The goal of this study is to identify the links between metacognitive mediation and students' self-efficacy belief for the performance of complex tasks. We worked for three weeks with six human sciences classes of secondary 1 students who all performed the same tasks. In two classes, metacognition was initiated by the teacher, in two others, activation of metacognition occurred between peers and in the remaining two, there was no metacognitive mediation. Our goal is to compare the three conditions in terms of changes in students' self-efficacy belief through a pretest-posttest comparison among the 85 students who took part in the study. In addition, we interviewed four students in each condition where metacognition was used. The quantitative results showed that self-efficacy improved in classes where metacognition was introduced, which was not the case in the classes where students did not receive any metacognitive prompting. In addition, the comparison of classes "with metacognition" showed that regardless of the agent (the teacher or peers), prompting of metacognition led to improvement in aspects related to active mastery experiences and psychological states of self-efficacy. The difference between these two conditions lay in the strategies available to students to succeed in the tasks.