There were other precedents for Dungeons and Dragons besides wargaming. Although fantasy sports leagues did not evolve into today's popular form until around 1980, an earlier type of game, called dice baseball, provided its own platform for fantasy and determined the outcome of events much as Dungeons and Dragons later would. In his 1968 novel, The Universal Baseball Association, Inc.: J. Henry Waugh, Prop., Robert Coover memorably described how this sort of game, in which a roll of the dice would determine chance events, could even allow a player to imagine an entirely different world.
The original Dungeons and Dragons rules were intentionally left incomplete so that different groups of people could adopt different styles of play. In a rare moment of lucidity, one Christian critic of the game noted that "Dungeon Masters play the game differently. Some dislike situations in which characters get killed. Others feel a game is successful only when half the players die in battle" (Robie 1991, 47).The extent to which chance, theater, or puzzle solving is involved depends on the particular group playing Dungeons and Dragons. Yet the general framework of the game (not just its particular setting and fantasy theme) was clearly important to what would follow.Aarseth (1997) writes that the "Dungeons and Dragons genre might be regarded as an oral cybertext, the oral predecessor to computerized, written, adventure games" (98).
EARLY COMPUTER GAMES
In 1912 Leonardo Torres Quevedo (1852-1936), a Spanish engineer, devised the first computer game. He constructed the first true chess-playing automaton-one that operated without a human concealed inside.This caused a stir when first demonstrated publicly in Paris in 1914 (Perera Dominguez 1997). The automaton was an electromechanical device with a vertical chessboard; the pieces were pegs. The machine played a KRK chess endgame, playing rook and king against a person playing a lone king.The machine did not just signal its move: It physically moved its pieces with a mechanical arm. An illegal move was indicated by the illumination of a light bulb. Although the automaton would sometimes exceed fifty moves (thus giving the player a draw), it would always be able to eventually mate the human opponent's king on the first rank. This first machine had a fairly easy-to-use interface, but that was improved upon in 1920 when Torres built a second automaton with an ordinary chessboard. This one moved its pieces using electromagnets, and, hooked to a gramophone, it employed speech synthesis: It verbally announced check ("jaque al rey") and checkmate ("mate"). This second automaton is now housed in Madrid at the Colegio de Ingenieros de Caminos, Canales y Puertos and still functions perfectly (Scientific American Supplement 1915; Lopez 1998; Randell 1982; Atkinson 1993, 22).
There were certainly limitations to Torres's machine. It was not developed to work generally, even for the reduced problem of the KRK endgame. According to Atkinson (1993),
Torres swiftly discovered that progranuning is not an easy task... . he decided to force the lone King to the first rank for the mate ... he simplified the problem further by assuming that the two Kings were already on opposite sides of the rank controlled by the Rook. His final algorithm assumed a fixed starting position for the automaton's King and Rook, but allowed the human opponent's King to be placed on any unchecked square in the first six ranks. (21)
With these restrictions on the setup (defining a subset of possible KRK chess endgames, and allowing for 61 rather than 50 moves without the capture of a piece) Torres's algorithm guarantees a win. It does so without, requiring that the machine's king be moved diagonally-allowing for the device to be mechanically simpler (Atkinson 1993, 21).
Perhaps Torres's invention of the first computer game is overlooked because it seems a minor achievement of this important figure, who, in addition to devising a new type of semi-rigid airship and a remote-control boat, did pioneering work in the development of the modern computer by making advances in cybernetics and by developing calculating machines (Randell 1982). It may instead be that this device, coming so far before general purpose digital computing was invented, is not considered a computer game. More likely, it has been overlooked by the English-speaking world because Torres himself has been overlooked in English-language scholarship on science and invention. Despite the neglect this father of the computer game has been shown, Torres's invention has had some influence: Many researchers working in chess automation, for instance, have been aware of the two automata Torres created and how they worked, and a photograph shows Torres's son demonstrating the second machine to pioneering cyberneticist Norbert Wiener in 1951 (Eames and Eames 1973). Contemporaries recognized these chess-playing automata as early steps toward artificial intelligence: "the automatons of Torres ... attempt the accomplishment of things that have hitherto been reserved entirely for the human mind" (Scientific American Supplement 1915, 298).
