Previous 3D games like Quake III Arena, used separate lighting models for determining how a light would illuminate a character or a map. Lighting and shadow information for maps would be static, pre-generated and stored, whereas lighting and shadowing information for characters would be determined at run-time.
Doom 3 uses a unified model that generates the lighting and shadows for everything at run-time. This means that any lights will affect the whole scene and not only certain parts. This may include self-shadowing via shadow volumes, where characters can cast shadows on themselves, for example the shadow cast by the marine's chin onto his neck.