Parser Games

16 Jan

I’ll be diving into Zork in some detail in my next post, but before I do that I just felt behooved to return in a bit more rigorous way to a subject I broached in my first post in this series: how impressive Zork was in the microcomputer world of 1980-81. I have a point I’m driving toward, one which involves a little bit of theory (uh oh!). But first let me set the stage with a few choice quotes from Jason Scott’s Get Lamp project.

“There were two products that sold more computers than anything else: VisiCalc and Zork.” — Mike Berlyn

“We would go after school to this store and play whatever games were available, type games in, and I remember Zork coming out and playing it on an Apple II, and we were just completely blown away.” — Andrew Kaluzniacki

“People would see Zork and say, ‘I gotta have me one of them, that’s all. Who do I make the check out to?'” — Mike Berlyn

“I think there was a time period, probably ’80 to ’84 sort of range, where, for a lot of the machines, compared to anything else out there, there was just nothing that compared.” — Mike Dornbrook

Yes, Berlyn’s placing Zork on a pedestal with the industry-defining VisiCalc is a bit over the top, but you get the picture. Statements like these read as ironic and maybe even a bit tragic today. Within a few years after Dornbrook’s 1980 to 1984 timeline, interactive-fiction publishers and fans would be lamenting IF’s lack of immediate, obvious appeal as the main reason for the genre’s declining commercial fortunes, amidst plenty of griping about the adolescent illiteracy of the typical videogame demographic and the like.

So, what did those early players find so immediately appealing about Zork? Certainly its world was not only bigger but modeled in a more rigorous, sophisticated way than anything that had come before. Certainly its writing, while often necessarily terse due to space constraints, showed a wit and nuance and, well, attention to basic grammar and spelling that eluded its competition. And certainly its design was, if still beset by infuriating mazes and some more-than-dodgy puzzles, also fairer than the norm. But these are things that text-adventure afficionados notice, the sort of things that only become clear after spending a few hours with Zork and (at least) a few hours with other games of its period. As the quotes above illustrate, people were playing Zork for a few minutes in shops and buying it in awe — and perhaps, Berlyn’s hyperbole aside, sometimes also buying the Apple II system they needed to play it. Why? I think the answer is bound up with the adventure game’s love-hate relationship with the parser.

In Joysprick, a book about James Joyce, Anthony Burgess divides authors into two fundamental categories. (Feel free to insert your own “two types of…” joke here; I’ll wait. Ready? Okay…) Class One authors are concerned exclusively with the storyworld — the virtual reality, if you will — that lives “beneath” their words. “Content being more important than style, the referents ache to be free of their words and to be presented directly as sense data.” “Good” writing, under this rubric, is writing that exists solely to serve the setting and the story it reveals, that evokes them as vividly as possible but that also gets out of the way of the reader’s imaginative recreation of the underlying virtual reality by diligently refusing to call attention to itself. Class Two authors, meanwhile, are concerned about their language as a end unto itself. Their books are “made out of words as much as character.” Sometimes, as in the case of Finnegans Wake or the “Siren” chapter of Ulysses, language seems like all there is — the writing is all “surface.” Some might say that being successful on this second level, or at least striving to be, separates “literature” from mere “fiction.” But let’s stay away from that can of worms. In fact, let’s try not to make any value judgments at all as we apply some of this to interactive fiction.

I don’t want to apply these ideas so much right now to the text that an IF game outputs to the player, but rather to the text that the player inputs — to the parser, in other words. One way to approach IF is as a rich virtual reality to be inhabited. In this view, that of the Class One player, the parser exists only as a conduit for her to inject her choices into that world, just as a Class One reader views the text as a window — hopefully as transparent as possible — through which she views the action in the storyworld. This has always been my basic approach to IF as a player and a writer. Since I seem to be indulging in a lot of direct quoting in this post anyway, let me get a bit pretentious and quote an earlier version of myself. I wrote the following as a comment on Mike Rubin’s blog back in 2008:

I think many people, myself included, did indeed play Facade as a comedy, trying ever more outrageous actions to see what happens, and, indeed, at some level trying to “break” the system. I would say, though, that when a player begins to do this it’s a sign that the game designer has failed at some level. I began to play Facade for laughs after trying several reasonable approaches and having the game respond either not at all or in a way that was clearly inappropriate to my actions. The mimesis broke down for me then and I began to treat the system as a clever toy rather than an immersive interactive narrative. There’s no shame in Facade’s failure, of course. It’s a revolutionary conception, and bound to need many more iterations before even approaching complete believability.