After this spectacular beginning, there was little work done in computer game development until after World War II. With the arrival of the generalpurpose computer came many opportunities for game making. During the early 1950s, before the advent of batch processing, Eric Solomon (1984) explains that though "machines of this period were slow, their immediacy made it possible for the adventurous to program the odd game," which might take the form of "reaction testing" (2-3). There were official and more elaborate efforts at computer games, too. In 1951 a computer game was displayed by Ferranti at the Festival of Britain. It was a special-purpose system that played Nim, in which the players alternately take one or more matches from different piles, and the one forced to take the last match loses (Bennett 1990, 283-284).The first academic paper on computer chess (Shannon 1950) led to the first new developments in that field since Torres. A remarkable checkers program was devised by Arthur Samuel at IBM that learned from the games it played and, by 1962, was able not only to defeat its creator (Spencer 1968, 14) but also to best a human champion (Samuel 1963, 103-104).The checkers programs even learned more rapidly than human players did, making greater progress over the same number of games (Williams 1972, 72). Beginning around 1950 computer war games were also developed in the United States, serving, as one early writer on computer games explained, "to simulate activity ranging from ... tactical action to a large fall-scale war." One was the 1955 HUTSPIEL, which ran on an analog computer and was for two players, Red and Blue, representing the USSR and NATO; this may have been the first two-player computer game. Other programs included the tactical simulator CARMONETTE and the theater-level war game THEATERSPIEL, both running on Univac computers (Spencer 1968, 12).
The batch-processing era was not as productive. Solomon (1984) writes: "The programmer found his access to the machine barred by a high counter, air-conditioning doors, and personnel in white coats....These were the `dark ages' for computer gaming" (3).When timesharing was introduced in the middle of the 1960s, working on games became easier. Still, those who were billed for computer time could not afford the activity.A few institutions where resources for game making were available would lead the way during this time. Star Trek was developed around 1967, offering a universe grid to explore, but the canonical version was written later, in 1972, by Mike Mayfield. That was also the year Gregory Yob wrote Hunt the Wumpus (Nelson 2001b, 344), notable for having a world that was not a simple grid but a dodecahedron. This feature of Hunt the Wumpus inspired the irregular worlds of interactive fiction; different rooms, from the very beginning, could be interconnected by the designer in any way.
One center of early computer game development-and one that would become important to the history of interactive fiction-was MIT. The playful approach to computing pioneered by the self-proclaimed "hackers" of the Tech Model Railroad Club (or TMRC, founded in 1947) gave rise to many early games and other computer recreations (Levy 1984, 17-38). One early development on the TX-0, a prototype computer to which the group was given access, was Mouse Maze, which animated a mouse's traversal of a maze; the mouse ate cheese (and, in a later version, drank martinis that caused it to perform more poorly) along the way (Oberg 2001; Levy 1984). Games were also used as testbeds for artificial intelligence research and in other ty
pes of work. As one of the creators of Zork said, "MIT was a place where you could tell your boss `I'm going to spend some time writing a game,' and that was okay" (Anderson 2001).
A notable computer game developed at MIT by Steve Russell and others was Spacewar! (Levy 1984,59-69; Graetz 1981). In the 1962 Spacewar!, the ancestor of all modern video games, two players used special controls to manipulate their spaceships in a battle that was depicted on a CRT display. Although video games later became an important industry, spawning arcades and home game consoles, the influence of these graphics-based, actionoriented computer games on interactive fiction actually turned out to be rather slight, in the long run.Video games showed themselves to be a different branch of recreational computing.
