GAPS, MAPS AND PERCEPTION: WHAT HYPERTEXT READERS (DON'T) DO

J. Yellowlees Douglas

      [In discourse]... all the logical small change is in the interstices.
                                   - Barthes, The Pleasure of the Text (9).
The essence of cinema, it has been suggested, is the cut (Gianetti 113). What is a film, after all, other than a succession of strips of celluloid, a series of various shots all linked together with editors' tape and shown as one continuous strip? As we sit in darkened cinemas watching features by the likes of Hitchcock and Sirk or even Coppola and Scorsese we no longer see the cuts or the edits: we see the connections. If we see a shot of an unshaven man half-heartedly slurping soup in a grimy diner, followed by a shot of an impeccably groomed croupier presiding over a craps table in a crowded casino, followed by a shot of an elderly woman lying amid a forest of tubes, drains, and drips in an intensive care unit, we do not perceive these as part of some sequence of random shots or images. Instead, we see this sequence as either a piece of scene-setting, establishing locations which will later play critical roles in the narrative as it unfolds, or as an introduction to the narrative's dramatis personae. We also tend to assume that these shots are joined in roughly chronological time, that we are dipping into these characters' lives at approximately the same time, and that the coupling of their lives in this short sequence of shots anticipates real, physical interactions and encounters to be revealed during the course of the film.

Are these assumptions learned, our expectations the natural outcome of our being thoroughly schooled in the grammar of cinema, or are they endemic to our way of perceiving the world? In the fledgling Soviet Union, due to an acute shortage of film stock, pioneer film theorist Lev Kuleshov found himself able to create films only by slicing and splicing existing film footage from wildly different sources. Using a single, lengthy take of an actor's face with a neutral expression, Kuleshov sliced it into three equal segments and intercut these with images of a bowl of steaming food, a dressed out corpse and a child playing. When he projected his completed film to an audience barely conversant with the new medium, however, they praised the actor's subtle ability to convincingly portray the emotions of hunger, grief, and joy (Gianetti 132). Our ability to perceive connections such as these, clearly, has little to do with our knowledge of the grammar of shot-reaction shot (which dictates that each shot following the close-up of the actorØs face reveal the objects he sees) and everything to do with the nature of human perception.


Perception, Gaps, and Connections

Narrative, Jerome Bruner has argued, can be defined by a single, serviceable criterion: narrative is that which deals with the 'vicissitudes of intention' (Bruner Actual Minds 17). The reason for the role of intention in defining narrative is clear, he continues,
[I]ntention is immediately and intuitively recognizable:  it seems to 
require for its recognition no complex or sophisticated interpretive 
act on the part of the beholder (17).
Narratives, in a sense, are about connectedness, sequence, and order qualities which are inextricably linked to the way we view the world around us. We might even say that narratives represent a reflection of our tendency to perceive the world in terms of intentional or causal states, albeit a reflection which produces an orderly, predictable, and complete world within a static structure. But if this mode of perception is part of what makes narratives so attractive to us and, apparently, so important a component of our lives (Brooks 21-22), it is also integral to our being able to read, interpret, and understand narratives at all.

Most theories of reading have virtually ignored the biases of human perception toward seeing causality and intention even where these qualities may not exist. In a study of perception conducted more than twenty years ago in Belgium, subjects asked to watch projected images of moving, animated objects persistently saw causation connecting movements. Invariably, subjects discussed the movements they had watched in strictly causal terms, with the objects perceived as dragging, or deflecting one another (Michotte 349). Although the subjects were conversant with physical laws which could explain the movement of the objects they saw, the presentation of two or more objects invariably prompted the subjects to see their movements as strictly caused (Michotte 25, 375 ). In a similar experiment, psychologists Fritz Heider and Marianne Simmel also projected an animated film featuring a small moving triangle, a small moving circle, a large moving square, and an empty rectangle to viewing subjects who unanimously described their movements in animate, causal, or intentional terms (Bruner Actual Minds 18). More recent studies have shown that subjects even arrange the space-time relationships between simple figures to reinforce their readings of intention or simply as part of their inability to perceive events independent of intentional states (Bruner Actual Minds 18). And research conducted on six month-old babies has demonstrated that infants react with changes of facial expression, heart rate, and blood pressure when cinematic sequences representing non-causal relationships appeared sandwiched between sequences portraying causal connections (Bruner Actual Minds 18). Our tendency to perceive the world according to causal or intentional states seems endemic to our being human.

Human perception, however, also seems to work in the opposite direction, erasing noise when a strong signal is present, enabling us to hear only what we perceive as meaningful and to ignore anything extraneous which intrudes. Subjects in an experiment conducted by psychologist Richard Warren listened to spoken sentences recorded on tape where a few phonemes had been erased and replaced by the sound of a cough. When asked to repeat what they had heard, the subjects overwhelmingly reported having heard the missing phonemes as precisely as if they had been physically present on the tape. When, however, an undoctored version of the sentences was played back to them, none of the subjects could distinguish the sounds they had imagined they had heard from the sounds they had actually heard nor could they pinpoint the cough's location amid the sentences. As the cough existed outside any meaningful sequence for these listeners, their perceptions excluded it, but when a brief silence replaced the cough, all the listeners were able to fix its exact location (Campbell 267-8). The brain, apparently, does not preserve sequences step by step but, instead, recognizes overall patterns. Our not necessarily conscious but prior recognition of these patterns enables us to unconsciously synthesize gaps such as missing phonemes without noticeable effort (Campbell 269; Schank 14).

It is our perceptions, then, and not the hand of a Scorsese or a Hitchcock which creates the illusion of continuity, sequence, and causation as we sit in a darkened cinema, showing us a steady sequence of moving images where, in fact, only flickering, still images exist. As any cinema theorist will quickly inform you, nearly half the time we spend at the movies is, in fact, spent in the blackness between frames. Working together, the human eye and brain play off what Gestalt psychologists call the phi phenomenon to produce the fluid, tangible reproduction of life to which we have become accustomed onscreen (Monaco 73).

What probably began as interpretive skills needed to ensure our survival in a highly competitive, natural environment now also endow objects and actions in the world around us with a continuity and a richness of meaning that enables aesthetic objects to exist and to retain meaning and significance for us.


