Dogs were tested (1) in a two-way choice experiment, where the experimenter indicated a baited bowl by pointing; and (2) in a task where the owner was asked to command the dog to execute simple obedience tasks. In expt 1 dogs (n = 10) were tested at first in the presence of the experimenter (three dimensional condition, 3D), that was followed by a series of pointing trials when the life-sized image of the experimenter was projected onto the wall by the means of a video-projector (two dimensional condition, 2D). Dogs performed correctly more often than expected by chance in both 3D and 2D conditions. In expt 2 the commanding owner was either present in the room (3D), or her/his image was projected on the screen (2D), or only her/his voice was audible for the dog through a speaker (0D). The performance of the dogs (n = 10) decreased to great extent comparing the 3D and 0D condition, as the number of different actions the dogs obeyed was significantly less in the 0D condition. However, there was no difference in the number of different actions obeyed in the 3D and 2D conditions. Our results show that dogs readily obey life-sized, interactive moving image in various communicative situtations. We suppose that the difference between 2D and 3D conditions can be attributed partially to the lack of some additional communicative signals as sounds (verbal cues) and odours (from the human), and to some changes in the context.