We compared the materialization of knowledge integration processes in class discussions that followed troubleshooting (TS) and problem-solving (PS) tasks and examined the impact of these tasks on students' conceptual understanding. The study was conducted in two sixth-grade classes taught by the same teacher, in six lessons that constituted a third of a unit on simple electric circuits. In these lessons, one class was assigned PS tasks where students were asked to solve conceptual problems. Later they were asked to share their work in a class discussion. The other class was assigned TS tasks where students were asked to identify, explain, and correct the mistakes in "teacher-made" erroneous solutions to these same problems. They were also engaged later in a class discussion. We found that students' performance on subsequent transfer problems was significantly higher for the TS class, in particular for students with low prior knowledge. We account for the difference in learning outcomes by the differences in the learning process: the TS tasks elicited more naïve ideas both in students' worksheets as well as in class discussions, and the TS discussions involved more episodes where students developed criteria to discern scientifically acceptable and naive ideas. We also found differences in the format of students' participation, where lower-achieving students participated in the TS discussions. The implications of these findings for future research are discussed.