Seeing Connections in Text and Hypertext

Noting that Douglas Hofstadter has suggested the perception of relatedness, or connections, is perhaps the single, definitive characteristic of intelligent behavior, media theorist John Slatin argues that the link is the defining characteristic of hypertext:

      [E]verything in hypertext depends on 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... the link simulates  the connections in the mind 
      of the author or reader...  (Slatin 877).

Hypertext links or connections, of course, bridge the very physical gaps yawning between the text of nodes separated by virtual, three-dimensional space. Yet these have no content themselves and few cues which might prompt readers to see them as anything but a physical connection between two segments of text. Although defaults, links, and paths labelled with names can all appear to readers in either browse menus or on cognitive maps, no naming convention exists to guide readers as to their significance. The gaps in hypertext space jar and bewilder us as readers wholly inured to the printed page, prompting hypertext theorists such as George Landow, director of the IRIS Intermedia project, to call for coherent guidelines for the creation of links, lest readers find themselves lost in hyperspace (Relationally Encoded Links 333; Rhetoric of Hypermedia 83).

Yet, if we recall the discoveries of the response theorists and psycholinguists of reading discussed in Chapter Two, we are confronted daily with gaps everywhere in the texts we read:

      Holes or gaps are so central in narrative fiction because the 
      materials the text provides for the reconstruction of the world (or a 
      story) are insufficient for saturation.  No matter how detailed the 
      presentation is, further questions can always be asked; gaps always 
      remain open (Rimmon-Kenan 127).

As Wolfgang Iser has succinctly noted, narratives represent opportunities for us to bring into play our own faculty for establishing connections for filling in gaps left by the text itself (Iser Reading Process 285). Gaps are precisely what enables the act of directed creation so lauded by theorists from Sartre to Barthes, by leaving readers with what Laurence Sterne in The Life and Opinions of Tristram Shandy called something to imagine (127). The reason why we so seldom glimpse these gaps except in student writing, perhaps, or in our own writings in progress is, as we have seen, largely a function of human perception and only secondarily of literary convention. Conventions, as we shall see, are often little more than labels which we give to connections already, even inevitably, perceived between two objects or segments of text.

Text into Hypertext

Marc Saporta's Composition No. 1 , a fictional narrative consisting of about 150 unnumbered, loose sheets of paper, is the first documented published work of fiction which required readers to compose the narrative themselves by shuffling the pages prior to reading them (Bolter Writing Space 140). More radical in composition than Julio Cortazar's Hopscotch, which was to follow three years later, Saporta's text demanded that readers shuffle the pages of the narrative like a deck of Tarot cards, cut, and then read the cards to determine the fate of the protagonist. Similarly, in Hopscotch, Cortazar establishes two orders for reading his text, obliging readers to work out possible connections between chapters, such as location, chronology, and causality. Saporta's narrative, however, provides no such obvious script for readers to follow, leaving Bolter to conclude, as he completes his assessment of the text, that to call the work fragmented is to assume that it was originally whole, that the fragments belong in one order (Bolter Writing Space 141).

Bolter's perception that the fragmented Composition No. 1 may, in fact, never initially have been a single, whole narrative owes less, perhaps, to what he sees as the elliptical character of the narrative than it does to the cuts introduced physically into the printed text. To study the degree to which print environments encourage us to bridge gaps in texts, I physically cut a short story into forty segments ranging from one word to three paragraphs, shuffled the fragments, and distributed them in envelopes to a class of twenty graduate students in English and English Education at New York University. Organizing the students into ten pairs, I asked the class to place the fragments of the story in the order in which they thought it originally had existed, obliging them to discuss their decisions between themselves and to arrive at consensus for their choices within the space of an hour and a half of class time, spread out over two consecutive weeks.

As the pairs bent over their scraps of paper and began sorting through them, a variety of strategies for assembling the text became apparent. Some teams began grouping similar actions they found in the text of the strips together ("There's an examination going on here and here: we have to put this one with that one"), then grouping together fragments which had similar subjects or words in common. One group placed three fragments together, all mentioning the names of drinks or showing characters drinking. Most of the groups began by reading each of the fragments and attempting to categorize the pieces according to literary conventions, with pieces sorted out according to:

  1. their appearing to present beginning, or background information;
  2. their status as kernels, or satellites (structuralist concepts only recently learned during the course);
  3. their presentation of action sequences;
  4. their status as exposition or descriptions of characters or settings;
  5. their appearing to supply transitions between sequences;
  6. the voice of the two characters in the story;
  7. the perceived tone of the narration.

The sorting of narrative segments in nearly all groups occupied the largest portion of the class time and the pace of the assembly of the narrative, in most instances, quickened in direct proportion to the number of narrative fragments placed down in a narrative order agreed upon between the pairs. The more pieces a pair assembled, the more rapidly they proceeded to place the remaining fragments into a complete narrative order.

Most strikingly, all the pairs involved appeared to rely upon similar strategies for assembling their respective narratives. First, they read through the individual fragments and attempted to articulate from them a global view of what the narrative might look like as a whole. Next, they attempted to find causal connections between actions or events from among the fragments to establish sequences or chronologies for what had happened. Finally, they tested these between themselves according to either their own life experience or their knowledge of other narratives.

In the initial stages of assembling their narratives, nearly all groups attempted to articulate between them some concept of what the work looked like as a whole, as what van Dijk has called a macrostructure, a global view of the narrative's themes and meaning (9- 11). For example, a number of groups saw a fragment which read, "Two scenes ensue and we follow them both, you and I, ma voyeuse charmante." as a thesis statement, prompting them to organize the narrative into two scenes. Other pairs perceived a pattern of alternation between the presentation of characters, and ordered the narrative fragments strictly according to whether they featured the male protagonist or the female protagonist in them. And one pair decided that the story structure involved first a presentation of the male protagonist's problems, then the female protagonist's problems, followed by the interaction between them and, finally, exposition.

