Sponsored By

Educational Feature: A History and Analysis of Level Design in 3D Computer Games - Pt. 2

Indiana University graduate student Sam Shahrani presents a detailed history of the evolution of level design and in-game interactivity from the inception of 3D engines to the near present, in the conclusion of this exclusive two-part Gamasutra educational feature.

Sam Shahrani, Blogger

April 29, 2006

43 Min Read

[Note: The following is the continuation and conclusion of author Sam Shahrani's article. The first half is available here.]

Doom created a sensation in the gaming community and popular media, but it was far from being the only title pushing the boundaries of technological innovation. In March of 1994, Looking Glass released System Shock, a science fiction title built on a modified version of the engine used in the Ultima Underworld titles. The gameplay of system shock is that of a first-person shooter merged with an RPG and an adventure game, much like that of Ultima Underworld, but with an enhanced role playing system. Indeed, much of the success of a System Shock player centers on the ability to make wise choices when literally upgrading and modifying the player’s avatar. Since the player is a hacker that has been turned into a high-tech cyborg, the player has a number of abilities and skills that can quite literally be upgraded, as well as allow the player to interface with a virtual reality cyberspace set inside the game, a sort of world within the world. The antagonist of the game, an amoral female artificial intelligence known as SHODAN, routinely taunts the player from displays and interfaces, as well as sending cyborgs, mutants and robots to attack the player. The game is not a fast paced title, with designers choosing instead to emphasis story and character development, as well as providing a complex mystery for players to unravel. This type of gameplay is a marked contrast to that of Doom and Wolfenstein 3D, which emphasized a faster paced, higher-body count approach to immersion.

System Shock’s engine had many graphical features in common with Doom, but was designed to create a much more detailed environment, as well as for a slower pace. A purely singleplayer game with no multiplayer capability, the emphasis in System Shock was not on “run and gun”, but instead on slowly unraveling the mystery of what had transpired on the Citadel Station space research and mining facility. The engine supported almost all of the features present in Doom, many of which had been present in the earlier Ultima titles. System Shock supported higher resolutions than most other games, allowing up to 640x480 resolution, which was necessary for the full amount of detail included in many of the textures to be completely visible. These abilities came at a price, however, as many computers couldn’t run such a complex game at a reasonable speed. Conversely, Doom was engineered to run very quickly on as wide a number of systems as possible. Since the engines were designed for games with two completely different approaches to interactivity, comparing the two on merits of mere speed is unfair, and any comparison must take into account the different approaches to gaming.

The creepy, almost oppressive atmosphere of System Shock was enhanced by the utter lack of non-player characters to speak with. All humans encountered in the course of the game are corpses, whose bodies can be rifled through. Many of the bodies contain data discs with audio or text messages that provide the player with clues as to what happened on the station, as well as information on how to defeat SHODAN. The original release of the game provided these logs and messages as merely text, but a later CD release of the game added an extensive amount of audio to the title, heightening the immersion and fear factor of the title significantly. Ambient audio combined with the vocal performance were an integral part of the game, providing clues as to hidden enemies, as well as allowing SHODAN to harass the player as they moved throughout the station.

The level design in System Shock emphasized giving the player choices and rewards for thorough exploration of the station. The levels varied between the computerized corridors of Citadel Station to hydroponics bays filled with mutant creatures and plants run amok, orange tentacles creeping across the walls and integrating with the digital systems. In certain cases, the player actually had to jack into a representation of cyberspace in order to achieve goals such as unsealing doors or repairing systems. The need for the player to balance choices, as well as having to actually interact with computer and security systems in the game were innovative features in the genre, and significantly increased the direct influence that players could have on the game world besides merely butchering enemies and throwing switches.

System Shock’s design choice to eschew non-player characters in favor of using logs and messages left before their death is an interesting choice from a game design standpoint. In a postmortem on System Shock 2, Irrational Games developer Johnathan Chey notes that System Shock made this decision primarily because the computer technology of 1994 “was simply inadequate to support believable and enjoyable interactions with them” (Grossman, 12). While the decision was made out of necessity, it served to greatly improve the feeling and immersion of the title, and was a decision that was carried through in the August 11th, 1999 release of System Shock 2 by Irrational Games and Looking Glass.




While System Shock and Doom took a grim and serious tone towards their gameplay, other titles such as Apogee’s 1994 Rise of the Triad took a somewhat more light-hearted approach to the violence that was such an integral part of FPS titles. With a design team led by former id software member Tom Hall, Rise of the Triad, or RotT, used a modified version of the Wolfenstein 3D engine. Since Apogee had been the distributor of Wolf3D, they had the rights to use the engine; Doom was made and distributed by id software itself, meaning that Apogee would have had to license the Doom engine if they wanted to use it in a product, a costly proposition.

