ABSTRACTMany thermodynamic calculations relating to metamorphic rocks hinge on the thermodynamic parameters of garnet. Though some models are widely used, it is not clear whether their underlying premise is correct: that a single set of equations can be written for the activities of the end‐members of garnet covering the whole compositional range. A voluminous body of data can be used to constrain the thermodynamics of garnet, namely Fe–Mg exchange experimental data involving garnet and another mineral, particularly clinopyroxene, orthopyroxene and olivine. However, examination of these data reveals inconsistencies, apparently stemming from differences between the thermodynamics of low‐Ca and high‐Ca garnets, with a boundary of aboutXgCa= 0.15. In the two regions, for the highP–Tof the experimental data, the thermodynamics follow the regular model, with values for the interaction parameters in the low Ca region of aboutwgFeMg= 50RandwaFe–wgMgCa=– 1300R, in whichRis the gas constant, and in the high Ca region of aboutwgFeMg= 1100RandwgCaFe–wgMgCa=– 2200R. Using the subregular, rather than the regular, model does not remove the discrepancy. The cause of the discrepancy needs to be identified if reliable calculations on roc