Smooth interaction between human and robot is necessary for coexistence of human and robot. Humans usually interact with each other smoothly by using gestures, thus we focused on them. We conducted two experiments in order to investigate a relationship between human attention and robot’s posture. According to the result, there is a difference between robot’s point of gaze and position of human’s attention. Then we proposed a computational model of robot’s posture that provides appropriate rotation angles of motors of arms and head to draw human’s attention considering the difference. At last, from the experiments, we showed relationships between the rotation angles of motors and position/size of human’s attention. The relationships contribute to develop a computational model of robot’s posture that enables robot to smoothly interact with humans.