This study compared undergraduate students' learning gains attained through student-led discussions with those acquired through teacher-led discussions in students who had previously engaged in a different learning activity. A total of 305 undergraduate students participated in the study. It was found that prediscussion activities yielded differential student outcomes depending on the applied combination. Self-study groups achieved higher scores through student-led discussions, whereas, lecture groups attained better scores through teacher-led discussions. To substantiate the joint effect, we attempted to elucidate students' outcomes through moment-by-moment analyses of cognitive-behavioral engagement observed during discussions. Students assigned to the self-study and student-led discussion groups displayed consistently high engagement throughout their discussions, however, the engagement dropped sharply in teacher-led discussion. Meanwhile, student engagement increased over time in the teacher-led discussion group after the class listened to a lecture. The results of moderated mediation analysis showed that discussion formats moderated by prediscussion activity affected students' cognitive-behavioral engagement, which led to their learning. Therefore, our findings contribute to a new perspective of engagement as a student learning in discussions and imply that teachers can induce students' engagement leading to superior learning outcome by combining appropriate instructional strategies and classroom discussion.