This study selected the vocabulary teaching session of a demonstration video of a Chinese elementary-level comprehensive course. Based on the framework of multimodal discourse analysis, it used the software ELAN to annotate the multimodal symbols and constructed a multimodal Chinese teaching corpus. Through quantitative and qualitative analysis, this paper explored the use of multimodality in vocabulary teaching, the selection of modalities, and the cooperation relation between different modalities each stage of vocabulary teaching. Through data analysis, it found that during this vocabulary teaching session PPT was the most often used modality; the auditory modality played a major role; the multimodal coordination relation of different stages is diversified in order to adapt different teaching contents and achieve best teaching effect. It further discussed the characteristics and precautions of multimodality use in teaching. This study enriches the research of teaching analysis based on a multimodal corpus and provides reference for carrying out multimodal teaching.