The first first-person shooter

It started as an experiment. Steve Colley had just figured out how to rotate a cube on the screen when Howard Palmer suggested they could make a three-dimensional maze.

The year was 1973. They were high school seniors in a work-study program with NASA, tasked with testing the limits of the Imlac PDS-1 and PDS-4 minicomputers. Their maze program flickered into life with simple wireframe graphics and few of the trappings of modern games. You could walk around in the first person, looking for a way out of the maze, and that’s about it. There were no objects or virtual people. Just a maze.

But Maze would evolve over the summer and the years that followed. Soon two people could occupy the maze together, connected over separate computers. Then they could shoot each other and even peek around corners. Before long, up to eight people could play in the same maze, blasting their friends across the ARPANET — a forebear to the internet. Two decades before id Software changed the game industry with Wolfenstein 3D and Doom, Colley, Palmer and MIT students Greg Thompson and Dave Lebling invented the first-person shooter.

This is the story of Maze, the video game that lays claim to perhaps more “firsts” than any other — the first first-person shooter, the first networked multiplayer game, the first game with both overhead and first-person view modes, the first game with modding tools and more.

Greg Thompson was always more interested in electronics than medicine, which his anesthesiologist father had tried to push him into. He had just the right mentor, John McCollum, who was the electronics teacher at Homestead High School and who Thompson believes is the unsung hero of Silicon Valley. Perhaps most notably, Apple co-founders Steve Wozniak and Steve Jobs (who was a year above Thompson) learned the skills they required to build the Apple I computer from McCollum.

In the summer of 1972, McCollum got Thompson into a college-level work-study program at NASA’s Ames Research Center. There Thompson met fellow Bay Area high schoolers Steve Colley and Howard Palmer. They worked in the computer graphics lab at NASA Ames, which was doing research into computational fluid dynamics to help design future spacecraft.

The lab had several Imlacs. These were programmable minicomputers with graphical displays — sort of the high-end personal computer of the time with enough memory to handle one person running what passed then for one decent-sized program. Colley, Palmer, Thompson and the rest of the group were tasked with exploring the capabilities of these systems — to experiment and push the boundaries. Being teenagers, this, of course, led to them mostly making games. They started with 2-D games like Pong and duplications of other early arcade titles such as Drop Zone, which they regularly played at their local Togo’s sandwich shop in hopes of getting a score high enough to win a free sandwich.

In 1973 Colley figured out how to rotate a wireframe cube on the screen without the lines at the back showing. Up to this point, Colley and Palmer, the senior members in the group, had refrained from making any games. Their focus was on serious, “useful” software. But one day while thinking about how to make something fun, they came up with the idea of creating a networked, multiplayer maze game. Palmer thought it would be cool in 3-D, but Colley doubted it was possible until Palmer noted that it could work if the maze had nothing but 90-degree angles. Colley got excited and returned the next day with an implementation that Thompson says was, effectively, a bunch of cubes stacked next to each other. The cubes that were “there” formed the walls while the rest acted as hallways. From that, Thompson continues, “You basically got a maze. You could see corridors you could walk down.”

The maze was small, sized 16 squares by 32. Thompson says that its graphics provided a sense of being in a space, and the three of them enjoyed wandering around its halls looking for the way out. But Colley, writing in a retrospective for the DigiBarn Computer Museum, notes that it got boring soon enough. You could solve the maze, and that was it. Then one of the trio had an idea: What if there were something else in that maze? Colley added a little extra code and connected two Imlacs directly with a serial cable. “It just seemed natural that you could potentially communicate to another machine,” Thompson says when asked how they came up with the idea.

“It just seemed natural that you could potentially communicate to another machine”

Suddenly you could have two people walking around the maze together each with a one to eight character floating username and a simple graphical indication to show each person’s location and orientation — two dots for eyes if they’re facing you, arrows if they’re looking left or right, nothing if they’re moving away.

To make things even more interesting, they added bullets. Now you could not only see a second person but shoot the person as well. And as a final contribution, Colley added the ability to peek around corners before stepping into a new corridor. Your viewpoint would shift around the corner so long as you held the key down, but you couldn’t do anything else while peeking.

Maze wasn’t drawing much attention outside of its three creators at this point, but that quickly changed after they graduated high school. In the fall of 1973, Colley went to Caltech and Palmer to Stanford while Thompson took Maze and the rest of their Imlac programs with him to MIT. And at MIT he got involved with Project MAC, which later became the college’s computer science laboratory, and he met the dynamic modelling group — which was making games of its own.

