Chapter 8: Selfish Interaction or Subversive Texts and the Multiple Novel

Interactive fictions are largely figments of our imaginations.

I am not sure what to make of this sentence. It has aspects of an unrealized pun, a bad koan, a polemical claim, and a manifesto. I want to doubt it, but I believe it utterly. Though we can point to an Infocom game here, and a Source serial there, when we are honest with ourselves we know that no truly interactive fictions exist.

Partly, of course, this is because, among the things we are likely to think sloppily about, interactivity ranks right next to expert systems and natural language parsers. There are grounds for arguing that no truly interactive system of any sort exists-except perhaps implanted pacemakers and defibrillators-since true interaction implies that the user responds to the system at least as often as the system responds to the user, and, more importantly, that initiatives taken by either user or system alter the behavior of the other. Video games from the glory days of the Atari renaissance, or Flight Simulator and its clones, or certain operating systems offer us at least a glimpse of what interaction might be. However, try as we may to believe they truly interact, we see them branching off below us at some transparent level of the successive planes of software and hardware. We know, as certainly as we know Eliza (or her erstwhile offspring Mom and Murray) that someone has been there before us.

And so we try to outguess her (or him). Ironically, however, in this process of attempting to accommodate our thinking in response to the demands of an application’s control and data structures, true interactivity does, of course, exist. In this sense, text editors or databases could be said to be among the most successful interactive fictions, especially during the early stages of the learning curve as we come to use them. For during that time we convince ourselves that we know the story of our own thought at least as often as the application reminds us that we do not know its representation.

Likewise, we imagine we give structure to a formless conceptual space, only to discover that the space itself is a labyrinth of glass walls within which we unravel skeins of our thought in order to find our way. An error message or a dialog box at such times becomes an utterance from an offstage demon. We accommodate our thought to the system, and the system accommodates our thought – we interact.

This paper is an attempt to explore, along these lines, the forms that true interactive fictions might take in the corning years. It proceeds from nearly three years of research (with J. David Bolter) into developing StorySpace, a structure editor for creating interactive fictions (the development of which was supported in part by a grant from The John and Mary H. Markle Foundation), but it is also the result of sixteen years of writing traditional novels which were nonetheless imagined as interactive fictions without the benefit of either appropriate tools to create them or a system to present them. (See Figure 8.1.)

My suggestion is that future interactive fictions, in order to be more open, will appear more closed, i.e., more like current printed fiction than the computer programs we currently consider interactive. The model here is, of course, Umberto Eco’s, wherein “an open text, however ‘open’ it be, cannot afford whatever interpretation.

An open text outlines a ‘closed’ project of its Model Reader as a component of its structural strategy.” [Eco, 1979, p. 9] As a consequence, I also argue that we are more likely to experience satisfying interactions through the play of mind than through playing within texts, no matter how theoretically compelling the latter activity may seem.

The first level of interaction precedes the creation of any text.

We live in a time when we are able to assemble more information than at any previous age in history. Even so, we sometimes seem to live as much in fear of fragmentation as in hope for coherence. Inquiry into how we process and transfer our knowledge to create coherent visions of ourselves and our worlds assumes increasing importance and is enhanced by our growing awareness that media themselves intertwine and interact and threaten to become more ether than pathway, more chaos than cosmos.

We know many things but are uncertain how to use what we know. We have tools but we are uncertain what tasks to put them to, since tools by their nature alter our vision of the tasks. This process is often recursive: tools alter tasks alters tools and so on.

For a fiction writer, the dynamic relationship between tools and tasks is a familiar reality. In fact, the most compelling aspect of computer tools is that they promise fiction writers a means to resurrect and entertain multiplicities that print-bound creation models have taught them to suppress or finesse. That is, where James Joyce’s Finnegan’s Wake, Julio Cortazar’s Hopscotch, or-for that matter-Laurence Sterne’s Tristram Shandy created multiplicities as intricate as any of those we envision for interactive fictions, those texts were bound at very least by the static nature of their presentational systems  Chapter 7: Toward a Paradigm for Reading Hypertexts: Making Nothing Happen in Hypermedia Fiction.

