposes. Each page of Composition No. 1 is a topographic unit, like an episode of "Afternoon" as it appears on the screen. In both fictions, the burden of constructing the text is thrown back upon the reader. And in both cases the reader struggles to make narrative sense of the episodes as they present themselves: to construct from these disordered episodes a story in which characters act with reasonable and explicit motives. In Composition No. 1 we may resort to stacking the pages in piles. In "Afternoon" the medium itself resists that solution, since we cannot get at the episodes except by typing responses, and even then the episodes appear only one at a time.

If we put Composition No. 1 and "Afternoon" side by side, we see all the more clearly how our desire for a story has been fostered by thousands of years of reading in manuscript and print. When we read and write in these older technologies, we are compelled to narrow the possibilities into a single narrative. We have not yet learned to read and write "multiply," as Stuart Moulthrop has described it [in his "Text, Authority, and the Fiction of Forking Paths," unpublished manuscript], in part because our technologies of literacy have suggested just the opposite: that we must read and write linearly. To read multiply is to resist the temptation to close off possible courses of action; it is to keep open multiple explanations for the same event or character. It is an almost impossible task for the reader to remain open in a medium as perfected as that of print. Saporta must tear his novel apart in order to resist the perfection of print, and still our first impulse is to try to put the novel back together.

"Afternoon," on the other hand, does openly and with ease what experimental writers in print could do only with great difficulty. It offers a narrative that encompasses contradictory possibilities. In "Afternoon" an automobile accident both does and does not occur; the narrator does and does not lose his son; he does and does not have a love affair. The story itself does and does not end:

Closure is, as in any fiction, a suspect quality, although here it is made manifest. When the story no longer progresses, or when it cycles, or when you tire of the paths, the experience of reading it ends. Even so, there are likely to be more opportunities than you think there are at first. A word which doesn't yield the first time you read a section may take you elsewhere if you choose it when you encounter the section again; and sometimes what seems a loop, like memory, heads off again in another direction.

There is no simple way to say this.

There is no simple way to say this in the linear writing of print. But "Afternoon" embodies the idea of openness with utter simplicity. What is unnatural in print becomes natural in the electronic medium and will soon no longer need saying at all, because it can be shown.

All our topographic writers in print … are "difficult" writers, and the difficulty is that they challenge the reader to read multiply. They call to the reader's attention the painful contrast between the temporal flow of narrated events and the interruptions and reversals that the act of writing imposes upon those events. All their experimental works are self-consciously concerned with the problem of writing. That concern is shown by the difficult relationship between the narrator and text, between the text and its reader, or both. The difficulty in writing (and therefore in reading) appears with Tristram Shandy, who is always getting further behind experience as he writes, and again and again in the 20th century. Saporta and Cortázar must give special instructions to their readers. Finnegans Wake may well be the most demanding book that has ever been published and critically received. In each case, the printed fiction must work against its medium in order to be topographic. There is a conflict between the printed volume as a frame and the text that is enframed. The frame is not adequate to contain and delimit the text, which is constantly threatening to spill out of its container.

By contrast, the computer provides a frame that gives way as the text strains against it; the stubbornness of the printed book suddenly disappears. In "Afternoon" the margins "yield" to the reader. The elements of the text are no longer fragments of a prior whole, but instead form a space of shifting possibilities. In this shifting electronic space, writers will need a new concept of structure. In place of a closed and unitary structure, they must learn to conceive of their text as a structure of possible structures. The writer must practice a kind of second-order writing, creating coherent lines for the reader to discover without closing off the possibilities prematurely or arbitrarily. This writing of the second order will be the special contribution of the electronic medium to the history of literature.

The computer gives the reader the opportunity to touch the text itself, an opportunity never available in print, where the text lies on a plane inaccessible to the reader. Readers of a printed book can write over or deface the text, but they cannot write in it. In the electronic medium readers cannot avoid writing the text itself, since every choice they make is an act of writing. The author of "Afternoon" wrote the prose and poetry for each episode, and he fashioned electronic structures of expectation and fulfillment on analogy with the static structures of printed fiction. But in giving the reader a role in realizing those structures, he has also ceded some of his traditional responsibility as author. This sharing of responsibility points the way in which electronic writing will continue to develop: we can envision an electronic fiction in which the reader is invited to alter existing episodes and links and add new ones. In this way the reader becomes a second author, who can then hand the changed text over to other readers for the same treatment. Electronic fiction can operate anywhere along the spectrum from rigid control by the author to full collaboration between author and reader. The promise of this new medium is to explore all the ways in which the reader can participate in the making of the text. Stuart Moulthrop suggests that the reader should be "… invited not just to enter a garden of forking paths, but to expand and revise the ground plan at will. This would cause a substantial shift in the balance of authority, one with enormous implications for the idea of literature itself; but it would nonetheless be a logical development of writing in the electronic medium."

