When speaker IDs, musical lyrics, and sound descriptions have their own distinctive stylistic treatments, they can be extracted from closed caption files and studied as separate units of discourse.