This paper examines the rating scale data of oral proficiency tests analyzed by a Rasch Analysis focusing on an item map and factor analysis. In discussing the item map, the difficulty order of six items and students' answering patterns are analyzed using descriptive statistics and measures of central tendency of test scores. The data ranks the items from easiest to most difficult. The factor analysis shows that one factor should be person-related, and the other should be linguistics-related. The result of the Rasch analysis suggests that there needs to be three tests to make a more precise measurement of the students' communicative language ability. The first is a test for monologue ability, the second for multilogue ability, and the third for dialogue ability. To truly understand the students' language ability it must be examined from several different viewpoints. (Contains an abstract in Japanese.) (KFT)