The reader's intervention may come at any level of the electronic text. We have been considering fixed episodes and their connections, but the reader could also intervene to change the text itself. Such intervention could begin with simple letter substitution and typographic changes, perhaps under constraints programmed by the author. Letter substitutions can change words into other words or into nonsense that may suggest several words at once. Typography can be altered to make the text visually more exciting. Such changes too have been prefigured by writers in print. Mallarmé with his spatial poems such as "Coup de dès" and Apollinaire with his calligrammes were followed by the Dadaists, by LeMaître, and more recently by the writers of "concrete" poetry, who deployed letters and words in defiance of the conventions of lines and strophes. Concrete poetry too was an expression of the growing dissatisfaction with the medium of print in the 20th century. Concrete poetry too belongs in the computer; indeed, the computer makes possible truly kinetic poetry, a poetry in which letters and words can dance across the screen before the reader's eyes.

The poet William Dickey has already begun to create interactive poems. In "Heresy: A Hyperpoem," words, images, and icons compete on the screen for the reader's attention. The poem is a network of many screens, and the reader moves from one screen to others by activating one of the snowflake-shaped icons. Each screen is a different arrangement of verbal text and image; each vigorously asserts its visual identity. The typography of this hypertext encourages the reader to examine and savor each screen before activating an icon to move on. Sometimes the verbal text penetrates the graphic image…. Sometimes the spatial arrangement of text against the white back-ground reminds us of the experiments of Mallarmé or LeMaître.

In "Heresy," as in "Afternoon," each textual unit is static; the reader sets the hypertext in motion by moving between fixed units. In future kinetic poems, however, the reader might also intervene to control the speed and direction or the way in which words coalesce and divide. He or she may participate in the reordering of the ultimate constituents of writing—letters, words, and graphic symbols. The reordering of such symbols is writing: it is all that any writer can ever do. This insight, which has been laboriously demonstrated and violently contested in the printed literature of the 20th century, becomes a simple and unavoidable fact in the electronic medium. Multiple reading passes inevitably into multiple writing.

Electronic fiction is technologically complicated in that it requires a computer and the sophisticated arrangement of text and graphics on a videoscreen. But it is conceptually simple—simpler than writing for print, where the writer must always force his or her text into a single line of argument or narrative. The computer frees the writer from the now tired artifice of linear writing, but the price of this new freedom for the writer is that the writer must allow the reader to intervene in the writing space.

Harry Goldstein (review date March-April 1994)

SOURCE: "The Changing Shape of Fiction," in Utne Reader, No. 62, March-April, 1994, pp. 131-32.

[Below, Goldstein comments on the new field of hyperfiction and reviews several works in the genre.]

Despite its dramatic name, hyperfiction—the latest craze in hightech literature—is basically software: You load it onto your computer and use certain tools to navigate through it. And what a ride it is.

In hyperfiction, you don't just read, you make choices. It's the classic existentialist dilemma manifested in the reading process. You're at least partially responsible for creating (by choice or actual writing) what you read, so you can never just snuggle up with your laptop and get lost in the story. Rather, you can get lost in the story, but it's your responsibility to find the way out, if indeed you want to get out, which is all up to you. There's no comfort in watching the pages between you and the end of the book evaporate in the sure knowledge that the end is in sight. And there's no such thing as reading ahead to the last page to find out who killed whom. There is no closure to a work of hyperfiction. When you stop reading, that's the end—of that particular reading of the fiction.

Some works, such as Quibbling or Hypertext Quarterly, give you a reading map of the hypertext—a sort of topographical representation of the blocks of text, or lexias. By clicking on links—highlighted words or phrases—you can zoom into another set of lexias (and characters and story-lines). Other hyperfictions allow you to type in a word, which may lead you anywhere. And almost all hyperfictions allow you to point the mouse, click on any word that might strike your fancy, and see where that takes you. Several hyperfictions indicate, by means of underlining, boxes, or boldface, certain words or phrases that have a link to another part of the hypertext.

Most readers, conditioned by novels and short stories, have traditional expectations when they first encounter hyperfiction: They want to know who did what, and why. Hyperfiction plays with these expectations, especially with the need to perceive an established order. You still read for plot, for characters and their motivations, and for the key to the mystery, but hyperfiction demands a spatial rather than a linear reading. Each lexia consists of several surfaces that interact differently with other text blocks and links, resulting in an almost boundless web of interconnections. Not only is there room for a vast range of interpretations, but hyperfiction actively encourages multiplex reactions and interactions.

Hyperfiction constitutes a radically new genre, quite distinct from the so-called expanded novels (published by Voyager)—novels adapted for computers and souped up with hypertext components, but which you read more or less as you would a book between covers. Hyperfiction empowers readers and overturns traditional notions of author and reader. Stuart Moulthrop, hypertext critic and author of Victory Garden, distinguishes between two kinds of hyperfiction—exploratory and constructive—in his essay "Polymers, Paranoia, and the Rhetoric of Hypertext" in the spring 1991 issue of Writing on the Edge. According to Moulthrop, an exploratory hyperfiction—which includes most commercially available hypertexts—allows the reader to "'transform' the textual body by following alternative paths or linkways" while the hyperfiction "retains its fundamental identity under all transformations…. The maze may have many permutations, the circuit many switchings, but in all of them the user still circulates through the same mechanized volume." Constructive hyperfiction, which arrived in the commercial sphere with the November 1993 publication of Deena Larsen's Marble Springs, allows you to add your own words to the hypertext and send your additions back to the publisher to be included in subsequent editions. Marble Springs points to the future of constructive hyperfiction, which lies in the proliferation of computer networks where many reader/writers will collaborate on projects made open-ended by virtue of their virtual existence as malleable data in cyberspace.

