28.
Multivariant Narratives
Marie-Laure Ryan
Media theorists divide the history of writing into four periods delimited by technological innovations: the oral age; the chirographic age (manuscript writing); the print age; and the digital age. The material support of language passed from unique to freely copiable; from restricted to a live audience to widely distributed; and from evanescent to durable, only to return to a strange combination of evanescence and durability: though digital texts can be stored in a wide variety of memory devices, computer systems become rapidly obsolete and their archives unreadable.
We know that the invention of writing and of the printing press had major consequences for textuality and narrativity. In oral cultures, as Walter Ong (1982) has shown, narrative was used as a mnemonic device for the transmission of knowledge; its memorization was facilitated by prosodic features – meter, rhyme, alliteration – as well as by fixed formulae and standardized images; and the limitations of memory were compensated by a relatively free episodic structure which allowed, within limits, permutation of its units. The invention of writing made it possible to shape the flat line of epic plots into the curve of dramatic form, a much more condensed narrative structure that allows a tighter management of emotional responses. Writing also froze the free order of plots into a fixed sequence. The printing press increased the length of narrative, revived the episodic pattern of epic poetry to fill the generous frame of the book, rendered mnemonic devices obsolete, and led to the birth of the novel, a relatively unconstrained narrative form that took plot to unprecedented levels of complexity: framing, embedding, branching, digressions, disruptions of temporal sequence, and multiple plot lines. Thanks to the spatiality of the page, images found their way into texts, and visual presentation was eventually recognized and exploited as an expressive device. After these three stages, is there anything left for digital media to develop in the narrative territory?
This question would not be worth asking if the computer merely served as a medium of transmission for print texts (as it does when a digitized version of a Stephen King novel is sold online) or as an instrument of production for texts to be experienced in print: most novels, after all, come out of word processors. A truly digital text, or narrative, is one that cannot be transferred into the print medium without significant loss. It depends on the computer as a sustaining environment, and it uses the screen (or any other display device) as a stage for performance.
What, then, are the properties of digital media, and, by extension, of digital texts, that bear upon the development of narrative? Several new media theorists (Murray 1997; Manovich 2001) have offered their own lists of distinctive features or essential properties of digital systems. The list proposed below is a distillation of the features I regard as the most relevant to the issues of textuality and narrativity:
• Algorithm-driven operation. Computers are machines that can run a variety of programs, through which they can perform a variety of tasks. The behavior of digital objects, such as texts, images, and sound, is therefore regulated by an invisible code, the machine-language instructions of the supporting software.
• Reactive and interactive nature. This property is a direct consequence of the preceding one. Computer code is based on conditional statements (if… then) that execute different instructions, depending on the state of the system or on external input. I call a system reactive when it responds to changes in the environment or to non- intentional user actions; it is interactive when the input originates in a deliberate user action. (Interactivity does not necessarily mean, however, that the system will act in the way intended by the user.)
• Performantial aspect. Another consequence of the first property. A digital text is like a music score or theater script: its written inscription is meant to be executed, either by the underlying code alone, or through a feedback loop that leads from the user to the underlying code to the display, and back to the user. Digital texts thus present the same contrast as the classic performing arts between the invariability of the script and the variability of its execution.
• Multiple sensory and semiotic channels, or what we may call multimedia capabilities, if we are not afraid of the apparent paradox of talking about multimedia media. Digital environments can combine text, sound, still pictures, and animations.
• Networking capabilities. Digital media connect machines and people across space and bring them together in virtual environments. This opens the possibility of multi-user systems and live (real-time) as well as delayed (asynchronous) communication.
• Volatile signs. Computer memory is made of bits whose value can switch back and forth between positive and negative. The value of these bits determines the display. This means that unlike books or paintings, digital texts can be refreshed and rewritten, without having to throw away the material support. This property explains the unparalleled fluidity of digital images.
• Modularity. Because the computer makes it so easy to reproduce data, digital works tend to be composed of many autonomous objects. These objects can be used in many different contexts and combinations, and undergo various transformations, during the run of the work.
The output of a hidden program, digital narrative is shaped not only by the general properties of its "material" medium (i.e., silicon chips), but also by the specific affordances of the system through which it is created and executed. An Infocom interactive fiction of the 1980s or a Storyspace hypertext narrative of the early 1990s differs significantly from a Flash game or Director "movie" produced in this new century. The code of authoring programs is a second-order means of expression, and the various software supports should, therefore, be considered the submedia of digitality – just as clay tablets, papyrus scrolls, and codices are the submedia of manuscript writing.
To complete the preliminaries of my discussion, let me briefly define what I mean by narrative. I endorse a medium-free, semantically based definition, according to which narrative is a type of meaning, or mental image generated in response to certain stimuli. A narrative text is an artifact designed to bring this meaning to mind. But the cognitive construct specific to narrativity can also be formed in response to stimuli not expressly designed for this purpose, for instance as an interpretation of life itself. This does not make life into "a" narrative, but it means that life may possess narrative potential – what we may call "narrativity." A narrative script (as I will call the relevant cognitive construct) pictures a world situated in time and populated by intelligent agents. The time span framed by the representation encompasses a series of different states mediated by accidental happenings and deliberate actions. To understand the sequence as narrative means to be capable of reconstructing the motivations of the agents and the causal connections between events and states. As a mental representation of a temporal sequence of events, narrative is not only linear – or multilinear, when it follows several parallel or interwoven destinies – but vectorial: a plot must be followed in a specific direction, from birth to death, beginning to end.
