Narrative as landscape.
By Bob Hughes.
Presented at the MENO Workshop on Hypermedia and Narrative, Brighton UK April 1997:
I WANT TO MAKE A CASE for thinking of narratives not as a paths, but as three-dimensional spaces, or landscapes, through which we can take paths. I don't claim that this is a solid, scientific theory - more of a bit of sturdy bricolage that serves me remarkably well. And maybe it has some scientific basis too, which I'll come to anon.
First, though, I want to explain what I mean by "narrative" - because I don't just mean "novels", "anecdotes" etc. My feeling is, these are particular manifestations of narrative: in the same way horse-chestnut trees are manifestations of biology, and not biology itself.
1. Narrative could be everything.
William Calvin  has said that the human mind has a fundamental "what-happens-next orientation" on which an awful lot - if not all - of its operations are based. Daniel Dennett's "Multiple Drafts Model" (MDM) seems to describe consciousness itself as a stream of narratives, if I understand him right.  Many if not all experiences have a characteristic "beginning-middle-and-end" structure: an onset that sets something in the mind guessing what will happen next, a middle section where it observes the event and compares it with its prediction, and a closing phase where the whole sequence is "wrapped up" as if it were a little object. Steven Pinker  shows how we do this when we parse sentences and individual words. It seems very like what happens when you read a story, but on a much smaller scale.
I believe that, in many cases, unless you can wrap something up in this way, you can't properly experience it - and that this applies at all scales: from the word, right up to a lifetime.
2. We need a Particle Physics of narrative.
How short is the shortest possible narrative? Hugh Miller (in this symposium) annexed the word "micronarrative" for anecdotes - so what I'm talking about here must be "piconarrative". It seems to me there is a whole fauna of "piconarratives" waiting to be discovered, literally under our noses: tiny little narratives that last from about 50 milliseconds to 3 seconds. According to Miroslav Holub , this timespan is "the duration of the present moment".
Why is this important? Because stories depend for their success on "the way you tell 'em". To the chagrin of students of form, this is often nothing to do with plot or character. It's to do with "style" - which seems to hinge on little atoms of behaviour and utterance, and these do indeed seem to last for three seconds or less: things that usually escape our conscious attention, and have certainly escaped deliberate study and analysis - except by philosophers, mystics and neuroscientists.
The "great blue whale" of the sub-three-second world is the line of poetry, which turns up in all cultures, and always lasts for around three seconds . Below that you have words, whose fascinating, action-packed lives unfold, develop and end in less than a second. These live alongside (and perhaps off) an unknown number of even smaller entities, of which some examples might be the "tick-tock" effect and the "colour phi" phenomenon (described by Dennett).
I think we need a good grasp of these things in interactive media because the most obvious action here is precisely in the sub-3-second domain - in the way things respond to the user's actions, screen transitions are made, and information is fed back to the user. Good writers, raconteurs etc. are certainly masters of the domain, but they master it in a somatic way, by sheer practice, instinct, by trial and error - the same way ballet-dancers and cyclists master Newtonian mechanics.
When you work on the computer, the tiny little nuances, asides and pauses that make a writer or a raconteur so fascinating have to be built up in code: a long job. An event that lasts 1 second can take days to build. So you need to be able to explain and justify what you're doing and pin it down with logic. You can't work in the spontaneous, care-free way a stand-up comic or preacher does, or an intense writer like Jack Kerouac does - and you certainly can't do it in real time.
I think it would be useful to make a study of these, and see how they map onto our subjective experience of good narrative.
3. Big narrative emerges from little narrative.
My feeling is that big narratives, like epics, jokes and novels, emerge from these little ones - and the forms they take depend on local conditions. Novels for example can only happen under very special local conditions - a particular technology and a particular socio-economic regime. You cannot necessarily predict what kind of "big narrative" will emerge in our new medium, any more than Gutenberg could in his. The lesson here is that we shouldn't try and force the medium to do what other media do, but play with it and see what emerges: as I think Sterne and Swift did with print.