RotT featured several innovations for the Wolf3D engine, including adding the ability to move vertically. The game added a number of both humorous and deadly methods of interaction for the player, including “jump pads” that could launch players and enemies high into the sky, razor sharp spinning blades that could eviscerate unwitting gamers, weapons that could leave bullet marks on walls and the introduction of explosive deaths for all enemies. In RotT, when an enemy character was hit with a rocket they would frequently be reduced to a shower of digital meat, completely obliterated, seeming to fly out towards any nearby player. This shower of exploded body parts included an eye, bloody skull and, occasionally, a severed arm with its middle finger upraised. This was a graphical advancement over Doom, which simply showed a shredded pile of an enemy after a rocket hit them. While a small addition, it made for some truly amusing kills in multiplayer, called Comm-Batt.

RotT’s deathmatch also introduced a variety of inventive new ways of dispatching enemies, including homing missiles, heat seeking missiles, flame wall bombs, fire jets, floor and ceiling spikes, and weapons such as the Excalibat, a cursed Louisville slugger. These weapons and innovations allowed players, who were frequently in the same room or near one another on a Local Area Network, to truly embarrass their opponents as they beat them, as well as pulling off impressive feats of acrobatics.

Other technical innovations included walls that could move inwards and crush players (a feature not present in Doom, where walls, ceilings and floors could only move vertically), poison gas that required a gas mask to evade, fireproof jackets to ward off flame-based weaponry, and enemies that could steal a players weapons and also feign death. While seemingly superficial additions, these ideas were innovative and forced RotT players to be more aware of their surroundings.

While Apogee was busy with RotT, Volition software was busy with their space-combat FPS, Descent. Released on March 17, 1995, Descent was the first PC game to feature a full three dimensional environment as well as fully three-dimensional enemies.
The engine was not completely three dimensional, as it still used sprites for doors, pilots to rescue and item pickups, but was a significant improvement compared to Doom.

In Descent the player flew an upgradeable space-fighter through narrow twisting corridors of a robot-infested mining colony. The goal was to clear out the robots in a given mine and then locate the reactor for that mine and destroy it. After destroying the reactor, the player had a set amount of time to reach an emergency escape door before the reactor went super-critical and destroyed the mine.




Descent’s level design was intriguing because it blended the narrow corridors of Doom with the spacecraft-based combat of the earlier Wing Commander and X-Wing games. The 1993 release of LucasArts’ X-Wing featured three dimensional ships like Descent, but X-Wing was set in deep space, and the ships were simple colored polygons, similar in nature to the walls of Hovertank 3D. LucasArt’s 1994 sequel to X-Wing, Tie Fighter, would add polygon shading but few other graphical enhancements. Again, faithful to the Star Wars movies, all combat took place in deep space.

Descent on the other hand, featured fully three dimensional ships with texture maps applied to them, allowing a greater level of detail. The various colors helped players to quickly identify the types of enemy robots they were engaging, even from a distance. Descent also took place exclusively inside the mines, though 1999’s Descent 3 would add the ability to leave the mines and do battle outside using its Fusion rendering engine.

Since the environment of Descent was fully three dimensional, that meant shafts could connect at unusual angles, requiring players to look up, down and to both sides when moving through the levels. Making it to the escape hatches after destroying a reactor either required extraordinary luck, or carefully pre-planning a route of escape before trying to detonate the reactor. It also meant that level design could be challenging, since the 3D engine had very specific requirements about how levels could be constructed.

Descent was also an innovator in its lighting. Where Doom’s lighting was relatively static, Descent had a dynamic lighting system that enabled the use of flares to light areas, as well as laser blasts and explosions. The dynamic lighting also allowed more gradations of light in the mines, which gave a more natural and realistic appearance to in-game lights.

While Volition was adding three dimensions to its world and characters, Apogee and its sister company, 3D Realms, would continue their more humorous take on the First-Person Shooter genre with their next title, the January 29th, 1996 release of Duke Nukem 3D or Duke3D for short. Based on the Duke Nukem side scrollers produced by Apogee in the early 90’s, Duke Nukem 3D was the first commercial implementation of a new engine known as BUILD, developed by Ken Silverman. A self-taught programmer, Silverman became a contract programmer for 3D Realms during his freshman year of college. His BUILD engine matched and, in several cases, surpassed the Doom engine in technical achievements. Set in a near-future science fiction world, Duke Nukem 3D places the player into the boots of world-renowned hero and tough guy Duke Nukem. Duke is essentially a caricature of the stereotypical macho action hero, spouting one-liners throughout the game and generally fulfilling the stereotype. The game was a huge hit, not merely because of the never-before-seen attitude that Duke displayed but because Duke 3D and BUILD had solid technical and gameplay advantages over the games that had come before.