In contrast to some of the narrower definitions endorsed by narratologists, the present approach does not limit narration to the re-presentation of past events by a narrator, but accepts various modalities: narrative scripts can be told (diegetic mode), they can be shown (mimetic mode), or they can be enacted, not for the benefit of an audience, as is the case in drama, but as a self-rewarding activity. In this last mode, which prevails in computer games and participatory environments, the interactor is also the beneficiary of the text. Rather than being necessarily pre-encoded in a semiotic body, moreover, narrative scripts can be dynamically generated during the performance of a text through a simulation of actions and events. In contrast to scholars such as Espen Aarseth, who want to maintain a strict distinction between computer games and narrative (a concept implicitly narrowed down to its literary manifestations), my definition regards games and literary texts (or rather, those games and those literary texts that satisfy the proper semantic conditions) as two different narrative modalities. In the discussion below, I will therefore consider texts conceived as literature as well as texts regarded as games, but I will exclude the rich field of electronic poetry, as well as some innovative digital texts that rely too much on aleatory principles to fulfill the strict requirements of narrative coherence.
The development of digital textuality hit the literary scene at a time when "serious" literature – in contrast to popular culture – was in the grip of an epistemological and aesthetic crisis that challenged the closure, stability, and vectoriality of narrative form. The epistemological issue was a critique of the alleged blindness of classical narrative to the complexity of the problem of truth: facts are asserted by authoritative narrators as a matter of absolute knowledge, and the reader is asked to take the text as the canonical version of the world. For the historian Hayden White (1987), narrative is not a reliable way to gain knowledge about the past, because it always involves a fabrication. Reality, he claims, does not offer itself to perception in the shape of a story. And for the discourse analysts Elinor Ochs and Lisa Capps, a truly "authentic" narrative of personal experience would not be a stable reconstruction of the past but "the airing and evaluating of alternative possible understandings of past events" (2001: 17). The aesthetic issue arose from a similar desire to open up the text to multiple variants. Throughout the twentieth century, as Umberto Eco has shown in The Open Work (1989), artists in many media have been obsessed with the idea of making the text endlessly self-renewable, of capturing infinity in its necessarily bounded body, of turning the work of art from a static self-identical object into a matrix of virtualities. Modern science has recently come up with some models and names for this aesthetic ideal: emergence, complexity, distributed intelligence.
Through its permanent inscription and linear reading protocol, its bound spine and bounded content, the book was perceived as an obstacle to the dream of the text that keeps giving. If the problem came from the limitations of print, the solution might come from the affordances of digital technology. Thanks to the properties of reactivity, interactivity, volatility, and modularity, every run of a digital text can be turned into a performance of different virtualities. Out of a limited number of elements, the computer allows the creation of a vast number of versions. Janet Murray (1997: 155–62) calls this property the "kaleidoscopic" nature of the medium.
Twentieth-century writers did not await the advent of the digital age to develop a kaleidoscopic form of textuality. They did so through the technique of physically "chunking" and rearranging the text. We find the principle at work in the combinatorial algorithms developed by the members of the Oulipo literary movement: novels written on decks of cards which could be shuffled to produce different texts (Marc Saporta's Composition No 1, 1961); or sonnets obtained by selecting each verse from a matrix of poems (Raymond Queneau's Cent mille milliards de poemes). In these last two examples the combinatorial principle is aleatory, and it does not guarantee the production of meaning, especially not of narrative meaning. We can make a picture out of any arrangement of pictorial fragments; and perhaps even a "text" in the same way, since non-sense can have a certain poetic force, but certainly not a story. Consider for instance what would happen if we reshuffled the functions of Vladimir Propp's Morphology of the Folktale (1962 [1928]):
The villain is defeated. One member of a family either lacks something or desires to have something. The hero is recognized. The hero is married and ascends to the throne. The hero and the villain join in direct combat. A false hero presents unfounded claims. An interdiction is addressed to the hero.
An alternative to random sequencing, also practiced by print authors, is a combination of chunking and directed linking. After a given fragment of text the reader is given the choice between two or more alternatives. In "A Story As You Like It" by the French Oulipo member Raymond Queneau, the reader is asked: (a) "Do you wish to hear the story of the three big skinny beanpoles?" (yes – go to b; no – go to c); (b) "Do you want to hear the story of the three meddling mediocre bushes?" (yes – go to d; no – go to e). The expansion of each node creates a decision tree. Because there is only one way to reach a given node, whether terminal or medial, the author has absolute control over the information available to readers at every moment of their itinerary. This prevents the nonsense of the Propp example. But the ratio of number of paths vs. number of units in a tree-based system is not efficient; most works based on this algorithm offer a single-digit number of variants (two, for instance, in Julio Cortazar's Hopscotch (1966); about eight, in a typical Choose Your Own Adventures children's book).
