Mark J. P. Wolf
Looking at the collection of essays in this volume, each exploring a single, particular imaginary world, it is natural to consider to what degree it is possible to compare worlds, and, in order to do so, to measure them in some useful way, so that worlds, described by their measurements, can then be quantitatively compared. These measurements should allow one to compare and contrast examples, perhaps even to locate them along a spectrum or within a phase space (which is really a multidimensional spectrum), where one can plot the points representing a collection of entities for comparison, and immediately see how similar or different they are, and what kinds of distributions occur.1 Such tools are helpful even in areas which are more typically studied qualitatively; for example, in Film Studies, one can compare the lengths of films, the number of shots per film, film budgets, cast sizes, number of lines of dialogue, and so on. Similarly, a study of literature could compare short stories, novellas, and novels, looking at the number of words used, the number of characters, story events, and so on. Conceiving of entities as sets of interrelated parts is the basis of structuralism, and mythemes, narremes, and lexemes are examples of attempts to reduce something to individual units (in mythology, narratology, and lexical meaning, respectively). The term “narreme”, defined as the basic unit of narrative, was first coined by Eugene Dorfman in 1969, and was further refined as a concept by Henri Wittman in 1974. Roland Barthes attempted to examine the codes and units of narrative in S/Z (1970), and others since then have tried to precisely define what constitutes an individual unit of narrative. So far, however, no proposed unit has succeeded to the point of being widely accepted, apart from larger divisions like chapters, episodes, or scenes. Imaginary worlds, then, are even more difficult to divide into basic units (mundemes?), due to all the possible types of world data that can be used in the construction of a world.
Imaginary worlds are notoriously hard to quantify precisely, making measurements and comparisons difficult. First, there is the problem of form; the transmedial nature of many worlds means that they can be made up of words, images, sounds, moving images, and so on, and such elements are difficult, if not impossible, to translate from one form to the other. Too often, no amount of ekphrasis is sufficient to convey an image in verbal form (a picture is often worth more than a 1000 words), and likewise, verbal descriptions can be written that are impossible to concretely visualize. Second, there is the problem of content; even within a particular medium, world elements can be depicted with varying degrees of narrative density and resolution. A decades-long world war with millions of participants can be summarized in a page of text, while two characters conversing during a single dinner involving could take dozens of pages to describe fully. Finally, there are the related questions of cohesion and coherence. Cohesion is how well world data sticks together or works well together, whereas coherence is the ability of a collection of world data to form a coherent whole, a world that makes sense and seems feasible or plausible. Coherence, which has to do with meaning, differs from consistency insofar as it underlies the potential for consistency; without coherence, there is no way to consider whether or not a set of world data is consistent or inconsistent. The worlds of Tolkien’s Arda and Wright’s Islandia are both very coherent, and thus we can judge how consistent they are, whereas the collection of world data found in the Codex Seraphinianus (1981) has some aesthetic or stylistic cohesion, but there is little or no coherence binding everything into meaningful infrastructures, much less a coherent whole, and without such relationships, notions of consistency and inconsistency cannot be meaningfully applied.
At first glance, it may seem like there are some good places to start an attempt at measuring and comparing worlds, such as spatiotemporal size, the number of world data present, or the amount of time needed for a user to experience the whole world. So let us examine each approach, and the resulting promises, perils, and pitfalls they contain.
The most common way to compare locations in the empirical, Primary World is by size; the geographic sizes of countries, in square miles or kilometers; the surface areas and circumferences of planets, population numbers, and so on. Imaginary worlds, however, cannot be compared as easily, however, because they are often not described statistically by their authors; for example, Tolkien never specified the population nor the exact square mileage of Gondor (though cartographer Karen Wynn Fonstad has estimated it at 716,426 square miles).2 Beyond that, there is also the question of how much of a world is actually used, with story events occurring in it. A world can have vast deserts and seas, but few or no inhabitants, and likewise little or no narrative activity; or, on the other hand, one could have a very small area (like Barsetshire or Lake Wobegon) which has a rich geography and history, and events covered by multiple works. The world of Georges Perec’s Life: A User’s Manual (1978) is only an apartment block in Paris at 11 rue Simon-Crubellier, but it is described in hundreds of page of exhaustive detail; and although the descriptions all occur frozen in time on June 23, 1975, shortly before 8:00 pm, the events discussed in flashbacks span over a 100 years, with around 100 interwoven subplots based on the apartment building’s many residents. Likewise, Richard McGuire’s graphic novel Here (2014) takes place all in one location, a small plot of land on which a room of a house is built, but covers the events happening there from millions of years before humans appeared on earth to thousands of years after the present day, with all of its hundreds of individual moments presented out of order and often visually overlapping each other. Both Perec and McGuire present worlds which are very small spatially, but which contain a great degree of detail and history, whereas other worlds might be vast, stretching across galaxies, but still only be the settings for short stories of science fiction.
