"Anything you can imagine is real." - Picasso

Interactive Fiction:

Stories in Software

A man has been murdered. Anyone in the household could have done it -- or perhaps even someone from the outside. There are clues, but only the sharpest detective will be able to spot them and build them into a case against the murderer.

[editor's note - The above introduction was published in 1984 but is accurate enough to perfectly describe Malinche's own murder mystery Greystone released twenty years later in 2004. ]

Will the detective succeed? That is up to you -- for you are the detective. You are, in fact, reading a story; but the story is unfolding on the screen of your computer, in response to your commands. You are playing one of the most engrossing and demanding kinds of computer games, the text adventure game also known as interactive fiction.

Computer text adventures fall into the very broad category of simulation programs [editor's note -- also commonly referred to as role playing games today] . Anyone who has played video games has dealt with graphic simulations of missile attacks, jungle chases, automobile races, and the like. Interactive fiction also simulates tense situations, but they interact with the user through words more than through pictures and game controller movements. People may disagree about whether the computer should use graphics or words to describe the simulated situation, and many adventures use both. But for telling the computer what you want to do, there is no question that words you type on your keyboard are far more flexible than any mouse or game controller.

You, the player, are the main character in the story that an interactive fiction title lays out. You are presented with a situation and a goal. Sometimes the goal is a single object, such as catching a murderer or escaping with your life; at other times it's a matter of accumulating as many points as possible by solving problems and picking up treasures along the way.

The program guides you along the by giving you a text description of your character's surroundings. It might tell you that you are in a large cave, or a broom closet, or -- heaven help you! -- a maze of twisting, identical passages. It will single out interesting objects that you might be able to pick up or otherwise use, and it will point out directions in which you can go.

In response to each description, you type in a verbal-style command indicating the action you want your character to take. Interactive Fiction today is sophisticated enough to understand full English sentences (TAKE THE SWORD AND KILL THE TROLL) or short abbreviations (GET SWORD. KILL TROLL) or super shortcuts (GO NORTH or just N for short.) You can also get fancy by using adjectives, multiple objects and modifying phrases such as TAKE ALL THE BOXES EXCEPT THE RED ONE.

Your character can carry objects. When you give a command to TAKE an object, the program checks whether you can in fact pick it up, then adds it to your inventory if you can. Having the right object at the right time (for instance, a sword if an unfriendly troll shows up) can be vital to your success in interactive fiction. There are limits, of course, on what you can carry; the story will let you carry only so many objects (or, in some cases, only so much weight), and there are some objects that are just too big to pick up.

An adventure consists of a series of puzzles to solve. The most common challenge is simply figuring out how to get somewhere. You may have to find a secret passage, open a lock, get past a guardian or answer a riddle. Then, after getting to your destination, you might find more surprises. You could encounter a locked box, an enigmatic machine, or a dragon sitting on a treasure. In each case, the solution lies in having the right tools and thinking of the right action.

Unlike video arcade games or real time action adventures, some interactive fiction titles invite you to sit back and think instead of calling on you to make quick reflect actions. Most text adventure games let you take as long as you want to decide on a command; game time doesn't pass while you're making up your mind. Every interactive fiction title lets you save your current position in the story, turn off your computer, and come back later to continue right where you left off. But interactive fiction isn't for the passive-minded. You have to be observant and inventive to get through one.

At its best, interactive fiction is a literary form that is unique in requiring the reader's participation. In fact, some books have been written in imitation of interactive fiction. These "pick your own ending" books ask the reader to select at intervals one of several possible courses of action and then to turn to a corresponding page. But they don't really demand active participation; the only choices are the ones listed. Text adventure games don't enumerate your alternatives in advance the way an adventure book does; the right thing to do may be one that doesn't occur to you for a long time. The thrill of finally thinking of that solution and finding that it works is one of the special pleasures of interactive fiction.

Some people seem to think that enjoyment means turning off their minds and watching something passively or engaging in meaningless activity. But for others, enjoyment means giving their minds a change of pace and scene while still staying alert and curious. Interactive fiction offers this kind of change of pace -- the chance to experience a different kind of world and try solve the problems that it offers, as well as the chance to take risks without the unpleasant consequences of real-life failure. It is escapism, perhaps in the sense of an escape from the routine of daily life. But it's not an escape from thinking.

In the Beginning

WIll Crowther created the first adventure game -- the preliminary version of adventure -- in 1975. It was written in FORTRAN on a Digital Equipment Corporation PDP-10 computer, and its popularity spread quickly as Crowther made it available nationally through ARPAnet (ARPAnet was the ancient predecessor to today's Internet.)

