Theoretical introduction

Today, no general theory of the computer game is available, but as texts and as objects of study, they most obviously fit in with the study of non-linear texts and hypertexts (like the World Wide Web). In both games and hypertext, the reader/players efforts are assigned a role in relationship to the game, but it is not quite the same relationship: Hypertext theory focuses on verbally based texts consisting of small pieces of text, the reader must move between. Unlike this, the computer game is mostly graphical and based on the continuous combination of elements.

It is basically possible to plot the structure of a hypertext. The simplest model is that of the tree structure which continuously forks:

More complex hypertexts are usually not based on forking at regular intervals, but are better described as any number of nodes connected by any number of links. A computer game is a more complex phenomenon since it is usually a simultaneous combination of elements, which can then not be represented as nodes with connections. In a computer game, the reader/player is not at a single "location" as the reader is in hypertext, but one could rather see the game as a complex system in a given "state" at any time: the spaceship of the player at a specific position, the enemies heading at a specific position; every object with a velocity in a specific direction; the player has a score and a number of lives left. But both hypertext and computer games carry limitations and rules for movement between nodes of text or states of the game system.

Computer games

According to the game designer Chris Crawford, computer games have four basic characteristics (Crawford 1982):

1. Representation: A game is a closed formal system that subjectively represents a subset of reality. (By subjective, Crawford means that a game is not necessarily trying to represent reality.)
2. Interaction: The game acknowledges and reacts to the player. (Unlike a puzzle, which simply lies still.)
3. Conflict: A game presupposes a conflict. This can be either between several players or between the players’ goal and whatever prevents the player from reaching that goal.
4. Safety. The player is safe (in a literal sense) from the events in the game. (Gambling presents a special case, where the outcome of the game is designed to have impact in the real world.)

The most problematic point of Crawford’s definition is probably the first one, representation, since it does suggest that games have a mimetic relationship to the world. This is certainly not true for a game like Tetris. Starting form these four points, this is the definition of computer games I’ll be using in this text:

This definition explains the difference between game and laws of traffic (traffic does not ignore the rest of the world), between computer games and children’s play (children’s play is not based on formally defined rules, rather the rules are under constant renegotiation, and play does not necessarily entail an evaluation of the player), between game and stories (a story does not evaluate the efforts of the reader, and stories can hardly be described as formally defined).

Though gambling basically doesn’t evaluate the player, it is general of obsessive gamblers that they think themselves capable of feeling when luck is coming their way; they think that the outcome is an evaluation of their skills.

In this thesis I’ll be assessing the quality of many computer games. As with other cultural phenomena, there is no simple procedure for such assessments, but the computer game does differ from things like novels and movies in that there is no apparent conflict between "high" and "low" computer games; the game reviewers generally seem to enjoy the games that are also widely popular. Still there are constant discussions of what a quality computer game really is. In the professional magazine Game Developer, Tzvi Freeman has put forward an attemptive list of traits of good and bad games, the first three points being these:

1.	A good game empowers your imagination.	A bad game gets in the way.
2.	A good game makes you feel in charge.	A bad game restricts you with artificial restrictions.
3.	A good game is transparent. You only feel your own mind, the other players, the ideas.	A bad game keeps reminding you that a game is there.

Many similar lists have been created, but this one does give us a hint that freedom ranks high, and that limitations should be motivated.

Interactivity

Games and hypertexts share the trait of interactivity. I do not consider this an impossible-to-define term, but interactivity has been blindly and widely used to cover both phenomena that are necessarily interactive ("interactive games"), phenomena where the term is meaningless ("interactive discussions"), and as a general selling point for anything with the slightest relation to computers ("interactive education", "interactive exhibition"). This is a fate also shared by the terms hypertext, non-linear, virtual, cyberspace and multimedia: Excessive and uncritical use in both advertising and theory.