And then, in the medium of video games, there are procedurally generated worlds, like those of Minecraft (2009) and No Man’s Sky (2016), which are so large that a player could not even hope to explore them within multiple human lifetimes. It has been claimed that a Minecraft world can be made of up to “Two hundred sixty-two quadrillion, one hundred and forty-four trillion blocks”3 and that if each block is said to be a cubic meter, that the surface area of a Minecraft world is around four billion square kilometers (compared to the Earth’s approximately 510 million square-kilometer surface).4 No Man’s Sky is even bigger, with 18,446,744,073,709,551,616 procedurally generated planets that you can actually fly to, land on, and explore (meaning that if you visited one per second, it would take you 585 billion years to see them all). If we were to measure the land area of all these worlds, we would have to conclude that No Man’s Sky has the largest world, at least geographically; but the vast majority of these worlds also have no history or narrative associated with them, either. Dwarf Fortress (2006), on the other hand, procedurally generates landscapes along with characters who live there, generating histories for each of them which include such things as who they battled and where they traveled.
Compared to hand-crafted worlds, procedurally generated worlds are often criticized for being too repetitive, with world elements that are oversimplified, and little more than recombinations of the same elements. They reveal the value of hand-crafted worlds, where a human author makes things that have meaning and are interrelated with other objects and events. Authors write histories that are driven by causality, as opposed to being merely lists of disconnected events. While procedural-generation methods are ever-improving, it is still difficult even for human beings to create interesting stories and characters consistently over time; so it seems unlikely that such things will ever be automated well. And there is only so much that a given author can create within a given timeframe. That brings us to our next method of comparison, that of counting world data.
Procedurally generated worlds demonstrate that mere numbers of world data usually reveal little or nothing about a world. In a few kilobytes of memory, one could write a short program to generate random character names, and leave it running for weeks, generating a long list of a world’s inhabitants, but even with billions of names, there is really no world created as a result. Limiting our discussion to world data hand-crafted by humans does not help much, since humans can also employ mindless methods of procedural generation.
And what exactly is a world datum? A name, an object, a design, a location, a character, an event; world elements can be as varied as we like, and of course, any given imaginary world will have to have a finite number of them, however we define them. Sometimes it is obvious which worlds have more data, without making a count; we can all agree that the worlds of Star Wars or Star Trek are larger and more detailed than the world of Stanisław Lem’s novel Solaris (1961), or that the worlds of Myst (1993) and Grand Theft Auto V (2013) are larger than the worlds of Asteroids (1979) or PONG (1972), but not all comparisons are so obvious (for example, which is bigger, the world of Star Wars or Star Trek?).
No matter how we even try to define a world datum, or divide a world into individual data, we will run into problems; it would seem world data cannot be completely quantized. How many world data are there in an image of a dense cityscape? Or a multi-layered soundtrack of a location’s ambience? Where do we draw the line when considering what counts as world data? John Williams’s musical scores are an integral part of experience Star Wars films, but while some of the music is diegetic (like the Cantina Band music), much of it is not, so should we not include anything that the characters of a world could not see or hear, even if the audience is aware of it? Trying to actually count the world data of a world of any size or complexity quickly reveals the shortcomings, if not the impossibility, of applying such a method. But questions regarding the experiences of the audience who is vicariously exploring the world brings us to the next method of measuring user experience time.
