I examined how well a self-diagnosis activity engages students in knowledge-integration processes, and its impact on students' mathematical achievements. The self-diagnosis activity requires students to self-diagnose their solutions to problems that they have solved on their own--namely, to identify where they went wrong and to explain the nature of their own errors--and self-score them, aided by a rubric demonstrating how to solve each problem step by step. I also examined knowledge-integration processes and the impact on students' achievements in a traditional activity in which teachers solve, together with their students, problems that students have solved on their own. The two activities can provide students with opportunities to reflect on their own errors, which is assumed to promote learning. Two 5th-grade classes studying simple fractions with the same teacher participated. A pre-test/intervention/post-test design was employed. In the intervention, one class was assigned to the self-diagnosis activity and the other to the traditional activity. Results suggested that at least for a teacher who is not competent in managing argumentative class discussions, the self-diagnosis activity is more effective than the traditional activity in engaging students in knowledge-integration processes and enhancing their achievements. I account for these results and suggest possible directions for future research.