Don Woods of Stanford University made some major enhancements to Crowther's program, and the result was the program that is still regarded as the standard version of Adventure. Enhancements and adaptations have gone on, though, and today the game is available in numerous versions for every sort of computer.

Adventure has all the features that are now traditional parts of the interactive fiction genre. The player gives two-word commands to direct the character's actions (through an underground complex). The more treasure the character finds and brings to the surface, the more points he gets. And if he solves the entire labyrinth, he is finally carried away on the shoulders of the cheering elves.

Crowther's inspiration for Adventure, with its mixed bag of mythologies and its environment of mysterious underground passages, came largely from a noncomputer game, which was as revolutionary in its own field as adventure was in computer games: the role-playing game called Dungeons and Dragons.

Tactical Studies Rules (later known as TSR Hobbies Inc and out of business some years after that) published the first edition of Dungeons and Dragons in 1974. The rules in that edition were badly written and often grossly ambiguous, but players were thrilled by the concept. D&D is an open-ended game in which the players assume the roles of characters in a story and can have those characters attempt any action whatsoever. The game is controlled by a games-master (referred to as a dungeonmaster), who uses tables, dice and personal judgment to decide on the effect of a character's efforts.

[Editor's note -- Could that be why Infocom named the third installment of the Zork series "The Dungeonmaster"? We think so.]

Obviously, no text adventure game can be as completely open-ended as a game managed by a human being. But Adventure and the legendary Zork series and Scott Adams' text adventure games and others that followed did capture some of the spirit of D&D by not laying out a set of usable commands before starting the game. The player can try any command at all, and it just might work, or at least elicit an amusing response.

[Editor's note -- Years later, Infocom began including instruction booklets with their interactive fiction titles because they learned that while some people were entertained by trying all sorts of different commands, many more people were frustrated by not knowing what to do or try next. Malinche continues that newer tradition by making available extensive help facilities within our games as well as thorough documentation anyone can download.]

The second milestone in text adventure games was a program called Zork, designed by David Lebling, Marc Blank, Tim Anderson and Bruce Daniels. Zork also used a D&D-type milieu, but its authors had several advantages over Adventure's creators. First, the language they used was MDL (humorously called MUDDLE) a language similar to LISP that has much better facilities for expressing data relationships than FORTRAN does. Second, they were working just downstairs from MIT's Artificial Language Laboratory, so they had ready access to previous work in handling English-language input.

Zork's creators gave it a command-handling facility that went far beyond the original Adventure. They added prepositional phrases, indirect objects, adjectives, multiple direct objects, and compound commands. Prepositional phrases allow specifying an implement to be used in an action (ATTACK DEMON WITH TOOTHPICK). Adjectives distinguish between objects of the same category (OPEN LARGE BOX). Multiple direct objects are convenient when grabbing or discarding several things at once (TAKE DIAMOND AND AX AND HAT). Multiple objects are implied in the command DROP ALL, and further refinements come from combining adjectives, prepositions and multiple objects (TAKE ALL GEMS EXCEPT THE BROKEN ONE).

Zork also expanded the scope of the nonplayer characters controlled by the adventure game. The original Adventure has several dwarves and a pirate who move around randomly and take specific actions, such as throwing knives at the adventurer; it also has some characters in fixed locations who perform specialized functions, such as blocking a bridge. But in Zork, the pirate is upgraded to Thief, a character who performs a number of unpleasant functions besides taking treasure away from you. For instance, he may pick up objects and drop them off in another place to confuse you. There is also a robot, who can move at your command and who might be able to do jobs that a human could not manage. Other creatures (including a troll and a Cyclops) do not move, but their capabilities are more varied than their counterparts in Adventure.

Combat with these creatures is especially interesting, for the battle is not decided by a single blow. Instead, each fighter accumulates wounds, and several wounds are usually needed to kill. In each turn during combat, both the adventurer and his adversary can attack; the effectiveness of the attack is affected by the attacker's health and his choice of weapon, as well as by a random factor. One side will usually get the upper hand and then finish off his wounded opponent; but if the attacker is disarmed at a crucial moment, he can suddenly find himself in trouble.

Another feature found in Zork is the use of vehicles, which can take the adventurer to places unreachable on foot. You must BOARD the vehicle before using it, then issue an appropriate command, such as LAUNCH, to make it carry you. You don't necessarily have full control of the vehicle; this can be inconvenient if you are in a boat rushing toward a waterfall.