The digital medium offers a much more powerful engine for the creation of multi-variant narrative than either aleatory combination or unconditional branching because the passage from one chunk of text to the next can be meaningfully controlled by code. In a hypertextual system, for instance, the operation of clicking on buttons activates the execution of some machine-language instructions, usually a "goto" to a certain memory address and an instruction to display the group of data beginning at this address. But the "goto" can be embedded in an "if… then… else" statement that imposes conditions on the branching. From the node "hero arrives at castle", for instance, the computer could be instructed to display the nodes "marriage to princess" or "king gives hero mission to rescue princess", depending on whether or not the reader has already visited "hero defeats villain." In a computer game, similarly, the player's actions function as links between segments in the sense that they trigger the execution of code, which leads to changes in the display and in the global state of the system. What players can do at a given moment depends on what they did in the past. The placement of links or the timing of user action in a well-designed narrative or informational system cannot be a random act, unless the point is to signify randomness itself. The creation of multivariant narratives depends on the existence of protocols that maintain linear coherence on the cognitive level – for, as the Propp example suggests, you cannot freely permute events in your mental representation of a story without consequences for the construction of causal and temporal relations. Here I would like to explore some of the system configurations, linking strategies, and modes of user participation that enable digital texts to achieve the difficult task of combining variability with narrativity. My investigation will focus on three aspects of narrative: discourse, which is the presentation of the narrative script through a particular medium, point of view, and plot – the narrative script itself.
Variable Discourse
Most of the digital texts that implement this type of variability are hypertexts, a form of organization described as follows in the documentation for Storyspace, the authoring software published by Eastgate:
The word "hypertext" refers to a specific kind of writing. While customary text appears in a single sequence (e.g. page two always follows page one), hypertext presents the reader with multiple pathways through a document. These pathways are created by using hypertext links, a special kind of hypertext object that permits readers to actuate sequences, just like turning a page…. Simply put, hypertexts are branching textual objects that allow the reader to decide where to go next.
(Storyspace FAQ; available from <http://www.eastgate.-com/storyspace/Download.html>
By the admission of its developers, Storyspace was designed to handle "large texts" (novels and databases rather than poems and short stories) with heavily interlinked nodes. One of the convenient features of Storyspace is the automatic building of a graph that keeps track of the system of links. This diagram gives the writer an overview of the developing network; but it can also serve as a navigational tool for the reader.
Figure 28.1 shows the organizational map of Stuart Moulthrop's novel Victory Garden (1991). This map (inserted as art, rather than generated by the system) represents only the upper layer of the textual architecture: each named site on the map stands for a region of the text with its own finely grained system of links, and it would be necessary to zoom in to a larger-scale image to get a full view of the textual architecture. The configuration of the map of Victory Garden is what theorists would call a network, or unrestricted graph. The presence of circuits – the formal trademark of a network – means that there may be many different ways to get to the same node. The system designer can control the reader's itinerary on the local level (where to go from a given node) but not on the global level. This feature discourages what I call a "narrative" interpretation of the sequence viewed by the reader: an interpretation that narrowly associates the order of appearance of lexia with a chronological and causal chain of events in the reference world. This type of reading would frequently lead to nonsense. For instance, if we first visited a node where a character is dead, and then a node where the character is alive, we would have to imagine a miraculous resurrection to make a story out of the sequence. It would take superhuman intelligence, or a map with very few decision points, for the designer to guarantee that every path will form a logically coherent plot. The map of Victory Garden is not a projection of branching possible parallel times into a spatial image but just what it claims to be: the map of a garden, or labyrinth, or graveyard – in any case a purely spatial construct. Despite its allusions to J. L. Borges's Garden of forking Paths, where paths represent possible futures rather than routes in space, Moulthrop's Victory Garden does not outline multiple destinies for its heroes, but rather traces many pathways into a reasonably solid and chronologically organized narrative core. There is no necessary global parallelism between the progression of narrative time and the progression of the reader on the paths of the garden, though there can be partial relations: for instance a stretch of path without branches or with a default continuation that captures a sequence of events in chronological order. (Victory Garden is particularly rich in such linear paths: they are what enables the reader to get a general idea of the plot.)
Figure 28.1 The map of Victory Garden
A text based on a map as complex as figure 28.1 does not tell a different story for every reader, or with every reading session, it rather tells a story in many different ways, varying discourse instead of plot. Approaching the text like a jigsaw puzzle, the reader rearranges lexia mentally, so that a fragment encountered at Tl in the reading sequence may be assigned time slot T22 in the reader's final reconstruction of the plot. Similarly, the reader of an epic text that begins in medias res will reconstrue a chronological order of events that differs from the order of their presentation.
With its implication of a stable, determinate, complete, and coherent image to be recovered in its entirety, the image of the jigsaw puzzle admittedly fails to capture important aspects of the hypertextual reading experience. How does the phenomenon resist the metaphor? The reconstruction of narrative meaning is hardly ever complete because readers rarely visit all the lexia. Narrative is often only one of the threads in the textual web; following the postmodern aesthetics of the collage, most "classic" hypertexts interweave narration with metatextual comments, philosophical reflections, and inter-textual imports. There may be not just one but many stories in the textual network, just as a print text may tell a complex story with many subplots, or the text may present different versions of the same events without designating one of them as true. (Michael Joyce's afternoon (1987) circles, for instance, among several versions of an accident witnessed by the narrator.) Even in its print form, narrative discourse rarely allows the linearization of all its information. Islands of free-floating events (mostly of the mental type) usually co-exist with streams of temporally ordered states of affairs. Whether print or digital, literary texts may actively militate against narrativity and its linear organization, though many of the texts that sound their demand for a post-narrative mode of signification do so from a platform of narrative fragments that maintain the reader's interest.