The Danish theorist Jens F. Jensen has compiled a detailed historical description of different definitions of interactivity, some clearly too specific (the number of laser disc players in a system), some lost in fruitless discussion of whether email or solitaire card games are the most interactive. (Jensen 1998). In most cases the term - that originally was supposed to signify what is new about information technology - is hopelessly inflated and unusable. Some definitions describe interactivity as a social phenomenon between people, and Brenda Laurel claims it to equal the feeling of being present in a world (agency). For my purposes, it would be more useful to describe the structural properties of an interactive work. The fairly purist definition of Peter Bøgh Andersen states this:

So it is not enough to be able to stop or start a movie, since it has no meaning within the movie itself. The same goes for switching channels on a TV. In this definition of interactivity, there has to be some kind of signifying processing of the user’s input. An on/off button is thus not interactive. Hypertext is interactive when the choices in reading are meaningful in the world of the text, for example when the context plane of the text is different according to the choices made by the reader. Computer games are interactive because the actions of the player play a part in determining the events in the game.

In my experience, many people will protest if I say that the World Wide Web or an electronic encyclopaedia are not really interactive, though the same people would not describe a paper-based encyclopaedia as interactive. The above definition can therefore be extended to distinguish between interactivity on the level of the discourse, and interactivity on the level of the story. A hypertext (such as an encyclopaedia) where you can read about a subject in different levels of detail can be described as interactive on the discourse level. But this is neither as new or as interesting as interactivity on the level of the story. (Andersen 1997). I will only use interactivity for interactivity on the story level.

George P. Landow

An interactive text gives their reader a kind of freedom that the reader of a traditional text doesn’t have. This has been described as a situation where the reader assumes the position of the author, since the reader now determines the text. This idea of hypertext as a liberating form has for a long time been the dominant way of describing the domain: The most influential theory so far comes from George P. Landow’s book Hypertext: The Convergence of Contemporary Critical Theory and Technology (1992, revised 1997). Where the humanities have hitherto seen the computer as connected to a rational, centralist, modern project, Landow now claims hypertext to be a liberating wholesale confirmation of Derrida’s deconstruction and all of poststructuralism:

This means that Landow reasons by way of analogies that assume deep connections between phenomena that could well be claimed to have only superficial similarities: Derrida’s theory on decentering is about the collapse of hierarchies in all texts, not in hypertexts. And Roland Barthes’ distinction between lisible (readerly) and scriptible (writerly) texts is a distinction between different literary texts. Landow replaces the objects described by Derrida and Barthes with a new one, which he then claims to fit their theories even better:

Unfortunately, this is a quite unconvincing interpretation of Derrida. If one were to briefly sketch the method of Derrida, he is very critical of western metaphysics, that is, the metaphysics of presence and logocentrism: Derrida claims that western thought has privileged certain things above others. It has privileged the spoken word is assumed to be a direct reflection of a prior intention. (In speech, thought is present.) Contrary to this, Derrida notes that speech is also a form of writing, and that writing is always characterised by displacements in time and meaning (differance). Accordingly, philosophy has repressed that it is also writing; philosophy has been assumed to be a series of thoughts that were afterwards expressed in clear and transparent language. But philosophy is also determined by the language it tries to speak (and it’s lack of stability). It is subject to discussion whether Derrida denies the existence of something outside language. But no matter, the philosophy of Derrida does not contain an idea of actual, language-independent experiences, so hypertext obviously cannot correspond to them. Hypertext is clearly also writing, but Landow tries to assign hypertext a status as present to the thought, a status that is remarkably close to the privileging of speech critiqued by Derrida.

