The purpose of the research is to explore the validity of the procedure and the mechanism of self-image of morality rehearsal (SIMR) and the effect of the SIMR on self-image of morality (SIM) and moral behavior self-management (MBSM). With between-subject design, four groups of undergraduates at a university in China were at random subjected to four intervening experiments, including no treatment, model learning, empathy training and SIMR. The no treatment group was subjected to normal teaching，the model learning group was subjected to traditional model learning, the empathy training group, which was provided by Li Liao, has good validity for advancing the ability of adolescent’s empathy, and the SIMR group, which was developed by the researcher, was composed of three processes, including generating, selfhood, and emotional experience of moral image. Four teachers were picked out randomly to give instruction, too. The pretest and posttest were carried out for all the four intervening experiments that were implemented eight weeks and composed of four units. There were two weeks and six lessons for one unit. Using the scales of the SIM and the MBSM as instruments, we carried out the pretest and posttest with an interval of two months. SIM was measured by using ten 5-point items and the scale including Self Scale and Behavior Scale, which all referred sensitivity, emotion, value, vividness and operation, and the higher the mark, the better SIM of the subject. Unification between Moral Cognition and Behavior (UMCB) Scale, which was composed of thirty 7-point items, was used to assess the MBSM. UMCB is an operational concept which could imply the level for MBSM. The higher mark in UMCB, the subject was more apt to conform to the commands of the society. UMCB Scale and SIM Scale were developed and validated by the researcher. The statistics analysis show that there is no significant difference between the SIM and the MBSM of every two groups at the pretest (F=2.34, p> .05; F=2.55, p> .05) and the SIM and the MBSM at the pretest and posttest for the controlled group respectively (t= .543, p> .05; t=1.304, p> .05). The T test shows that there is significant difference between the pretest and posttest for the MBSM of model learning, empathy training and SIMR (t=6.935, p< .01; t=5.898, p< .01; t=9.803, p< .01), and for the SIM of model learning and SIMR (t=4.744, p< .01; t=5.874, p< .01). The F test shows that there is significant difference between the SIM and the MBSM of four groups at the posttest (F=6.67, p< .01; F=9.71, p< .01). The scores of the SIM and the MBSM are the highest at the SIMR group. There is significant difference between the SIM of the SIMR group and the controlled group, the model learning group, and the empathy training group (p< .01; p< .05; p< .01). The SIM of the model learning group was significantly higher than of the controlled group (p< .01). The differences of every comparison group are significant for the MBSM (ps< .01) except the comparison group between the model learning and the empathy training (p> .05). The experimental results show that Self-image of Morality Rehearsal has a significant role in SIM and MBSM. The SIMR is an effective method of model learning.