Other sorts of games were developed and played at MIT, however. The Artificial Intelligence Lab and the Dynamic Modeling Group were both active gaming spots around 1970. The Dynamic Modeling Group (DM) got its start, one DM member explained, "writing libraries of reusable software that would dynamically link together as needed when the client program ran" (Lebling 2002). "DM was in a very good spot for computer gaming," Tim Anderson (2001) said, with Imlac terminals capable of graphics instead of the standard text-onlyVT-52s used elsewhere on campus. These terminals had green-on-black vector graphics, not the raster displays that are ubiquitous today. The Imlacs had their own processors and 8k of memory. There were stand-alone versions of Hunt the Wumpus and of a game like Pong that ran on them without using any of the resources of the PDP-10 to which they were connected; there was even an Imlac implementation of Spacewar! (Lebling 2002). Anderson (2001) said, "At DM, because the terminals were so expensive, we didn't have terminals in the offices. There was a terminal room with four or five Imlacs and some cubicle-like areas that were partitioned off."
According to Anderson (2001), "The most interesting thing was a game called Maze-which was later seen as Mazewars on the Altos." Maze was a video game-a multiplayer video game, and in fact it was the first first-person shooter, ancestor of Doom, Quake, Unreal, and their brethren. Dave Lebling (2002), who programmed the Maze server in PDP-10 assembly and who had also coded the Imlac versions of Wumpus and Spacewar, called Maze "a computer version of kids running around shooting each other with toy guns." Players' avatars (which simply appeared as their login names) were situated in a maze; the screen displayed a three-dimensional view of this maze from the avatar's perspective. Over opponents' "heads" were arrows showing which way they were facing-"or a pair of dots indicating eyes if they were facing directly at you," Lebling (2002) said. Greg Thompson, an undergraduate, programmed the Imlac client in assembly. Up to eight people could play, and the computer could take the place of some players if there were fewer than eight who wanted to play: Lebling also implemented the behaviors of the robot opponents. Anderson (2001) described Maze as "all about speed and reflexes," noting that "DM was the only place where you could play it."
A less visually spectacular MIT game-but one that was available on the ARPANet and became quite popular-was a text-only quiz game, Trivia. Tak To (who had also coded some additions to Maze) wrote the first version of Trivia, which Anderson said was an "utter kludge."Around 1975 or 1976 Marc Blank built a type of database system, a message store, as part of his work on a DM project. As a test of this, he created a new version of Trivia. Those who played Trivia on MIT's ARPANet host 80 would be the first ones outside MIT to discover a new game on that computer in 1977. In that year, four programmers and writers from the DM-Tim Anderson, Trivia programmer Marc Blank, Bruce Daniels, and Maze programmer David Lebling-would install a work of interactive fiction called Zork. That work is the subject of most of chapter 4.
The idea of computer conversation was developed along with the generalpurpose digital computer; Turing (1950) presented it quite clearly in his influential paper "Computing Machinery and Intelligence," in which he established the well-known, if sometimes misapplied," Turing Test." He gave the example of a parlor game in which a person of one gender tries to imitate a person of the other, and then suggested a similar game in which a computer would imitate a human in conversation. The question "Can machines think?" could then be replaced by one that was easier to answer: "Are there imaginable digital computers which would do well in the imitation game?" (Turing 1950, 54-55). In formulating this test, Turing highlighted the importance of conversational ability to our perception of intelligence, at a time when computers almost exclusively processed numbers rather than language.
There was not a lot of good human/computer conversation early in the 1950s, but by the start of the next decade several interesting systems were being developed. One of many important early conversational programs was Baseball, which could answer questions like "Who did the Red Sox lose to on July 5?" and "How many games did theYankees play in July?" by searching through a store of this information, represented in English (Green et al. 1963, 211). A more sophisticated system was SIR (Semantic Information Retrieval), which was "capable of `understanding' statements dealing with set relations, part-whole relations, ownership, and certain special relations" and which maintained a dynamic model of its knowledge in a special format, using word associations (Raphael 1968, 33-44). While SIR worked well, Terry Winograd (1972) noted that "the types of complex information it could use were highly limited, and could not be easily expanded" (37).
