This paper describes a method of controlling a gaze of a mobile robot for interpretation of road intersection scenes considering the uncertainty of interpretation. From a monocular color image, candidate regions are extracted for possible objects such as curve mirror. The probabilities of the regions coming from the objects are calculated from the probabilistic models of the objects. From the probabilities of the regions and the relation between the objects and intersection types, the current probability distribution of intersection types is calculated. If the entropy of the probability distribution is low enough, the robot adopts the best hypothesis. Otherwise, the robot selects and gazes the part which minimizes the expectation of the entropy. These actions are iterated until the entropy becomes low enough. The experimental results are shown for actual intersection scenes.