The paper explores one episode from a larger study, where a multilingual student (10-year-old) described her understanding of what makes a shape 2D or 3D. Bakhtin's dialogic theory and Garfinkel's ethnomethodology inform the theoretical framework. Transcribed data of the episode is presented, which is analysed at micro-level and macro-level. The analysis revealed three major findings. First, the analogy of "flat vs fat" for describing shapes may not be useful. Second, the meanings of the terms such as fat are shaped by the interaction of unitary language and heteroglossia. Third, prosody embedded in utterances contribute to the meaning constructions during mathematical discussions. A few implications are presented.