Twenty one undergraduates, acquainted with one another, rated both their own personality and those of the peers twice at three-month interval, using the ACL (Adjective Check List) based on the five-factor model of personality (FFM). By using the member-recombination technique, various subgroups were formed out of the subjects. With respect to each personality factor, the variance components for raters, ratees, rater-by-ratee interaction, and residual were estimated within each subgroup. The relationships of the variance components with the size of subgroups, with the matrices of subgroups, and with the familiarity among the members of each subgroup were analyzed. In the last analysis, the subgroups were divided into three classes according to the familiarity, and the variance partitioning was compared among these classes. The results showed that, in the class of maximum familiarity, the ratee effect for agreement and neuroticism was larger and that for extraversion was smaller.