BUILD featured an editor that had a real-time What You See Is What You Get (WYSIWYG) interface, meaning that level designers could lay out a level in two dimensions, then immediately switch into a 3D mode to see what the level would look like. Previous editors and engines required the map to be compiled and then run in the game engine in order for level designers to see the progress of their work. This innovation significantly reduced the turn around time for level design, and also made the process much more intuitive.

Besides making level design easier, BUILD allowed Duke3D to have an unprecedented amount of interaction with the world. The game had the ability to give the illusion of dynamically altering portions of the level, allowing effects such as buildings exploding and collapsing, ground cracking in earthquakes, and certain walls that players could destroy with rocket launchers or explosive barrels. Most of these effects were accomplished with technical slight-of-hand in the engine and in the level design program, and didn’t mean that the engine was actually capable of changing level geometry. Duke 3D and BUILD allowed level designers to add in, for lack of a better term, special effects that gave the player the illusion that they were dramatically effecting or altering the game space, when in reality they were merely triggering the special effects that the level designers had pre-placed. This is in contrast to later games such as Volition’s 2001 release of Red Faction, a title in which the player could use explosives and other weapons to dynamically alter and destroy many walls and other surfaces in the game.

In addition to the influence players could have on the geometry of the level, Duke 3D also added in the ability to destroy or interact with a large number of in-game objects. Fire hydrants could be smashed, urinals interacted with, coke cans exploded, and so on. Practically any decorative object could be destroyed, resulting in a shower of debris, adding realism to the firefights. Glass also made one of its first appearances in Duke 3D, though another sprite-based version had also appeared in Apogee’s earlier Rise of the Triad. In addition to glass, Duke featured mirrors that reflected the architecture around them, as well as Duke. The glass and mirrors could usually be broken, adding yet another small touch to the worlds.




Duke 3D, for all of its technical innovations, was not a fully three dimensional world. Enemies were still sprite-based, as were all of the in-game objects, and the BUILD engine, much like that of Doom still did not support rooms-over-rooms. This made effects such as multi-story buildings or sewers running under a building impossible to do traditionally. Instead, Duke 3D leveraged an effect first seen in Doom: the teleporter. In Doom, teleporters were spaces, usually denoted by pentagrams, that when stepped on would immediately transport a player to another part of the level. The effect in Doom was primarily used to transport players from point to point or to teleport monsters into an area to attack the player. The effect of having monsters appear in this manner was referred to as “spawning”, a term still widely used in level design to refer to the appearance of enemies or objects in the game world.

While still not a completely three dimensional engine, Duke 3D found many innovative uses for sprites, allowing certain decorative sprites to be applied directly to wall surfaces. These sprites were commonly used for items such as signs, boards and calendars, though they were also used for blood spatter on walls, cracks, scorch marks and bullet holes. Such sprite based effects were first used in Rise of the Triad, but Duke Nukem 3D greatly expanded their use, and did so in highly creative ways. Minor effects such as blood from enemies splattering against a wall behind them helped to make characters seem more a part of the world.

In Duke Nukem 3D, developers took the idea of teleporting and used it to cover up the weaknesses of the engine, giving the impression of it being capable of more than it really was. An excellent example of this can be found in the Red Light District map, the second map of the first episode. After obtaining a keycard and destroying a building, one can find a manhole cover in the wreckage. If the player destroys the manhole cover with explosives, they can drop into the sewers. Looking more closely, though, one will note that the manhole pipe is actually a dead end; if you look down into the vertical drop, the bottom can be seen. By dropping into the hole, however, an invisible teleporter is triggered that moves the player to a different area of the level that looked like a sewer. The sewer was supposed to be immediately below the destroyed building, but since the BUILD engine couldn’t do rooms-over-rooms the level designer, Alan Blum III, chose to use an invisible teleport to move the player to a location not immediately underneath another room. Such techniques are used throughout Duke Nukem 3D to accomplish a number of effects, including any water in which the player can actually submerge themselves and swim in. Because of careful forethought and good map design, these effects are almost completely transparent unless you know what to look for.

These effects were a crude predecessor of the scripting languages now used to control many of the variables and effects in FPS titles. By altering values in the editor, known as “hi tags” and “lo tags”, level editors could assign certain actions to certain objects, as well as link a number of objects together to function as a single entity. These tags and links made extremely complex actions possible.

Unlike Doom and RotT, the levels in Duke 3D were usually built around a central theme, as well as sharing a thematic link via episode. For instance, many of the maps in episode one, L.A. Meltdown, and in episode three, Shrapnel City, are centered on recognizable city buildings such as a movie theater, sushi house, prison, and so forth. The second episode consists of more fanciful, but still recognizable, space-based structures. Again, all of these maps, while not linear in the way levels in Half-Life are connected, are still linked, giving the player the impression of a larger world. The fact that the game world was both easily recognizable and more interactive than ever before made Duke Nukem 3D an extremely popular title.