4. The Audience issue: a paradox within a paradox.
A lot of work that's been done on computer-generated narrative ignores the audience issue. It's assumed there will be an audience, but that's about as far as it goes. Scott Turner , for example starts with Propp's grammar , and tries to generate interesting tales from a set of rules. For my money he fails. Tales do indeed seem to contain rules - but they don't seem to be very good at following them. So far, rule-driven systems don't seem to produce anything very interesting (true in life as well as in art!). Are the rules just not good enough yet, or is there more to it than this?
Well, in other media, practitioners and teachers place enormous emphasis on "knowing your audience" - and they acknowledge an interesting paradox: works that are aimed at "everybody" please nobody. Conversely, works that have a very specific audience (eg Beatrix Potter's "Peter Rabbit" - done for a friend's son, in a letter) regularly turn out to have universal appeal. This is an important phenomenon: there's an audience within the text as well as the one outside it.
Generally, this supports people like Aleen Stein and Curtis Wong  (founder members of the hugely influential Voyager Company) who maintain that successful hypermedia comes out of a desire, a passion, to tell some particular story to some particular person. The technology comes a poor second to that. It also suggests an experiment: along with your story-grammar, include a "virtual audience model" against which all new story elements are checked before they are presented to the real audience (or user). It would be interesting to see whether this "Beatrix Potter system" would produce more interesting results than the grammar alone. Initially, the virtual audience could be a simple, static array of "likes and dislikes", although one can also envisage one whose tastes and interests change according to context, and develop as the story develops.
The Beatrix Potter Paradox sits, Chinese-box-wise, within a greater paradox: a lot of very great creative work is done with no regard for the audience at all. A painter, rock guitarist, pianist, dancer etc. is, at their best, so utterly absorbed in what they're doing that they just do not notice the audience - and the audience finds this utterly compelling. What is going on when this happens - and can a computer do it? What I think happens, is that we see a person exploring a very demanding "possibility space", with some personal risk as to the outcome. This brings me onto my main theme:
5. Narrative is not a path, but a landscape.
Most approaches to computer-generated narrative focus entirely on the creation of paths. But there is much more to narrative than this. To propose that the path is the narrative, is like proposing that the Pyg track is Snowdon, or the Pennine Way is England. Each path is chiefly a route through a particular terrain - and the terrain is the main thing. Each is popular because it is a very good route that begins promisingly, develops interestingly, and concludes satisfyingly. However, the Pyg track is no use without its surroundings and if you were to divorce it from the landscape, you would have nothing very much.
Also, if you were to tell someone to climb Snowdon by meticulously specifying every single step and change of direction, they'd be much less likely to get to the top safely than if you'd simply told them which way-marks to look out for.
If that analogy is any good, then the way to create computer narratives is to define the features of the landscape to be explored, and let those define the path. Before I go any further (so to speak) I want to explain how I came to believe that narrative really is landscape.
My first intimation came when I was an advertising copywriter. In advertising, you hardly ever tell a story just once: you nearly always have to tell it a number of different ways, of different lengths, in different media (5 different ads for the press campaign, 3 different TV commercials, plus all sorts of direct-mail and point-of-sale literature). In a really good campaign (such as David Abbott's classic press campaigns for Volvo in the 1980s) the start-point of each ad is different, so is the way it develops, but it is still somehow the same story, every time, however you tell it. You are taking the reader "over the same ground" again and again.
Good salesmen do this to an even greater extent: their "terrain" is the product, and they take every single customer through it by a different route, tailored exclusively for them, on the fly, in real time. A bad salesman, by definition, is one who only knows one route - his "sales script" - and gets "lost" when distracted from it.
My second insight came when I read Frances M. Yates's "The Art of Memory" . Yates explains the classical training system for orators, whereby they built up imaginary places for themselves - memory palaces, cities and theatres - in which to store the components of a train of argument. When preparing an argument, they would plan a route through the palace and place mnemonic objects along it. When the time came to speak, they would simply walk over their ready-planned route, presenting each item to the audience. I imagine the orator would be able to apply his full attention to each item, because he wouldn't have to worry at all about what was coming next: he wouldn't be able to miss it. Our conventional expressions: "in the first place", "in the second place", "commonplace" etc. came from this tradition. Yates conveys an infectious sense of wonder at the sheer, unsuspected scale of the art of memory: a vast, yet completely invisible industry that, for around 2,000 years, created the intellectual "home turf" of just about any thinker you care to name, up to and including Shakespeare and Newton.
