Learning to talk science is an important aspect of learning to do science. Given that scientists' language frequently includes intentions and purposes in explanations of unobservable objects and events, teachers must interpret whether learners' use of such language reflects a scientific understanding or inaccurate anthropomorphism and teleology. In the present study, a framework consisting of three "stances" (Dennett, 1987)--intentional, design and physical--is presented as a powerful tool for analysing students' language use. The aim was to investigate how the framework can be differentiated and used analytically for interpreting students' talk about a molecular process. Semi-structured group discussions and individual interviews about the molecular self-assembly process were conducted with engineering biology/chemistry (n = 15) and biology/chemistry teacher students (n = 6). Qualitative content analysis of transcripts showed that all three stances were employed by students. The analysis also identified subcategories for each stance, and revealed that intentional language with respect to molecular movement and assumptions about design requirements may be potentially problematic areas. Students' exclusion of physical stance explanations may indicate literal anthropomorphic interpretations. Implications for practice include providing teachers with a tool for scaffolding their use of metaphorical language and for supporting students' metacognitive development as scientific language users.