While Duke Nukem 3D was gaining fans with its tongue in cheek attitude to the game world and its technical innovations, id software, fathers of the PC First Person Shooter revolution, were not resting on their laurels. On October 10th, 1994 id released Doom II: Hell On Earth, the sequel to their smash hit. The game was a huge seller, but offered no major technical advancements over Doom. Indeed, the engine was exactly the same, featuring no improvement to graphics or to the gameplay, though there were several new enemies and a new weapon, the double-barreled shotgun. The game, while wildly successful, offered little more than its predecessor, but the gameplay of the original Doom and Doom II was so compellingthat it did not matter. Still, id’s John Carmack had a vision for the future, and that vision was a fully three-dimensional world (Kushner, 178-179).




Quake would be that next id title, and the realization of Carmack’s technical vision. Everything, from the environment architecture to the enemies and powerups would be polygon based, another first in the industry. Singleplayer gameplay and world detail, however, would suffer a severe decrease during the transition to full 3D, since the computing power needed to render the world meant that the pace of the game would be much slower than Doom. Worse still, since everything was polygon based, that meant that adding detail to an object meant adding polygons, and more polygons meant less speed (Kushner 216-217).

Released on July 22, 1996, seven months after Duke Nukem 3D, Quake featured next to no story, but like Doom chose to focus primarily on action. The game featured dynamic lighting, similar to that implemented in Descent, and a variety of enemies that ranged from towering lightning-shooting behemoths to twisted knights to zombies that would throw hunks of their own bloody entrails at the player. The game was extremely popular, and was a major software engineering achievement, but featured single-player gameplay that was almost exactly identical to that of Doom.

The levels in Quake were a mixture of the work of a number of level designers, all working on different themes. This led to an uneven tone in the level designs that id attempted to reconcile by making teleportation and inter-dimensional travel a core theme of the game. Nevertheless, compared to many other titles, particularly Duke Nukem 3D, the world had a very static feel. Combined with the dark color palette, Quake provided a singleplayer experience that, beyond the technical achievements of the engine, offered little new gameplay.

Like Doom, Quake was designed with modification in mind. This time, instead of simply relying on WAD files, Carmack developed a scripting language called QuakeC that allowed members of the mod community to drastically alter the game. Adding new weapons and enhancing player function became a relatively simple affair, and a number of popular modifications such as TeamFortress and ThreeWave Capture the Flag were a direct result of the power of the modding tools. These user-created modifications would help fuel the popularity of Quake as well as a growth in the popularity of modding games.
Multiplayer proved to be Quake’s strong suit, with the game featuring support for the TCP/IP networking protocol, allowing multiplayer games to now take place over the burgeoning internet. A later update to the game would add in a system known as QuakeWorld, which added client-side prediction to the game, greatly improving network performance on slow dial-up connections.

Curiously, the greatest achievement of Quake may not lie in its gameplay or its ease of modification, but in its use as a test bed in the evolution of 3D accelerator cards. Carmack used a modification of Quake known as GLQuake to allow the game to use the new consumer technology of graphics accelerators to add both new features to Quake, as well as improve its rendering of the world as it existed. In addition to increasing the speed of the game, allowing gameplay speed closer to that of Doom, GLQuake added graphical enhancements such as making water transparent, adding reflections and also adding shadows. Until GLQuake, water in Quake and most other titles had been essentially opaque, with no way to see what was in the water without jumping in. GLQuake made it possible to look right into the water, which not only allowed players to butcher their swimming opponents, but added another small touch of realism to the now fully three-dimensional world. The added shadows served an important function, giving game characters and items a greater appearance of being grounded into the game world. Use of 3D graphics acceleration is now common in the industry, and its adoption has shifted much of the graphics strain from the processor onto specialized graphics chips, allowing the computers main processor to devote it’s time to other tasks, such as artificial intelligence for non-player characters and physics calculations for game objects.

Id would follow up Quake with two official mission packs, the first being Scourge of Armagon, released on February 278th, 1997 and created by Ritual Entertainment. The second mission pack was Dissolution of Eternity, released on March 31, 1997 by the now-defunct Rogue Entertainment. While Scourge of Armagon received considerable praise for its excellent level design and inventive use of traps, as well as a cohesive series of levels with an overarching story, Dissolution of Eternity was somewhat less popular. The fact that Richard “Levelord” Grey, one of the founders of Ritual, had been intimately involved in the level design for some of the most memorable Duke Nukem 3D levels likely played a part in the inventive design of the Scourge of Armagon maps. In addition to new levels, both expansions added new weapons and new monsters.

