Theories of Interactive Narrative.

An Introductory Literature Review
by Zach Tomaszewski

for CIS 702, Fall 2004, taught by Dr. Martha Crosby


Interactive narrative is a relatively new field, developing mostly over the past 15 years. However, the basic components of interactive narrative--story narration, composition, interface design--have been around much longer. Here, we look at the inherent features of interactive narrative, as well as some of the current and historical theories that attempt to define the field.

Theories of Narrative

Narrative means simply telling a story, or giving an account of events (American Heritage 2000). Narrative is incredibly pervasive. We find it in books, movies, daily conversations, ads, magazines, posters, essays, newspapers, myths, and games. Whenever we are relaying our experiences to another person, we tend to use narrative to do it.


Aristotle offers the oldest definition of narrative in his Poetics. He claims that narrative is an imitation of an action, a whole complete in itself, with a beginning, middle, and end (Meadows 2003). As an imitation of an action, narrative imitates the events of life.

Since most narrative was conveyed through drama and oration during his time, Aristotle describes the elements that comprise a narrative drama. These are:

Moving from the top to the bottom of the list, we can trace the formal cause. (A formal cause is the design or goal-oriented purpose of something. For instance, the blueprints would the formal cause of a house.) An author determines a plot. This plot dictates what sort of characters are needed, which in turn have certain thoughts or motivations, which they express in certain language (or diction). These form certain patterns, such as songs, rhythms, or motions. It is these patterns that shape the spectacle that the audience perceives.

Moving from the bottom of the list upwards we can trace the material cause of the drama. (A material cause is the material substance that gives a thing its physical qualities. The bricks and mortar would be the material cause of a house.) The audience perceives the spectacle, through which they can perceive the patterns, language, and thoughts of the characters. Through understanding the characters and their actions, we discover the plot. (Aristotle, via Laurel, via Mateas 2000).


Gustav Freytag, in his Technique of Drama of 1863 describes his famous Freytag Triangle. Freytag graphs the plot of a story verses time. When graphing the plot, we are interested in its "density" or the "thickening" of the focus on the primary problem of the story. This should correspond with the reader's interest.

And so we learn that stories have three main parts (as Aristotle pointed out--a beginning, middle, and end.) The first is the Desis--complication or rising action. The second is the Peripeteia--the "thickest" part of the plot, the climax or crisis. And finally the Denouement--solving the problem or falling action (Meadows 2003).

Freytag Triangle - Standard.

In practice, Freytag triangles tend to be more complex. The Denouement is usually briefer than the other parts. The Desis does not tend to rise smoothly but may include a number of minor climaxes (Binsted 2003).

Freytag Triangle - More realistic.

Other elements

Though these are the two most commonly cited theories of narrative, they do have some failings. For instance, it is arguable whether plot really determines characters for all authors, as Aristotle implies. (Some authors report that they start with good characters and their interaction produces a plot.) As Meadows (2003) points out, mystery stories remove the desis from Freytag's triangle. Instead, they may begin with the Peripeteia--a murder, for instance--and the rest of the book is Denouement, or solving the problem. At the very end, the Desis is revealed--who did it and how.

But most importantly, these theories do not mention the myriad other elements of story besides Plot (the series of events that comprise the story). Yet most of these other elements are covered in high school English classes.

Characters. A story must have at least one character, or protagonist. There may be a number of other characters in different roles: antagonists, friends, love interests, mentors, sympathy characters, background characters, etc.

Conflict. There must be some sort of tension to resolve or problem to overcome. (This will generally correspond to Freytag's plot "thickness.")

Point of View. The narrator has a certain view and omniscience concerning the action, which determines her objectivity and what we know about the story. This includes 1st person, 2nd person, and 3rd person points of view.

Theme. Authors tend to have a reason to tell a story, or at least to tell it a certain way. They reveal a certain message or view about life that is encapsulated in the theme. (Meadows (2003) holds that this conveying of an author's perspective or opinion through a narrative is in fact the defining element of narrative, even more so than plot.)

Genre. Narratives can be categorized into a number of different types, such as mystery, fairy tale, swashbuckler, romance, western, fantasy, science fiction, etc. Each of these has a collection of rules or norms associated with it. For instance, a heroic fairy tale has the following basic formula: the innocent hero sets out on a quest, gains skills and friends through adventures along the way, vanquishes the evil he set out to destroy, and returns home wiser for his journey.

