From text to image
Adventure games began as pure text. Colossal Cave Adventure (1976) by Will Crowther and Don Woods described a world through prose: "You are in a maze of twisty little passages, all alike." Players typed commands — "go north," "take lamp," "xyzzy" — and the game responded with more prose. The entire experience existed in language. Zork (1977), MIT's expansion of Colossal Cave's concepts, refined the parser — the text-interpretation system that turned typed input into game actions — to a sophistication that allowed compound sentences and complex commands. Scott Adams developed the first commercially distributed text adventure games in the late 1970s, establishing that players would pay for the experience.
Mystery House (1980) added graphics to the formula — hand-drawn line art of each room, alongside the text interface — and demonstrated that players wanted to see as well as read. Sierra's subsequent AGI engine made the graphics interactive: characters moved through perspective-rendered environments, walked behind foreground objects, and interacted with a world that had visual depth. By 1984, with King's Quest, the graphical adventure had defined its conventions: perspective backgrounds, an animated player character, a text parser for commands, and puzzles that required finding, combining, and using objects in specific ways.
The text parser remained standard until LucasArts replaced it with the verb-based SCUMM interface in Maniac Mansion (1987). The transition from typing to clicking eliminated a significant frustration: the parser-driven game required players to guess the vocabulary the game accepted, leading to situations where the correct action was obvious but the correct phrasing was not. "Move log" might work where "push log" and "shove log" didn't. The SCUMM interface made all available actions explicit, which made design more demanding — every visible object needed a response to every applicable verb — but made play less arbitrary.
The two philosophies
Sierra and LucasArts, the dominant adventure game developers of the late 1980s and early 1990s, had fundamentally different philosophies about what adventure games should do to players who failed. Sierra's games were consequential: wrong actions produced death, unwinnable states, and the need to reload much earlier saves. King's Quest IV could be made unwinnable by using an item in the wrong situation early in the game. The standard Sierra advice — save frequently, maintain multiple files — was a direct response to the design's propensity for trapping players in situations they couldn't escape without losing hours of progress. The philosophy was inherited from text adventures, where getting stuck and dying were standard features rather than failures.
LucasArts' philosophy, codified in Ron Gilbert's 1989 essay "Why Adventure Games Suck," held that players should not be able to die, should not be able to reach unwinnable states, and should always have all the information they need to make progress — eventually. LucasArts games were explorable in every direction simultaneously; items obtained early were needed later, but the game tracked their necessity and wouldn't allow the player to miss them permanently. A LucasArts adventure could be navigated to completion by a player who was patient and observant without requiring them to manage save files defensively.
The commercial records suggest Sierra sold more games throughout most of the genre's peak period, which complicates the argument that LucasArts' forgiving design was objectively superior. The Leisure Suit Larry series, the Space Quest series, and the King's Quest series all generated substantial and loyal audiences who accepted and in some cases valued the punishing design. The Day of the Tentacle and Sam & Max Hit the Road were critical successes with smaller commercial footprints than Sierra's equivalent releases. Whether forgiving design produced better games or simply different games appealing to a different audience is not fully answerable from the commercial data.
The genre's peak and its structural problem
The early 1990s were the genre's creative peak. Monkey Island 2: LeChuck's Revenge (1991), Day of the Tentacle (1993), Sam & Max Hit the Road (1993), Gabriel Knight: Sins of the Fathers (1993), and Grim Fandango (1998) are routinely cited among the finest games of their era by players who experienced them. The CD-ROM transition enabled voice acting, allowing characters to speak with performances that matched their written dialogue. Full Throttle (1995) and The Dig (1995) moved the form toward cinematic presentation. Myst (1993) — technically an adventure game, though its puzzle philosophy was different from both Sierra and LucasArts — sold 6 million copies and became the best-selling PC game of its era by reaching audiences who would never have identified themselves as adventure game players.
The structural problem was that adventure games were primarily about puzzles, and puzzles had a specific relationship with narrative that became harder to sustain as narrative expectations rose. A puzzle exists to be solved; once solved, it is inert. The same solution can be applied at most once. A game built around puzzles consumed its own content as it was played — each solved puzzle was a piece of the game that no longer existed as a challenge. The more sophisticated the puzzle design, the more work was required to create it, and the more quickly that work was consumed by the player. The replayability of adventure games was low because the second playthrough required executing solutions the player already knew.
The genre also had a specific failure mode that players called the "moon logic puzzle" — a solution that required a non-obvious chain of reasoning that was internally consistent within the game's logic but felt arbitrary to players approaching it without the designer's perspective. The infamous "cat hair moustache" puzzle in Gabriel Knight 3 required players to create a disguise from a piece of tape, a cat hair, a travel poster, and a bottle of masking powder in a sequence that seemed comprehensible only after reading the solution. These puzzles were not common in the best adventure games but were common enough in the genre's output to become its defining critique.
The fall and the return
The genre's decline accelerated after 1996. Grim Fandango (1998) was LucasArts' final traditional adventure game — it sold poorly enough that LucasArts cancelled two in-development adventure games and shifted resources to Star Wars action titles. Sierra's parent company dissolved its games operations around the same time. The causes were multiple: the rise of 3D action games (Quake, Tomb Raider, Half-Life) that offered more immediate visceral engagement; the CD-ROM glut of mediocre adventure games that had trained players to expect low quality; the specific frustration of adventure games that had produced enough stuck-and-frustrated players to create a cultural association between the genre and irritation.
Telltale Games, founded in 2004 by former LucasArts employees, revived the genre by radically changing its relationship to puzzle-solving. Telltale's games — beginning with Sam & Max: Season One (2006) and achieving commercial success with The Walking Dead (2012) — subordinated puzzles to narrative, emphasising player choice and emotional consequence over mechanical challenge. The puzzle difficulty dropped to near-trivial levels; players who would have been frustrated by traditional adventure design found Telltale's games accessible. The genre had transformed from puzzles with narrative framing into narratives with minimal puzzle interruption.
Indie developers explored the same territory without Telltale's commercial constraints: Kentucky Route Zero (2013–2020), Disco Elysium (2019), and Norco (2022) produced adventure games that treated narrative, character, and setting as primary interests and puzzles as optional or incidental. The genre that had defined itself around puzzle-solving had reinvented itself as a medium for interactive fiction — a redefinition that the best text adventures had always implied and that point-and-click mechanics had only temporarily obscured.