Quake and it’s sequels Quake II and Quake 3 Arena would continue to push the boundaries of rendering technology, but would do little to advance the art of level design and storytelling. While Quake II’s release on November 30th, 1997 would be a significant cash cow for the company, its much-vaunted single player storyline would once again place the player in the shoes of a lone space marine against impossible odds. Technically, the game would add improved graphics and the ability to render colored lighting, allowing for much more dramatic graphic effects. Quake III Arena would enhance the engine technology by allowing rounded surfaces in games, meaning that more organic shapes could be constructed. Previously, almost all levels were constricted to more angular shapes. As Quake II Arena was essentially a multiplayer only title, little use was made of this technology, and even if it had been properly seized upon it is unlikely that players involved in intense multiplayer deathmatches would stop to admire the architecture.

Engine Refinements, Storytelling and Interactivity

The move into a fully three dimensional world with Quake was probably as momentous an occasion as the release of the original Wolfenstein 3D or Doom, a turning point in the development of three-dimensional first person titles. Many companies would license Quake engine technology in order to construct their own games around its powerful rendering technologies, just as companies did with Doom. In addition to permitting faster development of games, this licensing of engine technology had a second, less recognized effect. It allowed the licensees to concentrate more of their energies on the design of the actual game, instead of focusing as heavily on technical concerns. That is not to say that the engines were simply plug and play, but that programmers were spending more time modifying the engine to suit their needs, instead of designing whole new engines from the ground up. As the 1990’s came to a close, a slew of new titles arrived on the shelves, with many offering singleplayer innovation.

On May 28th, 1998, Digital Extremes and Epic Games released Unreal, a title that had been under development for four years (Grossman, 91). Unreal had impressive graphical capabilities, supporting very detailed textures, connected linear levels and fairly advanced artificial intelligence for the enemies. This resulted in moments where enemies would narrowly dodge projectiles at the last moment, a nasty surprise to players.

Level design wise, the game featured moments demonstrating nearly cinematic pacing, such as the players first encounter with a Skarrj warrior. Like Quake, Unreal featured a full three-dimensional engine, but supported more complex environments. Unreal also required levels to be constructed in a much different way than Quake engine titles. In Quake based titles, a level starts empty and must be assembled from various geometric shapes, called brushes. These brushes can be manipulated to alter size and shape, as well as other features, resulting in what can be called additive level construction. Unreal engine based projects, on the other hand, use a subtractive model, where the world starts full and level designers create empty spaces to serve as rooms, then add other geometry as details, much like a sculptor whittles down a block of clay or marble to create a sculpture. Level design for Quake engine titles were more akin to working with Legos that could be stretched and modified.

Unreal also featured much more natural environments. While Duke 3D did a good job of simulating cities and urban environments, Unreal was adept at creating believable and lush pseudo-tropical landscapes. The levels featured effects such as waterfalls, transparent water, colored lighting and greater interactivity with objects such as boxes, which could be pushed and used to create stairs. While the actual game offered little new, the impressive use of graphical effects served to add yet another layer of depth to the virtual world.

While the Unreal and Quake engines would become the two dominant engines used for the creation of First Person Shooters for computer games, they would not be the only engines developed. Several companies, such as Looking Glass, would continue to develop their own engines from scratch.

The Dec 3, 1998 release of Thief: The Dark Project and the August 1999 release of System Shock 2, developed nearly simultaneously, marked the first implementations of the Dark engine. Thief was best described as a First Person Sneaker, where the object of the game was not to loudly blast through enemies, but instead to avoid detection while pilfering valuable or interesting objects. The storyline was involved and played out in animated cut scenes before and after each level, setting the stage for the action to come. The cut scenes were well done, but it was the gameplay that was novel, encouraging players to hide in the shadows and use a variety of arrows to ease their path. Thief featured truly dynamic lighting, with almost every light source able to be doused, a vital component of the gameplay. Thief is, at the very least, the spiritual ancestor of popular modern titles such as the Splinter Cell series from UbiSoft. Thief also illustrated that there was a market for titles played from a first person perspective other than violent slaughter-fests.

Thief also relied heavily on audio as an element of player involvement. In most previous titles, enemies were essentially silent unless they were attacking the player. In Thief, one of the best ways to determine the location of an enemy was by their footsteps. Further, players could use the sounds made by the NPC’s to determine how aware or suspicious they were; casual whistling could indicate they were unaware of anything amiss, while yells for help would ensue should the player be spotted. Players could also use these aural capabilities to their advantage, throwing objects or using special noisemaker arrows to distract opponents. This use and recognition of audio as an important part of the immersive experience was a significant step forward, adding another vital element to level design; the placement and use of ambient audio. While ambient audio had been used in previous projects from Doom to Duke Nukem 3D and beyond, Thief was the first title to make audio a central element of the gameplay (Grossman 175-176).