The merit of George P. Landow is rather that he has pointed to hypertext and technology as something that is not in radical opposition to "culture" or "the book", but is something that can be studied by the humanities. He has made it clear that many of the characteristics of hypertext can be found, in actuality or prophesied, in earlier theory and in literature. He has collected the canon of texts, that are now considered central: Theoretically, Vannevar Bush’s 1945 article As we May Think (Bush 1945), where the concept of hypertext is introduced as a mechanical desk full of microfilm. Theodor Nelson’s essays from the 1960’s, where the term hypertext is introduced. On the literary level it is the electronic text Afternoon by Michael Joyce (1989), where the reader can click through a story using a mouse, and the Jorge Luís Borges short story The Garden of Forking Paths (1941, 1962)

Landow states that forking texts and stories assign new power to the reader, and in the most extreme cases makes the reader a new author. But this is not entirely true: A work has still been produced by one or more people. On the World Wide Web, the reader can only change his/her own text. And in a game or a story with many endings, the reader cannot make a new text, but only choose between the possibilities created by the writer/programmer. The story can be new every time, but only new within some predefined constraints. The reader has a new kind of influence, but is not in power.

Espen Aarseth

A newer and more practically useful description of the subject can be found in Espen Aarseth’s book Cybertext: Perspectives on Ergodic Literature (Aarseth 1997). Where Landow has a tendency towards the general, Aarseth is closer to specific texts and tries to differentiate between different types of texts. His main concept is that of cybertext:

A cybertext is both an ordinary text and something more, a machine capable of generating several manifestations of the same material. In relation to the cybertext, there is the ergodic:

A cybertext is defined by that the text can be combined (or can combine itself) in new configurations. The ergodic is characterised by that the effort of the reader is greater than that of interpretation; an ergodic text presupposes activity from the reader. These two terms are fairly broad in scope, including a long list of different phenomena from hypertext to computer games. Hypertext is simply a subcategory of the larger cybertext.

The starting point of Aarseth is to some extent the opposite of Landow's. Where Landow argues in analogies, with rhetorical similarities between technology and literary theory, with hypertext as a radical break with earlier forms of text, Aarseth experiences this kind of reasoning as a large didactic problem:

The ideological project of Aarseth is primarily that we should not assume that there are radical differences between print and electronic texts. There can be greater similarities between an electronic and a print text than between two print texts.

The problems in terminology also apply to a word like "labyrinthine". Many hypertexts and games are labyrinths in a literal sense; as a reader you much search for the exit. At the same time it is not uncommon to use the term as description for "difficult" texts like Ulysses or just large novels like In search of Time Lost. In the literary sense, "labyrinthine" is a metaphor, in hypertext it is literal. As a way out of this problem, Aarseth suggests the terms unicursal/multicursal. A unicursal labyrinth is characterised by having only one route from entrance to exit; the multicursal labyrinth contains several possible paths. When Ulysses is characterised as labyrinthine, it is in the unicursal sense. Hypertexts or games are basically multicursal.

A theory like this, based on textual phenomena, inevitably faces some problems in relationship to the more visual and plastic form of the computer game. The unicursal/multicursal pair carries the assumption that the player always is in one specific "place" in the "text". But computer games are unfortunately not fixed sequence, but combinations of different materials. Accordingly, Aarseth makes a distinction between textons (the pieces of text in the text) and scriptons (the pieces of text presented to the user/reader). In a more advanced hypertext (like the game Adventure), the user is introduced to scriptons that are combinations of the implicit textons of the game. On top of this are the traversal functions of the text; the mechanisms determining what scriptons the user is introduced to and how they are combined of the textons in the text.

In an attempt to create an overall strategy for the categorisation of texts, Aarseth then tries to categorise all texts according to seven parameters for their traversal functions:

1. Dynamics: In a static text, the number of scriptons is constant. In some texts, the contents of the scriptons change (the text is then intratextonic dynamic). In other texts, both content and number of scriptons change (the text is textonic dynamic).
2. Determinability: Whether the text develops according to fixed causality; with the same event always leading a specific other event. I.e. if there is chance involved.
3. Transiency: Whether the text develops regardless of user activity; if time is a factor.
4. Perspective: Whether the user is set as playing a role in the world of the text (personal) or not (impersonal).
5. Access: Whether access to different pieces of text is controlled; if the reader has immediate access to all pieces of text. (Which you have in print text, but often not in an electronic text.)
6. Linking: If there are links between different scriptons, if this link is explicit, if the linking is conditional (requiring, for example that you’ve read specific other pieces of text).
7. User functions: If the user/reader does more than interpret. In an explorative text, the user chooses a path. In a configurative text, the user explicitly chooses between scriptions. In a textonic text, the reader can add new textons or functions.

