Ron Gilbert: The Puzzle Grammar of Monkey Island

In December 1989, a 25-year-old designer at Lucasfilm Games published an essay in the Journal of Computer Game Design titled “Why Adventure Games Suck”. He had shipped three adventure games at that point and was about to ship the one everybody remembers. The essay is a list of things his own genre did that made him furious, and the fury is specific: adventure games kill you for exploring, then expect you to keep exploring.

Ron Gilbert’s career runs on that document. Everything he has built since is a working demonstration of an argument he wrote down before he was thirty, and the argument turned out to be right in a way that took the rest of the industry two decades to catch up with.

The rules, and why they are mechanical

The essay’s rules read like manners. They are engineering.

No deaths. No dead ends. No unwinnable states you cannot detect. Puzzles that advance the story rather than blocking it. Solutions that make sense in retrospect — the famous line about the player saying “of course” rather than “I would never have got that”.

Consider what each rule does to the state space. A game that can kill you has to be saved constantly, which means the player is holding a mental model of the save system alongside the fiction, and the fiction loses. A game with dead ends means every action carries unknown risk, so the rational player stops experimenting. Sierra’s parser adventures were riddled with both: you could miss an object in the second room and discover ninety rooms later that your last twelve hours were void. The player response was completely sensible — touch nothing, save everything, walk down the middle.

Gilbert’s rules delete that entire failure mode. If nothing can go permanently wrong, experimentation costs you nothing except time, and once experimentation is free the designer can make the puzzles genuinely hard. Difficulty without punishment is play. That is the trade, and it is why a LucasArts adventure can be more demanding than a Sierra one while feeling kinder.

The joke in The Secret of Monkey Island is that you can die in it, exactly once, if you stand underwater doing nothing for ten minutes. The rule is strong enough to be worth teasing.

SCUMM was a means

He got to Lucasfilm Games in 1985 by way of Human Engineered Software, where he had co-written Graphics BASIC for the C64. In 1987 he and Gary Winnick made Maniac Mansion, and scripting it in 6502 assembly proved intolerable, so Gilbert built a scripting language and an interpreter to run it. The Script Creation Utility for Maniac Mansion — SCUMM — meant a designer could write game logic without writing machine code, which meant designers could be writers and cartoonists instead of only engineers. I have taken that apart properly in the SCUMM doctrine, because the tool reshaped the studio around it.

What matters for Gilbert specifically is that SCUMM let him iterate. A puzzle grammar is not designed on paper. It is found by building forty puzzles, discovering that eleven of them are unfair, and rewriting them on a Tuesday. Every design rule in the 1989 essay is a rule he could afford to enforce because the tool made enforcement cheap.

Zak McKracken and the Alien Mindbenders followed in 1988 with David Fox, and Indiana Jones and the Last Crusade: The Graphic Adventure in 1989 with Noah Falstein and David Fox — the game that introduced the idea of multiple routes through the same problem, so the Grail could be reached by fists, wits or a forged pass.

The three trials are the thesis

The Secret of Monkey Island arrived in October 1990, designed by Gilbert with Tim Schafer and Dave Grossman, its premise borrowed openly from the Disney ride and from Tim Powers’ 1987 novel On Stranger Tides.

Look at the structure of Part One. Guybrush must master the sword, master thievery, and master treasure hunting. Three chains, available simultaneously, mutually independent. This is the single best structural decision in the history of the genre, and it is invisible because it works.

Being stuck is the adventure game’s terminal disease. Every other failure — bad writing, ugly art, a fiddly interface — is survivable. Being stuck for forty minutes with no available action is the point at which a player quits and never returns. The three trials mean you are essentially never stuck: you are stuck on one thing, and there are two other things you could be doing instead, and the walk between them is where your subconscious solves the first one. The design has built the player a place to put their frustration.

That is a systems solution to a psychological problem, and it is the reason a nine-hour game with genuinely obtuse puzzles never feels like homework.

Insult swordfighting is a lookup table with feelings

The famous bit deserves a mechanical description, because the mechanical description is why it works.