The amount of time needed to experience an entire world seems like a good place to start, particularly when you consider media usage. We can compare the numbers of hours of TV shows and movies which depict a world, or the numbers of pages of novels, or hours needed to complete the narratives found in video games, and all of these figures are usually used when describing various media experiences of worlds. Of course, reading speeds vary, and not all players will advance through a game at the same rate. Audiovisual media like movies, radio, and television may seem to provide the most similar and standardized experience for all audience members, as they have a set running time, so counting the hours and minutes would seem to be a good way to compare them. But all these media, especially the visual ones, can be more or less dense with detail; soundtracks can be more layered, and images more intricate, and packed with enough detail to require multiple screenings in order for all the detail to be noticed. Home viewing media also allows freeze-framing and re-viewing so that audiences can spend greatly varying amounts of time with the same movies or television show episodes.
User experience time also can vary based on the conditions surrounding the experience; watching something in a darkened movie theater is of course different than watching something on a cell phone screen in a noisy, busy public environment in daylight (can any horror movie be as effective in the latter environment as in the former?). Watching with distractions means interruptions, and possibly more re-playings, changing viewing time, as well as the whole experience itself. Also, a user’s experience of a world will depend much on the amount of time allotted for the completion of world gestalten, which may require some contemplation of the world data, allowing connections between them to be recognized by the user. Some presentations of world data, and the media in which they appear, promote reflection time, while others do not.
In any event, user experience time is still too variable and even more difficult to measure when it comes to transmedia worlds appearing across a variety of media.
Can worlds be compared in any useful way, or is the point of this essay merely to abandon any hope of finding a way to discuss worlds in relation to each other in some systematized way?
To some degree, we can still compare various aspects of imaginary worlds, such as invention (the degree to which a secondary world relies on Primary World defaults), completeness (how fully imagined a world is), consistency (the degree to which the parts of a world are in agreement), cohesion (how well the parts of a world are connected together), coherence (how well the parts of a world form a coherent whole), and the various amounts of media (words, images, moving images, sounds, etc.) that are used to convey all the world’s data to an audience. But can we compare the entire worlds themselves with each other? Or derive some statistical measures of them for the purpose of comparison?
Maybe… if we carefully build our conceptions of the worlds being measured from several different sources. Instead of trying to consider whole worlds all at once, we can first look at the individual infrastructures that make them up, but in an abstracted way.5 For example, starting with maps and locations, rather than thinking in terms of square mileage, we could consider what we might refer to as “salient locations”, regardless of their size, each of these being a place, treated as a single location, where some narrative event occurs, and the number of narrative events occurring at each location. To use an example familiar to a wide audience, consider the planet Hoth in The Empire Strikes Back (1980). While Hoth is an entire planet, the number of salient locations on it is rather small; there is Echo Base (which is made up mainly of the control room, the hallway, and the garage where the Millennium Falcon is parked), the Wampa’s cave where Luke is held prisoner, the generator (seen at a distance), the trenches where the rebels await the Empire’s troops, and the open, featureless land where much action occurs (the ATAT attack, the shooting of the probe droid, Han Solo opening the dead Tauntaun with a lightsaber, and so on). From an extradiegetic standpoint, this is like making a list of the sets needed for the Hoth sequence of the film. While the entire planet of Hoth has a relatively small number of salient locations, other places might have a large number of them, even in a much smaller geographic area, such as the towns of Twin Peaks or Wayward Pines found on the eponymous television shows. Thus, a small town may have more salient locations than an entire planet, and perhaps even seem more expansive, at least narratively speaking. The concept of salience applied to locations also allows one to deemphasize places which are only seen on a map but never used, or merely referred to but never visited; and likewise, salience itself could be seen as a spectrum, since some locations may contain a majority of a story’s action while others are only visited momentarily (thus, one still has to set a bound as to what will count as “salient”).
The notion of salience can be applied to other infrastructures as well, usually using narrative importance to determine their salience (admittedly, nonnarrative worlds may not work as well, though they may have other ways of indicating salience). Once the salient elements are identified, we can determine the overall shape of each infrastructure, and the connections between them. This, of course, still becomes more difficult the larger a world is, but we could compensate by simplifying our scales, by selecting only those elements with a higher degree of salience, or by chunking elements together into larger groups (for example, considering Hoth’s Echo Base as a single location, rather than dividing it into control room, hallway, and garage, or even, at a still broader scale, considering Hoth as one location among many other planets). Naturally, changing the scope and scale used to measure a world means parsing a world at different resolutions, some more coarse than others, but we may want a lower-resolution conception for the purpose of a comparison, especially for larger or more complex worlds.