The reconstruction of a reasonably consistent narrative world from a scrambled discourse would quickly become a tiresome activity if the reader's role were exhausted by the metaphor of the jigsaw puzzle. Any picture can be cut up, boxed, and sold as a puzzle. A narrow mapping of hypertext onto puzzles would therefore mean that the significance of the reader's involvement is independent of the narrative content of the text. From a literary point of view, the best hypertexts are those that manage to present the reader's activity of moving through the network and reassembling the narrative as a symbolic gesture endowed with a meaning specific to the text, a meaning which cannot be predicted by reading the medium as a built-in message. The hypertextual mechanism does not make a text automatically innovative and significant; it is up to the author to put it in the service of a unique textual idea, of a metaphor that gives meaning to the reader's activity. In Victory Garden, this activity is framed as an exploration of space. With its 2,804 links connecting 993 lexia through a variety of paths planned over long stretches rather than from node to node (so that the narrative retains considerable linear coherence), Victory Garden takes Daedalian architecture to a level of complexity that probably no reader has fully appreciated. Moulthrop's metaphor of the Garden of Forking Paths was so suggestive of a radically new mode of reading, and described the structure favored by Storyspace so well, that it has become something of a theoretical cliche. In Michael Joyce's afternoon, the often frustrated search of the reader for explanations and narrative continuity emulates the frantic calls of the narrator to find out whether or not the accident he witnessed involved, and perhaps killed his ex-wife and son (Bolter 1991: 126). And in Shelley Jackson's Patchwork Girl, the reader is made to stitch together a text out of heterogeneous fragments, some recycled from other texts, just as the narrator-character Mary Shelley assembles a monster (allegory of a multiple, decentered subjectivity) by sewing together body parts collected from different women, and just as Shelley Jackson constructs a narrative identity for the monster from the stories of these women.
Variable Point of View
When the idea of interactive television was first introduced to the public, the advantage of the new technology was presented as an opportunity for spectators to view live broadcasts from a variety of perspectives. The capture of different cameras would be shown in small windows on the side of the screen, and by clicking on one of these windows the viewer would expand the image to the entire screen. Interactive movies, a genre that never really got off the ground, also played with the idea of variable point of view. In the exposition of I'm Your Man (1998), a DVD interactive movie playable on personal computers, we are given the choice between three options, each bearing a character's name. If we choose Leslie, the heroine, we will hear about Leslie's mission to give some important computer disk files to an FBI agent at a party. If we choose Richard, the villain, we will find out about his sinister plan to pose as a fake agent in order to get the documents from Leslie and then to kill her. (This is a very cheesy plot.) If we choose Jack, the fool-turned-hero-in-spite-of-himself, we will simply accompany him on his way to the party, looking alternatively at Jack and with Jack at the derrieres and decolletes of female passers-by. At every decision point we can switch to the corresponding moment on one of the other two narrative tracks. The switches have no impact on the plot, or on the order of its presentation. Time moves inexorably forward, and by selecting a point of view we miss information that will only become available when we stop the movie and start from the beginning again. Because of the temporal structure of film, it takes several passes through the movie to tie the three strands together into one coherent plot.
Purely textual environments can vary point of view without running into this problem because a written text normally exists simultaneously in all of its parts. (Digital texts may change this, by destroying some of the links after the reader has visited them.) In the hypertext short story "A Long Wild Smile" by Jeff Parker (found at <http://www.hypertxt.com/parker/magnetic>), the reader can move back and forth between two narrative strands: one narrated by a woman's fiance and another by her lover. Each node has links to the partner narrative, enabling the reader to follow both versions in parallel. Some links make very quick incursions into the other perspective, highlighting micro-level conflicts. In the following sequence, as Parker explains, the underlined words (which also function as links to the other passage) represent the interpretation by two different characters of the same string of sound:
(Node 1: Fiance's perspective) When I awake, she's joined him in the kitchen. She stares at the fridge while he stares at her. She says, "Did you ever ride ponies?" He says, "No."
(Node 2: Lover's perspective) "Do you ever write rhyming poetry?" she said, taking a big sip of water and passing the cup to me. I drank and passed it back to her.
This is a subtle authorial way to suggest misunderstanding (most likely the fiance's) behind the back of the two narrators. Locked in their own perspectives, the lover and the fiance quote what they hear, and they are not aware that the other interpreted the same words differently. But this kind of effect could have been achieved in print by juxtaposing the two passages and distinguishing them with indentation or with different fonts.