There are additional, minor elements that affect or influence the above. These include timing (the pace or speed of the plot), setting (the location of the action), mood (the general "feel" of the story conveyed through the setting, timing, and characters), and symbols (objects or characters in the story which stand for larger, abstract qualities). All of these elements are shaped by the narrative's medium, whether this be oral, print, visual, or multi-media (Kardan 2004, Binsted 2003).

Alternative Models to Narrative

Though narrative seems the logical starting place for understanding interactive narrative, it may not be the foundation most able to support interaction. Alternate foundations for interactive narratives are drama, improvisational theatre, and traditional role-playing games (Louchart 2002).

Though very similar to narrative, Brenda Laurel claims there are a number of differences between drama and narrative. Drama relies on enactment and action to reveal the text; narrative relies on description. Drama offers an intensified experience by condensing time and emotion around an event. Narrative extensifies by expanding time in order to provide multiple interpretation or perspectives on events. Drama is more likely to maintain a "unity of action", where all events are casually linked to a central action or theme. Narrative frequently employs an episodic structure where many incidents are causally unrelated (Mateas 2000). In short, drama tends to be a more focused and immediate experience, whereas narrative tends to be more exploratory. There are a number of scholars who focus on interactive drama rather than interactive narrative.

Related to drama is improvisational theatre, or "improv". Depending on the particular "game" or rule set of the moment, improv actors produce stories on the spot from random gestures or poses, audience-contributed sentences, genre and setting pairs, etc. Though very free-form, there are some guidelines to produce better improv (Binsted 2003):

Many of these guidelines and many of the improv formats could provide a basis for automated story generation.

Similar to improv are children "pretending". This subtle art is usually grows rusty as we grow older, but much like improv, children can assume roles and generate stories spontaneously. These can develop into episodic stories or games that extend over a period of weeks or months.

Finally, the study of traditional role-playing may offer some insights. Though table-top role-playing games (such as Dungeons and Dragons and GURPS) and live-action role-playing (such as White Wolf's Vampire) are basically as old as interactive narrative itself, they have been successful for nearly 30 years. While there is little theory, there is much practice. And while they do restrict the game to the bounds of the rule set (unlike pure "pretending"), they do not restrict the story to the bounds of some system implementation. You are constrained only by the flexibility of the game master (GM) leading the session and the imagination of the players.

GMs usually have a general plot in mind for the players to follow. What is interesting is how a GM can choose to handle actions proposed by a player that he did not anticipate. He could:

There are no pre-written rules explaining how to handle unanticipated user actions. The details depend on the particular GM.

Nature of Interaction

The other half of interactive narrative is interaction. There are a number of theories of interaction.


Meadows (2003) characterizes interaction in terms of the following principles:

Interaction comprises of four steps for Meadows:

Though he praises open, less predictable system, Meadows also calls for enough redundancy in the interface to allow the user to determine recurring contexts that reveal what decisions are currently possible. Our actions should have predictable outcomes; yet an engaging interactive narrative will hold some surprises for us, as good narratives usually do.


Another theory of interaction comes from Donald Norman (1990). A system generally operates based on the designer's model of it. Ideally, the system interface should communicate the system's affordances (what can be done to the system) and its internal state to the user. Based on this interface information, the user develops her own conceptual model of the system. Ideally, the user and the designer have the same conceptual models, allowing the user to use the system as intended.

There exists a gulf--a cognitive distance--between the system and the user model. This gulf comes in two flavors--the gulf of evaluation and the gulf of execution. The gulf of evaluation affects how easy it is to determine the state of the system. The gulf of execution affects how it easy it is to purposefully change the state of the system to reach some end.

For Norman, there are 7 stages of action when interacting with a system:

Interaction Basics

These two theories give us a sense of interaction. Meadows's principle of Input/Output implies Norman's discussion of a system interface revealing its state and affordances. Meadows's Inside/Outside distinction is part of Norman's model of a designer's mental model being transferred to a user through a physical system interface.

When it comes to interacting with a system, we see similarities in the process as a whole: determine what can be done (possibly through experimentation), read what the system is ready for, and intentionally change that system through its interface controls. This process changes the system, and so the user is changed as she begins to explore the new state. It is a cyclic form of user-system communication.

Normally in discussions of interactive systems we mean "computers", but this need not always be the case. Even when interacting with other people, we determine what state they are in (mood, wakefulness, etc) and how we might interact with them (can they see and hear us). As we communicate, they change and respond to us. As their state changes, so does our own. This reciprocal change is essential to an engaging interaction.