This does raise a point, though: I don’t think games can maintain their mimesis by scolding the player, telling her in no uncertain terms that she shalt NOT when she attempts to eat her sword or hit her friends. Rather, we should strive to make our writing so good and our environments so believable and our interactions so smooth that our player is drawn into our story, and it never occurs to her to eat her sword or hit her friends, any more than it would to her avatar. In other words, we must enable her to truly BECOME her avatar for the little while she plays.

As soon as the game starts to break down, so to speak, for the player… that’s when she remembers it’s just a silly text adventure, and that’s when she starts playing it for laughs and trying to break the system even further. I do it every year with at least a dozen of the Comp games, PURLOINING doors and buildings and generally running amok through the storyworld. Entertainment is where you find it, after all.

Some players will of course come to every game determined to break it. Some might find IF in general more interesting as a system to be played with than as a story, although I think other genres of gaming would scratch this particular itch much better. To those players, I say, fine, have your fun. However, I think most people who play IF do come to it wanting to be immersed and to experience a storyworld and, yes, a coherent story through someone else’s eyes for a while. The rewards of that must be far greater than those of trying random actions to see where the boundaries of the simulation are (entertaining as that can be).

(Did we say something about not making value judgements? I forget…)

Still, those folks who marveled at Zork in computer stores were not responding to it as a deep and immersive piece of fiction, nor even as a really sophisticated adventure game. Their awe was all bestowed at the level of the parser itself, as an object — a toy — unto itself. For all of the space restrictions they were laboring under, Infocom reserved room for witty rejoinders to the sort of crazy or nonsensical inputs people might walk up and enter in a computer store.


This is playing Zork as Eliza: seeing what response this or that input gives, and of course probing for the limits. As some recent experiments have demonstrated, this mode of interaction is still pretty much the default when the uninitiated are confronted with a work of IF for the first time. Back in 1981, when computers were not so well understood and for most people still seemed vaguely magical (if not sinister), the idea of typing something, especially something off the subject or just plain inappropriate, and being understood was a much more powerful one, bringing to mind HAL and the Enterprise‘s talking computer.

All of which is apropos of… what? I’m not sure there are any grand lessons to take away here. After a pretty short while, toying with the parser and trying to break things loses its appeal, and the player either starts to engage with the storyworld and its fiction or just goes on to something else; thus the impatience I express above with players who just can’t seem to get past their triumph that, yes, they can break the parser and probably even the world simulation without too much effort. Over a decade after Zork, a graphical adventure called Myst became for some years the bestselling computer game of all time. It was often labeled the least-played bestseller ever. People bought it to show off their new graphics cards, sound cards, and CD-ROM drives in the midst of the “multimedia PC” boom of the early 1990s, but I’d be shocked if even ten percent seriously engaged with its intellectually intricate puzzles or made a real effort to finish it. Similarly, I suspect that plenty of copies of Zork existed more as something to pull out at cocktail parties than an abiding passion.

But let’s not start printing our “I appreciate Zork on a much deeper level than you” tee-shirts quite yet, because it’s also true that none of us ever wholly become Class One players. One more Get Lamp quote, this time from Bob Bates, captures some of the back and forth that forms a big part of the delights of the text adventure:

“A lot of games only program the ‘if,’ which is that main path I was talking about earlier. If the player does this and everything’s right, then you do this and the game goes on. But there’s always that ‘else.’ What if the player doesn’t do what you expected? What if he comes up with this weird idea or that strange input or that other, off-the-wall thing that he wants to try, just to see if the game breaks. Just to see where the edges are. That’s part of the fun of playing a text adventure, and that’s part of the fun — a great deal of the fun — that I had in creating them, in that imagined dialog with the player, so that at the end of the day when the player does this very weird thing, and he says, ‘Oh, nobody would ever think of trying this,’ he says, ‘Oh, my goodness! There’s a non-default response there! The author actually thought about that!’ That helps form that bond between you and the author. ‘That guy’s just as strange as I am. He and I think the same way.'”

So, a motto for text-adventure success: attract and charm them with your parser, retain them with your storyworld. In the spirit of the latter, we’ll put our Class One players’ hats on and venture into the Great Underground Empire next time. No, really, this time I promise.


Tags: , ,

16 Responses to Parser Games

  1. Vyacheslav Dobranov

    January 16, 2012 at 3:22 pm

    For last several months (after reading “Let’s Tell a Story Together”) I wait your every post with growning impatience. Thanks for this pieces of history.

