A method for transmitting a three-dimensional (3D) caption signal and caption display method are disclosed to three-dimensionally display caption text such that it fits a 3D image in a 3D display device. 3D caption box configuration information and caption text are generated to display a 3D caption within a 3D image, and the 3D caption box configuration information and the caption text are inserted into a video picture header region to code the image signal, and then, the image signal is transmitted. In the caption display method, a broadcast signal including a 3D image signal is received, and 3D caption box configuration information and caption text included in the broadcast signal are acquired. A caption image including 3D caption text disposed within a 3D caption box is then generated based on the 3D caption box configuration information and the caption text, and displayed.