Variations in point of view have been so successfully achieved in "old media" (think of William Faulkner's novel The Sound and the Fury, or of Akira Kurosawa's film Rashomon) that they seldom carry digital texts all by themselves. The online soap operas The Spot and The Lurker Files, which consisted of interwoven first-person narratives that represent the perspective of different characters, have now disappeared from the Web. Interactive film and TV will have to find other selling points to succeed commercially – such as the possibility for the latter to play any document from an archive. The linking design of Parker's text pursues a wider variety of literary effects than merely switching back and forth between two perspectives. Even a work as crude as I'm Your Man combines variable point of view with other types of choices, such as deciding what the characters should do next. It is perhaps computer games that show the most efficient use of variable point of view in a digital environment. Many games enable players to switch from a god's eye perspective, third-person display, through which they see their character as a moving object on a map of the playing field, to a first-person, horizontal perspective that shows the game world to the players through the eyes of their avatar. One view is for the planning of strategy in a suspended time, and the other for the execution of moves in the heat of the game-action.
Variable Plot
The simplest way to implement variability on the level of plot is to arrange multiple narrative developments along the branches of a decision tree. We see this principle at work in Choose Your Own Adventures children's stories, in the Queneau text quoted above, and in postmodern narratives with two or more endings or with different reading paths (Hopscotch). All these examples come from print texts, and all are fairly static. The various narrative lines are pre-inscribed within the database, limited in number (it would take God-like vision to control the combinatorial explosion of an arborescent structure with more than three levels), and the reader activates them through a purely selective type of interactivity. As long as this scheme is maintained, the digital medium does not really facilitate the creation of multiple plots. It makes travel along the branches a little bit easier, since clicking brings text to the reader, while print makes the reader go to the text, but the author still has to write the individual stories. But digital media present one significant advantage over print. They allow the user to interact with the text not just selectively but also productively. It is only when the user contributes elements to a developing story, allowing plots to be dynamically generated at run-time, that a system's narrative productivity can be raised above the level reachable by print media. In this type of system the player's actions perform an individualized narrative by responding to the affordances of the textual world. Since no two players will take the same moves, no two runs of the program will produce the same narrative trace. This productive type of interactivity is found mostly in computer games, although it also occurs in chatrooms, MOOs (Multi-User Dungeons, Object-Oriented), dialogue systems (the famous ELIZA program), virtual reality installations, and projects in Interactive Drama (Ryan 200la). Here I will examine its effect on plot in three types of games: interactive fiction, first-person shooters, and simulation games.
The lowest degree of plot-level variability occurs when the player's actions fill in the blank spots in a pre-determined narrative script. This situation occurs in interactive fiction (IF), a text-only digital genre that flourished in the early 1980s, when computer graphics were too primitive to create an immersive environment. The genre nearly disappeared in the 1990s, displaced from the game market by fancy graphic interfaces and exiled from the front stage of digital literature by hypertext, a genre which quickly became the darling of theorists. But IF survives in small Internet niches, practiced without commercial and theoretical pressures by enthusiastic writers and programmers who freely share their work with a devoted community.
In an interactive fiction, we impersonate a character within the story world. We specify the actions of this character by typing instructions to the system in a simplified version of a natural language on an old-fashioned command-line interface. A parser analyzes the input, and the system responds with appropriate, or sometimes comically inappropriate actions. Since most IPs are designed as adventure games, the user's character has a task to perform, and the plot only reveals itself fully to the user who is able to complete the task by solving all the problems.
The downside of a system that grants free speech to the user is the impossibility of incorporating every verbal action into a coherent dialogue. IF systems understand only a limited vocabulary – typically 200 verbs, and as many nouns as there are different objects in the game-world – and they reject any input that cannot be parsed by their limited syntactic and semantic competence. This means that while the user can technically perform an unlimited number of actions, only some of these actions yield events in the developing story. The following examples, all taken from Spider and Web by Andrew Plotkin (Plotkin, website), illustrate various relations between input and plot. In the first example, player input counts as a turn in a conversation between the player's character and another member of the story world, and it reproduces exactly a verbal event within this world (> represents the output of the system):
Interrogation chamber [You are imprisoned in a chair]
"Don't be absurd", [the interrogator] says. "… if you'd had enough sense to walk away from that door, you wouldn't be here. You don't and you didn't and are; we caught you. And you're going to start by telling me how you got through that door. Do you understand me? Player: Yes.
>The man nods briefly – a man satisfied with the least important detail.
In the next example, the player and the system takes turn at producing the story, though the player's input is not an act of narration but a command to the system. The system responds by updating its knowledge base, by narrating the result of the action, and occasionally (though this is not exemplified in the passage below) by retelling the action to the player in more elaborate language:
>You are carrying nothing of importance, except a lockpick.
Player: Put lockpick on plate.
>The pick locks itself rigidly to the black plate by the door. Its status dot flashes to yellow.
Player: Wait.
>Time passes.
The door slides neatly open.
In the last example, the player's actions cannot be executed because they cannot be parsed by the system. Their event-status within the narrative is uncertain: do they count as futile attempts to solve problems, or are they altogether external to the story world? The response of the system is clearly external, and since it is situated on the same ontological level as the player's attempted action, the whole dialogue may be regarded as taking place outside narrative space and time:
>A plain metal door faces you to the east, near the alley's end. It's firmly shut.
Player: Open door.
>You don't see how.
Player: Scream in despair.
>That's not a verb I recognize.
Interactive fiction thus keeps the plot on the projected track by expelling the actions that do not fit within the script to a sphere that narratologists would call the extra-diegetic or para-textual.