You need to beat the Sword Master. Swordfighting on Mêlée Island is conducted by insult: your opponent insults you, and you must answer with the correct comeback. You do not have the insults or the comebacks. You get them by fighting pirates on the road and losing — they insult you, you note it down, and now you own that insult. Fight another pirate, use your new insult, and their comeback goes in the book too.

So: losing is the acquisition mechanic. The fight is a lookup table, the table is populated by failure, and the failure is funny because both parties are calling each other rubbish in increasingly baroque terms. The insults themselves were sharpened with Orson Scott Card, brought in for exactly that.

Three things happen at once. You are learning a system, you are being entertained, and the entertainment is the feedback signal that tells you the system is being learned. Comedy and mechanic are the same event. That is a very small club — I made the same argument about the wine-in-the-time-capsule puzzle in Day of the Tentacle, which is Gilbert’s studio culture executing on Gilbert’s rules after he had moved on.

Monkey Island also brought the branching dialogue tree into SCUMM, and it is a quieter revolution than it looks. A parser asks the player to guess the writer’s vocabulary. A tree hands them a menu of things worth saying, which means every line can be written to be good rather than merely parseable. The genre’s dialogue got funny the moment it stopped being a typing test.

The grammar is a literal grammar

The word in the title of this piece is not decorative. A SCUMM game is built on a sentence line: a verb, an object, sometimes a preposition and a second object. Use bottle with grog. Give map to pirate. The interface is a constrained natural language with a vocabulary the player can see.

That constraint does an enormous amount of quiet work. A parser adventure asks you to guess what the writer typed; the failure state is a game that understood the world perfectly and did not understand you. A verb list removes the guess. The set of expressible sentences is finite and visible, so when a puzzle beats you, it beats you on the puzzle rather than on your vocabulary.

The consequence for design is that the puzzle space is now enumerable. Every solvable problem is some sentence in the verb grid crossed with the inventory, and a designer can reason about the whole space rather than hoping. Gilbert kept trimming that grid for a decade because every verb removed makes the remaining ones sharper, and by the time you reach the nine-verb games the interface has almost disappeared into the fiction.

Monkey Island 2: LeChuck’s Revenge took the argument one step further and shipped with a reduced mode that strips out puzzles for players who wanted the story at a gentler grade. A 1991 adventure game with a difficulty setting, from the studio whose entire identity was hard-but-fair puzzles. He was still worrying about the same thing the 1989 essay worried about: the player who puts it down.

What he did with the rest of it

Monkey Island 2: LeChuck’s Revenge shipped in December 1991 and is the harder, stranger, better-looking one. Gilbert left LucasArts in 1992 and co-founded Humongous Entertainment with Shelley Day, where he spent most of a decade making Putt-Putt, Freddi Fish and Pajama Sam — adventure games for small children, built on the same fairness rules, which is the least surprising career move on this list. His rules were always about not punishing a curious player, and nobody is more curious or less deserving of punishment than a six-year-old. Humongous also spun off Cavedog, which produced Total Annihilation in 1997, so the man’s ledger includes one of the best RTS games ever made almost by accident.

Then a long stretch in the wilderness, and a return: Thimbleweed Park in 2017, reuniting with Gary Winnick and funded by people who wanted a 1987 game made properly in 2017. And Return to Monkey Island in 2022, with Dave Grossman, thirty-one years after he last touched the series. Both games are built on the 1989 essay. Neither has moved an inch.

The verdict

The thing Gilbert understood before anyone else is that an adventure game is a directed graph and the player is walking it blind. Every design decision he is known for follows from that: fairness because a blind walker cannot avoid a hidden pit, parallel chains because a blind walker needs alternative edges, dialogue trees because a blind walker should be shown their options.

The genre spent the nineties chasing cinema instead and largely broke itself doing it — I traced the wreckage in why point-and-click died and what replaced it, and the graph is still the thing that survives when the FMV rots. Modern games rediscovered his rules under other names. The generous checkpoint is no-death. The open objective board is the three trials. The quest marker is a dialogue tree for the whole world, which Gilbert would probably say goes too far.

Start with The Secret of Monkey Island on whatever you own; the special edition’s redrawn art toggles back to the 1990 version mid-scene, and both are the same game underneath. Then read the essay. It takes fifteen minutes and it is still the best thing anybody has written about why people put a game down.