Teaching and learning new songs in class is an integral cultural practice. This paper focuses on a singing lesson and studies the complex relations among the target song, children’s song acquisition and the teacher’s instructions. However, instead of simply examining pitch accuracy or ratings, this paper proposes a novel methodology for analysing song learning in class. First, a transcription method is used to document a video-recorded lesson. Second, syllables are identified as both the essential components of the grammar of songs for children and the fundamental units for analysing singing in terms of pitch, timing and articulation. The individual singing of three children, divided into the syllables of the target song, is examined for these three properties. Third, a comparison between the song model and the children’s singing reveals how rule-breaking in the song corresponds to the children’s difficulties in singing. This detailed and subject-related analysis exemplifies the complex dynamic among the teacher, the object and the learner and is a contribution to the research of subject-related didactics.