As a result, Sterne, Joyce, and Cortazar create unique – and illustrative – solutions to the static, linear presentational models open to them. Since metaphors and meta-languages alike inform through juxtaposition, it might be useful to consider these authors’ solutions in terms of what we might – by Joycean “license” – call “computerese”.

Sterne (who wrote Tristam Shandy in 1760), exploits the decorative and self-referential qualities of the Gutenbergian medium in what we might call a screen-based mode. Like John Barth after him, and the medieval copyists before him, he recognizes that the graphical coherencies and conventions of the printed page can be conveyed in the textual linearity. In this sense, Sterne anticipates graphical user interfaces.

James Joyce attenuates language in what we might call the line editor mode, recognizing that imprinting is in printing and thus an eidetic image. Like the Lewis Carroll of Jabberwocky, he forces a “what you see is what you get” environment to yield what we would call virtual windowing, where the in-printing seems to expand and disclose a momentary flicker of other words and other languages. His process thus is also graphical, allied as much in his time with collage or Ezra Pound’s notion of ideograms, as with Donald Knuth’s metafont or Macintosh’s Cairo font in our own time.

However, Joyce compounds and parallels these eidetic qualities of print with narrative macros which summon overlays from what we could call, quite rightly in Joyce’s case, libraries. These interlace successive text segments and cause side effects which extend the flicker of a line to a cyclic pulse of whole pages.

Where the effects of Sterne and Joyce reside largely in the record fields themselves, the effects of Cortazar are strictly relational. Hopscotch comes with what we might call subroutines (or perhaps “shellscripts”) which are not at all unlike Infocom branching structures or children’s Choose Your Own Adventure books.

Where Joyce and Sterne overtly reference a system library of parallel texts and thus make their language what we might call declarative, Cortazar provides a procedural language which readers may use to reference their own libraries. That is, Cortazar calls upon us to recall not the subject but the act of our previous readings, and invites us to read Hopscotch in a successive, literally programmed iteration. This second reading depends upon our decaying, yet dynamic, memory of the master text.

Whatever their differences, each of these three novelists surely preprocessed interactions with their readers, anticipating in their input stream the peculiar limitations of the batch-processed, linear output stream of a static presentational text. While this kind of preprocessing is, jargon aside, surely an aspect of any novelist’s creation of an ideal reader, these novels are distinctive to the degree that their texts overtly confirm correct traversals.

In other words, the preprocessing results in the creation of an intricately networked novel-as-knowledge-structure which both simultaneously invites and confirms reader interaction. (e.g., The so-called Gorman-Gilbert Schema for Ulysses provides graphic substantiation of such preprocessing, especially since Joyce originally circulated it on a not-for-publication basis among friends and potential critics. See the Appendix to Richard Ellmann’s Ulysses on the Liffey [Ellmann, 1972, pp. 186-187] for a concise history and reproduction of the schema.) Moreover, the text itself carries a syntax for such interaction (an implied labeling of network arcs) which depends upon a reader’s familiarity with the operating system of literacy. That is, we readers are invited to confront these novels, even on first reading, in the same fashion that we have previously flipped through, browsed, reviewed, or recollected other, more linear novels.

This invitation to confront is what I believe instantiates the interactivity of these texts, offering, as it were, the structural strategy of Eco’s open text. Each of these novels is what Eco calls work in movement. As such, “the possibilities which the work’s openness makes available always work within a given field of relations… we can say that the work in movement is the possibility of numerous personal interventions, but it is not an amorphous invitation to indiscriminate participation. The invitation offers the performer the chance of oriented insertion into something which always remains the world intended by the author.” [Eco, 1979, p. 62]

Therefore, it is perhaps not too much to say that, if truly interactive computer fictions are to exist, they will require tools for simultaneously creating and recovering such oriented insertions within given fields of relations. That is, interactive fictions require text processing rather than word processing tools.

For text processing, unlike word processing, is a method of intellectual and artistic discovery and presentation. If word processing may be thought of as a tool for thought (making any written text available for emendation, elaboration, and restructuring at any point along its length), text processing may be thought of as a thoughtful tool (making a master text and variations of it available to readers along a presentational path predetermined by the writer and selected or influenced by the reader). Text processing makes texts transparent, inviting readers to consider parallels, explore multiple alternate possibilities, and participate in the uncertain process of discovery and creation.

