Interactive fiction is a story that answers back. You read a situation in prose, you express an action, and the prose changes. Almost everything else about the form — what you have to implement, what a reader can do wrong, where the thing gets published — follows from a single split: whether you express that action by typing it or by picking it.
In parser work you type. The game describes a room; you write take the lamp or ask the ferryman about the bridge, and it either understands you or
tells you it does not. The line runs back to Colossal Cave Adventure — Will
Crowther's 1976 cave simulation, expanded by Don Woods the following year — and
on to Zork, which put a monster in the dark and a warning into the language: you
are likely to be eaten by a grue. (Dave Lebling took the word from Jack Vance.)
Infocom sold the form commercially until graphics took the shelf: it went to
Activision in 1986 and shut down in 1989. The decline took about four years, not
an afternoon.
In choice-based work you pick. The passage ends, two or three links appear, you take one. Its ancestor is the printed gamebook; its second life is the web, where a hyperlink is the entire interface. Most new interactive fiction is written this way; most of the works people cite as classics sit on the other side of the split.
| Parser | Choice-based | |
|---|---|---|
| How you act | Type a command in free text | Click one of the links offered |
| What can go wrong | The game cannot parse what you meant | Nothing is refused, so a thin piece has nowhere to hide |
| What the author writes | A simulated world: objects, rules, and replies to things most players never try | Passages, the graph joining them, and the variables it sets |
| Where it's published | A story file run in an interpreter; Inform still compiles to Infocom's Z-machine | One self-contained HTML file, on any host |
| Who it's for | Players who want to propose an idea | Readers who want the whole surface legible |
Parser work's real failure has nothing to do with difficulty. It is knowing
exactly what to do and failing to find the phrasing the author had in mind —
pull lever against push lever, look behind the curtain against search curtain. That is not a puzzle. It is a vocabulary quiz you did not enrol in,
and it has turned more people away from the form than any actual challenge.
The fix is not a longer verb list, it is depth. A parser game is worth playing in proportion to how often it has an answer ready for something you tried on a whim. Nobody had to write what happens when you smell the rug. The ones that did are the ones people still recommend. Guess-the-verb is what it feels like when that budget ran out three rooms ago.
Choice-based work fails in the opposite direction: with the guessing gone, a weak piece is nakedly a series of paragraphs with buttons. The answer is to make each option express a value — what you will spend, who you are prepared to disappoint — so picking is characterisation rather than a coin flip with extra reading.
All three are free, which is not a coincidence: when the money left, the form went amateur, and then it got good.
Inform (Graham Nelson, 1993) is the parser standard. Inform 7 source reads
as English sentences — The brass lamp is in the Cave. — and compiles to
Infocom's Z-machine, or to Glulx when a game outgrows it, so something written
this year runs in the same interpreters as work from the 1980s.
Twine (Chris Klimas, 2009) is the choice-based one. You lay passages out as boxes and links as arrows, and it exports one HTML file you can put anywhere — no server, no store, no build step, which is why so many Twine authors are writers who never set out to make a game. ink, inkle's scripting language, is the same idea for people who did: it drives 80 Days and drops into a game engine.
IFComp has run every year since 1995. Entries are free and judges score each one after at most two hours, a rule that set the length of the modern canon. It is where the form renews itself.
Continuous space and timing. A chase, a crowd, the felt geometry of a room: these are laborious in prose and trivial in almost any other genre. The form is extraordinary at interiority and terrible at choreography. If the best moment in your idea is a spatial one, you are in the wrong medium; if it is a thought your character cannot say out loud, nothing else comes close.
It is also why interactive fiction thrives inside games that never claim the name — the event text in a colony sim, the encounter prose in a roguelite. Add portraits and a soundtrack and you have a visual novel, which trades the parser's open proposal for a fork you can see coming.
Type: Interactive fiction — parser (you type) or choice-based (you pick)
Core loop: Read the situation, form an intent, express it, read what changed
Perspective: Text, second person; illustration optional and usually absent
Win / lose: Usually which ending; parser work keeps deaths and dead ends as well
Session: A passage on a phone or a whole arc at once — the state is the text, so it pauses anywhere
Progression: Story state and hidden stats: trust, money, how ruthless you have been
Name the tradition in the first line of any pitch. By then you have already decided what has to be implemented, what your reader can get wrong, and whether the thing ships as a story file or a web page.