With the existence of thermodynamic data for a wide range of end-members in rock-forming minerals, thermobarometry now involves combining many equilibria to find the pressure and temperature (P-T) of formation of a rock. We reiterate that this task need only involve an independent set of reactions representing all these equilibria. In finding a P-T of formation, there is an implied displacement of the equilibria to coincide with this P-T. These displacements are mainly made by varying the activities of the end-members of the minerals, in proportion to their uncertainties. As a consequence, the equilibria are constrained to move in a more or less highly correlated way because the equilibria involve overlapping subsets of the end-members. These essential correlations should be included in any thermobarometry calculations. Of the three thermobarometry approaches in use, the TWEEQU approach of Berman (1991), the individual species approach of Gordon (1992), and our average P-T approach, only the last two are optimal on this basis. In addition, such optimal approaches allow P-T, their uncertainties, and a range of diagnostics for outlier identification to be calculated in a computationally straightforward way.