Our program, Storyspace, originated as an attempt to develop a text processing tool to enable writers of interactive presentations to exploit multiplicities. Storyspace depends upon a decisional order rather than a fixed order of presenting material. A fixed order presentation may be thought of as a road map, within which the author may present side trips or forks (some of which the traveller or reader may be allowed to choose among). The order of presentation, however, remains linear. The trip (or text) proceeds from point A to point B, however various the digressions.

A decisional order, however, may be thought of as a series of locales, some of which are linked by linear progression or argument, but with others determined by allusiveness, resemblance, evocation or unexplained and “intuitive” parallels determined as often by the author as by the reader. The journey along a decisional pathway may continue to proceed from point A to point B, but the traveller will not only be invited to dream or recall other journeys (some as yet unmade), but also will be confronted with unexpected or surprising turns, detours, and compelling alternate routes. More importantly, each alternate reading will cause the text itself to degrade or reform, so that no successive reading will ever again substantially parallel a traversal of the initial master text. In such a way, the reader’s creation of an implied text could become physically enacted on the screen or in instantaneously laser-printed (and, one could imagine, perfect-bound) hardcopy.

This kind of subversive text, or multiple novel, would offer its readers the opportunity to explore the cohesion of a work in a way analogous to successive readings separated by time. The reading would, in this sense, encompass in its interaction either the kinds of imaginary dialogues we often conduct with author or character after reading a story, or the reveries of succeeding (or “what if”) episodes which either interject or follow.

Rather than speculating on a character’s deeper motivations, for example, the story itself might-in successive readings-offer greater or lesser explorations of motivation. As such, the text would overtly offer in its structure that which Eco claims for every work of art. It would be “effectively open to a virtually unlimited range of possible readings, each of which causes the work to acquire new vitality in terms of one particular taste, perspective, or personal performance.” [Eco, 1979, p. 63]

To enable potential interactions of this kind, the Storyspace text processor preserves both a decisional pathway (the choices made by a creator of a work, i.e., the interaction which precedes text) and a presentational pathway (sequences of nodes and their links available to the audience/reader, i.e., the interactions which follow from texts).

The decisional pathway provides the creator of a work with a repertoire of associations and linkages, some designed solely to serve as mnemonic, architectonic, and editorial stimuli and utilities, and others designed to directly interface with the presentational pathway. This latter pathway provides a means for presenting multiple, interactive versions of a work, with the range, liberty and syntax of interactivity determined either by the creator or by the reader. The text thus may be accessed either directly or through predetermined associative scripts designed by the author but transparent to the reader.

Storyspace might thus be used to create a novel as supple and multiple as oral narratives, but with the referential and coherent richness of in-printed, relational works such as those of Sterne, Joyce, Cortazar, and others. These computer-enhanced, subversive texts, would merge process and product much in the manner of Baroque themes-and-variations or jazz improvisation, in which the colorations or embellishments are ephemeral, often depend on audience or occasion, and usually resist static apprehension or capture. The intellectual basis for such work involves both a natural collation of trends in twentieth century thought concerning the transitory and multiple nature of human experience; and also a reflection of our widespread cultural urges and individual longing for methods to identify and represent the perceived order and complexity which underlies a mass of information.

To the extent that the decisional pathway would also be available for readers to use, we might also expect that they would template story generation scripts which would offer them opportunities to participate in the narrative itself in the fashion of a hypertext.

Certainly, the narrative and the technological models exist for readers to do so.

Yet would they? I think not. For a fiction is essentially a selfish interaction for both its author and its reader.

It is likely that no one interrupted Homer.

Or, perhaps there were drunken hecklers even then, in the courtly supper clubs of Asia Minor. Almost certainly there were the Attic equivalents of piano bar patrons who politely request “Melancholy Baby” or “Steadfast Penelope.” Even so, we can feel reasonably certain that no outsiders interrupted Homer in the sense that they established a priority in his narrative, or that their language subsumed his.

Or to put it differently, a question we might ask ourselves in our enthusiasm about interactive fictions is: Since the technology has existed for some time, why don’t people write alternate chapters in the blank spaces of bound novels, or alternate sentences in the blank spaces between printed sentences?