Dave Lebling may be the last person you’d expect to have been involved with inventing the first-person shooter. He was co-founder of Infocom, which was a company that made adventure games for personal computers before adventure games had graphics. Infocom’s imaginative text-only games dominated the industry in the first half of the 1980s, before graphic adventures rose to prominence. And Lebling co-wrote the first and perhaps most famous of those: Zork, an adventure that he’s surprised and delighted to see is still a beloved source of inspiration for many creative minds. But before he did any of that, Dave Lebling helped expand Maze into a livelier experience.

Lebling doesn’t recall the version of Maze that Thompson brought to MIT having shooting in it. “You wandered around in a maze, and that was it,” he says. Regardless of whether the shooting was added before or shortly after Maze reached MIT, Lebling and Thompson soon teamed up. Thompson focused on the Imlac side of things since he knew the machine better, and Lebling said that he would work on the PDP-10 mainframe — the big computer upstairs that the Imlacs were all connected to.

At NASA Ames, each Imlac ran its own version of Maze and calculated positions locally — a quirk that could lead to discrepancies over whether someone was shot or not. With the mainframe, they could offload this work to a central machine that would coordinate all moves and broadcast them to all players.

“You wandered around in a maze, and that was it”

“It took a certain amount of debugging,” Lebling recalls, “but it worked, and it was really cool.” With Lebling’s help, Maze gained the ability to handle up to eight players on a network. Thompson added code to keep track of scores and graphics that would show missiles travelling down a corridor. And Lebling also wrote a maze editor, which allowed players to create new maps.

“It was very simple,” he explains. “You called up a text editor and you created a grid of — I forget what character we used, but it may have been any character other than space. You just created a grid of text characters that had 1s where there were walls and 0s where there weren’t once it was processed.”

“People would try various bizarre ideas for mazes,” Lebling says. But the standard maze got the most use. “It had some very twisty parts and it had some long hallways, so you could potentially duck out into the hallway and shoot at someone who was quite far away and then duck back and they would die.”

Word got around in the lab, which had eight Imlacs. “People started playing and it was an enormous — it was a viral thing in today’s sense,” Lebling says, “because we weren’t the only research group in the lab.”

People came from other levels in the building just to play Maze. “The Imlacs were sufficiently scarce that we actually had to sign up to use them,” Lebling says. “So people would come in the middle of the night to play and they would shoot each other and run around and there’d be a running score and so forth.” Thompson says that any MIT student or professor could set up an account on the system, so word quickly spread around campus. Twenty years before Doom death matches took over local networks in college dorm rooms, MIT students were doing the same thing on computers meant for cutting-edge research.

Spurred on by this popularity, the pair kept adding new features. They enjoyed playing Maze, but their true passion was making it better, learning more and impressing their friends. Thompson’s grades started to slip, too. “I probably spent a lot of time coding and getting it working because I was intrigued by it,” he says. “I wanted to, and I was learning.” Lebling’s driving motivation differed slightly: “Writing [games] and watching other people enjoy them was, I think, more fun than playing them,” he says.

They added a top-down view that you could switch into to see the whole maze and better plan your routes, but the caveat was that, while you were looking at that, someone could sneak up on you fairly easily. Thompson added cheats that made it possible to pass through walls while Lebling wrote robot players.

“People started playing and it was an enormous — it was a viral thing in today’s sense”

“There weren’t always enough people to make it interesting,” Lebling says, “so I decided to write a little really stupid — artificial stupidity rather than artificial intelligence — robot program. You could start up a bunch of robots. If you wanted to play by yourself with seven robots you could have done it. And people would do that and the robots would wander around the maze and shoot you.”

The robots were too good for many players. “We had to make them move slower than they normally could have,” Lebling says. “They adaptively decided they had beaten you too much and made themselves stupider, or if you beat them too much, they would make themselves smarter. There was a bit of feedback going on there.”

Nevertheless, Lebling notes, the robots were no substitute for human players. They were simply a fun option for if you were “bored and wanted to play by yourself,” he says.

The very best Maze player, Ken Harrenstien, soon rewrote much of the program on the mainframe to make it more efficient, while Charles Frankston did the same on the Imlacs portion. This made it more feasible for two or more simultaneous Maze games to run through the PDP-10 mainframe.