System Shock 2, developed by both Irrational Games and Looking Glass Studios, was a sequel to the innovative, if overlooked System Shock. System Shock 2 continued the story of System Shock, with the player taking the role of the sole survivor of a terrible disaster aboard two ships deep in space. The player awakens with no knowledge of past events, and through audio logs and e-mails must piece together what happened aboard the ships.

Like it’s predecessor, System Shock 2 was a difficult title to categorize, having elements in common with role playing games, action games like Doom and adventure games. More generally, the game could be categorized as an action horror survival game, as the player had no idea why the crews of the ships were dead, and seldom enough ammunition to simply blaze through any opponents. Item placement was a critical element of level design in System Shock 2, as designers were careful to never give the player an overwhelming amount of resources. Players were required to carefully horde ammunition and supplies, as well as manage various ammunition types. As in System Shock, certain weapons and ammunition types worked best against certain enemies, so players had to be aware that they could encounter any of a variety of enemies at any time, and that using a more effective ammunition type would help conserve their meager resources.

Problems or “puzzles” in System Shock 2 frequently had multiple solutions that would depend on the various skills of the player character and their playing style. Enemies could be killed or snuck by, doors opened by finding a key code or by hacking the lock. Players could disable cameras by shutting down a security system, destroying the camera or merely sneaking by it when the camera was pointed elsewhere. As in previous games from Looking Glass, players were usually rewarded for careful play and exploration of the world, receiving upgrade chips that could be spent to improve character abilities in an RPG style system. The game also allowed characters to do research on enemies using a variety of simple chemicals. This research would then yield distinct knowledge or combat advantages over opponents.

System Shock 2 also made extensive use of scripted sequences, a concept that would be fleshed out more fully in Half-Life. As opposed to pre-rendered movies advancing the story, System Shock 2 chose to display almost all events inside the game engine itself, helping to maintain player immersion which could easily be broken by the interjection of pre-rendered movies. Many of these events were highly unexpected, such as the player’s first encounter with a “ghost” of a crewmember. While the models of characters and objects would later be criticized by some players as primitive, the attention paid to character and level design, as well as the vital role of sound effects and spoken dialogue made System Shock 2 a highly successful and critically acclaimed title. The game is still considered by many to be one of the best examples of the genre and of game story in general.

System Shock 2 is joined in this pantheon by another game that has direct ties to
Ultima Underworld and System Shock; Ion Storm Austin’s Deus Ex. Released in late June of 2000, Deus Ex was set in a dystopian future where conspirators and terrorists have turned the United States into a fractious, diseased and crumbling nation. Levels were set to resemble recognizable locations such as Liberty Island and the Statue of Liberty, Battery Park in New York and other areas throughout the world. The player Avatar, J.C. Denton, was a nano-augmented agent for a United Nations anti-terrorist group.

The game, designed by former Looking Glass developer Warren Spector, had much in common with titles like System Shock, System Shock II and Ultima Underworld. The player’s character could define an early set of skills and abilities that later could be modified through a combination of experience points and “augmentation canisters”, which would add new functions to a player, such as the ability to increase their strength or to become resistant to radiation. Augmentations could also be upgraded using upgrade canisters, a separate system from the experience or “skill” points system.

In addition, Deus Ex allowed players to use a variety of play styles and tactics to achieve in-game objectives. Many objectives had several different approaches that would all be suitable, allowing players to exercise their discretion and giving the impression of a great deal of freedom in what was still a largely linear game world. For instance, when confronted with a locked door in most games, players would know they would have to find the key or a switch to open it. Deus Ex could allow players several options, such as destroying the door with explosives, picking the lock, hacking the security system to open the door or finding a way around the door, typically through a ventilation or sewer system, or by navigating other nearby rooms.

Naturally, such freedom came at a considerable cost for level designers, necessitating massive amounts of pre-production and planning for level design and other systems (Grossman 200-201, 205-206). Level designers would have to take into account the various augmentations and skills that a player might have and provide a sufficient variety of tools for a player never to become completely stuck in a dead-end merely because they didn’t have the requisite skill level to hack a computer or pick a lock. This meant that other solutions had to be found, such as key rings containing necessary keys for players to use.