Why don’t most readers write at all? Why don’t most readers at least write the beginnings of sequels to the novels they love? (Some readers, of course, do. The Baker Street Irregulars are perhaps best known, although science fiction fans seem even more devoted to creating an alternate canon. The Friends of Darkover, “a non-sectarian, non-sexist, and non-profit group of science fiction and fantasy lovers… has come into being with no purpose except the discussion of secondary universes, primarily Darkover.” [Marian Zimmer Bradley and the Friends of Darkover, 1980].

Bradley mentions similar alternate worlds created by Star Trek and Tolkien fans. Friends and colleagues of mine report similar alternate versions created by fans of the Dr. Who television program and readers of The Witch World of Andre Norton. Bradley notes that the majority of alternate Darkover stories are written by women, and suggests-much in the same vein as Sherry Turkle-that women, who are trained as children by role play rather than fantasy, tend to feel more comfortable at first in someone else’s world.

In the case of those readers who don’t, we would like to respond that the medium was not appropriate, since we would like to believe that the computer medium is appropriate. Yet another series of questions follows upon these proposed beliefs.

Why don’t most readers of current interactive fictions write alternatives within them, or sequels without them? Certainly, the medium offers opportunities. Users of multitasking systems could input alternatives in one window as the interactive story scrolls in another. Users of other systems could summon text-editors running under the application, or-as an emergency expedient-use break keys and input alternate lines to an uncompromising operating system or monitor. (They might also use Storytree, a simple and quite fascinating structure editor, published by Scholastic Software.)

To a certain extent, of course, users already do these things. Even the most experienced users of interactive adventure or mystery games sometimes find themselves inputting commands which exercise the parser beyond its limits and cause it either to admit ignorance or output default, Eliza-like solecisms. And few honest computer users can swear that they have not explored the variety of canny, canned responses which applications or systems programmer’s provide in anticipated response to hallowed Anglo-Saxon expletives and imperatives.

The two preceding cases suggest a paradigm for what I want to call selfish interactions. In the first, readers either become too enveloped in the language of the narrative, or-paradoxically-fail to become fully cooperating participants in the linguistic subset which the parser demands. That is, they either lose themselves in the story world, or experience the loss of the story lexicon in the parser. In the second, users acknowledge the linguistic subset by seeking ironic or hortatory confirmation of it. That is, they explore the limits of phatic communication which underlie a highly codified linguistic subset.

In each case, the attendant pleasures are ones which involve confirmation of interaction. Yet what is confirmed are limits, or the traversal of boundaries between procedure and declaration, i.e., between the act of reading in a certain way and the text which anticipates and signifies successful reading in that certain way.

Thus, we are unlikely to write intertextual, parallel, or detached versions of an interactive fiction because we selfishly seek confirmation of our alternative choices within the text. Likewise, we are more likely to write (or utter) alternatives to a fiction either within a computer conference, where other readers can assume author or character roles to confirm our choices, or in adventuring or other groups, where conversation and mutual fantasy supply confirmation outside the software or text.

Yet the communal aspects of these conference or group readings are likely to be unsatisfied to the degree that any participating role-player moves the text beyond the constraints we imagine the author imagined. Or, to state the case positively, we are satisfied and delighted when the text reconfirms our variations. Thus, an interactive mystery which presents a taunting prompt, i.e., “Too bad you dropped the phaser, you could sure use it now,” delights us even as we are reduced to carbon atoms.

It is unsurprising that the overtly intertextual nature of most current interactive fictions is so attractive to literary theorists and narratologists. That is, we often read interactive fictions in parallel streams consisting of:

1. the narrative stream, or the story itself, which echoes, self-reflexively quotes, and anticipates its variations within a relatively unconstrained lexicon, and
2. the game stream, or the “story” of the narrative flow chart and program syntax, which not only enables confirmations within the first stream, but also provides confirmations of its own, but within a highly constrained, formulaic lexicon.

We are as pleased to recognize the structure of the latter as we are to experience the structure of the former.

