Geometrical relationships involving inclusions and partial inclusions in metamorphic microstructures can be inadequate for inferring an order of crystallization and hence a metamorphic reaction. Unique spatial and/or chemical relationships need to be defined for mineral inclusions, in the context of a reference paragenesis, commonly the matrix assemblage. Corona microstructures are reliable indicators of metamorphic reactions, but require considerable care when used to infer reactions or changes in P-T conditions, owing to kinetic problems, as well as to changes in the effective reaction volume during changes across relatively broad P-T stability fields of assemblages. Mineral equilibria models, most commonly implemented through P-T pseudosections, may allow the order in which different minerals become stable along a given P-T path to be inferred. However, the order in which two minerals become stable may be different from the order in which two grains of these minerals nucleate. Furthermore, such diagrams cannot make predictions about which minerals will form porphyroblasts and which minerals will form inclusions in porphyroblasts. An evaluation of three examples from the Australian Proterozoic shows that modelling, in combination with inclusion-host relationships, is a powerful tool for understanding the metamorphic evolution of a rock, but involves considerable uncertainty.