The final session of the final day of last week's GDC 2007 was a cross-disciplinary take on level design. Brian Upton, a senior designer for SCEA and the lead designer for Rainbow Six and Ghost Recon, called on theme park design, landscaping, and city planning as parallels for level design, explaining that they all work under the same principles.
The main concern with all of these disciplines is human psychology - an understanding of how people orient themselves within, organize, and think about the space around them. Since theme park designers, landscapists, and city planners have been doing their job a heck of a lot longer than game designers, Upton suggested looking to these older fields not only for technique but for terminology with which we might describe and define level design.
Upton described Main Street USA, the long, shop-filled entryway to Disneyland, as perhaps the most popular videogame level ever. In the distance is Cinderella's castle, luring in the visitor, yet for ten full minutes of walking, there is nothing immediately entertaining. "The first thing you'd think is, 'this is really crappy level design,'" Upton said. "Why design a park like this - this long run-up? How is this functioning?"
Upton compared the structure to Hyrule Field, from The Legend of Zelda: The Ocarina of Time - and then pulled out Joseph Campbell's The Hero With a Thousand Faces, one of the more popular references at GDC this year. On the one hand, Link is an easy parallel to Frodo Baggins or any number of other heroes - yet in a sense, so is the random Disneyland tourist. The reason Upton gave is that the design of the space around us, and the way we interpret it, is in parallel with narrative structure.
Main Street USA, or the Kokiri Village, can be seen as a hyperealized home, or normality. The castle off in the distance serves as a call to adventure. Reaching the end of the street, or the drawbridge of the castle, is a form of crossing the threshold into act two, at which point the hero undergoes a series of heroic adventures. Finally, at the end of the day, the hero returns home - away from the fantasy of his quest to the nomality that he left behind, changed by his experience.
The design of Main Street USA puts people into the mindset of a hero, setting the tone for the entire experience of the park. Without the dramatic beats and anticipation, there would be no payoff, no sense of discovery. There would be no adventure.
In order to deconstruct level design and speak about it in the abstract, Upton said, we need a critical language that defines our relationship to space - something that does not currently exist in the game industry. He offered a 1960 text by Kevin Lynch, titled The Image of the City, as a decent model. The book was actually based on empirical research, Upton explained. Rather than come up with a set of original theories, the author went around several North American cities and asked people a series of questions about their understanding of city structure.
The result of this research was a set of five key concepts: landmarks, paths, edges, nodes, and districts. Thanks to games like Ocarina of Time, we have the concept of gameworld landmarks, in the form of "weenies". Landmarks attract us, and we use them for triangulation. They lend a feeling of uniqueness to an area, and a sense of scale. "If you create something interesting," Upton said, "it tends to draw people's attention. If you attract people's attention, people tend to move toward it."
Upton stressed that not every street is a path; rather, a path is a route that people normally think of as bridging two locations. If another route exists, yet people never think about it as a passage, then it is in a sense wasted space. Paths can be explicit - trails, corridors - or implicit, suggested by gaps and obstacles. Even a tiny gap tends to attract the player, while even a small obstacle tends to cause a detour. They can also be defined by momentum, as people tend to keep moving in the direction they have been aimed unless pulled off course by something interesting.
Edges are the boundaries between areas - bridges or train tracks. Once you cross this line, you are somewhere else. In games, they may simply be game state transitions. As far as level design goes, edges might be defined by doors or corners. The idea is that every time the player crosses an edge, he is introduced to something new. Crossing a threshold can reveal a vista, an ambush, or a view of where the player has been. By managing edges, a designer can control the scope - the moment-to-moment emotional feel - of traversing the gameworld.
Nodes are waypoints, destinations - places for gathering, resting, thinking about where to go next. Upton described intersections, the Boston Commons, and stations all as nodes. Nodes are defined by an organization of space, bordered by "conceptual walls". They have a sense of place to them that lends the locations a sense of dramatic weight. "Things tend to happen in nodes," Upton said.
Districts are places like Little Italy or Chinatown - more of a change in overall theme and set dressing than in geometry. Upton described a district as a "localized atmosphere".