Combining Interaction and Narrative

So far, we have seen that narrative is series of events, taken as a whole with a beginning, middle, and end, as conveyed from a certain perspective. Interaction is a reciprocal communication between a system and a user. Yet plots are generally static once created by the author. How can we give a reader the ability to affect a narrative?

Traditional Controls

In the past, the audience has had some small element of control over a story. Particularly with live performances, such as theatre or storytelling, the audience's response can affect the actors or storyteller. The presenter of the tale can slow down or speed up, skip over parts, or linger over other parts. Yet these changes remain in the hands of the teller. The audience's control is only indirect.

The audience can participate in the telling of the tale by booing, hissing, cheering, clapping, or singing a refrain. An obvious example of this is showings of the movie Rocky Horror Picture Show, with its accompanying live performance and audience participation at different parts of the film. Certainly participation in the telling can increase enjoyment of a narrative. It requires closer attention to participate at the right times. It involves the user emotionally, since the experience as a whole changes depending on their response. However, the narrative itself is largely unaffected. Participation is, at best, only a poor man's interaction.

With print and other static medium, the user has been able to control the timing of the narrative. A reader can skim parts, reread others, put the book down and come back to it later. In a museum, a viewer can glance at a painting as they walk by, or study it for half an hour. As we move to modern media, we should remember to maintain this user control of timing.


The most important and obvious way to interact with a narrative is to change its plot. However, this is very tricky to do. As Aristotle suggests, the characters, thoughts, actions, and other "material" elements of the story build on each other. As authors, we have a story we want to tell, a perspective to share. We want to allow readers the chance to interact with the story, to be changed by it and make it their own. But we don't want to give up our story in the process. If we are telling a heroic fairy tale, we may not want to allow the reader/hero to kill the princess and marry the dragon. This may make for an interesting story, but as authors, it may not be the story we want to write. Additionally, if we give the reader complete control over the plot, then we find that there is no plot unless the reader creates one. The reader may not care to be left filling the "author" position as well.

Meadows (2003), citing Raph Koster (lead designer of Ultima Online), places interactive plots along a line between impositional and expressive designs. Completely impositional plots impose themselves on the reader, allowing little or no interactivity. Expressive plots provide hardly any plot at all--practically only a setting and system through with the reader can express herself.

Meadows gives us three example categories of plots along this continuum. The first, a nodal plot, consists of a number of nodes in a line, from the beginning of the story to the end. This mirrors traditional, static plots. However, each node represents a place where the user can control the interaction. In a nodal plot, the reader aims to uncover the plot, i.e., by completing each task or level needed to advance to story. Failure on the part of the reader frequently leads to character death in modern computer games, which means the user has to try solving that level or problem again.

A linear nodal plot structure.

A modulated plot still supports a dramatic arc to some degree, but allows the reader a number of choices at various points along the way. They can complete sections out of order, skip sections, or make choices that change the outcome of the story. However, the more options the author gives to the reader, the more possible plots (seen as a path through all the possible story events) there are, and the more difficult it becomes to tell an engaging or coherent story.

A branching modulated plot structure.

Finally, there is the open plot. Here, there is very little plot at all. Instead, there is usually a simulation of a world according to some rule set. Instead of following a story, the user develops a character or the environment itself. Example games include Ultima Online, Sims, and Age of Empires.

An open plot structure.

When giving a user control over the plot, the difficulty is giving them enough to be satisfying without giving them so much control as to destroy the story you want to convey. This is more of an art than a science at this point.

Point of View

There are other avenues into interactive narrative besides the plot, however. One of these is by letting the user control the point of view. For example, a story may contain a number of characters. Each character has its own plot line--the series of events that happen to that character. The user could choose which character to follow and experience the story through. The user could even switch characters part way through the story. The experience could be very different depending on which character perspective the reader is using.

Even when the user is following only a single character, this character does not always need to be the hero. A very different story might be experienced by seeing events from the view of the hero's sidekick or of the villain (Kardan 2004).

Another option is controlling how much of the scene the user sees. The user may decide whether to see the world through the eyes of a character (1st person), over the shoulder or near a character (2nd person), or divorced from any particular character (3rd person).

Depending on the story, this limited form of interaction--choosing a point of view--may still be rewarding and engaging.

Character and Other Parameters

An essential part of any plot is the characters. Even if the user can't control the decisions of the plot, she may be able to design one or more of the characters. The plot may vary depending on these initial character settings. For example, the reader may select the ethical alignment of the protagonist, and the story may have very different endings depending whether the protagonist is good or evil.

This extends to controlling other initial narrative parameters, such as theme, setting, or genre. Given an expert story generating system, requesting a particular type of story be generated would be a kind of interaction.