  2. matt w

    January 16, 2012 at 7:46 pm

    It seems like, when you first get into text adventures, the parser messages are a source of delight — there’s a smartass response to “dig” even if digging doesn’t do anything on that particular occasion. But the more you play, the less charming they get — if only because it becomes apparent that it’s just a default response, and the game hasn’t really translated your input into an action. This is exacerbated because, thanks to Inform 7, a lot of the smartass responses are now the same from game to game. (When I first betatested a game I wrote something aggrieved to the author about getting “What a wonderful idea!” in response to “think.”)

    Maybe instead of “neutral library messages” we should have “funny the first time library messages”?

    • Jimmy Maher

      January 17, 2012 at 8:50 am

      Yes. I wouldn’t be the first person to point out that Inform’s default messages, with their dryly droll English humor, sometimes clash horribly with the rest of the game they live within. Infocom never developed a standard library, but rather developed each game from the bones of the previous game to which it was “most similar.” What with the default responses in that skeleton already presumably somewhat appropriate for the genre and with them not being secreted away in a separate library (and thus more accessible and noticeable to authors), the Infocom games tended to do better in this area. And then Infocom, for all their experimentation and genre-hopping, did have a sort of house editorial style that made their games differ much less wildly stylistically than those of, say, the typical annual IF Competition.

      • matt w

        January 18, 2012 at 3:38 am

        But I think the problem exists even when the default messages are lovingly handcrafted. If you’ve played IF for a little while, you’ll experience a thrill of delight the first time you see (say) Broken Legs’ “There’s nothing like that here. You’d think Bridger would keep things around.” So in character! The second time you see it you’ll process it just as if it were “You can’t see any such thing.”

        My idea is that the bloom comes off the rose of even hand-crafted default messages very quickly; at the beginning it’s exciting that they thought enough about it to program a response beyond “I don’t understand that sentence,” but pretty soon you start wishing that they’d programmed an action instead of simply a response. which is to say, “attract and charm them with your parser, retain them with your storyworld.”

  3. Jason Scott

    January 16, 2012 at 9:45 pm

    What part of Berlyn’s statement is over the top? People certainly bought machines just to play Zork and they certainly bought machines to be able to use Visicalc.

    • Jimmy Maher

      January 17, 2012 at 8:00 am

      Oh… you didn’t read that quite the way I intended it, which is probably my fault. I’m going to make a minor edit in the post itself to make it more clear.

      I have no doubt that people bought machines just to play Zork. It’s just that many, many, many more bought them just to run VisiCalc. The problem with the statement is that it rather overstates the importance — great as it was, as I’ll be the first to argue — of Zork.

      VisiCalc was simply HUGE. It sold over 700,000 copies in less than six years. As your sales chart shows, that’s about twice the numbers of the bestselling single Zork game. Indeed, it’s an astonishing number considering the size of the PC industry of the time and the fact that VisiCalc retailed for at least $99 (and usually more). It was all over the press, and — literally, even though “revolutionized” is an overused word — revolutionized the way companies did business. In a very real way, it gave these little microcomputers for the first time a reason to exist in the eyes of those who were not entranced by them as ends unto themselves. It’s the urtext of the commercial business/productivity software market in the same way that Adventure is of digital ludic narrative.

      Zork, important as it was to ludic history, simply doesn’t compare. If you must compare a game to VisiCalc a better choice would actually be Wizardry, a game I’ll be getting to shortly, which outsold Zork and caused even more of a stir upon its initial release. But much better would be to simply say games in general, which were the great hidden driver of PC sales for years. (Many for some reason didn’t want to admit that they spent all this money and time playing games; the U.S.’s collective Puritan guilt complex in action, perhaps.)

      In short: Zork was big, but placing it on a pedestal with VisiCalc as a driver of Apple II sales is… over the top. Videogame history has generally been written from a fan’s perspective, which leads to a lot of hyperbolic statements just like this passing unscrutinized. So, when they come up, I try to administer a little corrective. It’s not meant to take anything away from the real achievements of Infocom or Berlyn himself.

      • DZ-Jay

        May 14, 2023 at 11:04 am

        It is funny to see the following two statements so close to each other:

        >> If you must compare a game to VisiCalc a better choice would actually be Wizardry

        >> Videogame history has generally been written from a fan’s perspective, which leads to a lot of hyperbolic statements just like this passing unscrutinized.

        The juxtaposition is made even more delicious after one has read your blog for much longer, and gotten to some of your accounts of genres and specific games for which you really show your personal bias.


  4. Jason Scott

    January 17, 2012 at 2:35 am

    From Visicalc and the Rise of the Apple II:

    “The program went on sale in November of 1979 and was a big hit. It retailed for US$100 and sold so well that many dealers started bundling the Apple II with VisiCalc. The success of VisiCalc turned Apple into a successful company, selling tens of thousands of the pricey 32 KB Apple IIs to businesses that wanted them only for the spreadsheet.”