My third insight came when I read Brenda Laurel's description of Gustav Freytag's analysis of drama  - specifically farce I believe. Freytag had the idea of drawing a diagram of a play's action: in a well-made play, you have a period of rising expectation and complication, a crisis-point, and an enjoyable downhill stretch where all the complications are resolved. It's a simple isosceles triangle (Figure 1).
Fig 1 - the basic Freytag Triangle.
Later writers (Laurel doesn't give their names unfortunately) observed that Freytag's triangle fits many other kinds of drama and literature, if you add subsidiary peaks as in Figure 2.
Fig 2 - the modified Freytag Triangle.
On the way up, you have "false summits", on the way down you have glimpses of the goal, but there are still intervening problems to be overcome. (Laurel also had the sublime idea that what we have here is a fractal curve that has self-similarity at every scale, right down to the sentence level.) It looks like a cross-section through a mountain landscape.
Figure 3 - the actual situation?
This, plus my earlier experiences, has led me to think that, for example, "War and Peace" really is a landscape, and the text Tolstoy has given us is simply a route he found through it; his other works traverse different parts of the same landscape, by different routes, of different lengths.
I certainly get a similar experience from all Tolstoy novels - and indeed from all Tolstoy sentences. Which is not unlike one's experience of walks in a real-world landscape: whether it's an afternoon stroll or a whole day on the peaks, you still get the "Snowdonia experience".What's more, Tolstoy behaves as if he is in a landscape. He is famously apt to indulge in what we call "digressions", where he "leaves the main path" of the tale to explore some philosophical or historical issue at length - just like some good mountain guide who takes you off the track for a while to show you some ruins, or an interesting geological structure, or a fantastic view, or to pick bilberries. (And he shares this tendency with just about every other story teller, from Milton - king of the extended metaphor - to stand-up comics like Eddie Izzard, to you or me). This tendency makes it very hard for a "path-centric" narratologist to write a rule for writing Tolstoy novels: like trying to program an anti-aircraft gun to shoot down flies.
(And here's a digression, while we're on the subject: we recognise great work, to some extent, by the density of detail found in the "narrative landscape". "Bad fiction" is perhaps a path through a poorly-realised landscape; successful bad fiction perhaps traverses a very good landscape that exists in the reader's mind already, requiring the merest of cues from the writer to bring it into play: an Enid Blyton does not create a landscape in the painstaking way a Tolstoy or a Dickens does; she "cheats", by invoking one that she and her readers share already; it is not to be found in the text; it is a sort of conspiracy between author and reader.)
The way we talk about narrative is riddled with spatial metaphor: we talk about "going over the same ground", "what the author springs on us around the next corner", writing being "a bit of a slog", authors who "lead the reader up the garden path", "exploring the possibilities" etc. etc. We find ourselves using terms like these (oops!) "at every turn". Other arts also make explicit use of spatial metaphor: music, painting, and especially mathematics.
What if this is more than a metaphor? What if we really are "doing spatial stuff" when we write or tell stories and jokes or explain things to others? I mean: what if the whole narrative business (or even some of it) really is built upon deep mental mechanisms for finding our way around - the way computers are built on basic Von Neumann procedures, that then radically colour everything they do, however sophisticated they get?
6. Could there be anything to this?
All my "evidence" so far is anecdotal. Is there any solid, scientific basis for claiming that a narrative is really landscape-driven?
There is quite a bit of evidence that spatial processing lies, literally, at the root of mental activity - and this is not too surprising, given that we've been doing spatial processing since we were chordates. But would that affect higher-level thinking?
James Newman  points out that all our higher mental processes (in the cortex) have extensive connections with two central structures in the thalamic system (part of the brain stem), which "contain a representational map of the'three-dimensional spatial envelope surrounding the organism'".