Most interactive fictions trace a narrowly scripted trajectory. Players must visit a number of locations, either in free or fixed order, and they must solve a problem to get to the next location. In Spider and Web, for instance, player A may try the blaster and the lockpick to enter a space before succeeding with the scan scrambler; player B may never solve the problem and be caught by the guards in the hallway (a fate which can be averted by consulting the solutions posted on the Web); while player C may succeed on the first try. The only room for variation resides in the player's unsuccessful attempts. Spider and Web creates an original variation on this pattern by presenting the user's failed actions not as actual events but as lies told by the player's character to an interrogator who wants to know how the character has managed to infiltrate a secret war laboratory. Some interactive fictions have two or three endings, but in general, the variety of the player's input does not translate into an equal variety on the level of plot. Though they can develop very imaginative scripts that spice up the solving of problems with narrative interest, IF texts offer little incentive to re-enter their world once the game has been beaten.
But endless replayability is not a reliable sign of high narrative variability. Consider the case of the so-called "First Person Shooter" or FPS (Wolfenstein, Doom, Quake, Unreal Tournament, etc.), a type of game that is played over and over again by its fans. Here also the user is cast as a character situated in both the time and space of the fictional world. The actions of the user determine the fate of the puppet character, and by extension, the fate of the fictional world. The script of the game simulates a live encounter with opponents, who can be manipulated either by the computer itself, or, if the game is played in a networked environment, by other human players. Whereas IPs allow users to type whatever they want, but filter out a narrow range of acceptable inputs, FPS games accept all of the user's actions as part of the game, but they limit these actions to moving around the game world, gathering ammunition, selecting weapons, and aiming and firing them. If we regard the player's moves as the writing of a "life story" for the character, every run of the system produces a new life, and consequently a new narrative trace. This narrative is created dramatically by being enacted, rather than diegetically by being narrated. Because the speed of the processor and the number of variables involved in the execution of the game script make the system's reaction partially unpredictable to the player, FPS games are inexhaustible matrices of different lives for the player's character. The stories of these lives remain in a virtual state until they are mentally replayed. When players tell about their game experience, they do so by producing a standard "diegetic" narrative. For instance: "I was out of ammo, and Tyrus and Jason were surrounding me, and I thought I was dead, but I made this cool jump, and landed on the other side of the wall, and I found a cache of new ammo. I got back to the dungeon and killed them both, and we won the game." But while FPS games are played over and over again, players rarely "replay" in their minds the story of a game, because these stories only differ from each other in who shot whom and who died when. This monotonous diversity is only compounded by the fact that all FPSs implement the same narrative archetype – the story of the quest – through the same motif: a physical confrontation between hero and villain. Though FPSs have a narrative base, like almost all recent computer games, the stories they generate are worth experiencing in the first person but rarely worth telling to a third party.
For a combination of replayability and narrative diversity, no formula can presently rival the type of game known as simulation or as "god game" – the latter name due to the fact that the player manipulates the objects of the game world from an external position of power. Simulation games create a dynamic model of a complex entity, such as a city (Simcity), an empire (Caesar), or a family (The Sims). An emergent process, the global evolution of this entity is the product of an algorithm that computes the interrelated consequences of the actions of many individual agents and continually feeds this output back into its own engine. In The Sims, for instance, a game whose master plot is a pursuit of happiness that requires a steady climb up the economic and social ladder, the well-being of the family of Bob and Betty Newbie is a balancing act through which this couple of suburbanites must manage the resources of time to satisfy bodily demands (eating, sleeping, using the bathroom), emotional demands (their relation to each other), entertainment demands (satisfied by buying the proper commodities), social demands (getting along with neighbors), and economic pressures (they must earn money to fill their house with pleasure-giving commodities). The game has been criticized for its capitalist philosophy, but the comic texts that ridicule some of the objects available for purchase can just as easily be read as a satire of consumerism.
In keeping with its conception of life as a story that is constantly being written by the interaction between individuals and their environment, The Sims presents each object within the game world as an opportunity for action. A computer, for instance, affords the actions of playing games (good to cheer up depressed characters) or of finding a job (good to expand buying power). Characters are like objects: they too offer opportunities for actions by other characters. By mousing over Betty Newbie, the user playing Bob will for instance discover that Bob can kiss her, hug her, tickle her, or talk to her. The possibilities of action evolve during the run of the program, and since affordances are determined by the global state of the system, as well as by the nature of the objects, the user's choices will always produce a coherent narrative development.
The object of the game is not to win, since life never runs out of problems to be solved, but to manage the life of the characters as successfully as possible according to cultural standards (make them rich), or to the player's personal idea of dramatic development (for instance, create interpersonal conflicts and situations leading to catastrophic events). Some players (the novelists) want to produce specific scenarios; others (the virtual historians) are more interested in discovering how certain actions will affect the fate of the family. It is the same curiosity that makes us wonder "what would my life be like now if I hadn't met this stranger in a bar" or that inspires practitioners of virtual history to write essays about "what would have happened in US foreign policy if JFK had not been murdered." The motivation of both types of players is much more narrative than that which drives the players of FPS. In contrast to the lives of FPS players, the narratives produced by The Sims can be enjoyed retrospectively as well as during the run of the program. The system allows users to retell the story of their game, or to invent their own Sim stories, by taking snapshots of the screen and by complementing their pictures with text. Many of these cartoon-form stories are posted on the game's website. Whether or not they make interesting reading, and whether or not they chronicle actual games, the urge of players to share them with other players provides ample evidence that computer games can produce a genuine narrative interest in their fictional worlds.