The issues faced by Deus Ex serve as both an example of how good planning can result in better level design, as well as a cautionary tale about the difficulties of giving players choices. While many players clamor for more inventiveness and freedom in games, implementation of such abilities presents serious challenges for designers, necessitating, as was done with Deus Ex, early functional prototyping of levels and other resources. Deus Ex was richly rewarded for its efforts, garnering a great deal of praise both for its comparatively open-ended gameplay and its ability to allow players to play the game in a manner that fit their personalities. The game also received considerable praise for its conversational system, allowing players to choose from a number of pre-scripted conversational choices, each of which would affect the course of the conversation with an NPC. This furthered the sense of immersion and the impression that player choices would have tangible effects on their ability to progress, as well as NPC’s perception of them. This system, allowing players to actually select from conversational choices was an ideal method for exposition and character development, but not the only approach to player and NPC interaction.



A company that took a different approach nearly two years before Deus Ex was Seattle based Valve Software. Released on November 20th, 1998 after more than a year delay, Half-Life put players in the shoes of Gordon Freeman, a research scientist at a top secret government facility in the fictional location of Black Mesa, New Mexico in the United States. Half-Life is remarkable in many ways, but one of the most obvious is the method used to introduce the player to the world. Typically, players are thrust into their characters immediately after a disaster has occurred rendering all other friendly non-player characters dead or dying, or at the very least in need of help. This is a storytelling device that serves to cover up the fact that the technology for players to interact believably with Non-Player Characters was, at best, limited. Indeed, this idea of the limited capability for players to interact with “friendly” characters had become something of an accepted fact in many titles.

Half-Life took a different and arguably more cinematic approach to their storytelling. Players began on a highly detailed tram ride into the Black Mesa Research Center, with the tram ride serving as an introduction to Black Mesa at the beginning of a normal work day. As would be expected, the player is completely unarmed throughout this portion of the game, a dramatic difference from practically all other titles. The player would then have to follow verbal prompts and instructions from Non-Player Characters in order to achieve their goals. The characters featured a form of lip syncing, similar to the appearance of a puppet, that caused their mouths to move in approximations of the proper shapes for certain sounds, giving the impression that the characters were actually human and speaking to you. Players would then proceed down to a test chamber where they themselves would become responsible for the initiating event that would lead to the disaster at the facility. The concept of showing players the world before the disaster, letting them become familiar with it in its natural state, served to give players a reference point by which to compare the following chaos and disorder.

The player would then have to move through the facility, frequently relying on Non-Player characters to open doors and provide medical attention, as well as supply advice and hints as to the next course of action that the player should take. Valve, also realizing that the technology was not yet sufficient to allow back and forth conversation with NPC’s, chose to make Gordon completely mute, and simply have characters speak to him directly. With careful writing the designers could give the impression that the conversation was at least a natural one, if decidedly lopsided.

Half-Life also featured an excellent implementation of level transitions, similar to those used by Unreal. Instead of an intervening screen between levels, Half-Life would load the next level dynamically when the player reached the end of one map, displaying a small “loading” graphic before resuming the game. The transitions were as seamless as possible, allowing for next to no pauses in gameplay. While the level transitions typically required a reasonable amount of pre-planning on the part of the level designers, the seamlessness gave players the feeling of truly being in a continuous world. Additionally, players could backtrack over considerable distances in the game, allowing them to go back for items or equipment that they may have missed or wanted to save.

Half-Life was based on a heavily modified version of the original Quake engine, providing the game with a fully three-dimensional world, but the additions made by Valve made the singleplayer game many times more advanced that that of Quake. Colored lighting, the use of scripted animated sequences to advance story and heighten tension and the construction of both impressive indoor and outdoor environments made Half-Life a hallmark of the industry. Combined with the intriguing plot and the addition of an endgame choice, the game was a wild success. Further, the release of level design and other tools, called a Software Development Kit or SDK, turned Half-Life into a success in the online gaming world, spawning a number of third party modifications such
as Counter-Strike, Natural Selection and Day of Defeat.

The Future

While a number of titles have been released since Half-Life, including its widely acclaimed sequel, Half-Life 2, there has been surprisingly little advancement in the field of level design since Half-Life. Many other titles have adopted features that were present in Half-Life and made iterative improvements, while some titles have updated older methods of interaction, such as Deus Ex or System Shock 2. Still, the question remains regarding what level design and level designers are becoming.

The release of mapping tools to the general public has allowed the creation of hundreds of thousands of maps and collections of missions for a variety of FPS’s, beginning with Doom and continuing on with titles like Doom 3, Half-Life 2, Star Wars Jedi Knight II: Jedi Outcast and Halo 2. First Person Shooter titles have branched out from personal computers and onto popular consoles, with games such as Goldeneye for the Nintendo 64 andthe Halo series for Xbox, but the gameplay model has, by and large, remained the same. Some of the more popular modification teams have even been hired to do commercial work, such as the poorly received Gunman Chronicles, the product of a total conversion for Half-Life.