I believe Joseph LeDoux and his co-workers have established that our understanding of "context" arises in the hypothalamus, that these "lower-brain" functions are activated earliest in the cognitive process, and then maintain some kind of dialogue with the higher brain systems (at least, that's my understanding of it). 
Then, Nicolaas de Bruijn has pointed out that the brain cannot "type" its data the way a computer can  - which is why we can represent anything in terms of anything else - hence the fundamental role played by metaphor, and maybe the start of a proper explanation for my theory:
If the "carrier wave" from the lower brain is a spatial one, then spatial metaphor will suffuse the current subject of thought - which becomes a sort of "landscape pseudomorph". Then, perhaps, the spatial metaphor may be passed from thought to thought like a relay-runner's baton. The Russian folktales ("skazki") that Propp studied may illustrate this:
In a Russian tale, the physical landscape is always vague, but the people and objects within it are spectacularly solid. For example, Ivan the Fool lives in "a certain village in a certain Tzardom in a certain realm"; we have no idea precisely where it is. The true mountains in this "realm" are like e.e. cummings's "delectable mountains" (in "The Enormous Room"): they are people, and as solidly defined as any mountain range (unlike real people). The major, eternal features are Baba Yaga the witch, Ivan's unappreciative parents; the avaricious Tzar etc. - each of whom has to be negotiated with care and skill. It's as if the reality has been sucked right out of the landscape, and transfused into its inhabitants. Maybe we see the same thing happening at the other end of the "literary scale": in Becket's "Waiting for Godot" we still have a landscape through which the audience travels, but it is a landscape built from expectation itself. If spatiality is integral to narrative, how does that help the creator of computer-generated narrative? Two "routes" (so to speak) suggest themselves to me:
i.) We could start by generating a simple computer landscape, and let a program traverse it, building a sequence of actions in response to what it finds. One could do this in a very simple, literal way by mapping a landscape onto a numeric grid, or array. On "idle", the computer picks a path through the array - this might be a short path through one corner of the grid or a long one that goes through its heart. Whatever its length, the path would describe a "Freytag curve" characteristic of the grid as a whole.
At the least, this might give us some nice musical and visual effects, which would have a more natural-seeming shape than random ones. However, I feel it might be hooked up to a language generator, and/or made to refer to itself recursively to build up bigger and more densely-detailed sequences.
ii.) A strong contender is the semiotic route - where we define the objects of the landscape in terms of their "affordances", and let the computer take it from there. I am only aware of a few people who have tried this. Peter Andersen and Berit Holmqvist of Aarhus University tried it with an interactive erotic novel . David Zeltzer of MIT  has used affordances in a VR agent interface. I have a sort of a notion that a Russian academician, Gennady Uzilevsky , may also have tried to convert Propp's grammar into set of semiotic objects. There must be others.
Finally, if there is any truth in this idea, it has some very practical, explicit implications for the way we do hypermedia - and many other things too.
a. If narrative really is landscape, then there is no really radical difference between the kinds of narrative you can deal with in linear and non-linear media: any well-enough realised landscape should be traversable by a variety of routes. It happens that a novelist can only manage one at a time (it's terribly hard work, writing a novel!) - but that doesn't rule out the existence of others. In principle, a multilinear "War and Peace" is possible.
b. However, the sheer quantity and density of the detail involved might make the project intractable. Maybe novels are best left to humans - but for strictly practical reasons. c. One need not feel obliged to provide an infinite variety of routes: no landscape worth bothering with allows unlimited access. We are all perfectly happy with the Everest massif, which has only two approach routes.
d. And there again, no landscape worthy of the name should deny a ray of hope to the Reinhold Messner within each and every one of us, who wants to ascend via the north face, without oxygen.
e. If narrative is really a process of exploration, rules will help only if we accept at all times that they are derived ex-post-facto. They are "for guidance", they have got to evolve, they cannot be graven on stone tablets.