These concepts of the macrostructure of the narrative enabled the groups to then establish scripts, or structures describing a sequence of events appropriate to a specific context again, based upon their knowledge of human behavior and of literary conventions which enabled them to sort the fragments before them into some sort of causal order and, after that, to establish narrative sequences (Schank and Abelson 37-9). As the pairs shuffled their fragments into sequences, they tried the order out on each other by telling narrative sequences aloud and applied their final, acid test to the narrative sequence or sequences after they told it. Each telling was subjected to the criterion of plausibility and/or a knowledge of narrative expectations based upon other works, as evidenced by some of their

discussions:


Bridging Gaps and Reading Singularly

After completing their narratives, I provided each pair with a copy of the short story in its original form and solicited for their responses to the differences between their concepts of the narrative order and that of the real short story. Several pairs asserted that it had been difficult to read the story and to read it in different ways, or what Jay Bolter has termed reading multiply which requires that we:

resist the temptation to close off possible courses of action; 
[and]...keep open multiple explanations for the same event or 
character (Writing Space 143). 

The task, he goes on to note, is nearly impossible when we read print narratives because the medium itself, with its printed letters and words locked uniformly into lines, encourages us to think of it as a changeless, closed, and authoritative way of framing the events we see as the only way of representing the events and characters we see within it (144). Even though Saporta's Composition No. 1 is delivered to us in printed cards, piled in a box, our first instinct, ironically, is to find a way to put it back together again, to recover its original order. The static, fixed nature of the printed page and its austere linearity make it ideal for the presentation of a single order. Attempting to read a print story multiply even one deliberately cut into fragments, or even, in the case of Saporta's narrative, overtly created as contradictory fragments can never be more than an endeavor to fly in the face of the essential predispositions of the medium.

Our perceptual bias toward seeing connections and causation enables us to create certain narrative sequences, as evidenced by the ease with which all groups were able to arrange narrative fragments into causal clusters. The greatest difficulty facing the pairs, however, was in finding what seemed to be the definitive sequence which securely locked all the elements into place. If the print medium encourages us to see narrative events in a linear, singular, and definitive order, why did this last task prove so elusive to these groups?

First, narratives themselves are filled with gaps. As we have seen both above and earlier, in Chapter 2, gaps or indeterminacies in narratives are what enable us as readers to engage in transactions with texts, to bring them to life. Since written language is itself indeterminate (Iser Act of Reading 24), there is no way for even the most realistic of narratives to represent characters and events completely concretely and wholly determinately. Many of the readers openly despaired of ever putting the story together, claiming that the original story itself must have been excessively, even incomprehensibly fragmented only to openly express their surprise at how coherent and even obvious the narrative sequence seemed, once they had read the original story. Good writing , as John Slatin has noted, involves presenting in print what seems to be not simply the most obvious, logical, and convincing order but, ultimately, the only possible order of events one which seems somehow inevitable (872).

Our tendency to perceive connections between objects and events in the text generally focuses on the way that the lines of print before us follow one another apparently seamlessly, locked into place. The physical substance of the page and the authority of the printed word combine to make these links seem definitive and to obscure any gaps yawning between actions or inferences. But when the text has been cut up, however, the gaps can threaten to engulf any glimmerings of coherence the narrative may originally have had.

Second, all narratives have multiple connections, sequences, and orders other than the linear, syllogistic or sequential order endemic to print narratives. The very austerity of print with its singular order and relentlessly syllogistic, sequential connections (Slatin 871), makes reading easier for us, in that it makes our predictions about where the text is heading and what to expect in the next sentence and paragraph fairly straightforward affairs. Other associative or thematic connections are driven beneath the surface of the text by its conventional linearity. Although these alternative connections are available to us in retrospect, when we read print narratives, we do not have to contend with them. Nor are we obliged to grapple, as we read, with the way in which these other, subversive orders may affect our perception of the narrative as a macrostructure (Bolter Writing Space 22). When these linear linkages are broken, however as they were in the fragmented short story readers are confronted with a plethora of possible and probable connections, boosted by our perceptual tendency to see links between characters and events which may, in fact, be entirely unrelated.

Third, the fragmented print text prevented readers from locking into a single script or perceiving the narrative as a single, global macrostructure. The openness of the fragmented narrative, with its many possible connections made overt by the physical cuts between paragraphs, made it difficult for readers to latch on to a specific script or schema enabling them to assemble a singular and conclusive version of the narrative. Throughout the process, they were haunted by other possible connections. Two of the pairs insisted that the fragments were drawn from at least two stories, making it impossible for them to create a single narrative order. One pair argued that the fragments they were unable to reconcile with the narrative order they had assembled actually belonged to another story or stories and had mistakenly been included in their envelope. And no fewer than four pairs claimed that crucial fragments had been left out of their envelopes, leaving them with gaps in their respective narratives which were physically unbridgeable. Like the readers of Composition No. 1, each of the pairs assumed that there was one specific order for the narrative fragments included in their envelopes although one pair declared that their version of the narrative seemed more satisfactory and probable than the real, original short story.

For these readers, then, the advantage of the print narrative was that it held all the disparate connections and probable links between items, events, and characters in a single, tidy order which constrained ambiguity and supplied a limited and determinate closure for the events it described. But the print narrative also closed off other possible orders and alternate connections which clearly existed in the text, orders which were apparent when the narrative had been cut into fragments. The pairs of readers could only attempt to lay the narrative fragments out in long chains of text (more than half the pairs actually spread the pieces out in strips on desks, tables or the floor), since they were working with a print narrative, however fragmented. Print technology, as one of its chief rules, did not allow them to include the same fragment in more than one narrative sequence nor did it permit them to organize the fragments into mutually exclusive narrative sequences. Although many of the pairs were obviously reading multiply (Moulthrop Containing the Multitudes 42) as their complaints about having fragments from other stories in their envelopes attest to read multiply in a print setting, to see an array of mutually exclusive possibilities for assembling the text, can only be disabling, a way of perceiving print text which does not lend itself to the discovery of the right narrative order. What is most interesting, perhaps, about the readers' experiences here is the ability of nearly every one of them to see multiple connections between blocks of text where no physical link joined them. In fact, many of the readers commented on the austerity of the connections in the uncut version of the story probably because it represented a considerable reduction of the number of connections they themselves had perceived during the assembly of their respective versions of the story.


Two Different Kinds of Reading

      In conventional narratives, readers are asked to imagine a world of 
      multiplicity from within an overwhelmingly linear and exclusive 
      medium.  For hypertextual readers, the situation is reversed given 
      a text that may contain almost any permutation of a given narrative 
      situation, their task is to elicit a rational reduction of this field of 
      possibilities that answers to their own engagement with the text 
      (Moulthrop Reading from the Map 125).