The next section discusses how the further development of knowledge representations, and attempts to understand sentences based on syntax, led to dramatic general advances in the understanding of language by computers. The most interesting development of the mid-1960s was no doubt that a researcher taking an entirely different approach-using a very unsophisticated representation of knowledge and a technique that could not be applied very generally-created a conversational system that has nevertheless been recognized as "astonishingly persuasive" U. Murray 1997, 69).
That system was ELIZA. It was programmed in MAD-SLIP (a language similar to LISP) during 1964-1966 by Joseph Weizenbaum at MIT's Project MAC. ELIZA is a language analysis program that can run many different scripts and have different sorts of conversations; the most famous script by far was called DOCTOR, which Weizenbaum (1976) explained was "designed to permit it to play (I should really say parody) the role of a Rogerian psychotherapist engaged in an initial interview with a patient" (2-3). The system, running this script, became known as the first computer conversationalist, the first chatterbot. To simplify ELIZA's procedure-but not to simplify very much-the program would first search for keywords in input. If some were found, the program would transform the input according to a rule and print out the transformed sentence. If none were found, it would provide a default (usually noncommittal) output. Here is a bit of a conversation with ELIZA/DOCTOR, with the computer's reply in all capital letters:
Although ELIZA was later used to do more sophisticated types of natural language understanding (Weizenbaum 1967; Winograd 1972, 38), Weizenbaum (1976) made it clear that it in no way offered a general solution to the problem of natural language understanding. Nevertheless, users were quite affected by it. Weizenbaum wrote that he "was startled to see how quickly and how very deeply people conversing with DOCTOR became emotionally involved with the computer" (6). Janet Murray (1997) described him as being "so disconcerted by his achievement that he wrote a book warning of the dangers of attributing human thought to machines" (71).
While other sorts of literary machines had been devised previously, ELIZA/DOCTOR may have been the first piece of interactive software to exhibit literary qualities. Certainly, there is no similar early program that is so widely known. One historian of the chatterbot writes that "anecdotes about Eliza instantly became part of computer science's folklore" (Leonard 1998, 42). Some psychiatrists thought ELIZA/DOCTOR should be used in therapy; people took contradictory positions on big questions in artificial intelligence based on ELIZA/DOCTOR; and the program inspired its own branch of computer literature based on conversation with fictional characters U. Murray 1997, 214-247; Leonard 199
8). ELIZA/DOCTOR has been widely distributed and implemented on different systems. While it was not immediately embraced as literary, by referring to it as "parody"Weizenbaum himself suggested it may have an element of literary art in it.As early as 1984 there was a suggestion that the system be considered by literary theorists (Neisz and Holland 1984, 118-119). It has no doubt been an inspiration to interactive fiction creators throughout the history of the form, and Janet Murray (1997) has spoken of the system's influence even more strongly: "Weizenbaum stands as the earliest, and still perhaps the premier, literary artist in the computer medium" (72).
It is well established that the first work of interactive fiction is Adventure. However, when considering the formal definition of interactive fiction that has already been advanced, one realizes that another computer program should actually be assigned this honor: Terry Winograd's SHRDLU, which he programmed in LISP at the MIT Artificial Intelligence Laboratory during 1968-1970 (Winograd 1972, 5). Although SHRDLU has been mentioned in online discussions and given as an example of a predecessor to interactive fiction in a book (McGath 1984, 11-14) and in Nelson's "Short History of Interactive Fiction" (2001, 334), no one has yet assigned it such importance.
This program (named after the second row of keys on a Linotype machine; see Winograd 1999)) carried on a dialog with the user via teletype, also displaying an image of a tabletop with blocks on it. It could respond to natural language commands in real time, in five to twenty seconds (Winograd 1972, 7-8). It would move the simulated blocks around if asked to; it could also answer questions about them. According to Winograd,
Twisty Little Passages Page 10