The purpose of this study was to investigate the effect of the jigsaw learning method on children's academic performance and learning attitude. The jigsaw learning requiring children to work together and mutually teach each other in order to have children depend on each other to accomplish their learning goals, used one of small group learning methods based on coopration and peer teaching. In this learning method, classroom children were divided into several jigsaw groups made of five or six children each. Each counterpart group composed of one child drawn from each jigsaw group was given one segment of the learning materials divided into as many parts as the number of counterpart groups (counterpart session). After learning of the counterpart session, children had to return to each jigsaw group and teach one another what they had learned, and hence they learned the total learning materials (jigsaw session). Subjects in the experiment were four hundred and fourty-one fourth, fifth and sixth graders. Subjects were divided at random into the four conditions as a unit of a classroom. The conditions used were as follows: At the first, the traditional whole-class learning method (control condition); second, the jigsaw method taken place for six or seven hours per week for six weeks (short condition); third, the jigsaw method for twelve weeks (long condition); and fourth, the jigsaw method for twelve weeks during which half the members of counterpart and jigsaw groups were changed at the beginning of the seven weeks (rotation condition). Jigsaw and counterpart groups were controlled in terms of sex, the relation of sociometric choice and degree of academic performance, respectively. To testify the jigsaw learning method, the subjects of national language and social studies were used. The results were as follows: (a) As concerns the academic performances of social studies, it was found that the children in short, long and rotation conditions showed significantly more improvement in the performances than did the children in the control one. In particular, the performances of children in rotation condition scored higher than those of children in short and long ones. Further, the performances of children with low performance of the first semester were higher than those of children in middle and high ones. About the academic performances of national language, however, there was no significance between the four conditions.(b) In terms of the frequencies of positive perception responses on the classroom teacher's leadership behavior, classroom children's and self's learning attitude, each positive perception responses about three measures in short, long and rotation conditions were observed more frequently than in the control one. The observed frequencies on each measure in the second session were significantly greater than in the first one. In addition to the above, we divided the positive perception responses on classroom children's and self's learning attitude into the positive learning attitude and the development of socialization responses according to the contents of positive perception responses. At the first and second sessions, the positive learning attitude responses of low performance children were observed more frequently than those of middle and high ones. In terms of the frequencies of development of socialization responses, the observed frequencies of middle performance children at the first session and those of high ones at the second session were significantly greater than those of low ones, respectively.