along, starting withAMFV, and then another pretty major revamping around ’87 with
the introduction of an entirely new, much more powerful parser. And then, of course,
there was a major overhaul for the introduction of graphics in ’88.
A lot of effort was put into the Infocom parser, and it was well respected as the
best in the industry. Did it ever get so good that you thought it couldn’t get any
better?
Certainly, by the time of the new from-the-ground-up parser circa 1987, I thought we
had a parser that, while it could certainly be improved, was about as good as we’d ever
need for a gaming environment. After all, we weren’t trying to understand all natural
language, just present-tense imperative sentences. The only area where I would have
liked to see continued improvement was in the area of talking to NPCs. But the main
problem with making NPCs seem more deep and real wasn’t due to parser limitations,
it was just the sheer amount of work needed to give a character enough different
responses to keep that character from seeming “canned,” even for a short while.
I personally loved and still love the text-based interface, both from a player and a
game writer point of view. But I don’t mind either reading or typing, and some people
dislike one or the other or both, and that tended to limit our audience, especially as
non-reading, non-typing alternatives proliferated. But I find the parser-based input
interface to be by far the most powerful and flexible, allowing the user to at least try
anything he/she can think of, and allowing the game writer to develop all sorts of puz-
zles that wouldn’t be possible with a point-and-click interface. So many point-and-click
adventure games became a matter of simply clicking every object in sight in every pos-
sible combination, instead of thinking through the puzzle.
What do you say to criticisms that the parser interface often proved more frus-
trating than intuitive, and that though the player may know what they want to
do, he or she may have trouble finding the correct words for that action?
I think that’s simply a poor parser. I can remember playing one Sierra game where
there was what I thought was a horse on the screen, and I was trying to do all sorts of
things with the horse, and it later turned out it was a unicorn. In those days, when the
resolution was so grainy, I was simply not noticing the one pixel that indicated a horn.
And so when I was saying stuff like, “Get on the horse,” it wasn’t saying, “There’s no
horse here,” which would have tipped me off that maybe it was a unicorn. Instead it was
responding with, “You can’t do that” or something much less helpful. So to me, the fault
wasn’t that the game had a parser interface; the fault was that the game was not well
written to begin with or well tested.
Certainly when someone sits down with even the most polished Infocom game,
there tends to be, depending on the person, a one-minute or a half-hour period where
they’re kind of flailing and trying to get the hang of the syntax. But for most people,
once they get past that initial kind of confusion, a well-written parser game isn’t partic-
ularly frustrating. Even in the later Infocom games, we were starting to introduce some
things that were really aimed at making that very initial experience less difficult: trying
to notice the sorts of things that players did while they were in that mode, and make
suggestions to push them in the right direction. The game would try to catch if they
Chapter 10: Interview: Steve Meretzky 183