How do readers accustomed to print narratives, where textual order is singular, definitive, and conclusive, respond to interactive narratives, where they are required to read narratives without pages, with multiple episodes of closure and with conscious decision points punctuating each narrative segment? At present, no exhaustive study of how readers read interactive narratives exists, although brief discussions of the interactions of individual readers with a single aspect of interactive fictions have recently found their way into print. These studies focus on readers' difficulties in navigating through Moulthrop's Forking Paths (Moulthrop Reading from the Map 119-132; Douglas Plucked 1-22), the problem of closure in Joyce's Afternoon (Kaplan and Moulthrop 1-28; Douglas Is There a Reader 38-50) and the embedding of readers' responses within the structure of Forking Paths (Moulthrop and Kaplan 105-132).

The scarcity of such studies is hardly surprising: few published examples of interactive narratives themselves are widely available, and, moreover, most writers discussing reading and writing in hypertext environments tend to concern themselves with the larger issues of orientation (Landow Relationally Encoded Links; Rhetoric of Hypermedia), ownership (Moulthrop Containing the Multitudes), aesthetics (Moulthrop Fiction of Forking Paths) and epistemology (Joyce Siren Shapes; Bolter Shapes of WOE). Indeed, the differences between readers' experiences of print narratives and their experiences of interactive narratives are so great and touch on so many different areas that it would be virtually impossible to encompass them within a single study. By combining the observations of individual readers encountering hypertext narratives, however, it is possible to map the areas in which reading hypertext differs from the experience of reading print and to conclusively discuss the implications of those differences.


Two Sets of Forking Paths: Borges' The Garden of Forking Paths and Moulthrop's Forking Paths

While teaching a writing workshop in the Macintosh Writing Lab of the Expository Writing Program at New York University in 1987, I provided a class of fourteen undergraduates with two texts to read: one half of the class were given a print version of the Jorge Luis Borges short story The Garden of Forking Paths; the others were asked to read a copy of Stuart Moulthrop's electronic Forking Paths loaded on diskettes in their Macintosh. Upon completing their readings of the print or electronic texts, I asked the readers to write a brief account of what they had read, to tell the story back in several paragraphs.

The Borges short story The Garden of Forking Paths is, among other things, a meditation on the plurality of possibilities inherent in every motion and act. Master spy Yu Tsun must find a way to signal the whereabouts of a strategic British munitions site in France to the Germans before his British pursuer, Richard Madden, reaches him. He selects a name from the telephone directory and meets Stephen Albert, a British scholar of Chinese culture. Albert has, coincidentally, unravelled the secret of Yu Tsun's grandfather Tsui Pn who had many years earlier abandoned fame and political power to create a book and a labyrinth, neither of which were ever discovered. Tsui Pn's labyrinth, discovered by his descendants, seemed nothing more than a manuscript heaped with conflicting drafts. As Albert has discovered, however, the labyrinth and the manuscript are the same, a Garden of Forking Paths, the embodiment of an infinite series of times....a network of diverging, converging and parallel times...embrac[ing] every possibility (91).

Although Yu Tsun feels the truth of his ancestor's discovery as related to him by the British Sinologist, the arrival of Madden forces him to kill Albert to transmit the location of the strategic ammunition site in Albert, France to Germany ironically reducing the infinite possibilities hinted at in the story to a single, sordid conclusion with the deaths of Albert and, later, Yu himself, by execution. With its metaphor of the strictly infinite garden of Tsui Pn and its multiplicity of narrative possibilities, the Borges story presents an ideal beginning for the Forking Paths fantasy, from which Moulthrop creates both a figurative and literal labyrinth springing from the original story (Moulthrop Fiction of Forking Paths 4). In Moulthrop's narrative, there are no fewer than twelve permutations on the ending to the Borges tale, as well as retellings of strands from the original story framed from alternative points of view, complete reversals of character traits and motives, and even metatextual commentary on the nature of interactive text itself.

The instructions I gave the readers of both texts were deliberately kept brief. Readers of the Borges short story were simply asked to read the text carefully and re-tell, in a short piece of writing, the narrative as they understood it; I made no comments on the narrative itself, nor on its relationship to the electronic texts. I quickly informed the readers of Forking Paths that they could navigate through the text by clicking on words in the text of each window or place, or by using a carriage return. At the time, the read-only component of the beta-version of Storyspace in which Forking Paths was written, then called ReadingSpace, did not include the means for readers to examine the cognitive map of the hypertext structure, nor did it make available to readers a menu of paths branching out from each node properties now readily available in some versions of interactive narratives written in Storyspace.

Provided with this brief introduction to navigating in hypertext space, the students facing the electronic Forking Paths began by attempting to move through the narrative by clicking on words in the text, testing their choices to see if these triggered links with other segments. When, however, the students' original choices failed to connect them to new segments, they took to first clicking on words in the text in a nearly random fashion, in an attempt to see which words in the text might prove to be linked to another node. Words which these readers thought likely to yield nearly nearly invariably did not, they soon discovered, nor did a carriage return in any place produce a default. Instead, one half of the lab rang with a continual cacophony of Macintosh beeps usually indicative of an action or command beyond the perimeters of the actions available to the user. In this instance, the beeps signified either that no default existed for a particular node or that the reader had selected a word which did not belong to any link.

Unknown to either myself or my students at the time, Stuart Moulthrop was still in the process of writing full-blown instructions for readers engaging Forking Paths in the Storyspace ReadingSpace, which he considered crucial to the reading of his interactive narrative (Reading from the Map 125). As he grappled with Forking Paths, however, Moulthrop himself had forgotten that ReadingSpace contained a set of control buttons enabling readers to move up, down, left, or right within the structure of the interactive narrative. Further, Forking Paths had originally been created in a rudimentary text-linking program of Moulthrop's own devising which permitted only linked connections between nodes. With the transfer of Forking Paths to Storyspace, Moulthrop had deliberately left default connections out of the hypertext structure, intending to create a text which obliged readers to assume the mantle of authorship and expand the existing structure (Reading from the Map 126).