The 7 parameters lead to a total of 576 different types of text (3 x 2 x 2 x 2 x 2 x 3 x 4). This gives us the hope of being able to sketch some clear genres. Aarseth attempts this by categorising a total of 21 texts, such as Moby Dick, Afternoon, and Adventure. The problem with this attempt is that most texts do not fit neatly into one category or the other. The game Witness (Infocom 1983) has an identical introduction every time (static), but much of the game is intratextonic dynamic. In regards to time, a modern action game like Doom II (ID Software 1994) is transient (time matters) when you are faced with opponents, but intransient once these opponents are defeated. Regarding access, a game is always a balance between controlled access and random access. Otherwise it s not a game. (This is connected to the previously mentioned definition of Chris Crawford, that a game has to have a conflict.)

It is my opinion that Aarseth’s text categories do not work as a general categorisation of texts. But the model is useful as a list of seven possible perspectives on texts. I do think that Aarseth is right in describing these seven parameters as interesting and central to texts, they are just not either/or questions. I have chosen to use them as qualitative characteristics.

Literary precursors

Even if this thesis has the computer game as its central field of study, it must be said that many of the techniques in the computer game; non-linearity, combinatorics have been utilised long before the advent of the computer. The most popular example is the short story The Garden of Forking Paths by Jorge Luís Borges (In Borges 1941, 1962). During World War I, the main character Yu Tsun visits the Sinologist Stephen Albert. Albert tells Tu Tsun about his grandfather Ts’ui Pên: Ts’ui Pên had once declared that he wanted to 1) write a book and 2) build a labyrinth. Nobody every found the labyrinth, just a book that seems a mess of contradictions. Stephen Albert then tells Yu Tsun that the book and the labyrinth are the same thing. In the book, every chapter is followed by "every" possible continuation:

This is connected to a discussion of time as such, described as a plethora of parallel and possible futures in simultaneous existence. Regardless of this story’s fame, it is still just a description of a non-linear text.

Another story by Borges, An Examination of the Work of Herbert Quain (In Borges 1941, 1962) is not as famous but quite relevant: The narrator tells us of the newly deceased author Herbert Quain and his work. The most interesting work by Quain in this context is the novel April March, which is split in 3 levels and 13 parts. Each part is followed by 3 other parts. The novel can be read in a total of 9 ways, each written from a unique viewpoint: Psychological, communist, anti-Communist etc.:

In the 1960’s many of these ideas were realised by the French OuLiPo-group (Ouvroir de Littérature Potentielle). The group (counting famous members like Raymond Queneau, Georges Perec and Italo Calvino) worked by creating mathematical, logical or language systems that were used in the writing process. This lead to Raymond Queneau’s classic Cent Mille Milliard de Poèmes (1961), where 10 sonnets can be combined on a line-by-line basis to form a total of 10¹⁴ formally correct sonnets. Another story, Un conte á votre façon, begins with the question: "Would you like to hear the story of the three alert peas?" (OuLiPo 1973, p.273). The text continues according to the reader’s choice, and we can choose whether the peas should dream, what colour their gloves should have. So this text is a staging of a situation with a slightly dishonest narrator and an audience.

There is no easy way to demark this area; large amount of marginal phenomena do something fairly similar in slightly different ways. Among movies I’d like to mention Aki Kurosawa's Rashomon (1950), where the same story is told three times from different points of view; Kieslowski's Blind Chance (1982) where the same situation – a young man trying to catch a train – is continued in three different ways. In Harold Ramis’ popular movie Groundhog Day (1993) the main character relives the same day countless of times until he finds true love (and inner peace). In Peter Howitt's Sliding Doors (1997), a woman tries to catch a train, which leads to two different parallel continuations.

