← All Essays
Design 12 min read

The Grammar of Stealth

How Thief, Metal Gear Solid, and Splinter Cell built game tension not from action but from information — and why "can they see me?" is one of design's most effective questions

The problem stealth solves

Most action games resolve threat through direct engagement: combat systems allow players to eliminate threats by overpowering them. The limitation of this approach is that it creates a specific type of player agency — agency through force — and a specific type of game economy in which the player's resources are health and ammunition and the relevant question is whether those resources are sufficient to win each encounter. The emotional register of this type of game is combat-adjacent: excitement, adrenalin, the satisfaction of winning an exchange.

Stealth games resolve the same problem — the player wants to pass through an environment containing threats — through a different approach. The player avoids encounters rather than winning them, moving through the environment in ways that keep threats from initiating combat. The relevant question shifts from "can I defeat this threat?" to "does this threat know I'm here?" The emotional register shifts from combat excitement to tension: the fear of detection, the relief of passing unseen, the anxiety of near-misses. The stealth game is producing a different set of psychological states than the action game, using the same basic situation of a player in an environment with threats.

The design implication is that stealth games must make the state of NPC awareness visible and manageable. A player who can't tell whether a guard can see their position can't make informed decisions about movement. The communication of NPC awareness — through sight cones displayed on a minimap, through the guard's posture and attention animation, through audio cues that indicate suspicion — is the primary communication challenge of stealth design. Everything else in stealth game design serves the management of this information gap.

Thief and sound as detection

Thief: The Dark Project (1998), developed by Looking Glass Studios, was the first commercially successful stealth game and the one that established most of the vocabulary subsequent stealth games used. The player character Garrett was a thief in a gothic-industrial fantasy setting who avoided combat because he was ill-equipped for it: a thief, not a fighter. The design consequence was that avoiding detection was not an optional playstyle but the game's intended mode — players who relied on combat found their resources depleted faster than the level structure accommodated.

Thief's central design innovation was sound-based stealth. The game modelled the sound surfaces produced by different floor materials — stone, wood, carpet, water — and adjusted the noise Garrett's movement generated based on which surface he walked on. Walking on stone in hard shoes produced noise that guards could hear across large distances; walking on carpet in soft shoes produced minimal noise that required close proximity to detect. The player who understood the sound system managed their movement through a level as a sound-management problem: seeking soft surfaces, avoiding noisy ones, crouching to reduce sound output on surfaces that couldn't be avoided. The mechanic was intuitive — players understood why carpet was quieter than stone — but required continuous spatial awareness to execute, because the player had to track both their current position and the sound properties of the surface they were on.

Thief's light system — a gem in the corner of the screen that darkened as Garrett entered shadow — provided the second major stealth mechanic: visibility management. Guards who didn't see Garrett in light could be stood next to without detection if Garrett was in shadow. The darkness that the player sought, the light the player avoided, and the surfaces that generated sound created a three-dimensional stealth environment where movement decisions involved simultaneously managing visibility, sound, and proximity to guards. The combination produced a game that rewarded careful observation and planning more than reflexes.

Metal Gear Solid and enemy AI states

Metal Gear Solid (1998), directed by Hideo Kojima, brought the stealth genre to a mainstream audience that Thief's PC-exclusive release and gothic setting hadn't reached. The PlayStation release, the cinematic presentation, and the action-game framing — the player character Snake was a soldier who could fight if necessary — expanded the genre's accessibility while maintaining the core mechanic of avoiding detection. Metal Gear Solid's most significant design contribution was the formalisation of enemy awareness states: a guard system with discrete alert levels that communicated clearly to the player where they stood in the detection sequence.

Guards in Metal Gear Solid existed in one of three states: normal (no suspicion), caution (aware of something unusual but not specifically aware of the player), or alert (aware of the player's approximate location and actively searching). Each state had distinctive music, distinctive guard behaviour, and a defined path back to the previous state. A guard who heard a noise moved to caution, investigated, and returned to normal if nothing further happened. A guard who saw the player moved to alert, called for reinforcements, and maintained alert status for a defined period even if the player hid. The clarity of the state system allowed players to understand their risk level at any moment and to make recovery plans when detection occurred.

The exclamation point that appeared above a guard's head when they spotted the player — accompanied by a distinctive sound cue — was the system's communication design solved in a single design decision. Players who heard the sound and saw the icon knew immediately that detection had occurred, who had detected them, and approximately when the guard had detected them. The communication was so effective that the exclamation point became the genre's cultural shorthand for detection across games that had nothing to do with Metal Gear. When a visual indicator needs to communicate "you have been seen," the exclamation mark is still the default choice in stealth games over twenty-five years later.

What stealth asks of the player

The cognitive demands of stealth games are specific to the genre and distinct from action game demands. A stealth game player who is managing a level successfully is tracking the positions and patrol routes of multiple guards simultaneously, monitoring their awareness states, planning movement routes that satisfy time constraints (passing through a patrol gap) and environmental constraints (avoiding light or sound surfaces), and maintaining contingency plans for detection events. This is a spatial and temporal planning task — modelling the environment as a dynamic system with predictable behaviour — that action games don't require in the same form.

The stealth game's punishment for detection is typically reset: reloading from a save point or checkpoint, retrying the section from a position before detection occurred. This punishment structure is more severe than most action game failure modes, because it resets not just resources but the planning and execution of a sequence that may have required several minutes to reach. The emotional weight of a detection-triggered reset in stealth games — the specific frustration of being forced to redo careful work — is a design tool: it makes successful non-detection feel meaningful in proportion to the cost of failure.

Dishonored (2012), developed by Arkane Studios, and the broader immersive sim genre's relationship with stealth demonstrates the genre's design flexibility. Dishonored made lethal action and non-lethal stealth equally viable, but tracked the player's approach through an "chaos" system that affected the game world's tone — more death produced a darker, more plague-ridden world. The moral weight given to stealth versus action was an explicit design statement: the game preferred the stealth approach and communicated that preference through systemic consequences rather than through explicit instruction. Players who chose stealth were not just taking an alternative path; they were making a moral choice the game acknowledged. The extension of stealth design from pure detection avoidance to moral consequence tracking represented the genre's maturation from a mechanical exercise in information management into a design tool for communicating values.