Convinced that a hypertext structure such as Forking Paths would invite readers to become co-authors, Moulthrop had assumed that readers would become engaged by the text as structure at the level of words and phrases within each node (Reading from the Map 126), following the contentions of Russian Formalists and other theorists of reading as succinctly formulated by Peter Brooks in Reading for the Plot:

      [W]e read the incidents of narration as promises and annunciations
      of final coherence, that metaphor that may be reached through the
      chain of metonymies:  across the bulk of the as yet unread middle
      pages, the end calls to the beginning, transforms and enhances it (93-4).

The chief difficulty with Moulthrop's chain of metonymies lay in his choices of words for providing links to other narrative segments. Where other writers of interactive narratives have specified words with texture as ideal candidates for links between nodes (Joyce Afternoon read at depth) and use unusual or striking words which seem to suggest a layer of meaning beneath them for links, Moulthrop's linking words in Forking Paths are not always the most obvious choices for readers to engage.


Metonymies, Metaphors, and Maps

As a result, initially, few of my students found themselves able to move through the narrative at all. Part of the reason behind Moulthrop's refusal to include defaults in his Storyspace structure stemmed from his desire for readers of Forking Paths to know when they had engaged a link. Had he incorporated defaults into the hypertext connections, readers would have moved through the text regardless of whether they had activated a link or not and they would also have found it virtually impossible to determine whether they had moved via default or through the activation of a link (Moulthrop Reading from the Map 263). Further, the "no-default" condition, according to the likes of Andy Lippman, head of MIT's paperback movies project, is integral to the definition of true interactivity, depriving readers and authors alike of a singular, primary pathway through the text which deprives readers of feeling their reading of the text growing, as it were, on the fly (Brand 46). With its no default condition, however, Forking Paths demonstrated to my students over and over again that they had not activated a link and also, that, having not fulfilled the specifications of an unseen author for movement, they were not entitled to proceed through the narrative.

Initially, the readers of the electronic text became intensely frustrated with their inability to discover the links buried beneath the text of the nodes in Forking Paths, and complained that, as a consequence, they could not read the text at all. Soon, however, one or two of the readers discovered that directional up, down, left, and right buttons enabled them to move through the virtual, three-dimensional space of the hypertext structure and that the right button invariably yielded a new node for reading. As one reader noted:

      with very few exceptions, right was the only choice one could 
      make in terms of movement within the story.  The up option 
      always took you back to the beginning, which was frustrating....  
      It was an interesting experience, and if there were more travel 
      options (other than just right), I would have enjoyed it more.

Eventually, all seven of the readers given the interactive Forking Paths used the right option to move through the text, and two out of the seven readers gave complete priority to the sense of the text they had derived from their navigating through the hypertext structure over their actual sense of the narrative itself. Confused by a multiplicity of narrative strands in which they could encounter a character dead in one place and very much alive and ambulatory in the next, the readers of Forking Paths lacked any tangible sense of a macrostructure which could give significance to the elements they encountered in each individual narrative segment. Only by using their sense of the narrative as a structure could any of the readers arrive at a sense of the relationship between individual narrative places and their relation to the narrative in its entirety.

When I supplied Stuart Moulthrop with my students' responses, he went on to write of his discovery in Reading from the Map: Metonymy and Metaphor in the Fiction of Forking Paths of the inverse of Brooks' construct, where readers of print texts become entangled in a metonymic web of language which ultimately leads them to metaphors for the text as a whole. Instead of reading for the plot by charting metonymies through the text, the readers of Forking Paths had attempted to plot their readings of what Moulthrop has called a cartographic space (Moulthrop Reading from the Map 128) by seeking to plot a design which might enable them to confer significance on the individual nodes within it. This map became, in a sense, a metaphor for the totality of the narrative's possibilities inverting the relationship between metonymy and metaphor suggested by Brooks. Where Brooks sees the text as a metonymic flow of language which leads readers to grasp the text as total metaphor (108), the readers of Forking Paths, instead, saw the metaphor of the text-as-map as a means of conferring meaning on the chain of metonymies stretching throughout the text.

But if these readings of the interactive narrative upset what is perceived as the conventional relationship between metonymy and metaphor, they did not move entirely outside the set of behaviors normally expected of print readers. Their plottings of the text as metaphor appears to be the beginning of what might have become, with additional reading time, a continual movement between microstructures and macrostructure, a hopscotching back and forth between global structure, genre, and local meanings:

      ...a reader goes from stones to arches to the significance of 
      arches... goes back and forth between them attempting finally to 
      construct a sense of the story, its form, its meaning... As readers 
      read, as they begin to construct a virtual text of their own, it is 
      as if they were embarking on a journey without maps... (Bruner Actual 
      Minds 36).

Where print readers search through a rich repertoire of previously encountered texts, genres, and literary contexts and use maps or Schank's scripts (Schank and Abelson 36) derived from them to endow meaning on local structures and episodes in the text, the readers of Forking Paths were actually creating mental models of the narrative as map. From a reading log hastily scrawled on a scrap of paper, one of the readers of the interactive narrative was able to determine the precise physical center of the narrative, the place labelled 076:

      After reading about 15 cells, I got to closure, but I was not 
      satisfied that it was over, so I continued reading.  I made it to  
      the center, cell 76, and that wasn't any big revelation, either...

This, he anticipated, would somehow represent the solution to the narrative, much in the way that finding the exact centre of medieval topiary mazes constituted a victory of sorts, a triumph over the devious artistry of the designer of the labyrinth. It is no coincidence, after all, that he chose to label his reading of Forking Paths with the title Maze. But such a construct is bound to disappoint this reader, as the narrative has several episodes of closure, none of which he finds particularly satisfying. In the end, when he decides he has completed his reading of the text, it is based upon his sense of having visited the majority of the points on his map of the text as structure. Like the readers attempting to reassemble Composition No. 1 into a single, conclusive order, this reader arrived at a sense of an ending which reaches back to print narratives, where we only expect to find one order and certainly not that the text we encounter in one reading may convey an entirely different meaning when encountered in another order, on a second read-through.


Endings, Closure, and Satisfaction

      ...a digression on each point in order to reach the center.
                          Pascal, Penses  (qtd in Brooks xv). 