The original version of Zork was created in 1977; the name incidentally, is just a nonsense word of the sort popular in MIT's computer labs for spur-of-the-moment generic names. A couple of years later, its creators implemented the Z-machine. The idea of the Z-machine was to write Zork for one imaginary computer and then write programs for real computers that would let them run as if they were Z-Machines. (This is the same concept as the P-code system that lets compiled Pascal programs be moved from one computer to another.) The Z-machine software emulator was created for the TRS-80 Model I, followed shortly by an Apple version. The original Zork was reduced to a smaller program, with the intention (since fulfilled) of releasing the remaining parts known as Zork II and Zork III.

[ editor's note -- Malinche's modern interactive fiction titles are based on the exact same Z-machine technology. It is this model that allows Malinche to launch text adventure games that work on Windows, any Mac, all Windows Mobile devices, all Palm handhelds, etc. etc.]

Zork's creators formed Infocom to sell their adventure games. Originally the marketing was done through Personal Software; later, Infocom set up its own marketing and direct sales operations.

Since releasing Zork, Infocom has continued to enhance the command-handling capability of its new adventures. In addition, starting with Deadline, it has introduced real plots into its adventures. True, Zork has some rudimentary elements of plotting, including the adventurer's ongoing war with the Thief. But Deadline and subsequent Infocom adventures have tied all the events into a single, overriding purpose -- catching the murderer, stopping a plague, or whatever.

As microcomputers grew in popularity, Scott Adams broke ground in a different direction with his Adventure International series. Many hobby computers have [ HAD, actually, twenty years ago. - editor] no more than 16 kilobytes of storage and can input programs only from ordinary audio-cassette recorder. This puts programs like Zork and even Adventure out of their reach. But Adams showed that with clever programming techniques, enough material can be crammed into that amount of storage to make an exciting adventure game.

Adams's first adventures used an interpreter program, written in BASIC, that would read a data file and spin out the adventure in accordance with the data in that file and the player's input. This method provided much more effective use of data storage than straightforward programming of the adventure in BASIC would have provided.

Since then, Adams had rewritten his interpreter to run in the machine language of a variety of computers. This rewriting enhanced the speed and compactness of the adventures while greatly reducing the amount of time they took to load from cassette tape. For those interested, the original BASIC version of Adams's Pirate's Adventure was published in the December 1980 issue of Byte.

All 12 text adventure games in Adams's series used a standard screen format for presenting information. The top part shows the character's current location ("I am in..."), followed by a list of "visible items." The remainder of the screen, which may vary in size depending on the size of the item list, is a scrolling area for commands and responses.

The descriptions and responses are terse; Adams preferred having more rooms and a larger command vocabulary to giving length descriptions. [editor's note - thanks to 21st century technology, Malinche's Howard Sherman is able to write interactive fiction delivering the best of both worlds; very descriptive, rich prose coupled with incredible power and an expansive set of areas to explore.] The same phrase would often be incorporated into several different responses, at a savings in memory over having each response be different. For instance, the phrase "won't let me" appears in the reactions of several different guardians.

As memory becomes cheaper and home computers with disk drives [250 gigabyte hard drives these days! - ed] become more common, the restrictions that affected Adams's earliest adventures have become less of a problem. [no problem at all - see editor's note in preceding paragraph.] Adventure game publishers are looking for ways to take advantage of the additional capability of expanded memory. [Malinche does this constantly with its interactive fiction publishing. - ed] One path that several publishers have pursued is the Zork approach, which has more descriptions, a larger vocabulary, and more complex command handling. [This is now called The Malinche Method to keep up with the times.] The first adventures to take this approach wee Cyborg and Empire of the Over-Mind who have gone this way to varying degrees. Another probable route, though it has not always been put to the best possible use, is the addition of graphics to adventure games and interactive fiction.

Graphics, when used to complement a text adventure game and not just to illustrate it, can add a great deal to an adventure game. One of the best examples is Prisoner 2, in which the situation is presented largely through graphics, and commands directly affect the picture. Other adventure games provide animated sequences [and full motion video these days! - ed] to get from one location to another. Too often, though, graphics are used just to illustrate the text with still pictures. Since the best home-computer graphics today [translation: 20 years ago] fail to live up to the quality of a Saturday-morning cartoon, this is a rather ineffective use of computer storage. If the author of a text adventure game really wants to provide illustrations, he or she would be better off to include a booklet and say, "Look at page 23" (though this does have the disadvantage of letting players peek before they've gotten to their destination.) Avalon Hill's G.F.S. Sorceress makes good use of this method.