"True" Interaction

Even when we strike a balance between the impositional and expressive and create a modulated plot, we still have a tension between the author and the reader. At some point, the reader will desire an action the author's plotline does not support, even though the system mechanics might otherwise support that action. This is one of the reasons why table-top role-playing games can be so much more engaging and memorable than computer-based RPGs--they are much more open. You are constrained only by the flexibility of the DM, even if you think of something they hadn't anticipated. Without the author present to add to the plot, players quickly discover the bounds of a static system.

But what if the narrative system included an author?

Dynamic Story Generation

One option is that we teach systems how to tell stories. Rather than have one human author try to satisfy multiple readers, we could have the system generate a story for each user.

Computer-generated stories are likely to be poor, especially at first. However, some scholars (Grasbon & Braun 2001) have attempted a computational theory of narrative based on Vladimir Propp's analysis of Russian fairy tales. Propp discovered a number of "functions" within a story that occur within a certain order or certain combination. A computer can model certain plot sequences--revenge, conflict with antagonist, rescue--and attempt to string these together into a coherent story. (Young 2003)

Post-facto Stories

Raph Koster (in Meadows 2002) mentions an interesting alternative to generating plots for users to explore: let user's explore and generate a plot to explain it. This matches real life, where we live our lives day-to-day, but when looking back we are able to generate a structure of causal links to explain where we are now. Again, the system would have to be familiar with plot sequences in order to highlight them. And this may be less fulfilling that pre-generated stories, since much of the attraction of fantasy games (the bulk of interactive narratives today) is fulfilling quests and prophesies.


We have seen that there are a number of elements to narrative, though no classic theory--such as Aristotle or Freytag--includes or relates them all. It may be that narrative--an inherently structured and linear format--may not be the most useful foundation model for building interactive narratives. Drama or improv may be more helpful guides. Also, table-top role-playing games are successful examples of interactive narrative without the use of an static plotline or a complex computer system. Perhaps this form of interactive narrative has enough practice to provide the elements of a theory all its own.

Interaction, in the abstract, seems to have a less disputed definition than narrative. However, the specifics will likely be harder to implement correctly--bridging the gulf between the story and the user through the design of the interface.

Another important consideration is why we want to add interaction to a narrative at all. Generally, it is to make a story more engaging to a reader. (In an interactive narrative, the terms "reader" and "user" are synonymous.) Yet there are many avenues to pursue to make stories more engaging and intimate. Some of these methods don't include interaction. A story might be made more exciting by conveying it non-linearly, by using images or live action, by using multiple concurrent media, or by allowing the user to participate.

Even if we determine that interaction is the way to go, there are a number of avenues to pursue. We may want to let the user affect the plot. This raises a number of issue of balancing the roles of author and reader--of still creating viable narratives complete with true interaction. We may want to allow the user to explore multiple pre-designed plotlines, or we may find that it would be more useful to have the system itself generate the plotlines on the fly. We may decide that other forms of interaction are also viable, such as allowing a user to determine the point of view or the nature of the characters.

With so many options to pursue, creating interactive narratives is a broader endeavor that it may first appear.

Works Cited

American Heritage Dictionary of the English Language. 4th Ed. Houghton Mifflin Company, 2000.

Binsted, Kim. Lecture for ICS491-2: Interactive Narrative. University of Hawaii-Manoa. Fall 2003.

Grasbon, Dieter, and Norbert Braun. "A Morphological Approach to Interactive Storytelling." Digital Storytelling Department, Computer Graphics Center, Darmstadt, Germany, 2001. < cast01-proceedings/pdf/by_name/Grasbon.pdf >

Kardan, Kaveh. Lecture for IS315: Narrative Game Design. University of Hawaii-Manoa. Spring 2004.

Louchart, Sandy. "Narrative Theory and Emergent Interactive Narrative." Ruth Aylett Centre for Virtual Environments, University of Salford, 2002. <>

Mateas, Michael. "A Neo-Aristotelian Theory of Interactive Drama." Computer Science Department, Carnegie Mellon University, 2000. < mateas00neoaristotelian.pdf>

Meadows, Mark Stephen. Pause & Effect: The Art of Interactive Narrative. Indianapolis, IN: New Riders, 2003.

Norman, Donald A. The Design of Everyday Things. New York: Doubleday, 1990.

Young, R. Michael. "Steps Towards a Computational Theory of Interactive Narrative in Virtual Worlds." Department of Computer Science, North Carolina State University, 2003. <>