Seen from another perspective, the readers of Forking Paths may have read their way through the narrative according to their virtual maps of it as a way of gauging its dimensions much in the way that readers turn to the last page in a book to find its page number and determine just how much reading lies before them. Confronted with a plethora of narrative strands before they could establish any semblance of a horizon of expectation, they found themselves also facing mutually exclusive narrative episodes, multiple episodes of closure and, even, a place which read, simply <> in some instances, less than ten nodes into the text. In one sense, the act of navigating through the text based on an intuited map of its structure approximates readers' flipping quickly to the end of a particularly problematic narrative to discover how it ends, in an attempt to make sense of its middle. The sense of a beginning...must in some important way be determined by the sense of an ending, Peter Brooks writes in Reading for the Plot:

      We might say that we are able to read present moments in 
      literature and, by extension, in life as endowed with narrative 
      meaning only because we read them in anticipation of the 
      structuring power of those endings that will retrospectively give 
      them the order and significance of the plot (94).

If, as Brooks argues, that the sense of adventure... plotted from its end is part of a pattern of anticipation and completion which overcodes mere succession, then our pre-conception of the probable end of a narrative, of the direction in which we believe it to be taking us, acts as a metaphor which controls how we read the text from its beginning onward.

So the readers of the interactive Forking Paths may have jumped to reading the narrative from their virtual maps of it to assuage their sense of confusion at the textual jumble of narrative strands, episodes of closure, and mutually exclusive representations of single sets of events and not simply as a means of escaping from the network of relatively inaccessible links connecting narrative episodes. If our sense of an ending acts as a metaphor enabling us to read progressively through the text, then, in a sense, the readers of the interactive narrative used their sense of the text as structure to approximate an ending of the seemingly endless interactive text.

Significantly, all of the readers of Forking Paths began their re-tellings of the narrative with several sentences discussing their confusion at mutually exclusive occurrences. This was a confusing story. People were killed, and then there they were talking in the next cell of the story, wrote one reader, while another complained: I found dead people coming back to life. It is perhaps, particularly revealing that three of the readers able to accurately represent the contents of most of the narrative strands in Forking Paths elided episodes of closure in their representations of the narrative much in the way that Richard Warren's subjects heard the missing phonemes which had actually been replaced by the sound of a cough. Instead of seeing an episode of closure as an end to one narrative strand, these readers seem to have used these endings as a means of forming transitions to other branches of the narrative. This may have been due, in part, to their navigating through the narrative outside its authored links, which would have confronted them with a dizzying series of confrontations, resolutions, deaths, and what must have seemed like resurrections all without benefit of sequential narrative episodes.

But these smooth elisions over episodes of closure may also represent a strategy for understanding a text which seemed to transgress all the familiar rules of narratives. By juggling narrative order in their re-tellings, these readers manage to simply incorporate mutually exclusive events most notably Yu Tsun's bungled suicide, which renders him virtually brain-dead into a seamless, more or less sequential narrative which comes to closure without any visible struggle:

      Yu Tsun later shot himself between the eyes but survived, [sic] he 
      later became the ward of Herr Ignazius Baumgartner.  Suddenly, he 
      found himself in a garden maze, [sic] Madden and Stephen were 
      also present in the maze, [sic] it is at this point, he had difficulty 
      identifying his allegiance, was it to Elisa or Germany? 

By creating a single, relatively straightforward narrative order out of a multiplicity of narrative orders, these readers manage to arrive at readings of the text which roughly correspond to their readings of conventional print narratives. This effort to bring disparate elements of the narrative into line with their expectations of print texts is, again, as theorists of reading have noted, an activity nearly endemic to the act of reading for most readers. As Frank Kermode has noted, readers pursue readings of narratives expecting the satisfaction of closure and the receipt of a message (Secrets 84) because this method of reading the text mirrors our interpretive strategies at work in everyday face to face communication:

      To attend to what complies with the proprieties, and by one means 
      or another to eliminate from consideration whatever does not, is a 
      time-honoured and perfectly respectable way of reading novels... 
      (88).

In contrast, two of the readers of Forking Paths attempted to direct their readings solely through their sense of the narrative as what Jay Bolter has called a structure of possible structures (Bolter Writing Space 144) via their virtual maps of the text. One reader writes of his frustration with the narrative when the story contained in a node didn't follow the narrative structure he had anticipated, based on his map of the text, when I did actually figure out where I was, where I had been, and where I was going, the story said otherwise and ends up concluding that the narrative isnØt readable, since the story doesn't hew to what he perceives as the structure of the narrative. The reader who discussed his navigation through Forking Paths (quoted at length earlier above), however, appears to have based his interpretation of the narrative solely on his perceptions of where each episode exists in the narrative's structure:

       At first, the stories did not seem to interconnect at all, but they all 
      began to relate and make more sense the deeper into the story I got.  
      After reading about 15 cells, I got to ÓÔ, but I was not 
      satisfied that it was over, so I continued reading.  I made it to the 
      center, cell 76, and that wasn't any big revelation, either... 

Significantly, of all the readers of Forking Paths, this reader is alone in seeing narrative episodes as connected, logical, and, clearly, part of some larger structure and his understanding is apparently tied to his sense of the narrative as a structure existing in virtual space. He sees himself reading deeper into the text as he proceeds, and his decision that the first episode of closure he encounters isn't sufficiently satisfying seems based on his sense of not having traversed enough of the narrative space of Forking Paths to have arrived at a sense of an ending. Finally, his recognition that the place numbered 76 lies at the center of Moulthrop's textual labyrinth leads to this reader's anticipation that the text of this place will satisfy his desire to reconcile the occurrences in the disparate strands of Forking Paths into a single, conclusive episode.