    This clip from Triumph of the Nerds makes the effect of the program and the purchasing of Apple IIs related to it very clear, especially after the third minute.

    Finally, as this sales chart shows, many, many copies of Zork I, II and II were sold, to great effect. How many caused purchases can’t be gleaned from here, but it’s significant.

  5. David Cornelson

    January 17, 2012 at 6:06 am

    I too am excited to see each blog post about the history of Infocom in an almost diary fashion. You can almost see these guys coming together and then thinking about how the z-machine came to life is really an astonishing achievement. All for a silly text adventure.

    The one argument about purchased PC’s and dust-collecting Zork boxes is that people continued to buy each Infocom game as they came out. I think this means people really did play the game and it wasn’t just a conversation piece. I remember it was hard to find the early games. It wasn’t until later that they didn’t quite fly off the shelves as quickly. But even the later Zork games always seemed to disappear, even with Atari and graphical PC games on the rise.

    I do agree with the assertion that the MIT folks implemented default responses better than we do even today. We can thank the easy development tools for that. We’re not _forced_ to actually write the code for default responses, which lends itself to laziness. One of the reasons web-based IF is so interesting to me is that if we can develop a following, we can use command tracking to review default responses and user’s reactions to them….and adjust accordingly.

    • Jimmy Maher

      January 17, 2012 at 8:54 am

      “The one argument about purchased PC’s and dust-collecting Zork boxes is that people continued to buy each Infocom game as they came out.”

      On the other hand, each succeeding Zork sold fewer copies than its predecessor, and, with the exception of Hitchhiker’s (a special case of its own), no other Infocom game ever came close to Zork’s sales. Based on all this, I think a fair number were probably sold as essentially novelty items. Hopefully, of course, some came for the parser novelty and ended up sticking around — and buying more — for the meat.

  6. wrm

    January 17, 2012 at 7:32 am

    Which reminds me of a quote from (google goole) (on beta testers)

    > Give them a fork in the road, and they will likely take
    > neither path, but attempt to EAT SPAGHETTI WITH FORK.

  7. Gravel

    January 17, 2012 at 10:09 pm

    This is still an awesome series that is awesome.

    (But my joke wasn’t a “two types of”. I was too busy sniggering about “Joysprick”.)

  8. Michael Waddell

    April 10, 2013 at 5:57 pm

    By neglecting to sell “I appreciate Zork on a much deeper level than you” t-shirts, you are passing up a small fortune, my friend. (This assumes that the wrath of Disney and Onion lawyers is not roused…)

    • Mike Taylor

      October 3, 2017 at 4:12 pm

      That was my immediate thought, too!

  9. Jeff Nyman

    June 27, 2021 at 3:47 pm

    “Their awe was all bestowed at the level of the parser itself, …”

    This feels like reading the intent of masses of people from a historical distance. Which is fraught with peril. No doubt the parser in Zork was better. Historically it can be hard to tell, however, if people just saw this as a spectrum along the lines of what was possible. So was it “awe” at the parser? Or was there something else going on?

    I think much more historically accurate is to recognize that Zork was one of the first games that referenced itself, the most obvious being the leaflet. There was a self-awareness to Zork that certainly guided how people explored it. Zork provided an odd comedic aspect to the adventure, and some ironic elements in its responses, that purposely asked the player to join in that comedy or irony. This allowed for a self-awareness that was not present in other games.

    Did the parser enable that? Possibly. But CRPGs at the time could have done something similar; they just didn’t.

    Beyond this, as you explored more about Zork, you got the feel of a lived in world, as it were. And that lived in world had some history but that history could seem a little comedic in terms of its inhabitants or the incongruity of the rooms you stumbled into.

    All of *that*, to me, seems much more relevant in discussing Zork and may be only tangentially related to the fact that Zork used a parser. Zork provided a world you felt you were stepping into that had a history and a persistence distinct from you, but bounded by self-awareness of playing a game (such as the response to typing “Win” or “Lose” or swearing). “Zork” — the “Dungeon Master” as code — was along with you on your journey and commenting on it. A clear nod to how people tended to play RPG style games.

    I think more of what made this possible was not so much the parser but rather the approach of the Z-Machine which allowed space to carry out these ideas. Most of the *ideas* in Zork could have been conveyed even with a two-word Scott Adams-like parser and certainly all the *descriptions* were distinct from a specific parser. What allowed it all to be possible was the space provided.


Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.