The relationship between language abilities and modeling effects in children were examined. At first subjects performed a pretest after being given an instruction for task structure. Next, they observed a model demonstrating correct responses with an oddity task and then performed an imitation test. This observation-imitation phase was repeated three times. Before this experiment, the subjects' language abilities had been measured by the Illinois Test of Psycholinguistic Abilities. The results indicated that the subjects who were put into the high reception group on the basis of ITPA scores showed better performance in the imitation tests than the subjects in the low reception group. These results were discussed from the view point of subjects' levels of comprehension of the task structure.