I have finally finished writing my history of Zork for GamaSutra. Despite some initial difficulties, I was able to secure interviews from a variety of important figures in the Zork and larger IF world, including Steve Meretzky, Marc Blank, Dave Lebling, Nick Montfort, and Howard Sherman. Everyone of these gentlemen has been extremely helpful and deserves applause for taking the time to help me research such an important topic. After the article appears on GamaSutra, I plan to publish the entire interviews on here as a sort of "round table" discussion.
At any rate, doing the work for this article has certainly caused me to think more about the relationship between text and graphical adventure games. Indeed, I'm wondering if the two can even be considered similar, much less part of the same genre. What's nice about the Zork series is that you have the whole spectrum from pure text to Myst-clone. Imagine the transition from text to full motion video. But which, ultimately, offers a more compelling experience--graphic or text? Is it ultimately a matter of personal preference?
One thing that came out in the interviews is that a text parser offers (or at least seems to offer) more freedom than any kind of menu or graphical interface. Games like The Secret of Monkey Island "channel" the player by offering up menu of possible actions (even though you can see a decreasing number of menu options in that series as well; compare Maniac Mansion with Loom, for instance). On the other hand, even if you can theoretically type anything into a parser, it can only "make sense" of a finite number of commands. It seems to me that the only way a command will work is if the programmers anticipated it and implemented a logical response. I know we've had this discussion before, but some linguists (Chomsky, in particular) argues that language is infinite (generative grammar). Thus, if you have a finite set of responses, but an infinite number of possible responses, you'll always have a problem. Possible solutions are the "adventurese" type pidgin language of text adventures, where players quickly learn what words and syntactical structures they share with the parser. The simplest or most "primitive" text parsers are hardly more flexible than the menu-based interface of a game like Maniac Mansion. On the other hand, the most advanced parsers are capable of making sense of a much broader range of inputs; the effect can seem like magic (People have even been fooled into thinking Eliza was a real person, or frankly not caring that she isn't and still using the program for therapy).
Another interesting thing that came out of the interviews was the problem of identification. One of the neat things about text-based IF is that the player can be required to identify an object with an incomplete description. For instance, a knight could find a computer, but not be able to describe it very well to the player ("I see a strange box with a glowing mirror and a series of tiny squares with strange markings") or the like. This sort of thing would be difficult (if not impossible) with a graphical interface, since the player would see right away that the object was a computer. The only exception I can see here is if the game had either a "fog of ignorance" effect of some sort, or if the graphics were in fact so poorly resolved (as in The Wizard and the Princess) that you simply couldn't tell what it was.
Ultimately, I think the question of which one is better comes down to what type of game you're making, and what type of activities the player will be performing. If the game is focused on puzzles, then a graphical interface opens up many interesting possibilities and can greatly facilitate visualization. Compare, for instance, the sliding blocks puzzle in Zork III with the sliding rooms puzzle in Myst. I also can't imagine the many symbol and glyph-type puzzles commonly found in GAGs existing in text-adventures.
Where text adventures really seem to trump is with dialog and interacting with other characters. No simple dialog tree can compare with being able to type literally anything you want to say. Unfortunately, this is also the hardest thing to implement convincingly, and even cutting-edge games of this sort (say, Facade, or the much older Eliza) are hardly perfect. What we need goes far beyond "parsing." The ideal would be to have dialog so effective that the other characters would act just like real people, with real motives, vocabularies, memories, personalities, emotions and so on. Thus, telling a character "Please give me that wrench, sir" would have a different response than "Oh, please. Give me that wrench, sir." I can imagine some pretty wonderful possibilities if a game was truly able to recognize such nuances. We could take it a step further and even have it interpreting italics for emphasis and quotation marks for irony. In effect, I seem to be pushing more for what Brenda Laurel envisioned as interactive drama.
I suppose we already see this, not between computer characters, but rather between human role-players in games like WoW and Second Life. But will it ever be to the point where this type of rich interaction will be possible in a one-player game, particularly one in which some type of "adventure" was called for? I suppose the real difficulty lies in grafting some type of story or narrative onto it. If you need the player to do X, Y, and Z, that's difficult without putting in rails. On the other hand, you could have the "stuff" going on independent of the player's actions (i.e., the Japanese will attack Pearl Harbor no matter what the player does), which might lead to some very interesting scenarios (i.e., the player tries to tell everyone that the Japanese are about to attack; no one believes him, but after the attack, he's imprisoned for being a spy).
All we really need, methinks, is an AI smart enough to act like a real dungeon master in a classic game of AD&D. If you've ever played D&D with a group of friends, you know what I'm talking about. There's usually a great deal of play-acting and "in character" stuff that gets lost on the computer. A really good DM knows how to respond to unexpected situations and "go with" the flow. Often enough, a good game of D&D may end up far from what the DM or players expected, though the DM is able to referee and make sure what the players are doing is consistent with physics and the logic of the fantasy world, and so on. it need not necessarily be text, either; with advances in speech recognition and better speech software, I can imagine this all taking place without the need to type or read.
At any rate, what I wonder is what would games be like today if AI and Natural Language Processing had come as far as graphical technology? Heck, it seems we're really dragging our heels when it comes to stuff like speech recognition, and computers still don't seem to able to speak naturally (i.e., you can always tell right away that it's a computer). Is it just somehow easier to offer better graphics than better AI? I'm not sure.