The ludic pleasure of deciphering the logic of the system – what game designers call reverse engineering- cannot be separated from the narrative pleasure of watching the story unfold. Without playing skills, the player would be unable to create interesting stories. On my first attempt to play The Sims, for instance, the cooking range caught fire, and because I hadn't bought a phone I could not call the fire department. I watched helplessly my whole family die, and my only option was to let the last survivor mourn for his loved ones before being himself engulfed by the flames. At this point there was nothing to do but start again from scratch. It is only by learning how the system "thinks" that players can increase their authorial power over their family. But since the system throws in unexpected events, the player's control over the fate of the characters is never absolute. For narrative to be pleasurable, appreciators must be able to anticipate to some extent the development of the story, but the story must be able to fulfill expectations in a surprising way. This time-tested formula holds no less for simulation games than for novels and drama.
For all its narrative productivity, however, the formula of The Sims does not entirely satisfy the ambition of Will Wright, the designer of the game. He envisions a networked system that detects the most innovative subplots produced by users, updates the program with devices that facilitate their creation, and sends the new version to thousands of other users to find out if they, too, will attempt these subplots. "So in fact you [will] have the players… cross-pollinating their creativity", but the exchange will be mediated by the computer (Pearce 2002: 6). What Wright has in mind is a narrative variability of a higher power – a variability that affects the storytelling engine.
Beyond Narrative
In all the previous examples, variability remained within the bounds of narrative logic. Let me conclude my survey with an example that starts well within these bounds, but transgresses them during the run of the program, so that the text performs before the reader's eyes the dissolution of narrative. The text in question is The Impermanence Agent, by Noah Wardrip-Fruin and Brion Moss (2002). It consists of two windows on the screen. One of them contains the input narrative: a story inspired by the death of the author's grandmother, Nana, illustrated with family photos. The other window contains texts from various authors and memorial imagery from multiple cultures. The content of both windows scrolls down slowly by itself, then returns to the top, in an infinite loop. A reactive, rather than interactive text which takes full advantage of the networking capability of the medium, The Impermanence Agent modifies the content of each window by gradually integrating materials culled from the user's "scrapbook" – the file in which the Internet browser of the user's system stores text and images from the websites most recently visited. Whereas the original story reads "As a girl, at recess, Joan played punching games with the boys" it may become, "As a girl, at recess, Joan to follow in the Bowlers' footsteps played TeacberSource: Health and Fitness with the boys" (Wardrip-Fruin and Moss 2002: 14). A "lightweight intelligence model" selects "interesting" or frequent words from the user's scrapbook, and makes sure that they fit syntactically into the text, but the model does not check the output for semantic coherence. The visual material undergoes similar blending with pictures from the scrapbook. The authors claim that this algorithm customizes narrative discourse to the reader's interests, but their definition of narrative is so loose that it accepts any grammatical sequence of words. After several loops through the program, all that is left of the original story is bits and pieces of an aleatory collage, as the text is invaded by fragments of other texts that may point towards, but never fully tell their own stories. My point in presenting this example is not to deny artistic potential to this kind of project, but rather to locate the point where multivariant textuality leaves narrativity behind and becomes conceptual art.
The textual phenomena described in this chapter represent two extremes on the cultural spectrum. While computer games have taken popular culture by storm, generating a billion-dollar industry that rivals Hollywood and Disneyland, hypertext is an arcane academic genre read mostly by theorists and prospective authors. What remains to be conquered for digital textuality is the territory that lies between the stereotyped narrative scripts of popular culture and the militant anti-narrativity of so many experimental texts: a territory where narrative form is neither frozen nor ostracized, but recognized as an endlessly productive source of knowledge and aesthetic experiences. In the early 1990s, when theorists embraced hypertext as the genre that would carry the future of digital literature, the concepts of non-linearity and spatiality stood at the top of their list of aesthetic preferences. But narrative, as we have seen, is a fundamentally temporal, and consequently linear form of meaning. It was the great achievement of the twentieth-century novel to have created complex networks of internal relations, through which the present word, passage, motif, or episode activated the energies of other textual elements to form patterns of signification that transcended the linear development of the plot. This process of irradiation produced what critics have called a spatial form. But the greatest spatial novels (for instance Proust's A la recherche du temps perdu) can still be read for the plot, because their spatial organization complements, rather than destroys, the temporality of their narrative layer. Computer games, especially the much-maligned FPS, also present a very efficient use of time and space, if by space one understands the concrete geography of the game world rather than an abstract formal pattern. The pleasure of FPS consists in equal parts of moving through a three-dimensional environment that constantly updates itself to reflect the player's perspective, and of taking the right action at the right moment in a world relentlessly subjected to the ticking of the clock. For digital texts to establish themselves within the cultural middle ground – the narratives of the educated but not professional public – they must do the opposite of what the twentieth-century novel achieved, and perhaps learn a lesson from computer games, without succumbing to their propensity for repetitive themes and stereotyped storylines: naturally spatial, these texts must reconquer the narrative temporality that fuels the reader's desire.