Yet the current apparatus seems somehow makeshift. The pleasures of parallel reading are not quite compensation for the jolt of alternating lexicons. The continued development of more successful parsers cannot conceal the fact that true natural language front-ends are years away from development, nor can it convince even the most naive users that they have engaged in genuine interaction resulting in alterations to their reading behavior and the story’s text. We know that we are engaged in something more like a game than a reverie, and although there are undeniable pleasures in mapping the story and its branches while matching wits with its programmer within the game stream, we somehow know that in the narrative stream we are interrupting Homer, or, worse, both Homer and his programmer.

Ideally, of course, both the program map and the text it traverses would change with our changing interests, much in the way, for instance, that Joyce’s Ulysses changes style to suit sensibility, quarter of the city, and hour of that June day. (See the Appendix to Richard Ellmann’s Ulysses on the Liffey [Ellmann, 1972, pp. 186-187] for a concise history and reproduction of the schema.) That is, it is possible to conceive of an interactive computer fiction within which the branching schemes, alternate traversals, and so on both disclosed themselves as integral parts of the narrative and altered themselves according to our shifting interests and current states. In such an interactive fiction, the branching choices, attribute lists, and implicit/explicit menus would disappear into a seamless, but subversive, text which constituted both the story and the game stream. The immediate effect would be both to widen the fictional domain, or story world, of interactive fictions, and, more importantly, to involve a wider audience and thus a wider authorship.

A subversive text would restore the Homeric situation to the extent that its branching would be driven by: 1) previous conditions and 2) detection of audience interests, rather than by direct intervention of audience into story. As an example, consider the following, purely hypothetical, example of interaction we envision for Storyspace.

The interactive fiction, Emma, consists of a paperback book (or master text) and an accompanying computer disk. On the disk are the master text and subversive variations. The screen display is very similar to the Macintosh notepad in that it allows paging/browsing forward and back in the text. Each screen page is identical to a portion of the corresponding page in the printed master text. Thus a reader who preferred screens to printed pages would be able to read Emma through ­ once. Without browsing forward or back. But she or he would have no indication of this fact, except that when the reader flipped, for example, forward in the screen master text from page two to page two hundred, it would be impossible to return to page two. Or, more accurately, the page two the reader returned to would be utterly different from the page two she or he left.

Since the master text would still exist in the printed version (and would have existed on the screen for one continuous read-through, before it decayed), it would be possible to try to make some sense out of the second “page two” by comparing its intent, focus, style, point-of-view, etc., against the master “page two.”

Perhaps the reader could test a hypothesis (still very much in the game stream fashion) by browsing forward again from page three (using either the master text as a map, or using a search function within the program to choose the same char­ acter or setting that the previous browse to page two hundred resulted in). Or perhaps the reader would simply read sequentially through the new story, choosing to compare it or not with the master text.

Suppose, however, that the reader did choose to pause at different points and compare with the master text. Suppose further that the author had embedded in the decisional pathway a branching structure which depended upon the machine’s ability to calculate (and/or accumulate) eventless reading of certain pages with common themes or characters. Suppose that branching structure presented yet an­ other version of the multiple novel, with an ordering and structure quite unlike the hypothetical one the reader has decided to pursue.

Suppose that in the very next episode, a character seemed to recall an event very much like the one that used to be on page two and asked someone’s help in recalling it. And the screen display suddenly refused to move to the next screen page, but would move forward and back through everything that preceded the current page and followed the page after next.

Suppose the reader quickly gets frustrated, puts the disk aside, and does not use it again until months later. Perhaps winter has changed to spring in the interim. Suppose that the spring novel was entirely different from the winter novel, and that the computer’s clock triggered the new version.

Or, suppose that each time the disk was rebooted after a master reading, a blank screen appeared, or an old-fashioned A> prompt, or a ? cursor, or a polite request, “What would you like to read?”

Suppose that there really were natural language front-ends and expert systems which constructed reader agents which tailored texts to suit the input to such prompts. Or suppose a Boolean technological leap resulting in both faultless voice recognition chips and natural language processors so that random mumbles or di­rected beefs and bravos created such agents, and the resultant new texts. Suppose that input to the above noncommittal prompts caused the software to take over the system and dial up 800 numbers which downloaded new versions (or caused Videotex services to pump new text through the friendly TRINTEX or AT&T blackbox in the living room).

