The Boundaries:
Limits of a Nonlinear Narrative Structure


CREDITS: Copyright © 1995 by Michael St. Hippolyte. All rights reserved. Reprinted with permission of the author.

NOTES: This article was originally written with an intended audience of adventure game developers (think Zelda). I have inserted some text to address areas where ARGs are different from these types of adventures. I welcome any commentary or thoughts on the message forum.

SOURCE: The Boundaries: Limits of a Nonlinear Narrative Structure. Reprinted with permission at

The narrative structure of story is the way in which the story is told. Traditional books, plays and movies have a linear narrative structure. Although the contents of the story may be nonlinear (flashbacks, foreshadowing, etc.), the telling of the story is linear, from page to page or frame to frame in a fixed order. An interactive story, in contrast, has a nonlinear narrative structure: there are multiple paths that the telling of the story can take, depending on the actions of the viewer.

The most basic nonlinear narrative structure is the decision tree. At various points in the story a decision is made that sends the telling of the story down one path or another. The decision may be an event in the story itself ("Do you get on the plane with Laszlo or stay on the tarmac with Rick?"), or it may be an explicit interaction with the narrative structure ("We'll always have Paris."). In its most general form, a decision tree is open-ended: a decision may allow the story to branch in completely different directions, and these directions are completely independent of each other. This is what "real life" is like. If you have a child, move to Cleveland, or join the army, your life in each case will never be the same, and what might have been is lost forever as an option. Even the most trivial of decisions ("Chocolate or vanilla?") may in principle have such an effect, though it is impossible to know in practice because one particular limitation of the "real life" narrative is that the story is told only once.

A decision tree is an interesting but difficult object. The sharpest problem arises out of its mathematical properties, apart from any narrative content.

The Mathematical Limit

Inescapably, an interactive story that follows the open decision tree model quickly runs into the unforgiving mathematics of geometric progression. A simple example will serve to illustrate. Imagine writing a story that presents the reader with a choice at the end of each page, with each choice leading the story in a completely different direction; thus, at each decision point, the reader is selecting one of two mutually exclusive sets of pages to view subsequently. Then, for example, if you wanted the story that the reader ultimately reads to be a two page story, you would have to write three pages, the opening page and the two possible continuations. If you wanted to write a short fable of, say, five pages; you would have to write thirty-one pages, a not unreasonable task.

But suppose you wanted to present your reader with a short story of modest length, perhaps a dozen pages. Then you would have to write about four thousand pages. A slightly longer twenty page story would require you to write a million and some odd pages. And a forty page novella would need over a trillion pages of material! Clearly this model will not work for interactive stories of any complexity.

But with a minor adjustment the decision tree model can be made useful. We simply require many of the branches to converge: certain completely different sequences of decisions will leave you in exactly the same place in the story. No longer does every possible decision require an entirely new story line; with sufficient convergence even a large number of decisions can be handled by a reasonable amount of material. To avoid getting caught up in a geometric progression, the decision tree must be completely or almost completely closed, i.e., almost every path must converge. There may be two ultimate outcomes, or two dozen, but nowhere near the five hundred billion possible outcomes for the forty page novella described above.

Mathematically, what we have is no longer a tree; it has become a graph. But "structure" is perhaps a better term, since it is exactly that: the narrative structure of the story. And in reality, the essence of the problem is economic: loose branches are expensive, convergence reuses material and thereby lowers cost.

The next problem the interactive storyteller faces has more to do with artistic expression than economics, but poses no less of a burden.

The Interactive Dilemma

The interactive dilemma may be described as the unavoidable tension between the author's goal to express something specific and the viewer's freedom to choose among different narrative possibilities. This echoes the question posed at the beginning of this paper, and relates as well to much of the discussion among students of hypertext on the topic of reader-as-author, since it is the viewer's narrative choices that comprise her authorial contribution.

The solution to this problem almost invariably involves placing further limits on the viewer. There may exist messages that rely so little on the narrative itself that the viewer can be given the latitude to change the story at will, but in all other cases the viewer's choices must be carefully constrained so that he does not obstruct the author's message.

The final limit we will look at is an unavoidable consequence of forsaking linearity.

Editor's Note: This dilemma increases exponentially given the boundless medium of the Internet. ARG Game Developers need to not only consider the complexity of their story, but what other information the viewer will have access to and stumble across in the course of their experience. The ability of PuppetMasters to adequately identify in-game sites, contain unintended peeks behind the curtain, and keep the viewers focused on the central story are all paramount to good game design.

The Chronology Problem

Most stories have a defined sequence of events, or chronology, without which the story makes no sense. If there are any cause-and-effect relationships between events in a story (and there almost always is), then there is a chronology that cannot logically be reversed. If, as in most cases, there are many cause-and-effect chains in the story, then the chronology must satisfy a large number of dependency rules for the story to work.

If the telling of a story is linear, then that story has a default chronology, which is the chronology of the telling. That is, if nothing else is said, the first events described are the ones that happened earliest, the lasts events described are the ones that happened last, and everything that happens in between is described in order. This is simply the default; stories with flashbacks and foreshadows, i.e., events described out of sequence, can have a much more complex chronology. Many literary devices are available to bend time; but no device is necessary to describe a well-ordered sequence of events linearly: it is what happens automatically unless an effort is made to the contrary.

If the telling of a story is nonlinear, however, there is no default chronology. The telling may go along many paths, each with its own implicit chronology. But if there are chronological rules to be enforced, the author must assure that the chronology that emerges from each telling fulfills those rules. Naturally, this translates into limits on the viewer's choices.