The flat screen and no camera
Two-dimensional games don't have cameras in the sense that three-dimensional games do. The "camera" in a 2D platformer is the viewport — the portion of the level visible on screen at any moment — and its movement is typically a simple follow of the player character with some scrolling lag. Super Mario Bros. scrolls right to follow Mario, keeping him in roughly the left third of the screen when he is moving rightward; the "camera" decision is the scrolling speed and the character's position within the frame. The design is functional and invisible: players understand intuitively that the screen shows the area immediately around the character and that moving right reveals what is ahead.
The 2D camera's most significant design decisions were around what to reveal. Contra's auto-scrolling levels removed the player's ability to advance or retreat, creating urgency through the camera's movement rather than through enemy pressure alone. Metroid's fixed-screen sections — each room was contained on one screen, transitioning to the next screen when Samus exited the room — created a spatial memory challenge rather than a scrolling exploration, with the room boundaries defining the game's spatial structure. The camera decisions in 2D games were primarily decisions about what was shown and when rather than about a three-dimensional point of view.
Fixed camera and Resident Evil
Resident Evil (1996) used pre-rendered background images displayed from fixed camera angles — specific viewpoints that had been rendered in advance and displayed as static images behind the real-time character models. The fixed camera approach was partly a technical decision — pre-rendered backgrounds allowed visual quality that real-time 3D backgrounds couldn't match on the PlayStation — and partly a design decision: fixed camera angles could be chosen specifically to create tension and withhold information.
The famous corridor shot in Resident Evil — the player character walking toward the camera, a zombie crouching over another character's body in the background, both visible but the zombie facing away from the player — was a horror film framing that a free camera wouldn't have produced. The camera placement created suspense because the player could see the threat without knowing whether the threat was aware of them, and the angle made it impossible to see what was around the corner ahead of the player character. Fixed cameras could be art-directed for emotional effect in ways that follow cameras and free cameras couldn't: the director chose the angle that best served the scene's intended emotional quality, rather than the angle that best served the player's navigational needs. The price was occasional navigation confusion — players moving through spaces where the camera angle made their movement direction difficult to read — which became Resident Evil's most persistent design criticism.
The 3D platformer problem
Super Mario 64 launched the 3D platformer into a camera design problem that has not been fully solved. The game's camera system — called Lakitu's camera, after the cloud-riding character who was notionally operating it — attempted to follow Mario at a useful distance and angle while avoiding collision with the environment. The camera succeeded often enough to make the game playable but failed noticeably in specific situations: narrow corridors where the camera couldn't maintain its follow distance, complex geometry where the camera clipped into walls, and rotating platform sequences where the camera's movement created disorientation rather than clarity. Mario 64's camera problems were well-known at the time of release and were subsequently addressed by the inclusion of free camera control (allowing players to reposition the camera manually with the C buttons) that the game provided but required players to actively manage.
Crash Bandicoot's design decision — restricting Crash's movement primarily to a forward-toward-the-camera axis — was the alternative approach to the same problem: if the camera always shows the player from behind at a fixed distance, and if the player can only move forward into or backward away from the camera, there are no camera failure cases. The camera is never wrong because there is only one valid camera position. The trade-off was that Crash's game world couldn't include the spatial freedom that Mario 64's offered; the linearity of Crash's stages was the constraint that the camera simplicity required. The solution that worked for Crash didn't work for games that wanted non-linear environments.
Over the shoulder and the modern solution
Resident Evil 4 (2005) introduced the over-the-shoulder camera that became the standard for third-person action games for the following two decades. The camera positioned itself behind and above the player character, maintaining a fixed distance and angle that showed both the character and the environment ahead of them simultaneously. The design was not new — it had been used in games before — but RE4's implementation was sufficiently polished and the game's commercial success sufficiently large that the camera position entered the design vocabulary as a known solution rather than an experiment.
The over-the-shoulder camera's advantages were specific to the game type it served. For games with ranged combat — shooting at enemies at distance — the camera's position relative to the character made aiming intuitive: the weapon pointed in the direction the camera was facing, and a crosshair over the camera's forward view communicated the aim direction directly. For exploration — navigating environments with vertical and lateral movement — the camera's follow behavior needed additional sophistication, and the camera systems in open-world games that adopted the over-the-shoulder position (Grand Theft Auto III, The Witcher 3, many others) required complex collision avoidance and blending systems to avoid the camera problems that Mario 64 had encountered a decade earlier. The technical complexity of implementing a camera that worked correctly in all situations increased as game worlds became more geometrically complex, and the camera remained one of the most technically challenging and player-visible systems in third-person game development.