Complicated command handlers do take a lot of programming effort, and the publisher has to be able to justify the cost. On the other side of the ledger, too, computer-game publishers have been contributing to the slow but steady growth of more sophisticated adventure programs. This growth suggests that there are computer gamers who can handle the English language and who regard a game as a chance to think about something different rather than to stop thinking.

Origins

The English-language-handling capabilities of text adventure games, and especially of Zork and its descendants, owe a great deal to the work that was done in artificial intelligence (AI) in the 1960s and '70s. The researchers in AI put a great deal of work into handling natural language input, meaning input in English or other human languages, and some of the results are remarkable.

The best known of these AI programs is Eliza by Professor Joseph Weizenbaum at MIT. Eliza, named for the heroine of Shaw's Pygmalion, accepts input text in English and produces responses based on a script. The best-known Eliza script simulates a psychotherapist, and the program that uses this script is often known as Doctor.

According to Weizenbaum, who described the program in "Eliza - A Computer Program for the Study of Natural Language Communication Between Man and Machine" (Communications of the ACM, Volume 9, No.1 January 1966), "The gross procedure of the program is quite simple; the text is read and inspected for the presence of a keyword. If such a word is found, the sentence is transformed according to a rule associated with the keyword, if not a content-free remark or, under certain conditions, an earlier transformation is retrieved. The test so computed or retrieved is then printed out."

The program might turn I HATE YOU into YOU HATE ME, for instance, or cue on a word like BROTHER or MOTHER to evoke a standard question about the user's family. It is easy enough for the user to provide input that drives the program to unintelligible responses, but some sessions have let the program seem surprisingly human. A piece of sample dialogue, presented by Weizenbaum, follows, with the program's responses in uppercase:

I need some help, that much seems certain.

WHAT WOULD IT MEAN TO YOU IF YOU GOT SOME HELP

Perhaps I could learn to get along with my mother.

TELL ME SOME MORE ABOUT YOUR FAMILY.

My mother takes care of me.

WHO ELSE IN YOUR FAMILY TAKES CARE OF YOU

My father.

YOUR FATHER

You are like my father in some ways

WHAT RESEMBLANCE DO YOU SEE

Eliza shows that a computer program could scan English-language input and produce interesting results; however, its approach is entirely different from the method needed for any sort of command input. Rather than trying to exhaustively analyze the syntax of each input line, Eliza simply looks for key words and phrases and makes the best of those. It will never admit to not understanding a sentence. Weizenbaum has been somewhat embarrassed by the "aura of magic" that the program has acquired and stresses that it is "merely a translating processor."

The program's strategy of concealing its failures is appropriate to generating that aura.

"But to encourage its conversational partner to offer inputs from which it can select remedial information." Weizenbaum notes, "it must reveal its misunderstanding." A program that does this successfully -- that, for instance, can tell the user it doesn't understand CLOBBER THE DIRTY CROOK and indicate the exact point of failure -- has to take quite a different approach from Eliza.

In addition, Eliza retains only a little information from previous input. It does not change its model of the situation on response to every input line. Such information as it does retain has very little structure to it; it simply allows the program to make occasional references back to earlier parts of the dialogue. For instance, some time after the user mentions an illness he once had, the program might attempt to liven the dialogue with the line, "AWHILE AGO YOU MENTIONED YOUR ILLNESS. WHAT ELSE DOES THAT SUGGEST TO YOU?" Even the most elementary text adventure game has to do more than Eliza in this regard.

A more direct ancestor to text adventure games, particularly to Zork, was T. Winograd's SHRDLU. This program, created as part of his 1971 Ph.D. thesis, accepts complex commands for moving blocks of various shapes around. The program kept track of the blocks only in an internal representation; but with today's robotic technology, if not that of 1971, it would be feasible to have a mechanical arm that actually obeyed the commands.

You might, for instance, tell SHRDLU to pick up the big red block. This causes it to perform a whole sequence of actions to carry out the command. First it must determine which object is the big red block; then it must check if any objects are sitting on the block and move them to the "table" that supports all objects. After that, it can designate the indicated block as being picked up.

Patrick H. WInston, in Artificial Intelligence (Addison-Wsley, 1977) cites four categories that are significant in SHRDLU's "blocks world". These are objects (such as specific blocks and pyramids); relations between objects (such as IN-FRONT-OF and SUPPORTED-BY); relations between objects and intrinsic properties (such as color and and size); and actions (such as PICK-UP, PUT-ON and STACK).

