Taking a multimodal perspective, this study analysis the closing section of multi-person interaction in an institutional setting, where Japanese lessons conducted by three Japanese teachers and three Chinese teachers in Chinese universities are involved. Data were collected through the form of audio, video, and transcripation. Results of this study are as follows: (1) the closing section of the Japanese lesson is constituted with two stage structures, that is “pre-closing” and “closing”. Moreover, the components of these two stages are somewhat solidified, (2) in both the “preclosing” and the “closing” stages, the leader is the teacher, who practice a variety of verbal and non-verbal behaviors to guide the interaction with the students until the end of the lesson. Correspondingly, the students use a variety of non-verbal behaviors to indicate their agreements or disagreements with the teacher’s intention to end the lesson, (3) there are two main patterns of procedures used by teachers in the “preclosing” stage. One is creating a negotiation space by which students’ attitudes towards ending the lesson could be confirmed. The other one is to use “discourse marker” to unilaterally terminate the interaction with students, (4) utilizing various “discourse markers”, the “closing” is constructed by teachers into different phases. The teacher can appropriately adjust the progress of the lesson’s “closing” by evaluating students’ non-verbal behaviors in each phase.