Hyperfiction addresses a fundamental issue that faces all writers: In a society where people are more willing to consume images than read, hypertext seems like a wordsmith's dream. Finally, here's vehicle for writing that is as fast as the microprocessor that paints the words on the screens, the words taking on a visual quality that can compete with pictures. Ultimately, however, hyperfictions must still succeed first as writing: Are the characters worth following through the labyrinth of a hyperfiction? Is the narrative structure engaging, its complexity challenging or its simplicity aesthetically pleasing? Is the language fresh? Most importantly, does the writer have something new to say in addition to having a new way to say it?

The hyperfictions reviewed here run on an Apple Macintosh Plus and more powerful models and are available for IBM compatibles in Windows format except where noted. Although they will run on 1 MB of RAM, they run slowly, so 2 MB is suggested, and required where indicated.

"Afternoon, a story" by Michael Joyce. One of the very first hypertext fictions and still one of the most fascinating, "Afternoon, a story" doesn't have any mapping graphics that allow you to see where you are in the whole scheme of things, but is so engrossing that you can get lost in it and not care. The lyrical intensity of the language lures you through a structure that is simultaneously compelling and perplexing.

"Afternoon" affords a remarkable amount of freedom. If you find yourself pursuing a link that doesn't interest you and remember seeing one a few screens back that looked more promising, you can backtrack to find it. A cascade of returns can be very entertaining, the words dancing on the screen, revealing the hyperfiction not as printed matter but as something more ephemeral. And when you turn off the computer, you might find that the voices of "Afternoon" linger in your ear, speaking of the intrigue that you sensed rippling just beneath the surface, waiting to be uncovered the next time you venture into the fiction.

Ambulance: An Electronic Novel by Monica Moran. Produced by the prolific Jaime Levy, Ambulance is the latest in a series of cyberzines published by Levy's New York-based Electronic Hollywood. Monica Moran's very linear narrative tells the story of a group of twentysomethings who crash their car somewhere in the Hollywood hills. A serial killer spies the accident from his house and decides to collect our band of hapless travelers and take them back to his house, where grim mayhem ensues. Ambulance combines Levy's cinematic sensibility with hypertextual flashbacks and some engaging multimedia effects, including bass samples by alternative rocker Mike Watt and illustrations by Jaime Hernandez (of the Love and Rockets comic book series). By turns amusing and grisly, Ambulance requires less work on the part of the reader than hyperfictions more consciously concerned with putting into practice the theories of Derrida and Barthes and the fictional visions of Borges and Pynchon.

The Madness of Roland by Greg Roach, with animation by David Crossley. The most striking things about Roland are its animation, videos, sound, look—in short, everything but the text of this multimedia extravaganza whose list of credits resembles that of a movie production. As hyperfiction becomes more geared toward a primarily imageliterate audience, collaborative hypermedia efforts of this sort will undoubtedly increase. Unfortunately, Roland points to one of the pitfalls: Mediocre writing will never be able to compete with wow-'em special effects. Requires CD-ROM player for Macintosh or MPC machine and Video for Windows for IBM compatibles.

Marble Springs by Deena Larsen. The first to invite the reader to collaborate on subsequent editions by writing characters and creating links, this hyperfiction is very much an Our Town brought to hypertext. You enter the work via the Marble Springs cemetery to discover a ghost town where women have woven the fabric of the town's history. In addition to the usual assortment of text-based links, the map of Marble Springs provides one way of getting into the fiction; you choose certain geographical locations and discover—or write—the lives of the inhabitants. This is easily the most demanding hypertext on the market in the range of choices the reader is given, from perusing actual bibliographies detailing the research that went into reconstructing an Old West town to collaborating with Deena Larsen and other readers. Requires HyperCard 2.0 and 2 MB of RAM; available for Macintosh only.

Victory Garden by Stuart Moulthrop. Victory Garden is hyperfiction about the Gulf War by noted Thomas Pynchon critic Stuart Moulthrop. One of the most complex hypertexts now available, with an astonishing number of lexias and links, Victory Garden explores reactions to the war in letters and scenes from Saudi Arabia featuring college student and reservist Emily Runbird, media coverage of the war, military technobabble, and protests at the college Emily attended before she was shipped off. Moulthrop has taken a historical event and given it back to us through almost 1,000 blocks of text and 2,800 links; careful characterizations; numerous storylines, plots, and subplots; and an incisive critique of news media, the war, and the war's media audience. The Gulf War was presented to the public in fragments. Moulthrop has used hyperfiction to link real and imagined fragments, allowing the reader to do what the media and the government made impossible: to explore the story of the war, and of ourselves, for ourselves.