Meanwhile, another man in their research group, Tak To, wrote a program called Maze Watcher, which allowed people to watch matches on an Evans and Sutherland LDS-1 graphics display computer. “You could sit in the lab that the display for the graphics computer was in and you could cheer on your friends as they played,” Lebling recalls, mimicking “Go go go! Oh no! He’s around the corner!”

Maze’s popularity didn’t stay just within MIT’s walls. They shared the source code across the country via the ARPANET. Lebling believes that the first game of Maze played over the ARPANET was between students at MIT and the University of California, Santa Cruz. It was slow — the ARPANET of the time was only a 50-kilobit network, which is a few orders of magnitude slower than the megabytes and even gigabytes that a fast internet connection manages today. But it was wonderful.

Perhaps even a little too wonderful. Legend has it that Maze was later banned for a period by the Defense Advanced Research Projects Agency (DARPA), which ran the ARPANET, because half of all packets (data, essentially) being sent over the network each month were flying between MIT and Stanford. Lab directors J. C. R. Licklider and Al Vezza worried well before then that Maze would be seen as too frivolous. They discouraged Maze play in the lab, which was funded by DARPA, to keep things serious and to keep the Imlacs available for serious work.

“Maze clogged up the Imlacs, which were a very scarce resource”

“[Maze] clogged up the Imlacs, which were a very scarce resource,” Lebling says. “To run Maze, you had to load the Maze program in instead of the usual terminal program because these things only had 8K of memory, [so] people would load the Maze program in and then forget to reload the terminal program, or it would crash or something and they wouldn’t know how to reboot an Imlac, which wasn’t all that complicated but was more difficult than rebooting a personal computer today.”

To combat this, the directors banned Maze. (Despite that they had been spotted on multiple occasions playing and enjoying the game, too.) It was a futile endeavor. The computers were on a network, so people would simply copy it over from another machine or hide away some code that would download the program again. Things soon escalated, and someone added a daemon program — a program that runs in the background to look for instances of Maze and crash them.

This Maze Guncher could be easily defeated by either recompiling the Maze code with a different signature or by taking advantage of the fact that the system had no security whatsoever to disable the Guncher daemon. “It was just this hackers versus nonhackers thing,” Lebling says. “This little war going on back and forth.”

In the fall of 1976, Greg Thompson, Mark Horowitz and George Woltman enrolled in an electrical engineering digital hardware design class together. For their class project, Thompson suggested they make a system designed to run just Maze.

“I thought it would be fun, and then we started thinking about how to do it and build it. It seemed like a pretty complicated and exciting thing, but talking about it we thought we could figure out how we could actually implement it,” says Horowitz. “We told the TAs in the class that we were going to do it, and they told us that basically they didn’t think it was a good idea but if we wanted to we could go ahead and try.”

The trio determined that it would be easier to build the game if they made it more complicated. Whereas the original Maze was effectively a two-dimensional maze that was drawn in line segments, the new version was fully 3-D. In the hardware version of Maze, you could turn in any direction inside a 16-by-16-by-16 cube that had four floors in each dimension.

“The reason we did that was it made the game completely symmetric,” Horowitz explains. “It turned out, from a hardware perspective, it was way easier to do that because there were less specialized surfaces. So you were walking in a maze with no gravity, essentially. You could turn left, right, or you could turn up or down.”

It was about as confusing as it sounds, but they found it exhilarating. Then Woltman added robot players. “They played so much better than we did because they never were confused,” Horowitz recalls. “They just followed some pretty simple rules and just looked around for people, and when they saw them, they shot them.”

Horowitz, meanwhile, would be too busy trying to remember where he was in the maze. “[I was] trying to remember which end was up,” he says. “They didn’t care. They just walked through and found openings and continued moving forward.”

Like in the Imlac version of Maze, you could make the robots play better or worse, but in this case it was entirely a matter of changing the clock rate on the processor. “If you wanted it to play faster you’d just turn up the clock rate and run the whole system a little bit faster,” explains Thompson.

As for the display, without a graphics processor they had to get creative. “We ended up deciding to use oscilloscopes because they were in the labs to help you to debug your hardware,” says Horowitz. “We used it in X-Y mode and I used the display as a vector draw display. So instead of doing raster scan, which is how all displays are today, I actually drew lines on the screen.”