References for Further Reading
Aarseth, E. (1997). Cybertext. Perspectives on Ergodic Literature. Baltimore: Johns Hopkins University Press.
Aarseth, E. (1999). Aporia and Epiphany in Doom and The Speaking Clock: The Temporality of Ergodic Art. In Cyberspace Textuality (pp. 31–41), ed. M.-L. Ryan. Bloomington: Indiana University Press.
Aarseth, E. (forthcoming). Repurposing the Novel - Narrative Literature in the Turning Universe. In The Novel, ed. Franco Moretti. Princeton, NJ: Princeton University Press.
Bolter, J. (1991). Writing Space: The Computer, Hypertext, and the History of Writing. Hillsdale, NJ: Lawrence Erlbaum.
Cortázar, Julio (1966). Hopscotch (Rayuela), tr. G. Rabassa. New York: Pantheon.
Douglas, J. Y. (2000). The End of Books - or Books Without End? Reading Interactive Narratives. Ann Arbor: University of Michigan Press.
Eco, U. (1989 [1962]). The Open Work. Cambridge, MA: Harvard University Press.
Hayles, N. K. (2000a). Flickering Connectivities in Shelley Jackson's Patchwork Girl'. The Importance of Media-specific Analysis. Postmodern Culture 10, 2. At http://www.iath.virginia.edu/pmc/text-only/issue.lOO/10.2hayles.txt.
Hayles, N. K. (2000b). The Transformation of Narrative and the Materiality of Hypertext. Narrative 9, 1: 21–220.
I'm Your Man (1998). Dir. B. Bejean. A Choice Point Film. Presented by Planet Theory in association with DVD international. DVD edition produced by B. Franzblau.
Jackson, S. (1995). Patchwork Girl. Hypertext software. Cambridge, MA: Eastgate Systems.
Joyce, M. (1987). afternoon, a story. Hypertext software. Cambridge, MA: Eastgate Systems.
Joyce, M. (1995). Of Two Minds: Hypertext, Pedagogy and Poetics. Ann Arbor: University of Michigan Press.
Koskimaa, R. (2000). Digital Literature: From Text to Hypertext and Beyond. Unpublished PhD dissertation. University of Jyväskylä Available at: http://www.cc.jyu.fi/~koskimaa/thesis/.
Landow, G. (1997). Hypertext 2.0: The Convergence of Contemporary Critical Theory and Technology. Baltimore: Johns Hopkins University Press.
Manovich, L. (2001). The Language of New Media. Cambridge, MA: MIT Press.
Montfort, N. (2003). Twisty Little Passages: An Approach to Interactive Fiction. Cambridge, MA: MIT Press.
Moulthrop, S. (1991). Victory Garden. Hypertext software. Cambridge, MA: Eastgate Systems.
Murray, J. (1997). Hamlet on the Holodeck: The Future of Narrative in Cyberspace. New York: Free Press.
Ochs, E. and L. Capps (2001). Living Narrative: Creating Lives in Everyday Storytelling. Cambridge, MA: Harvard University Press.
Ong, W. (1982). Orality and Literacy. The Technologizing of the Word. London: Methuen.
Parker, J. A Poetics of the Link. Electronic Book Review 12. Accessed April 21, 2004. At http://altx.com/ebr12/park/park.htm.
Pearce, C. (2002). Sims, Battlebots, Cellular Automata, God and Go. A Conversation with Will Wright. Gamestudies 2. At http://gamestudies.org/0102/pearce.
Plotkin, A. Spider and Web. Interactive Fiction. Accessed April 21, 2004. At http://www.wurb.com/if/game/207. Solution to the game (by PJG, modified by Nils Barth) at http://www.ifarchive.org/if-archive/solutions/tangle.sol.
Propp, Vladimir (1968 [1928]). Morphology of the Folktale, tr. L. Scott, rev. L. Wagner. Austin: University of Texas Press.
Queneau, R. A Story As You Like It: Interactive Version. Accessed April 21, 2004. At http://www.thing.de/projekte/7:9%23/queneau_l.html.
Ryan, M.-L. (2001a). Narrative as Virtual Reality: Immersion and Interactivity in Literature and Digital Media. Baltimore: Johns Hopkins University Press.
Ryan, M.-L. (2001b). Beyond Myth and Metaphor: Narrative in Digital Media. Gamestudies 1. At http://www.gamestudies.org/0101/ryan/.
Saporta, Marc (1961). Composition No 1. Paris: Seuil.
Sloane, S. (2000). Digital Fictions: Storytelling in a Material World. Stamford, CT: Ablex.
The Sims (2000). Computer game. Designer: Will Wright. Maxis/Electronic Arts. At http://www.thesims.com.
Wardrip-Fruin, Noah and Brion Moss (with A. C. Chapman and Duane Whitehurst) (2002). The Impermanence Agent. Project and context. CyberText Yearbook, ed. Markku Eskelinen and Raine Koskimaa (pp. 13–58). University of Jyvaskyla: Publications of the Research Centre for Contemporary Culture.
White, H. (1987). The Content of the Form: Narrative Discourse and Historical Representation. Baltimore: Johns Hopkins University Press.