These capabilities are necessary to any simulation that involves manipulating physical objects in a realistic way. In a text adventure game, you need to know whether the dragon (object) is on the pedestal (relation between objects), which pedestal are you talking about (its intrinsic properties), and whether it is possible to open the pedestal (action) without doing something else first.

SHRDLU, in its restricted context, has capabilities that go beyond any text adventure game yet created (at the time this was written some 20+ years ago - editor). Its syntax handling allows very complex commands. Relations between intrinsic properties, such as "taller than," are within its capacity, as are alternatives such as "either a green cube or a pyramid."

On the other hand, SHRDLU deals with a very small class of objects, which occupy only one "room" and which always remain passive except when manipulated by a command. A block will not explode two turns after you put it in the box, nor will a second robot walk in and steal it. There is no question of what implement to use for an action, since the robot has only one hand. SHRDLY is an expert in a very narrow field; interactive fiction titles must be generalists by comparison (not anymore! - editor)

Computer Fantasies

When talking about Eliza, SHRDLU, or text adventure games, it is important to realize that these programs do not think. Each of these programs is set up to deal with its input and generate output according to certain rules that cannot be broken.

Some artificial intelligence programs, such as Eliza, have made a great impression on naive users, who have thought that the computer really "understood" what they were telling it. Weizenbaum has been honest enough to discourage this impression; unfortunately, creators of some other computer programs seem to seriously believe that what the programs are doing is "thinking" or "understanding."

Working with these programs is a good way to dispel any illusions about what they are doing. Enter commands which go beyond their designated capabilities, and the programs either return a stock confession of ignorance or issue nonsensical responses.

Deliberately playing a text adventure game in this way -- not trying to solve it, but to probe the limits of its present capabilities -- can be fun as well as help to break down the illusion that the computer "knows" something. With a two-word command handler and a tiny vocabulary, overcoming the program is so easy that it doesn't provide any satisfaction. But if the program is powerful enough that you can't be sure whether a command really is beyond the program, then testing it can become an enjoyable game in itself.

Player: Drop dead

Text Adventure Game: I DON'T SEE ANY DEAD HERE.

Granted, even the best text adventure games don't match the complicated conversational programs coming out of artificial labs. Still, the pattern is there. The program handles only a certain amount of subject matter and can handle it only according to certain rules. When the input touches on a borderline or unusual case of the subject matter, the success of the program depends on how complete the rules are in that area. If the subject matter is well delimited -- a chess game or the manipulation of blocks, for instance -- then the programmer doesn't have to worry very much about the borderline cases. But when it is wide open, as it is when the program is trying to simulate the actions of a human adventurer, then there will always be cases that the text adventure game can't handle. Discovering this truth helps to take much of the mystique away from complex interactive fiction titles.

A text adventure game can be played on a number of levels. You can treat it simply as a game of logic and try to solve the puzzles. You can look at it as an unfolding story in which your actions shape your character's fate. Or you can offer it to your own challenge and see how far you can push it before it reaches its limits.

Text adventure games also promote a special sort of interaction among people, even though they are played solitaire. In playing a text adventure game, you may find yourself stuck on some point; if you've tried your best, there's nothing wrong with asking a friend who's already played the game to offer you a hint. Experienced players delight in coming up with subtle, devious hints in response to inquiries. A good hint still lets the player the puzzle mostly on his or her own; it just gives another perspective on the situation. Pac-Man, on the other hand, doesn't lend itself to much conversation beyond asking what fruit is in the tenth maze.

Can a text adventure game be played only once? This is like asking if a novel can only be read once. Usually you would rather go on to a new book than reread an existing one; but if a particular book contains more subtleties than you could enjoy in a single reading, you may well want to read it over and over again. The same is true, in a somewhat different way, of a good text adventure game. Even after you've reached the highest score, you can go back and try courses you hadn't taken before just to see what might happen. If you got through a situation by lick, you might want to look again in order to see whether the author planted some subtle clue you could have used.

No other class of computer games offers such a multi-faceted source of excitement. Chess sharpens your analytic powers, war games test your sense of strategy, word games stretch your ability to recognize words, and arcade shoot-'em-ups challenge your coordination and reactions. But none of these offers the range of satisfactions to be found in a top-quality text adventure game.

Welcome, then, to the dungeon, to the scene of the crime and to other worlds.

Next Chapter: What Makes a Good Text Adventure?