There are certain pitfalls to level design, which are mostly defined by a failure in implementation of the above elements, resulting in confusing or uninteresting spaces:
"The Same Damn Corridor" problem is defined by a lack of landmarks, an over-use of paths (resulting in a sense that nothing ever really happens), the over-use of instanced geometry, or symmetrical or repetitive layout, leading to tedium. "If that's what you're going for," Upton said, "all right." Yet if the design is not meant to bore and disorient the player, this is a problem to watch out for.
"The Trackless Forest" is both similar and contrary, in that it is defined by a lack of paths, in an area with too many features of equal weight. There are no clear boundaries telling the player where he or she is going, and why. To avoid this issue, Upton suggested that designers organize their space - impose an artificial sense of structure.
"The Secret Passage" is a turn-off that the player will almost certainly miss, going by the tendencies noted above for people to walk in a straight line unless prompted otherwise and to walk toward landmarks wherever they might be visible. Upton blames the problem on "misleading weenies", improper path weights (in that the important path might not be emphasized enough), and edges located outside the player's "view cone".
"The Shotgun Shack" is a case of too many nodes, all loading into each other. The issue here is a lack of anticipation between events, leading to a flat cadence and an exhausting experience. Too much happening is the same as nothing happening.
"The Weak Corner" comes from not thinking of corners as edges, therefore as triggers for events. Elements are revealed in a gradual, piecemeal manner, resulting in a certain lack of drama and an ambiguity about the role of the upcoming location.
"The Wrong Doorway" is a scale mismatch between the edge or the path and the upcoming node. A door to a huge, important encounter might be too small, or the build-up might be too short, leading to a failure to foreshadow the scope and importance of the upcoming area. "The distance you move should be about the size of the target," Upton said. If an upcoming node is ten meters wide, the player should need to travel about ten meters to get there - depending on what experience the designer has in mind.
"The Road to Nowhere" is a case of missing nodes, resulting in a lack of a feeling of accomplishment and a poor flow to the level.
"The Missing Wall" is a poorly defined node, resulting in an ambiguous area, creating a sense of uncertainty in the player. It might suggest an unintended path, making the space hard for the player to organize, mentally. "What is this place?"
Upton wrapped up his lecture by defining a set of cadences - different styles of flow, that result in different emotional states for the player - and ways of creating emotion for the player, through level design.
A "Path-Dominated Cadence" emphasizes continuous motion. There are few edges, few nodes, and long, linear levels. Driving games are good examples.
An "Edge-Dominated Cadence" would be something like Resident Evil - a design defined by short sight lines, many corridors and doorways, and continual dramatic reveals.
To describe a "Node-Dominated Cadence", Upton put forward Ico - a game defined by open spaces, many elements to interact with within each space, and a general lack of time pressure.
To create tension, a level must press a conflict between safety and danger. "If there is no downtime, it's hard for there to be an up." The game must establish anticipation with consistent threat cues. Upton again referenced Resident Evil, by mentioning how at the beginning the game meticulously defined corridors as safe places - only to send a zombie Doberman crashing through a window a short while into the game, breaking the sense of security the player had built up and giving a sense that nowhere was absolutely safe.
Upton further discussed two of his Tom Clancy games, each with a conflicting approach to tension. Rainbow Six focused on claustrophobia - short sight lines, many corridors and doorways. Paths were generally safe; edges were generally dangerous. Ghost Recon takes place outside, so tension comes from agoraphobia - long sight lines, few buildings. Paths are generally dangerous, as enemies can come from anywhere, while edges are safer as they block out 180 degrees of surprise.
Creating wonder involves long paths, imposing landmarks, upward slopes, and transitions of scale. Creating triumph demands "terminal nodes" - nodes that serve as settings for big ending events - a bunch of foreshadowing, and a level-long approach to the target. To an extent, the plot in a videogame is mapped to the level geometry in much the same way as a music score. There are embedded arcs - swells and furrows, reflecting the hero's emotional state - building up to a climax.