While the availability of the tools has given rise to new ranks of level designers, the job is constantly increasing in complexity. Early titles could have their levels designed by only one person in a few hours, as was the case with Wolfenstein 3D. Games such as Half-Life and Half-Life 2 now require team efforts, with designers specializing in lighting, weapon and enemy placement and the creation and implementation of scripted sequences to make the world come alive. It is highly likely that in the coming years we will see the emergence of a division of labor very similar to that of the film industry, with certain designers laying out architecture while others apply textures and still others place enemies, items and monsters. Valve Software itself noted that it has had to change the design process for its own levels, laying out architecture with a flat default orange texture in order to test gameplay and level flow before dedicating the resources to applying the necessary texture maps, lighting and other small touches that truly bring levels to life.

Level designers have come a long way from the early days of the first person shooter, but with each technological leap the necessary time, preplanning and design required to create a level has increased significantly. It is highly likely that just as the auteur game programmer has become extinct, so too will the auteur level designer, replaced instead by what Valve software refers to as “cabals”, teams of designers working in concert to bring a level to life. This is not limited to just FPS titles, since the growing complexity and open-ended gameplay of games like Grand Theft Auto: San Andreas and World of Warcraft require level designers to expand their skills far beyond that of simple geometry creation and lighting.

These design challenges raise important questions for the game development community regarding the methods and technologies that are being used to develop content for titles. Certain designers, such as Maxis’ Will Wright, advocate the use of procedural generation technology to allow algorithms to handle the bulk of content generation, a technique he plans to use in his upcoming game Spore. Valve appears to advocate the cabal design process, wherein they recognized that level gameplay and flow is the primary issue. Because of this, they chose to use their technique of texturing prototypes in a flat orange color in order to concentrate fully on gameplay and not be distracted by graphical concerns, a process that appears to have worked well for Half-Life 2.

However, perhaps it is not an issue of team size, but an issue of tool improvement. The level design tools that we have today are advanced, but likely have not advanced at the pace of the rendering engines themselves, so there is likely room for improvement both in function and usability. Could we alter the way the levels are created so that rapid prototyping could be made even easier? Which approach to level design is more robust, the additive techniques used in the Quake and Doom 3 engines, or the subtractive methods used by Unreal engine titles? Is there a combination of the two techniques that would work best? These are questions that must be answered so that the pace and advances of level design can keep up with the requirements that are being placed upon the level designers, particularly with a new generation of consoles and other hardware nearly upon us.

About the Author

Sam Shahrani is an M.A. candidate at Indiana University in the Master’s in Immersive Mediated Environments program through the Department of Telecommunications. He can be reached via e-mail at [email protected].

Works Cited

Atari Corporation. “Operation, Maintenance and Service Manual Complete with Illustrated Parts List: Battlezone”. California, 1980. December 29, 2005.

Bowery, James. “Spasim (1974) The First First-Person-Shooter 3D Multiplayer Networked Game”. Jim Bowery’s Personal Website. April 4, 2001. December 28, 2005

Braben, David. “Elite Frequently Asked Questions”. Frontier Developments. 2003. December 30, 2005.

Dunnigan, James F. “TheComplete Wargames Handbook 2nd Edition”. New York, 1992. December 29, 2005.

Grossman, Austin. Postmortems from Game Developer: Insights from the developers of Unreal Tournament, Black & White, Age of Empires, and other top-selling games. San Francisco: CMP Books, 2003.

Hall, Tom. “RE: Questions about your career in level and game design.” E-mail to the author. 1 January 2006.

Handy, Alex. “The First First Person Shooter.” Computer Games Magazine. July 2005. October 5, 2005.

Kushner, D. Masters of Doom: How Two Guys Created an Empire and Transformed Pop Culture. New York: Random House, 2003.

Langston, Peter. “BALLBLAZER and Rescue on Fractalus!: The Lucasfilm Computer Division Games Project is born - A very brief personal history”. January, 2005. Peter Langston. December 28, 2005.

Rollings, Andrew and Dave Morris. Game Architecture and Design: A New Edition. Indianapolis: New Riders, 2004.

Romero, John. “RE: Wolfenstein 3D and Level Design.” E-mail to the author. 20 February, 2005.

Siegler, Joe. “A History of Wolf3D”. 3D Realms Website. 3D Realms, Inc. January 19, 2005

Thompson, Greg. “The aMazing history of Maze – It’s a small world after all”. DigiBarn.com. November 7, 2004. DigiBarn Computer Museum/Computer History Museum. December 30, 2005.





Read more about:


About the Author(s)

Sam Shahrani


Sam Shahrani is an M.A. candidate at Indiana University in the Master’s in Immersive Mediated Environments program through the Department of Telecommunications. He can be reached via e-mail at [email protected].

Daily news, dev blogs, and stories from Game Developer straight to your inbox

You May Also Like