There are, however, two main problems with the method of comparing interconnected infrastructures; the subjective process of dividing infrastructures into elements, and the reliance on narrative for the determination of salience. The first problem is one that is encountered in any kind of structuralist venture; defining boundaries between individual elements and precisely defining what constitutes an element. For example, in regard to spatial locations: one might have a chase which passes through an ever-changing landscape on a planet’s surface. The chase scene, which could be done in a single take with moving camera, could pass seamlessly through a variety of different environments, giving us two possibilities: we could try to determine some criteria for deciding where one location ends and the chase crosses over into another one (for example, using bridges or changes in terrain as boundaries), or we could consider the entire chase route as a single location, since the action is continuous and no part of the route seems more salient than the rest of it (if that is the case). No matter how you slice it, it is a subjective decision, and one which will likely depend on the scale of the world and the purpose of the comparison. Although it is subjective, a useful comparison can still result as long as the same method of division is applied to all the worlds involved in the comparison.
The other problem, the reliance on narrative for salience, becomes a problem when the worlds being compared differ in regard to their own reliance on narrative. Some worlds contain a straightforward narrative, others a branching or multi-threaded narrative, some (like video game worlds) may even rely on user input for narrative to emerge. Some may contain little or no narrative at all. Even if we consider only worlds with a single, linear narrative line in them, there can still be a varying relationship between story and world; in some cases, there is just enough world to support the story, while in others there may be much world data and material beyond what is needed for the story.6 In worlds with multiple storylines, something that is salient in one storyline may not be salient in another, although some things may be salient in multiple storylines. Storylines themselves can also vary in salience in regard to the world in which they appear (to use another Star Wars example, Anakin Skywalker’s storyline is far more salient than that of Greedo or Tion Medon). Overall salience, then, may depend on a combination of things, though it should not be difficult to determine what the main storyline is, and how important something is in relation to it; especially when narrative fabric is itself one of the infrastructures being considered.
I am aware that the preceding suggestions do not solve all the problems involved in measuring and comparing worlds, nor do they completely remove the subjective element that seems an inevitable part of such an activity. Worlds may take on many shapes and forms, and may even differ as to which infrastructures they rely upon, but these infrastructures, and even their presence and absence, do give some starting ground upon which measurement and comparison can be built. Still, comparisons of worlds will mostly likely be made using criteria specific to each comparison, on a case-by-case basis, while an objective standard of measurement applicable to any secondary world seems unlikely to be found. But the more worlds we examine, and the greater the variety of those worlds, the more we may able to discern what is essential to all imaginary worlds, and more firmly ground the basis by which we are able to measure and compare them. If anything, this essay was intended to at least raise some of the issues involved in the measurement and comparison of worlds as entities, and suggest directions for further work in his area. In the same way that David Hilbert’s famous list of unsolved problems in Mathematics inspired attempts to find their solutions, perhaps the difficult nature of these problems will attract more interest in them; but for now, an objective standard for the measurement of imaginary worlds remains one of the great unsolved problems in Subcreation Studies.
1 Websites like Gapminder.org are particularly good at visualizing data in this manner.
2 See Karen Wynn Fonstad, The Atlas of Middle-earth, Revised Edition, Boston, Massachusetts: Houghton Mifflin Company, 1991, page 191.
3 See Jeremy Peel, “Just how big is a Minecraft world? Big, as it turns out”, PCGamesN.com, February 3, 2013, available at https://www.pcgamesn.com/minecraft/just-how-big-minecraft-world-big-it-turns-out.
4 See Sarah Fallon, “How Big is Minecraft? Really, Really, Really Big”, WIRED.com, May 27, 2015, available at https://www.wired.com/2015/05/data-effect-minecraft/.
5 For a list of world infrastructures and their descriptions, see chapter three of Mark J. P. Wolf, Building Imaginary Worlds: The Theory and History of Subcreation, New York, Routledge, 2012.
6 This would itself be an interesting thing to try to measure along a spectrum, though it is beyond the scope of this essay.