Many studies have previously focused on how people with different levels of expertise solve physics problems. In early work, focus was on characterising differences between experts and novices and a key finding was the central role that propositionally expressed principles and laws play in expert, but not novice, problem-solving. A more recent line of research has focused on characterising continuity between experts and novices at the level of non-propositional knowledge structures and processes such as image-schemas, imagistic simulation and analogical reasoning. This study contributes to an emerging literature addressing the coordination of both propositional and non-propositional knowledge structures and processes in the development of expertise. Specifically, in this paper, we compare problem-solving across two levels of expertise--undergraduate students of chemistry and Ph.D. students in physical chemistry--identifying differences in how conceptual metaphors (CMs) are used (or not) to coordinate propositional and non-propositional knowledge structures in the context of solving problems on entropy. It is hypothesised that the acquisition of expertise involves learning to coordinate the use of CMs to interpret propositional (linguistic and mathematical) knowledge and apply it to specific problem situations. Moreover, we suggest that with increasing expertise, the use of CMs involves a greater degree of subjective engagement with physical entities and processes. Implications for research on learning and instructional practice are discussed.