f. So it follows that narratives (and the way we design them) cannot be dictated by some wise and important person, sitting in state and viewing the "big picture" from a distance. It is necessary to explore the terrain, try different routes, take risks, and trust people. i.e., the old military model of "managing creativity" is no good: with everything specified in advance by some all-knowing supremo who never gets his hands dirty. "Big Picturism" is a big problem in all human endeavours that get beyond a certain size: in industry, academia and politics. For me, one of the great joys of hypermedia is the way it causes the Big Picturists to fall flat on their faces. 
g. For me, the "landscape model" explains a lot about people, and our failures to understand each other. We can be sitting in the same physical space, in totally different worlds. You may be "on top of the world"; I may be deep within its deepest chasm, and you wonder why on earth I don't do anything; why I believe I have so few choices. We converse from different landscapes at crossed purposes. Our lingua franca is logic, and we see each other as illogical, mad or intransigent - neither can see where the other is "coming from" - unless we take some trouble to share our worlds. In the modern industrial world of work, we hardly ever do this. In hypermedia work, which depends so much on shared understanding, we have to make that effort, and organise ourselves in ways that make shared understanding easier.
h. If narrative is, to any degree, as fundamental and as fine-grained as I believe it is, then hypermedia is a revolutionary kind of tool and requires a new kind of working culture: one where we get to grips, collectively and consciously, with the "atoms of experience" that make us tick, and with the worlds in each other's heads.
1. William H. Calvin - "The Emergence of Intelligence", Scientific American October 1994. "Language, foresight, musical skills and other hallmarks of intelligence are connected through an underlying facility that enhances rapid movements." The ideas are developed in his latest book, "How brains think" (Basic Books, 1996).
2. Dennett "Consciousness Explained". Penguin 1991.
3. Steven Pinker: "The Language Instinct", Penguin 1994. Pinker does for Noam Chomsky what Richard Dawkins did for Charles Darwin in "The Blind Watchmaker".
4. Miroslav Holub: "The Duration of the Present Moment" (one of a collection of essays that has that as its title). A better source is probably Robert Efron's "The Duration of the Present" (Proceedings of the NY Academy of Science, 1967) - referred to by Dan Dennett in "Consciousness Explained".
5. The source here (given by Pinker) is D. E. Brown, "Human Universals" - McGraw-Hill 1991.
6. Scott R. Turner: "The Creative Process - a computer model of storytelling" Laurence Erlbaum 1994. ISBN 0-8058-1576-7.
7. Vladimir Yakovlevich Propp: "Morphology of the Folktale", Leningrad 1927. English translation Bloomington, 1958.
8. I cover the work and ideas of Aleen Stein and Curtis Wong in "Dust or Magic" (Addison-Wesley, 2000).
9. The Art of Memory, Frances M. Yates. Penguin 1969 - also "The memory palace of Matteo Ricci" - Jonathan Spence, Viking 1984.
10. "Computers as Theatre": Brenda Laurel - 1991, Addison-Wesley. Freytag's work is "Technique of the Drama", 1863.
11. Journal of Consciousness Studies, Vol 4, Issue 1. 1997.
12. Scientific American, June 1994.
13. "Can people think" by Nicolaas de Bruijn - Journal of Consciousness Studies Vol 3 Issue 5/6, 1996.
14. "Narrative Computer Systems: The Dialectics of Emotion and Formalism", Peter Bogh Andersen and Berit Holmqvist. Published in "Computers and Writing - State of the Art" (ed Holt and Williams) Intellect Books, Oxford 1992. ISBN 1-871516-20-X
15. Presentation at the National Gallery in 1993 - sorry, no reference but he's pretty famous. Search for him on the Web. He's probably "zeltzer@MIT.edu".
16. Uzilevsky is a mystery. I have one short, badly-translated paper of his from an HCI conference in Moscow in 1992 - and I can't find it at the moment. Any info welcomed.
17. For a harrowing example, see Fred Moody: "I Sing The Body Electronic - A Year With Microsoft On The Multimedia Frontier" Viking Penguin 1995.
Bob Hughes © April 1997. www.dustormagic.net (bob at dustormagic dot net)