Suppose that the text did invite you to write a version, an alternate sentence, a chapter, a sequel? And then never showed it to you again. Or showed only parts of it, intertwined with a version of the multiple tale appearing months after the read­ ing that prompted you to write in the first place.

Suppose somebody shouted, “Play Steadfast Penelope,” and the screen display altered to Greek fonts, or the voice chip began to declaim formulaics in a foreign tongue.

Suppose a text can anticipate unpredictable variations upon it.

I am not sure what to make of this sentence. It has aspects of an unrealized pun, a bad koan, a polemical claim, an oxymoron, and a manifesto. I want to doubt it, but I believe it utterly.

There is a phenomenon known variously to computer scientists but often called
interference, the concurrency problem, or software interaction, where, in complex systems and programs, levels upon levels of software functioning together exhibit unanticipated side effects causing unpredictable and usually nettlesome-but occasionally

felicitous-effects. For a debugger these interferences call to mind the swamp of The Big Two-Hearted River, where “the banks were bare, the big cedars came together overhead, the sun did not come through, except in patches (and) in the fast deep water, in the half light, the fishing would be tragic.”

In a similarly dark place, Eco considers a question (parallel to interference or the koan that begins this section) which exhibits itself in James Joyce’s pre-discovery of nuclear fission in the phrase “abnihilation of the ethym” in Finnegan’s Wake. According to Eco, “the poet anticipates a future scientific and conceptual discovery because-even if through expressive artifices, or conceptual chains set in motion to put cultural units into play and to disconnect them – he uproots them from their habitual semiotic situation.” (This and the following quotations, with the exception of the last line from “Big Two-Hearted River,” are from Eco, p. 86.)

The discourse here is also a swamp (and the fishing is dangerous), yet what comes through is a vision of interactive interference not only anticipated but, in a sense, forced by the networking of language. “Sooner or later someone understands in some way the reason for the connection and the necessity for a factual judgment that does not as yet exist,” suggests Eco. “Then, and only then, is it shown that the course of successive continguities, however tiresome, was traversable or that it was possible to institute certain transversals. Here is how the factual judgment, anticipated in the form of an unusual metaphor, overturns and restructures the semantic system in introducing circuits not previously in existence.”

This kind of fishing, of course, exists at the level of the linguistic microcosm, yet it does not seem so far removed from the kind of macrocosmic interaction of dual story streams which I have suggested as the most likely mode for coming interactive fictions. The linking arc in such a network is the selftsh reader, the someone who sooner or later understands the reason for a connection. What is critical, however, is that the arc itself be literally instantiated, as present in the creation as in the performance.

“The factual judgment,” says Eco, “draws, perceptively or intellectually, the disturbing data from the exterior of language. The metaphor, on the other hand, draws the idea of possible connection from the interior of the circle of unlimited semiosis, even if the new connection restructures the circle itself in its structuring connections.”

The multiple novel, likewise, will invite writer and reader to restructure the circle of a text in its ability to simultaneously, and subversively, present both the exterior and interior of language through successive, and shifting, story streams. Already the tools present themselves for use: the idea processors, hypertexts, and Storyspace’s. (See Figure 8.3.)

"(Nick) looked back. The river just showed through the trees. There were plenty of days coming when he could fish the swamp. " – 1/9/1989

About the Author

Michael Joyce

Michael Joyce is a prize-winning novelist as well as a teacher of writing. He has lectured and published widely on issues relating to hypertext and writing, and is part of the TINAC collective.

Together with Jay Bolter and John B. Smith, he is the co-developer of Storyspace, hypertext software for writers and readers published by Eastgate Systems. His interactive hypertext fiction, Afternoon, also published by Eastgate Press, has been called “an information age Odyssey” (by Pamela McCorduck writing in Whole Earth Review).

He holds an MFA from the Iowa Writers Workshop, where he was a Teaching Writing Fellow; and he has been a Visiting Fellow at the Yale University Artificial Intelligence Project. He is currently Professor of Language and Literature and Coordinator of the Center for Narrative and Technology at Jackson Community College, as well as Director for JCC’s National Community College Alliance charter project sponsored by Apple Computer and the League for Innovation in Community Colleges.

Michael Joyce may be contacted at the Center for Narrative and Technology, Jackson Community College, Jackson, MI 49201.