The purpose of this study is to analyze perceptions (emotional perceptions, cognitive identification with teacher and leadership perceptions) which have to do with modeling effects. When classmates and teachers in classes as continuously existing groups are presented as models, it is assumed that there will be different modeling effects, depending on how one perceives internal relations with them. The experiment was made up of two parts. We gave a pre-experiment survey which included near sociometric test, teachers leadership perceptions, cognitive identification with teacher, innate aggressiveness scores and composition-writing based on TAT diagram (13). One and a half days later, an aggressiveness test was given by means of language instruction in a language laboratory. Subsequently, we measured aggressiveness scores and figured out the differences. Used as subjects were 180 junior high school boy students, and the teachers were all male. The instructors were also male. As a result of the experiment, the following facts were found. 1. Positively-perceived classmates were found valid as models, but negatively or neutrallyperceived classmates and teachers in any emotionalperception groups were not found valid as models (Table 2, 3; Fig. 2). 2. Regarding both classmates and teachers, more significant modeling effects appeared in those models who were perceived to be significantly high cognitive identification with teacher than in those models who were not perceived that way. Incidentally, the correlation between cognitive identification with teacher than in those models who were not perceived that way. Incidentally, the correlation between cognitive identification with teacher and modeling effects was significant (Table 4). 3. No significant differences were found between teachers leadership perceptions and modeling effects (Table 5). However, significant differences in cognitive identification, leadership perceptions and modeling effects were found that PM, P and M types were superior than control group with significant of cognitive identification with teacher and pm type is superior than control group in those groups with no significant of cognitive identification with teacher.