The purpose of this study was to examine the changes and developmental differences of representation in the process of children's motor proficiency, and whether the changes of representation were constrained by developmental stages. The subjects used in this experiment were third, fifth and seventh grade children who were in different developmental stages. They were unable to perform "backward roll on the bar" motor task prior to this experiment. This experiment was composed of three main phases; demonstration VTR, trials and, introspection of representation, and they were repeated. The experiment continued until the subjects were successful in the task, i.e. for 4-7 weeks (22-54 trials in all). After the experiment, the representation of each subject was categorized into visual and motor representation, with the latter further divided into spatial, temporal and kinetic representations. the frequencies and contents of representation were then analyzed. The summary of the results is: the frequencies of motor representation were higher than the frequencies of visual representation, and the older subjects showed them earlier in the process of motor proficiency than the younger one. The frequencies of representation increased with motor proficiency but the contents of representation did not change with proficiency, further, such changes were found to be equal within each grade. But the frequencies and contents of representation in the same motor proficiency were different with each grade. These results suggest that the development of representation is composed of stages, and representation changes continuously within each developmental stage and that these changes are being different according to their respective developmental stages which in turn indicates the existence of constraint based on developmental stages.