The readers grappling with the interactive narrative seemed uniformly to grasp at some sense of closure which would, in turn, enable them to anchor their readings of the shifting and often contradictory multitude of narrative episodes between the beginning of the text and its ending. Ironically, the readers of the Borges print narrative The Garden of Forking Paths reacted in reverse. Although none of the students in my class had been briefed as to the nature of the texts before them, two of the readers confronting the print story questioned the order and outcome of many of the episodes in the story, feeling them to be over-determined:

  1. I figured out TsunØs [sic] motivation throughout this story after rereading the beginning of it, but that does not seem like the point that [sic] was really trying to get put across. For all we know, the particular end used in this story may have been only one of an infinite number of possibilities down the forking paths of time. Since, according to the all-wise Tsui Pn, all of these possibilities peacefully co-exist in parallel dimensions, isn't it true that we can reject this ending and make up one of our own?
  2. Tsun [sic] then kills Albert, and is captured by Madden to be sentenced to death. But in killing Albert he is able to convey the name of the city that the Germans has to attack to survive, which they did (attack, that is). But some questions remain: Why didn't Tsun just kill Albert and then leave, instead of remaining with Albert long enough to allow Madden to catch up with him? Why did Tsun only carry one bullet in his gun? Why did Tsun declare to Albert, just before he killed him, that I am your friendDÔ? How did the kids on the platform in Ashgrove know Tsun was going to AlbertØs house? In the definition of the book, is Albert still alive, just living one of his other lives?
Provided with tangible cues as to the length of the story, the genre to which it belongs, and its probable conclusion, these readers appear to feel that the narrative has violated its own precepts. Having opened the possibility for a universe in which multiple possibilities co- exist and have equal tangibility, The Garden of Forking Paths seem to them to have reneged on its promise without having provided what saw as the sort of closure appropriate to the macrostructure of the text a world of seemingly infinite possibilities where mutually exclusive developments could unfold simultaneously. As both readers have noted, the relentlessly conclusive spy story is only one of the stories which exists within the Garden of Forking Paths, the labyrinth created by Tsui Pn and rediscovered by both Stephen Albert and Yu Tsun. In the Borges narrative, the Garden is also a space which enables mutually exclusive possibilities to exist simultaneously, and it is the Garden which becomes the metaphor for the Garden of Forking Paths, not Yu's murder of Albert or his own subsequent execution. It is the metaphor of the Garden, with its plethora of possibilities, which enables its readers to decide that their own endings to The Garden of Forking Paths can assume equal importance to the ending conferred on the narrative by Borges.

The readers of the interactive Forking Paths appear to feel swamped by a multiplicity of endings and narrative possibilities; the readers of the print story The Garden of Forking Paths express their restiveness at Borges' singular and very limited endings to a narrative which concerns itself, nearly to its penultimate paragraph, with a universe of seemingly infinite possibilities. Perhaps the differences in their readings can be accounted for by the medium in which each group experiences the narrative. One group seems frankly bewildered by an environment which overturns their comfortable assumptions about what narratives should do and be, while the other is sufficiently well acquainted with the conventions of printed space. Yet the group of interactive readers included one who succeeded in creating a mental model of the structure of Forking Paths, and using this map to confer meaning upon narrative nodes according to their placement within its structural space hardly the response of a reader simply overwhelmed by the demands of a new and entirely foreign reading environment.

In fact, the reader of Forking Paths who stressed navigation, together with the readers resisting the ending supplied in The Garden of Forking Paths, seem to have arrived at a definition of closure which somewhat revises traditional definitions of closure as both the place where goals are satisfied the protagonist [can] engage in no further action (Trabasso et al. 87) and the point at which, without residual expectation, [readers] can experience the structure of the work as, at once, both dynamic and whole (Herrnstein-Smith 36). Where traditional definitions of closure generally give priority to our grasp of the structure of the work as a whole, these readers seem to have arrived at strong readings of the text which emphasize their sense of the text as an ongoing dynamic a quality which Nancy Kaplan and Stuart Moulthrop first identified in a study of readers of interactive narratives. As one of the readers in their study noted, closure really occurs

      when we have decided for ourselves that we can put down the story 
      and be content with our interpretation of it.  When we feel satisfied 
      that we have gotten enough from the story, then we are complete 
      (16-17).

Similarly, all three strong readings of both Forking Paths and The Garden of Forking Paths featured what we might call a reader-centered dynamic of reading. The reader navigating through the narrative spaces of Forking Paths (quoted above) decides, based on his sense of having traversed both the margin and center of the hypertext, that the text will not hand him any great resolution which will both unite and resolve the tensions and possibilities percolating in its plethora of narrative strands. Reflecting back on the dynamic universe of possibilities set up by Tsui Pn's Garden of Forking Paths, the readers of The Garden of Forking Paths feel that the narrative continues unfolding beyond the singular ending outlined by Borges. In the definition of the book, is Albert still alive, just living one of his other lives? wonders one reader, upon concluding her re-telling of the narrative. The other reader, however, perhaps realizing that his classmates reading the electronic text were, as he proceeded in his reading, possibly experiencing alternate endings to the same narrative episode, does not simply resist the singular closure of the print story he invents his own ending:

      I prefer a future in which Tsun [sic] murders Richard Madden and 
      lives peaceably with his mentor Dr. Stephen Albert, following in 
      the footsteps of his ancestor Ts'ui Pn and finishing his life work:  
      The Garden of Forking Paths.  

Two Kinds of Readers: Other-Directed and Inner-Directed

We seem to be dealing here with more than simply two different media environments, since resistant and reader-centered interpretations emerged from readings of both the interactive and print narratives. We appear to be dealing with two different kinds of readers: what David Riesman would call inner-directed and other-directed readers (Riesman 259). Inner-directed readers here are distinguished by their ability to redefine their roles as readers by either discovering a new way of navigating through narrative space or by revising the concept of closure. Other-directed readers, conversely, take their cues for reading from their knowledge of established reading practices and literary conventions, leading them to brand examples of narratives wildly divergent from familiar norms, such as Forking Paths as frustrating, nonsensical and, even, failures.

With their lack of conventional closure, their indeterminacies, ambiguities, and representations of mutually exclusive occurrences, interactive narratives will, clearly, do little more than befuddle other-directed readers. These new and alien narratives seem to require readers to both immerse themselves in the narrative webs of possibility and to extricate themselves from it, in order to grasp at a sense of the narrative as a totality, as a structure of possible structures (Bolter Writing Space 144). This dynamic of reading differs from conventional reading in that it seems to demand that readers physically engage in a game of interaction with the text as an author-constructed structure of links, paths, and yields and, at the same time, rely on their own, reader-centered judgments about meaning, significance, closure and, even, connections in the narrative. The demands particular to interactive narratives and to reading in any new environment lacking established reading and interpretive strategies seem to demand that we evolve into inner-directed readers, or readers who move beyond simply realizing an author's virtual text and resist authorial prescription to arrive at readings of our own.


