In social-psychological studies, it is often necessary to observe what role each member of a small sroup has and how the structure of the group is. SYMLOG (a SYstem for the Multiple Level Observation of Groups: Bales & Cohen, 1979) is a method for that kind of observation. But it is not appropriate to use SYMLOG translated directly into Japanese because of cultural differences. The purpose of this study is to make simple rating items of SYMLOG for the Japanese people. In SYMLOG, three dimensions for locating each group member are assumed, which are “Upward-Downward (dominant-submissive) ”, “Positive-Negative (friendly-unfriendly) ”, and “Forward-Backward (task-oriented-emotionally expressive) ”. First of all, it was pointed out that the definition of “F-B” in the Japanese version should be “in-context-out-of-context”, and then eighteen items were made which are constructed of sets of three items for each polar of each dimension. Observation 1 examined the validity of these eighteen items through factor analysis of ratings of SIMSOC (SIMulated SOCiety) interaction. The result showed the necessity to improve the items N, F, and B. The validity of the remade twenty four items was investigated in Observation 2 through factor analysis of ratings of discussions by five people. It yielded six factors that almost correspond to each polar in each dimension. Further factor analysis of twelve items extracted from the remade items made six factors, that showed the three-dimension-structure. Correlations between ratings by these twelve items and scorings (that is the other method in SYMLOG) were positive. Therefore these twelve items in Japanese Japanese improved version of SYMLOG-are valid and useful for observation of small group interactions.