In literature, The French Lieutenant’s Woman by John Fowles (1969) has three different endings, even they are presented in a fixed sequence (unicursally, like Blind Chance). Julio Cortázar's Hopscotch (1963, 1966) consists of number paragraphs and describes itself as two novels: One from paragraph 1 to 56, the second is a sequence of paragraphs in a quite complex sequence. Nabokov's Pale Fire (1962) is an introduction, a poem, and footnotes to that poem – and can thus be considered a borderline case between the unicursal and the multicursal: You can read it from beginning to end, but it’s unlikely that anybody would do this.

A Danish angle

There is a small Danish tradition in this area. Inger Christensen's Azorno (1967) does approximately what Rashomon does - In Azorno several voices claim to be the narrator of a story. 6512 by Per Højholt (1969) is a novel where the pages have been torn out and rearranged, meaning the order of reading in principle can be chosen freely. But the primary Danish practitioner is surely Svend Åge Madsen. Tilføjelser ("Additions", 1967) consists of 5 folders in a box, meaning that the order of reading is free. These folders are hardly about anything, but mostly comment each other from specific ideological standpoints. In the short story Den slette fortæller ("The evil narrator", 1970) a story is told where the protagonist Borg both stays on and leaves a public square:

Both paths are taken, and this is continued in 6 levels. (Meaning that the last level has 64 pieces of text.) The story forks, and it might reasonably be described as non-linear. Typographically, however, the text is set up for being read from top to bottom, unicursally. In the novel Dage med Diam ("Days with Diam", 1972), the forks are mapped on page 1, and each of the possible continuations have a separate chapter. So the text more openly invites the reader to make choices. In the short story Spor ("Tracks/traces") in Af sporet er du kommet ("You have come from tracks/traces", 1984) the choices have more of a game character; things can go well or less well for the protagonist. So Svend Åge Madsen initially uses the forks as a multitude of possible and parallel futures. Gradually, the role of the reader becomes more important, and the texts become more game-like. Spor does seem influenced by the Choose your own adventure series, where the reader has to make choices of the sort "Do you want to attack the dragon?" (This series was first published in 1979.)

Astrology and cut-up

Mathematical principles (such as chance) have been used in the writing process by writers such as William Burroughs in his cut-up strategies (I refer to his essays Les Voleurs and The Fall of Art (1986)). Espen Aarseth has noted that such techniques go much further back, as a minimum to the 3000-year-old Chinese oracle system I Ching, where a 6-line text is constructed by means of chance. It could be added that this is characteristic of most divination: the use of chance interpreted as signs from Divine Providence or similar. The cut-up methods, often interpreted as loss of meaning or coherence in the world, have parallels in some practices with the exact opposite worldview. Today, the by far the most popular religious combinatory technique is clearly astrology, where a number of pieces of text are combined according to mathematical principles with the time and place of birth as sole input. Astrology is not interactive, but I find it surprising that astrology shares the techniques with Raymond Queneau, William Burroughs and computer games like Quake or Myst.

On the other hand, this surprise builds on a common, but questionable opposition between the "sensitive" and "technical". For example, the Danish poet Inger Christensen has written both a long poem based on the mathematical Fibonacci series (Christensen 1981) and a sonnet (1991). The sonnet is in no way less of a system than the Fibonacci series, but the sonnet is not counted as a technical experiment since it is part of the literary tradition. Raymond Queneau is not oblivious to this connection, and in his essay Potential Literature (Motte, p.51-64), he examines the sonnet as one system among historical and disused systems like the sestina and the triolet along with newer inventions like S+7 (exchanging every substantive in a text with the seventh following in a dictionary) and the lipogram (texts where specific characters are forbidden).

This does not mean that multicursality and combinatorics can be used for any possible purpose. They do seem to lend themselves to many viewpoints and modes of thinking - it is hard to describe their ideology as such. They are fixed on more specific and technical levels.