The extent to which kinetic barriers to nucleation and growth delay the onset of prograde metamorphic reaction, commonly known as overstepping, is related to the macroscopic driving force for reaction, termed reaction affinity. Reaction affinity is defined in the context of overstepping as the Gibbs free-energy difference between the thermodynamically stable, but not-yet-crystallized, products and the metastable reactants. Mineral reactions which release large quantities of H2O, such as chlorite-consuming reactions, have a higher entropy/volume change, and therefore a higher reaction affinity per unit of temperature/pressure overstep, than those which release little or no H2O. The former are expected to be overstepped in temperature or pressure less than the latter. Different methods of calculating reaction affinity are discussed. Reaction affinity ‘maps’ are calculated that graphically portray variations in reaction affinity on equilibrium phase diagrams, allowing predictions to be made about expected degrees of overstepping. Petrological consequences of variations in reaction affinity include: (i) metamorphic reaction intervals may be discrete rather than continuous, especially in broad multivariant domains across which reaction affinity builds slowly; (ii) reaction intervals may not correspond in a simple way to reaction boundaries and domains in an equilibrium phase diagram, and may involve metastable reactions; (iii) overstepping can lead to a ‘cascade effect’, in which several stable and metastable reactions involving the same reactant phases proceed simultaneously; (iv) fluid generation, and possibly fluid presence in general, may be episodic rather than continuous, corresponding to discrete intervals of reaction; (v) overstepping related to slowly building reaction affinity in multivariant reaction intervals may account for the commonly abrupt development in the field of certain index mineral isograds; and (vi) P–T estimation based on combined use of phase diagram sections and mineral modes/compositions on the one hand, and classical thermobarometry methods on the other, may not agree even if the same thermodynamic data are used. Natural examples of the above, both contact and regional, are provided. The success of the metamorphic facies principle suggests that these kinetic effects are second-order features that operate within a broadly equilibrium approach to metamorphism. However, it may be that the close approach to equilibrium occurs primarily at the boundaries between the metamorphic facies, corresponding to discrete intervals of high entropy, dehydration reaction involving consumption of hydrous phases like chlorite (greenschist–amphibolite facies boundary) and mica (amphibolite–granulite facies boundary), and less so within the facies themselves. The results of this study suggest that it is important to consider the possibility of reactions removed from equilibrium when inferring the P–T–t evolution of metamorphic rocks.