Horowitz also drew bullets that you could see flying down the screen, gradually getting bigger as they came closer or smaller as they moved away.

The game was so impressive that the trio earned near-perfect marks for the project and the lab kept hardware Maze around for a year or two as an example game to show other students. And Horowitz, for his part, is surprised not only that people still care about the project today, but also that he can remember it all. “I have a really bad memory,” he says. “I don’t remember what I did two weeks ago let alone 40 years ago. But the truth is, I remember the maze game. I remember what it was like. What the graphics looked like and how it all worked and I just remember how exciting it was trying to build it.”

Maze’s many creators each went off in separate directions. Lebling and a few of his dynamic modelling group cohorts founded Infocom while Thompson flitted around the technology industry as a consultant, engineer and designer for the likes of Digital Equipment Corporation, nCube and Cisco. Steve Colley founded nCube, which became a leader in the early video on demand industry. Horowitz became director of Stanford University’s computer systems lab. And George Woltman went on to discover several of the largest known prime numbers.

Maze carried on, mutating and spreading throughout the computing world. Jim Guyton rewrote the Imlac version for the raster displays of Xerox’s Alto and Star personal computers, renaming it Maze War and adding the minimap to the bottom of the main display. Two members of the early Macintosh team at Apple saw this version and wrote their own for testing purposes — Gene Tyacke’s Bus’d Out, which was leaked out at an alpha level of completion, tested an early version of the AppleTalk networking protocol, while Burt Sloane’s Maze Wars demonstrated a file transfer and messaging server.

Sloane’s Maze Wars was reworked and expanded by MacroMind as the first commercial version of Maze with the 1986 Mac release Maze Wars+, and that later led to Callisto Corporation making a more sophisticated and fleshed-out (not to mention 256-color) version called Super Maze Wars in 1993 — also for the Mac.

Maze popped up on several other platforms as well. Xanth Software put it on Atari ST as MIDI Maze in 1987. Their title refers to the use of MIDI cables to string computers together for multiplayer matches. “We just passed joystick data around the ring,” programmer Michael Park explains via email, “then each machine updated its world state and redrew the screen, then repeated. Everyone had to wait on the slowest machine (whichever took the longest to render its display).”

They also wrote an Atari 8-bit version that was finished but never released, and a German group led by coder Markus Fritz reverse engineered MIDI Maze so they could update it and fix bugs for MIDI Maze 2. Fritz says that that version only sold around 200 copies, but it nonetheless developed a cult following around the world with players still getting together today for semi-regular tournaments.

Maze also made it onto Game Boy, retitled Faceball 2000, and several other early ’90s games consoles and handhelds, along with Palm OS, NEXTSTEP and yet more platforms.

Its original creators seem somewhat bemused by Maze’s longevity and legacy. “It’s just kind of a small contribution that I and the people around me helped to make in pushing technology into a fun area,” says Thompson. “We’ve had games since probably the days of the Stone Age. It’s just been implemented in different ways. And the nice things about computers and the network is that you can be connected; you can play with other people. You don’t have to be physically together, and you can play with an experience of kind of looking like you’re inside a building that you’re not actually in.”

Lebling was surprised to see Maze spread, and he believes that the first-person shooter genre that dominates games now is at least partially indebted to Maze. But “that doesn’t mean that they wouldn’t exist if Maze hadn’t existed,” he notes. “It’s like the thing they used to say in science fiction, which is that steam engines come when it’s steam engine time. There was a confluence of technical development and ideas and culture that made something like Maze seem like a good idea.”

Lebling and his Infocom cohort could well have become a first-person shooter company, too — rather than the text adventure company. They founded Infocom before deciding what to develop, and one idea was to make Maze for the arcades. Nothing ever came of it because too many people who heard about the idea as a way to raise money for the fledgling company raised concerns that “nerds have no friends.”

“I think it would have been a cool game to play,” says Lebling, noting that arcades were in fact very social places. “You could have bragging rights and fights and drunken brawls in the arcade with ‘You shot me!’”

In any case, nobody expected Maze to live on — let alone to foreshadow a revolution in gaming that would occur 20 years later. “I think it was just a combination of enough bright people had come across it at MIT and then reimplemented at other platforms,” says Thompson. “Because it’s fun, because it’s enjoyable and at the time it was pushing the state of the art. It was a good test tool. You could justify it however you wanted to justify it, but you kind of built it because you could show off some of the neat things you could do.”