Texts

Barthes, Roland. The Pleasure of the Text. Trans. Richard Miller. New York: Hill and Wang, 1975. 64.

Bolter, Jay David. The Shapes of WOE. Writing on the Edge. 2.2 (Spring, 1991): 90-91.

Topographic Writing: Hypertext and the Electronic Writing Space. Delany and Landow 105-118.

Writing Space: The Computer, Hypertext and the History of Writing. Hillsdale, NJ: Lawrence Erlbaum Associates, 1991.

Borges, Jorge Luis. The Garden of Forking Paths. Ficciones. Trans. Anthony Kerrigan. London: John Calder, 1985. 81-92.

Brand, Stewart. The Media Lab: Inventing the Future at MIT. New York: Viking, 1987.

Brooks, Peter. Reading for the Plot: Design and Intention in Narrative. New York: Vintage, 1985.

Bruner, Jerome. Acts of Meaning. Cambridge: Harvard University Press, 1990.

Actual Minds, Possible Worlds. Cambridge: Harvard University Press, 1986.

Campbell, Jeremy. Winston Chuurchill's Afternoon Nap: A Wide-Awake Inquiry into the Human Nature of Time. London: Paladin, 1989. (268-270)

Chatman, Seymour. What Novels Can Do That Films Can't (and Vice Versa). On Narrative. Mitchell 117-137.

Cortazar, Julio. Hopscotch. Trans. Gregory Rabassa. New York: Random House, 1966.

Douglas, J. Yellowlees. Are We Reading Yet? A Few Pointers on Reading Hypertext Narratives. Print introduction toVictory Garden. Cambridge, MA: Eastgate Systems, 1991.

Is There a Reader in this Labyrinth? Notes on Reading a Hypertext Novella.Computers and Writing: State of the Art. Eds. Patrik O'Brien Holt and Noel Williams. Oxford: Intellect Books (in press).

Plucked from the Labyrinth: Intention, Interpretation, and Interactive Narratives. Readers, Texts and Teachers II. Ed. Gordon M. Pradl. London and New York: Heinemann (in press).

Reading and Writing in Hypertext Spaces. Proceedings of the Fourth Annual Conference on Computers and the Writing Process. Brighton: University of Sussex, 1991. 49-60.

Gianetti, Louis. Understanding Movies. 5th ed. Englewood Cliffs, NJ: Prentice Hall, 1990.

Herrnstein-Smith, Barbara. Poetic Closure: A Study of How Poems End. Chicago: University of Chicago Press, 1987.

Iser, Wolfgang. The Act of Reading: A Theory of Aesthetic Response. Baltimore: Johns Hopkins University Press, 1978.

The Reading Process: A Phenomenological Approach. Tompkins 50- 70.

Joyce, Michael. Afternoon: A Story. Computer software. Cambridge, MA: Eastgate Systems, 1990.

Siren Shapes: Exploratory and Constructive Hypertexts.Ô Academic Computing 3.4 (November 1988): 10-14, 37-42.

Kaplan, Nancy and Stuart Moulthrop. Something to Imagine: Literature, Composition, and Interactive Fiction. Unpublished ms, 1990.

Kermode, Frank. .Secrets and Narrative Sequence. Mitchell 79-99.

The Sense of an Ending: Studies in the Theory of Narrative Fiction. New York: Oxford University Press, 1966.

Landow, George P. Relationally Encoded Links and the Rhetoric of Hypertext. Proceedings of Hypertext Õ87. Chapel Hill: University of North Carolina: 321-338.

The Rhetoric of Hypermedia: Some Rules for Authors. Landow and Delany 89-105.

Michotte, A. The Perception of Causality. New York: Basic Books, 1963.

Monaco, James. How to Read a Film: The Art, Technology, Language, History, and Theory of Film and Media. New York: Oxford University Press, 1981.

Moulthrop, Stuart. Containing the Multitudes: The Problem of Closure in Interactive Fiction. Association for Computers in the Humanities Newsletter 10 (Summer 1988): 29- 46.

Forking Paths. Unpublished computer software, 1986.

Hypertext and the Hyperreal. Hypertext 89 Proceedings. Pittsburgh, PA: Association for Computing Machinery, 1989: 259-267.

Reading from the Map: Metonymy and Metaphor in the Fiction of Forking Paths. Landow and Delany 119-132.

Text, Authority, and the Fiction of Forking Paths. Unpublished ms, 1987.

They Became What They Beheld: The Futility of Resistance in the Space of Hypertext Writing. Literacy and Computers. Eds. Susan J. Hilligoss and Cynthia L. Selfe. New York: MLA, 1991. 105-132.

Ong, Walter J. Orality and Literacy: The Technologizing of the Word. New York: Methuen, 1982.

Riesman, David. The Lonely Crowd: A Study of the Changing American Character. New Haven: Yale University Press, 1950.

Rimmon-Kenan, Shlomith. Narrative Fiction: Contemporary Poetics. New York: Metheun, 1983.

Schank, Roger C. Dynamic Memory: A Theory of Reminding and Learning in Computers and People. Cambridge: Cambridge University Press, 1982.

and Robert P. Abelson. Scripts, Plans, Goals and Understanding: An Inquiry into Human Knowledge Structures. Hillsdale, NJ: Lawrence Erlbaum Associates, 1977.

Slatin, John M. Reading Hypertext: Order and Coherence in a New Medium. College English, 52.8 (December 1990): 870-884.

Sterne, Laurence. The Life and Opinions of Tristram Shandy. 1759. London: Penguin Books, 1967.

Trabasso, Tom, Tom Secco and Paul Van Den Broek. Causal Cohesion and Story Coherence. Learning and Comprehension of Text. Eds. Heinz Mandl, Nancy L. Stein and Tom Trabasso. Hillsdale, NJ: Lawrence Erlbaum Associates, 1984. 83-107.

van Dijk, Teun. Macrostructures: An Interdisciplinary Study of Global Structures in Discourse, Interaction, and Cognition. Hillsdale, NJ: Lawrence Erlbaum Associates, 1980.