For realizing robots working in human society, interaction between people and robots is one of the important issues. We have developed a robot that behaves based on visual information and interacts with people. The robot controls its gazing direction for representing the internal states. By using the robot, we have evaluated impressions given by the robot based on psychological methods. SD method and factor analysis have been applied to evaluate the impressions. As a result of the experiments, impressions of the robot principally consisted of 4 factors: familiality, enjoyment, activity, and performance evaluation, and these have been effected by computer skills of subjects. In addition, we have verified that gazing direction control promotes interactions between people and the robot.